{"current_steps": 5, "total_steps": 42420, "loss": 10.0039, "lr": 9.42951438000943e-07, "epoch": 0.0023573785950023575, "percentage": 0.01, "elapsed_time": "0:00:01", "remaining_time": "3:46:58", "throughput": 2870.23, "total_tokens": 4608} {"current_steps": 10, "total_steps": 42420, "loss": 8.2661, "lr": 2.1216407355021218e-06, "epoch": 0.004714757190004715, "percentage": 0.02, "elapsed_time": "0:00:02", "remaining_time": "3:02:26", "throughput": 3781.25, "total_tokens": 9760} {"current_steps": 15, "total_steps": 42420, "loss": 5.7706, "lr": 3.3003300330033004e-06, "epoch": 0.007072135785007072, "percentage": 0.04, "elapsed_time": "0:00:03", "remaining_time": "2:48:52", "throughput": 4249.72, "total_tokens": 15232} {"current_steps": 20, "total_steps": 42420, "loss": 4.0092, "lr": 4.479019330504479e-06, "epoch": 0.00942951438000943, "percentage": 0.05, "elapsed_time": "0:00:04", "remaining_time": "2:43:31", "throughput": 4528.77, "total_tokens": 20960} {"current_steps": 25, "total_steps": 42420, "loss": 2.5553, "lr": 5.657708628005658e-06, "epoch": 0.011786892975011787, "percentage": 0.06, "elapsed_time": "0:00:05", "remaining_time": "2:39:11", "throughput": 4698.72, "total_tokens": 26464} {"current_steps": 30, "total_steps": 42420, "loss": 1.7038, "lr": 6.836397925506836e-06, "epoch": 0.014144271570014143, "percentage": 0.07, "elapsed_time": "0:00:06", "remaining_time": "2:38:49", "throughput": 4849.05, "total_tokens": 32704} {"current_steps": 35, "total_steps": 42420, "loss": 1.1179, "lr": 8.015087223008015e-06, "epoch": 0.0165016501650165, "percentage": 0.08, "elapsed_time": "0:00:07", "remaining_time": "2:36:01", "throughput": 4905.57, "total_tokens": 37920} {"current_steps": 40, "total_steps": 42420, "loss": 0.7665, "lr": 9.193776520509194e-06, "epoch": 0.01885902876001886, "percentage": 0.09, "elapsed_time": "0:00:08", "remaining_time": "2:32:55", "throughput": 4947.4, "total_tokens": 42848} {"current_steps": 45, "total_steps": 42420, "loss": 0.4354, "lr": 1.0372465818010372e-05, "epoch": 0.021216407355021217, "percentage": 0.11, "elapsed_time": "0:00:09", "remaining_time": "2:30:51", "throughput": 4990.43, "total_tokens": 47968} {"current_steps": 50, "total_steps": 42420, "loss": 0.4069, "lr": 1.155115511551155e-05, "epoch": 0.023573785950023574, "percentage": 0.12, "elapsed_time": "0:00:10", "remaining_time": "2:29:18", "throughput": 5028.01, "total_tokens": 53152} {"current_steps": 55, "total_steps": 42420, "loss": 0.436, "lr": 1.2729844413012731e-05, "epoch": 0.02593116454502593, "percentage": 0.13, "elapsed_time": "0:00:11", "remaining_time": "2:27:27", "throughput": 5045.08, "total_tokens": 57952} {"current_steps": 60, "total_steps": 42420, "loss": 0.4741, "lr": 1.3908533710513909e-05, "epoch": 0.028288543140028287, "percentage": 0.14, "elapsed_time": "0:00:12", "remaining_time": "2:26:25", "throughput": 5073.37, "total_tokens": 63136} {"current_steps": 65, "total_steps": 42420, "loss": 0.4324, "lr": 1.5087223008015088e-05, "epoch": 0.030645921735030647, "percentage": 0.15, "elapsed_time": "0:00:13", "remaining_time": "2:24:40", "throughput": 5075.5, "total_tokens": 67616} {"current_steps": 70, "total_steps": 42420, "loss": 0.3734, "lr": 1.6265912305516265e-05, "epoch": 0.033003300330033, "percentage": 0.17, "elapsed_time": "0:00:14", "remaining_time": "2:23:07", "throughput": 5079.54, "total_tokens": 72096} {"current_steps": 75, "total_steps": 42420, "loss": 0.3503, "lr": 1.7444601603017444e-05, "epoch": 0.03536067892503536, "percentage": 0.18, "elapsed_time": "0:00:15", "remaining_time": "2:22:10", "throughput": 5089.41, "total_tokens": 76896} {"current_steps": 80, "total_steps": 42420, "loss": 0.3371, "lr": 1.8623290900518624e-05, "epoch": 0.03771805752003772, "percentage": 0.19, "elapsed_time": "0:00:16", "remaining_time": "2:22:11", "throughput": 5107.61, "total_tokens": 82336} {"current_steps": 85, "total_steps": 42420, "loss": 0.3113, "lr": 1.9801980198019803e-05, "epoch": 0.040075436115040074, "percentage": 0.2, "elapsed_time": "0:00:16", "remaining_time": "2:20:50", "throughput": 5108.98, "total_tokens": 86688} {"current_steps": 90, "total_steps": 42420, "loss": 0.5756, "lr": 2.0980669495520982e-05, "epoch": 0.042432814710042434, "percentage": 0.21, "elapsed_time": "0:00:17", "remaining_time": "2:20:08", "throughput": 5117.35, "total_tokens": 91488} {"current_steps": 95, "total_steps": 42420, "loss": 0.4254, "lr": 2.215935879302216e-05, "epoch": 0.04479019330504479, "percentage": 0.22, "elapsed_time": "0:00:18", "remaining_time": "2:19:21", "throughput": 5116.78, "total_tokens": 96032} {"current_steps": 100, "total_steps": 42420, "loss": 0.3342, "lr": 2.333804809052334e-05, "epoch": 0.04714757190004715, "percentage": 0.24, "elapsed_time": "0:00:19", "remaining_time": "2:18:50", "throughput": 5127.12, "total_tokens": 100928} {"current_steps": 105, "total_steps": 42420, "loss": 0.4159, "lr": 2.4516737388024516e-05, "epoch": 0.04950495049504951, "percentage": 0.25, "elapsed_time": "0:00:20", "remaining_time": "2:19:19", "throughput": 5145.06, "total_tokens": 106720} {"current_steps": 110, "total_steps": 42420, "loss": 0.3511, "lr": 2.5695426685525695e-05, "epoch": 0.05186232909005186, "percentage": 0.26, "elapsed_time": "0:00:21", "remaining_time": "2:18:29", "throughput": 5144.6, "total_tokens": 111136} {"current_steps": 115, "total_steps": 42420, "loss": 0.5393, "lr": 2.6874115983026877e-05, "epoch": 0.05421970768505422, "percentage": 0.27, "elapsed_time": "0:00:22", "remaining_time": "2:18:41", "throughput": 5167.58, "total_tokens": 116896} {"current_steps": 120, "total_steps": 42420, "loss": 0.4394, "lr": 2.8052805280528053e-05, "epoch": 0.056577086280056574, "percentage": 0.28, "elapsed_time": "0:00:23", "remaining_time": "2:18:47", "throughput": 5182.74, "total_tokens": 122432} {"current_steps": 125, "total_steps": 42420, "loss": 0.3446, "lr": 2.9231494578029232e-05, "epoch": 0.058934464875058934, "percentage": 0.29, "elapsed_time": "0:00:24", "remaining_time": "2:18:12", "throughput": 5182.14, "total_tokens": 127008} {"current_steps": 130, "total_steps": 42420, "loss": 0.3559, "lr": 3.041018387553041e-05, "epoch": 0.061291843470061294, "percentage": 0.31, "elapsed_time": "0:00:25", "remaining_time": "2:17:59", "throughput": 5186.68, "total_tokens": 132000} {"current_steps": 135, "total_steps": 42420, "loss": 0.3621, "lr": 3.158887317303158e-05, "epoch": 0.06364922206506365, "percentage": 0.32, "elapsed_time": "0:00:26", "remaining_time": "2:17:24", "throughput": 5186.46, "total_tokens": 136512} {"current_steps": 140, "total_steps": 42420, "loss": 0.397, "lr": 3.276756247053277e-05, "epoch": 0.066006600660066, "percentage": 0.33, "elapsed_time": "0:00:27", "remaining_time": "2:17:04", "throughput": 5190.26, "total_tokens": 141344} {"current_steps": 145, "total_steps": 42420, "loss": 0.4877, "lr": 3.394625176803395e-05, "epoch": 0.06836397925506836, "percentage": 0.34, "elapsed_time": "0:00:28", "remaining_time": "2:16:48", "throughput": 5195.41, "total_tokens": 146272} {"current_steps": 150, "total_steps": 42420, "loss": 0.5641, "lr": 3.512494106553513e-05, "epoch": 0.07072135785007072, "percentage": 0.35, "elapsed_time": "0:00:29", "remaining_time": "2:16:56", "throughput": 5201.35, "total_tokens": 151648} {"current_steps": 155, "total_steps": 42420, "loss": 0.3575, "lr": 3.6303630363036307e-05, "epoch": 0.07307873644507308, "percentage": 0.37, "elapsed_time": "0:00:30", "remaining_time": "2:17:03", "throughput": 5211.85, "total_tokens": 157184} {"current_steps": 160, "total_steps": 42420, "loss": 0.3053, "lr": 3.748231966053748e-05, "epoch": 0.07543611504007544, "percentage": 0.38, "elapsed_time": "0:00:30", "remaining_time": "2:16:25", "throughput": 5203.89, "total_tokens": 161280} {"current_steps": 165, "total_steps": 42420, "loss": 0.4987, "lr": 3.866100895803866e-05, "epoch": 0.07779349363507779, "percentage": 0.39, "elapsed_time": "0:00:31", "remaining_time": "2:16:05", "throughput": 5204.38, "total_tokens": 165952} {"current_steps": 170, "total_steps": 42420, "loss": 0.3741, "lr": 3.983969825553984e-05, "epoch": 0.08015087223008015, "percentage": 0.4, "elapsed_time": "0:00:32", "remaining_time": "2:15:45", "throughput": 5202.33, "total_tokens": 170496} {"current_steps": 175, "total_steps": 42420, "loss": 0.3333, "lr": 4.101838755304102e-05, "epoch": 0.08250825082508251, "percentage": 0.41, "elapsed_time": "0:00:33", "remaining_time": "2:15:34", "throughput": 5202.99, "total_tokens": 175328} {"current_steps": 180, "total_steps": 42420, "loss": 0.3026, "lr": 4.21970768505422e-05, "epoch": 0.08486562942008487, "percentage": 0.42, "elapsed_time": "0:00:34", "remaining_time": "2:15:17", "throughput": 5202.44, "total_tokens": 179968} {"current_steps": 185, "total_steps": 42420, "loss": 0.5046, "lr": 4.3375766148043374e-05, "epoch": 0.08722300801508723, "percentage": 0.44, "elapsed_time": "0:00:35", "remaining_time": "2:14:50", "throughput": 5200.13, "total_tokens": 184288} {"current_steps": 190, "total_steps": 42420, "loss": 0.3443, "lr": 4.455445544554455e-05, "epoch": 0.08958038661008957, "percentage": 0.45, "elapsed_time": "0:00:36", "remaining_time": "2:14:53", "throughput": 5206.48, "total_tokens": 189600} {"current_steps": 195, "total_steps": 42420, "loss": 0.5145, "lr": 4.573314474304573e-05, "epoch": 0.09193776520509193, "percentage": 0.46, "elapsed_time": "0:00:37", "remaining_time": "2:14:43", "throughput": 5207.79, "total_tokens": 194400} {"current_steps": 200, "total_steps": 42420, "loss": 0.3713, "lr": 4.691183404054691e-05, "epoch": 0.0942951438000943, "percentage": 0.47, "elapsed_time": "0:00:38", "remaining_time": "2:14:24", "throughput": 5208.2, "total_tokens": 198976} {"current_steps": 205, "total_steps": 42420, "loss": 0.4361, "lr": 4.809052333804809e-05, "epoch": 0.09665252239509665, "percentage": 0.48, "elapsed_time": "0:00:39", "remaining_time": "2:14:15", "throughput": 5210.97, "total_tokens": 203840} {"current_steps": 210, "total_steps": 42420, "loss": 0.3484, "lr": 4.9269212635549277e-05, "epoch": 0.09900990099009901, "percentage": 0.5, "elapsed_time": "0:00:40", "remaining_time": "2:14:01", "throughput": 5208.12, "total_tokens": 208352} {"current_steps": 215, "total_steps": 42420, "loss": 0.4454, "lr": 5.044790193305045e-05, "epoch": 0.10136727958510136, "percentage": 0.51, "elapsed_time": "0:00:40", "remaining_time": "2:14:03", "throughput": 5214.31, "total_tokens": 213664} {"current_steps": 220, "total_steps": 42420, "loss": 0.4614, "lr": 5.162659123055163e-05, "epoch": 0.10372465818010372, "percentage": 0.52, "elapsed_time": "0:00:41", "remaining_time": "2:14:03", "throughput": 5215.94, "total_tokens": 218720} {"current_steps": 225, "total_steps": 42420, "loss": 0.4025, "lr": 5.280528052805281e-05, "epoch": 0.10608203677510608, "percentage": 0.53, "elapsed_time": "0:00:42", "remaining_time": "2:13:47", "throughput": 5214.24, "total_tokens": 223200} {"current_steps": 230, "total_steps": 42420, "loss": 0.4538, "lr": 5.3983969825553986e-05, "epoch": 0.10843941537010844, "percentage": 0.54, "elapsed_time": "0:00:43", "remaining_time": "2:13:53", "throughput": 5218.2, "total_tokens": 228544} {"current_steps": 235, "total_steps": 42420, "loss": 0.4303, "lr": 5.5162659123055165e-05, "epoch": 0.1107967939651108, "percentage": 0.55, "elapsed_time": "0:00:44", "remaining_time": "2:13:45", "throughput": 5218.56, "total_tokens": 233312} {"current_steps": 240, "total_steps": 42420, "loss": 0.4171, "lr": 5.634134842055634e-05, "epoch": 0.11315417256011315, "percentage": 0.57, "elapsed_time": "0:00:45", "remaining_time": "2:13:48", "throughput": 5224.19, "total_tokens": 238656} {"current_steps": 245, "total_steps": 42420, "loss": 0.4568, "lr": 5.752003771805752e-05, "epoch": 0.11551155115511551, "percentage": 0.58, "elapsed_time": "0:00:46", "remaining_time": "2:13:29", "throughput": 5221.86, "total_tokens": 242976} {"current_steps": 250, "total_steps": 42420, "loss": 0.3449, "lr": 5.86987270155587e-05, "epoch": 0.11786892975011787, "percentage": 0.59, "elapsed_time": "0:00:47", "remaining_time": "2:13:27", "throughput": 5222.4, "total_tokens": 247904} {"current_steps": 255, "total_steps": 42420, "loss": 0.3641, "lr": 5.987741631305988e-05, "epoch": 0.12022630834512023, "percentage": 0.6, "elapsed_time": "0:00:48", "remaining_time": "2:13:59", "throughput": 5233.06, "total_tokens": 254432} {"current_steps": 260, "total_steps": 42420, "loss": 0.349, "lr": 6.105610561056106e-05, "epoch": 0.12258368694012259, "percentage": 0.61, "elapsed_time": "0:00:49", "remaining_time": "2:14:05", "throughput": 5236.46, "total_tokens": 259808} {"current_steps": 265, "total_steps": 42420, "loss": 0.9591, "lr": 6.223479490806224e-05, "epoch": 0.12494106553512493, "percentage": 0.62, "elapsed_time": "0:00:50", "remaining_time": "2:14:27", "throughput": 5244.12, "total_tokens": 265952} {"current_steps": 270, "total_steps": 42420, "loss": 0.7705, "lr": 6.341348420556342e-05, "epoch": 0.1272984441301273, "percentage": 0.64, "elapsed_time": "0:00:51", "remaining_time": "2:14:27", "throughput": 5246.52, "total_tokens": 271136} {"current_steps": 275, "total_steps": 42420, "loss": 0.3494, "lr": 6.45921735030646e-05, "epoch": 0.12965582272512965, "percentage": 0.65, "elapsed_time": "0:00:52", "remaining_time": "2:14:38", "throughput": 5251.33, "total_tokens": 276800} {"current_steps": 280, "total_steps": 42420, "loss": 0.4664, "lr": 6.577086280056576e-05, "epoch": 0.132013201320132, "percentage": 0.66, "elapsed_time": "0:00:53", "remaining_time": "2:14:29", "throughput": 5251.25, "total_tokens": 281568} {"current_steps": 285, "total_steps": 42420, "loss": 0.3819, "lr": 6.694955209806696e-05, "epoch": 0.13437057991513437, "percentage": 0.67, "elapsed_time": "0:00:54", "remaining_time": "2:14:24", "throughput": 5251.27, "total_tokens": 286432} {"current_steps": 290, "total_steps": 42420, "loss": 0.4216, "lr": 6.812824139556812e-05, "epoch": 0.13672795851013672, "percentage": 0.68, "elapsed_time": "0:00:55", "remaining_time": "2:14:28", "throughput": 5255.97, "total_tokens": 291904} {"current_steps": 295, "total_steps": 42420, "loss": 0.2964, "lr": 6.930693069306931e-05, "epoch": 0.1390853371051391, "percentage": 0.7, "elapsed_time": "0:00:56", "remaining_time": "2:14:35", "throughput": 5261.0, "total_tokens": 297536} {"current_steps": 300, "total_steps": 42420, "loss": 0.453, "lr": 7.04856199905705e-05, "epoch": 0.14144271570014144, "percentage": 0.71, "elapsed_time": "0:00:57", "remaining_time": "2:14:47", "throughput": 5265.43, "total_tokens": 303296} {"current_steps": 305, "total_steps": 42420, "loss": 0.6835, "lr": 7.166430928807166e-05, "epoch": 0.1438000942951438, "percentage": 0.72, "elapsed_time": "0:00:58", "remaining_time": "2:14:51", "throughput": 5268.19, "total_tokens": 308704} {"current_steps": 310, "total_steps": 42420, "loss": 0.4571, "lr": 7.284299858557285e-05, "epoch": 0.14615747289014616, "percentage": 0.73, "elapsed_time": "0:00:59", "remaining_time": "2:14:26", "throughput": 5262.47, "total_tokens": 312512} {"current_steps": 315, "total_steps": 42420, "loss": 0.3496, "lr": 7.402168788307402e-05, "epoch": 0.1485148514851485, "percentage": 0.74, "elapsed_time": "0:01:00", "remaining_time": "2:14:21", "throughput": 5262.34, "total_tokens": 317376} {"current_steps": 320, "total_steps": 42420, "loss": 0.3271, "lr": 7.520037718057521e-05, "epoch": 0.15087223008015088, "percentage": 0.75, "elapsed_time": "0:01:01", "remaining_time": "2:14:57", "throughput": 5272.27, "total_tokens": 324512} {"current_steps": 325, "total_steps": 42420, "loss": 0.3412, "lr": 7.637906647807638e-05, "epoch": 0.15322960867515323, "percentage": 0.77, "elapsed_time": "0:01:02", "remaining_time": "2:14:49", "throughput": 5271.06, "total_tokens": 329216} {"current_steps": 330, "total_steps": 42420, "loss": 0.3748, "lr": 7.755775577557755e-05, "epoch": 0.15558698727015557, "percentage": 0.78, "elapsed_time": "0:01:03", "remaining_time": "2:14:33", "throughput": 5268.97, "total_tokens": 333536} {"current_steps": 335, "total_steps": 42420, "loss": 0.3804, "lr": 7.873644507307875e-05, "epoch": 0.15794436586515795, "percentage": 0.79, "elapsed_time": "0:01:04", "remaining_time": "2:14:48", "throughput": 5271.95, "total_tokens": 339424} {"current_steps": 340, "total_steps": 42420, "loss": 0.4602, "lr": 7.991513437057991e-05, "epoch": 0.1603017444601603, "percentage": 0.8, "elapsed_time": "0:01:05", "remaining_time": "2:14:36", "throughput": 5270.9, "total_tokens": 343968} {"current_steps": 345, "total_steps": 42420, "loss": 0.3956, "lr": 8.10938236680811e-05, "epoch": 0.16265912305516267, "percentage": 0.81, "elapsed_time": "0:01:06", "remaining_time": "2:14:42", "throughput": 5275.37, "total_tokens": 349632} {"current_steps": 350, "total_steps": 42420, "loss": 0.4638, "lr": 8.227251296558227e-05, "epoch": 0.16501650165016502, "percentage": 0.83, "elapsed_time": "0:01:07", "remaining_time": "2:14:41", "throughput": 5277.99, "total_tokens": 354880} {"current_steps": 355, "total_steps": 42420, "loss": 0.3612, "lr": 8.345120226308345e-05, "epoch": 0.16737388024516736, "percentage": 0.84, "elapsed_time": "0:01:08", "remaining_time": "2:14:21", "throughput": 5271.51, "total_tokens": 358624} {"current_steps": 360, "total_steps": 42420, "loss": 0.3684, "lr": 8.462989156058463e-05, "epoch": 0.16973125884016974, "percentage": 0.85, "elapsed_time": "0:01:08", "remaining_time": "2:14:12", "throughput": 5269.68, "total_tokens": 363200} {"current_steps": 365, "total_steps": 42420, "loss": 0.3841, "lr": 8.580858085808581e-05, "epoch": 0.17208863743517208, "percentage": 0.86, "elapsed_time": "0:01:09", "remaining_time": "2:14:12", "throughput": 5270.6, "total_tokens": 368352} {"current_steps": 370, "total_steps": 42420, "loss": 0.3386, "lr": 8.6987270155587e-05, "epoch": 0.17444601603017446, "percentage": 0.87, "elapsed_time": "0:01:10", "remaining_time": "2:14:16", "throughput": 5275.0, "total_tokens": 373952} {"current_steps": 375, "total_steps": 42420, "loss": 0.2795, "lr": 8.816595945308817e-05, "epoch": 0.1768033946251768, "percentage": 0.88, "elapsed_time": "0:01:11", "remaining_time": "2:14:10", "throughput": 5273.3, "total_tokens": 378624} {"current_steps": 380, "total_steps": 42420, "loss": 0.3299, "lr": 8.934464875058935e-05, "epoch": 0.17916077322017915, "percentage": 0.9, "elapsed_time": "0:01:13", "remaining_time": "2:14:42", "throughput": 5282.79, "total_tokens": 385952} {"current_steps": 385, "total_steps": 42420, "loss": 0.5293, "lr": 9.052333804809052e-05, "epoch": 0.18151815181518152, "percentage": 0.91, "elapsed_time": "0:01:14", "remaining_time": "2:14:44", "throughput": 5284.24, "total_tokens": 391296} {"current_steps": 390, "total_steps": 42420, "loss": 0.3825, "lr": 9.17020273455917e-05, "epoch": 0.18387553041018387, "percentage": 0.92, "elapsed_time": "0:01:15", "remaining_time": "2:14:48", "throughput": 5286.93, "total_tokens": 396800} {"current_steps": 395, "total_steps": 42420, "loss": 0.2793, "lr": 9.288071664309288e-05, "epoch": 0.18623290900518624, "percentage": 0.93, "elapsed_time": "0:01:15", "remaining_time": "2:14:43", "throughput": 5285.84, "total_tokens": 401600} {"current_steps": 400, "total_steps": 42420, "loss": 0.4471, "lr": 9.405940594059406e-05, "epoch": 0.1885902876001886, "percentage": 0.94, "elapsed_time": "0:01:16", "remaining_time": "2:14:43", "throughput": 5286.32, "total_tokens": 406784} {"current_steps": 405, "total_steps": 42420, "loss": 0.4286, "lr": 9.523809523809524e-05, "epoch": 0.19094766619519093, "percentage": 0.95, "elapsed_time": "0:01:17", "remaining_time": "2:14:45", "throughput": 5285.53, "total_tokens": 411936} {"current_steps": 410, "total_steps": 42420, "loss": 0.3427, "lr": 9.641678453559642e-05, "epoch": 0.1933050447901933, "percentage": 0.97, "elapsed_time": "0:01:18", "remaining_time": "2:14:42", "throughput": 5287.12, "total_tokens": 417056} {"current_steps": 415, "total_steps": 42420, "loss": 0.2254, "lr": 9.75954738330976e-05, "epoch": 0.19566242338519566, "percentage": 0.98, "elapsed_time": "0:01:19", "remaining_time": "2:14:45", "throughput": 5288.07, "total_tokens": 422432} {"current_steps": 420, "total_steps": 42420, "loss": 0.4173, "lr": 9.877416313059878e-05, "epoch": 0.19801980198019803, "percentage": 0.99, "elapsed_time": "0:01:20", "remaining_time": "2:14:42", "throughput": 5289.33, "total_tokens": 427520} {"current_steps": 425, "total_steps": 42420, "loss": 0.1681, "lr": 9.995285242809996e-05, "epoch": 0.20037718057520038, "percentage": 1.0, "elapsed_time": "0:01:22", "remaining_time": "2:15:06", "throughput": 5297.25, "total_tokens": 434560} {"current_steps": 430, "total_steps": 42420, "loss": 0.1926, "lr": 0.00010113154172560112, "epoch": 0.20273455917020272, "percentage": 1.01, "elapsed_time": "0:01:23", "remaining_time": "2:15:08", "throughput": 5299.12, "total_tokens": 440000} {"current_steps": 435, "total_steps": 42420, "loss": 0.32, "lr": 0.00010231023102310232, "epoch": 0.2050919377652051, "percentage": 1.03, "elapsed_time": "0:01:23", "remaining_time": "2:15:04", "throughput": 5299.59, "total_tokens": 445024} {"current_steps": 440, "total_steps": 42420, "loss": 0.2086, "lr": 0.0001034889203206035, "epoch": 0.20744931636020744, "percentage": 1.04, "elapsed_time": "0:01:25", "remaining_time": "2:15:14", "throughput": 5302.64, "total_tokens": 450976} {"current_steps": 445, "total_steps": 42420, "loss": 0.4404, "lr": 0.00010466760961810467, "epoch": 0.20980669495520982, "percentage": 1.05, "elapsed_time": "0:01:26", "remaining_time": "2:15:25", "throughput": 5305.71, "total_tokens": 457056} {"current_steps": 450, "total_steps": 42420, "loss": 0.2701, "lr": 0.00010584629891560585, "epoch": 0.21216407355021216, "percentage": 1.06, "elapsed_time": "0:01:27", "remaining_time": "2:15:21", "throughput": 5305.8, "total_tokens": 462016} {"current_steps": 455, "total_steps": 42420, "loss": 0.1272, "lr": 0.00010702498821310702, "epoch": 0.2145214521452145, "percentage": 1.07, "elapsed_time": "0:01:28", "remaining_time": "2:15:23", "throughput": 5307.27, "total_tokens": 467456} {"current_steps": 460, "total_steps": 42420, "loss": 0.5032, "lr": 0.00010820367751060821, "epoch": 0.21687883074021688, "percentage": 1.08, "elapsed_time": "0:01:28", "remaining_time": "2:15:07", "throughput": 5303.03, "total_tokens": 471360} {"current_steps": 465, "total_steps": 42420, "loss": 0.287, "lr": 0.00010938236680810938, "epoch": 0.21923620933521923, "percentage": 1.1, "elapsed_time": "0:01:29", "remaining_time": "2:15:13", "throughput": 5305.93, "total_tokens": 477152} {"current_steps": 470, "total_steps": 42420, "loss": 0.5424, "lr": 0.00011056105610561057, "epoch": 0.2215935879302216, "percentage": 1.11, "elapsed_time": "0:01:30", "remaining_time": "2:15:18", "throughput": 5309.03, "total_tokens": 482880} {"current_steps": 475, "total_steps": 42420, "loss": 0.2418, "lr": 0.00011173974540311175, "epoch": 0.22395096652522395, "percentage": 1.12, "elapsed_time": "0:01:31", "remaining_time": "2:15:04", "throughput": 5305.76, "total_tokens": 486944} {"current_steps": 480, "total_steps": 42420, "loss": 0.2452, "lr": 0.00011291843470061291, "epoch": 0.2263083451202263, "percentage": 1.13, "elapsed_time": "0:01:32", "remaining_time": "2:15:05", "throughput": 5306.41, "total_tokens": 492256} {"current_steps": 485, "total_steps": 42420, "loss": 0.2177, "lr": 0.0001140971239981141, "epoch": 0.22866572371522867, "percentage": 1.14, "elapsed_time": "0:01:33", "remaining_time": "2:15:05", "throughput": 5307.43, "total_tokens": 497568} {"current_steps": 490, "total_steps": 42420, "loss": 0.2655, "lr": 0.00011527581329561527, "epoch": 0.23102310231023102, "percentage": 1.16, "elapsed_time": "0:01:34", "remaining_time": "2:15:00", "throughput": 5306.94, "total_tokens": 502400} {"current_steps": 495, "total_steps": 42420, "loss": 0.2513, "lr": 0.00011645450259311646, "epoch": 0.2333804809052334, "percentage": 1.17, "elapsed_time": "0:01:35", "remaining_time": "2:15:06", "throughput": 5309.06, "total_tokens": 508160} {"current_steps": 500, "total_steps": 42420, "loss": 0.3417, "lr": 0.00011763319189061763, "epoch": 0.23573785950023574, "percentage": 1.18, "elapsed_time": "0:01:36", "remaining_time": "2:14:59", "throughput": 5307.04, "total_tokens": 512704} {"current_steps": 505, "total_steps": 42420, "loss": 0.2396, "lr": 0.00011881188118811881, "epoch": 0.23809523809523808, "percentage": 1.19, "elapsed_time": "0:01:37", "remaining_time": "2:14:52", "throughput": 5306.21, "total_tokens": 517376} {"current_steps": 510, "total_steps": 42420, "loss": 0.2225, "lr": 0.00011999057048562, "epoch": 0.24045261669024046, "percentage": 1.2, "elapsed_time": "0:01:38", "remaining_time": "2:14:52", "throughput": 5306.4, "total_tokens": 522560} {"current_steps": 515, "total_steps": 42420, "loss": 0.2145, "lr": 0.00012116925978312117, "epoch": 0.2428099952852428, "percentage": 1.21, "elapsed_time": "0:01:39", "remaining_time": "2:14:44", "throughput": 5304.21, "total_tokens": 527008} {"current_steps": 520, "total_steps": 42420, "loss": 0.2013, "lr": 0.00012234794908062236, "epoch": 0.24516737388024518, "percentage": 1.23, "elapsed_time": "0:01:40", "remaining_time": "2:14:48", "throughput": 5305.36, "total_tokens": 532544} {"current_steps": 525, "total_steps": 42420, "loss": 0.2378, "lr": 0.00012352663837812353, "epoch": 0.24752475247524752, "percentage": 1.24, "elapsed_time": "0:01:41", "remaining_time": "2:14:41", "throughput": 5304.13, "total_tokens": 537184} {"current_steps": 530, "total_steps": 42420, "loss": 0.603, "lr": 0.0001247053276756247, "epoch": 0.24988213107024987, "percentage": 1.25, "elapsed_time": "0:01:42", "remaining_time": "2:14:44", "throughput": 5305.64, "total_tokens": 542688} {"current_steps": 535, "total_steps": 42420, "loss": 0.3207, "lr": 0.00012588401697312588, "epoch": 0.2522395096652522, "percentage": 1.26, "elapsed_time": "0:01:43", "remaining_time": "2:14:33", "throughput": 5303.19, "total_tokens": 546912} {"current_steps": 540, "total_steps": 42420, "loss": 0.2114, "lr": 0.00012706270627062708, "epoch": 0.2545968882602546, "percentage": 1.27, "elapsed_time": "0:01:43", "remaining_time": "2:14:22", "throughput": 5300.22, "total_tokens": 550976} {"current_steps": 545, "total_steps": 42420, "loss": 0.2551, "lr": 0.00012824139556812824, "epoch": 0.25695426685525696, "percentage": 1.28, "elapsed_time": "0:01:45", "remaining_time": "2:14:30", "throughput": 5303.02, "total_tokens": 557024} {"current_steps": 550, "total_steps": 42420, "loss": 0.3513, "lr": 0.00012942008486562943, "epoch": 0.2593116454502593, "percentage": 1.3, "elapsed_time": "0:01:46", "remaining_time": "2:14:39", "throughput": 5304.67, "total_tokens": 563008} {"current_steps": 555, "total_steps": 42420, "loss": 0.0849, "lr": 0.0001305987741631306, "epoch": 0.26166902404526166, "percentage": 1.31, "elapsed_time": "0:01:47", "remaining_time": "2:14:41", "throughput": 5306.35, "total_tokens": 568512} {"current_steps": 560, "total_steps": 42420, "loss": 0.2114, "lr": 0.00013177746346063177, "epoch": 0.264026402640264, "percentage": 1.32, "elapsed_time": "0:01:48", "remaining_time": "2:14:39", "throughput": 5305.66, "total_tokens": 573504} {"current_steps": 565, "total_steps": 42420, "loss": 0.1594, "lr": 0.00013295615275813296, "epoch": 0.2663837812352664, "percentage": 1.33, "elapsed_time": "0:01:49", "remaining_time": "2:14:38", "throughput": 5306.42, "total_tokens": 578656} {"current_steps": 570, "total_steps": 42420, "loss": 0.3268, "lr": 0.00013413484205563415, "epoch": 0.26874115983026875, "percentage": 1.34, "elapsed_time": "0:01:49", "remaining_time": "2:14:30", "throughput": 5305.58, "total_tokens": 583168} {"current_steps": 575, "total_steps": 42420, "loss": 0.1603, "lr": 0.00013531353135313532, "epoch": 0.2710985384252711, "percentage": 1.36, "elapsed_time": "0:01:50", "remaining_time": "2:14:24", "throughput": 5304.4, "total_tokens": 587808} {"current_steps": 580, "total_steps": 42420, "loss": 0.114, "lr": 0.00013649222065063648, "epoch": 0.27345591702027344, "percentage": 1.37, "elapsed_time": "0:01:51", "remaining_time": "2:14:21", "throughput": 5303.17, "total_tokens": 592672} {"current_steps": 585, "total_steps": 42420, "loss": 0.4821, "lr": 0.00013767090994813767, "epoch": 0.2758132956152758, "percentage": 1.38, "elapsed_time": "0:01:52", "remaining_time": "2:14:20", "throughput": 5303.9, "total_tokens": 597792} {"current_steps": 590, "total_steps": 42420, "loss": 0.2133, "lr": 0.00013884959924563887, "epoch": 0.2781706742102782, "percentage": 1.39, "elapsed_time": "0:01:53", "remaining_time": "2:14:25", "throughput": 5305.41, "total_tokens": 603584} {"current_steps": 595, "total_steps": 42420, "loss": 0.0072, "lr": 0.00014002828854314003, "epoch": 0.28052805280528054, "percentage": 1.4, "elapsed_time": "0:01:54", "remaining_time": "2:14:29", "throughput": 5306.31, "total_tokens": 609120} {"current_steps": 600, "total_steps": 42420, "loss": 0.0764, "lr": 0.0001412069778406412, "epoch": 0.2828854314002829, "percentage": 1.41, "elapsed_time": "0:01:55", "remaining_time": "2:14:37", "throughput": 5308.62, "total_tokens": 615232} {"current_steps": 605, "total_steps": 42420, "loss": 0.3986, "lr": 0.0001423856671381424, "epoch": 0.28524280999528523, "percentage": 1.43, "elapsed_time": "0:01:56", "remaining_time": "2:14:35", "throughput": 5308.63, "total_tokens": 620224} {"current_steps": 610, "total_steps": 42420, "loss": 0.1338, "lr": 0.00014356435643564356, "epoch": 0.2876001885902876, "percentage": 1.44, "elapsed_time": "0:01:57", "remaining_time": "2:14:33", "throughput": 5308.39, "total_tokens": 625312} {"current_steps": 615, "total_steps": 42420, "loss": 0.2108, "lr": 0.00014474304573314475, "epoch": 0.28995756718529, "percentage": 1.45, "elapsed_time": "0:01:58", "remaining_time": "2:14:30", "throughput": 5307.3, "total_tokens": 630080} {"current_steps": 620, "total_steps": 42420, "loss": 9.0995, "lr": 0.00014592173503064594, "epoch": 0.2923149457802923, "percentage": 1.46, "elapsed_time": "0:01:59", "remaining_time": "2:14:24", "throughput": 5307.31, "total_tokens": 634880} {"current_steps": 625, "total_steps": 42420, "loss": 13.3552, "lr": 0.0001471004243281471, "epoch": 0.29467232437529467, "percentage": 1.47, "elapsed_time": "0:02:00", "remaining_time": "2:14:19", "throughput": 5307.03, "total_tokens": 639616} {"current_steps": 630, "total_steps": 42420, "loss": 13.1909, "lr": 0.00014827911362564827, "epoch": 0.297029702970297, "percentage": 1.49, "elapsed_time": "0:02:01", "remaining_time": "2:14:11", "throughput": 5306.82, "total_tokens": 644128} {"current_steps": 635, "total_steps": 42420, "loss": 10.8986, "lr": 0.00014945780292314944, "epoch": 0.29938708156529936, "percentage": 1.5, "elapsed_time": "0:02:02", "remaining_time": "2:14:08", "throughput": 5307.35, "total_tokens": 649152} {"current_steps": 640, "total_steps": 42420, "loss": 9.9493, "lr": 0.00015063649222065066, "epoch": 0.30174446016030176, "percentage": 1.51, "elapsed_time": "0:02:03", "remaining_time": "2:14:03", "throughput": 5307.45, "total_tokens": 653952} {"current_steps": 645, "total_steps": 42420, "loss": 8.6485, "lr": 0.00015181518151815182, "epoch": 0.3041018387553041, "percentage": 1.52, "elapsed_time": "0:02:04", "remaining_time": "2:14:00", "throughput": 5307.33, "total_tokens": 658880} {"current_steps": 650, "total_steps": 42420, "loss": 7.0079, "lr": 0.000152993870815653, "epoch": 0.30645921735030646, "percentage": 1.53, "elapsed_time": "0:02:05", "remaining_time": "2:14:03", "throughput": 5310.27, "total_tokens": 664672} {"current_steps": 655, "total_steps": 42420, "loss": 5.7073, "lr": 0.00015417256011315418, "epoch": 0.3088165959453088, "percentage": 1.54, "elapsed_time": "0:02:06", "remaining_time": "2:13:55", "throughput": 5308.86, "total_tokens": 669024} {"current_steps": 660, "total_steps": 42420, "loss": 5.0042, "lr": 0.00015535124941065535, "epoch": 0.31117397454031115, "percentage": 1.56, "elapsed_time": "0:02:06", "remaining_time": "2:13:52", "throughput": 5308.81, "total_tokens": 673920} {"current_steps": 665, "total_steps": 42420, "loss": 4.9141, "lr": 0.00015652993870815654, "epoch": 0.31353135313531355, "percentage": 1.57, "elapsed_time": "0:02:07", "remaining_time": "2:13:47", "throughput": 5308.71, "total_tokens": 678688} {"current_steps": 670, "total_steps": 42420, "loss": 4.4488, "lr": 0.0001577086280056577, "epoch": 0.3158887317303159, "percentage": 1.58, "elapsed_time": "0:02:08", "remaining_time": "2:13:38", "throughput": 5307.97, "total_tokens": 683072} {"current_steps": 675, "total_steps": 42420, "loss": 4.0059, "lr": 0.0001588873173031589, "epoch": 0.31824611032531824, "percentage": 1.59, "elapsed_time": "0:02:09", "remaining_time": "2:13:33", "throughput": 5308.0, "total_tokens": 687776} {"current_steps": 680, "total_steps": 42420, "loss": 3.5623, "lr": 0.00016006600660066006, "epoch": 0.3206034889203206, "percentage": 1.6, "elapsed_time": "0:02:10", "remaining_time": "2:13:27", "throughput": 5307.21, "total_tokens": 692320} {"current_steps": 685, "total_steps": 42420, "loss": 2.9386, "lr": 0.00016124469589816123, "epoch": 0.32296086751532294, "percentage": 1.61, "elapsed_time": "0:02:11", "remaining_time": "2:13:23", "throughput": 5309.12, "total_tokens": 697440} {"current_steps": 690, "total_steps": 42420, "loss": 2.7172, "lr": 0.00016242338519566245, "epoch": 0.32531824611032534, "percentage": 1.63, "elapsed_time": "0:02:12", "remaining_time": "2:13:35", "throughput": 5313.73, "total_tokens": 704256} {"current_steps": 695, "total_steps": 42420, "loss": 1.907, "lr": 0.00016360207449316361, "epoch": 0.3276756247053277, "percentage": 1.64, "elapsed_time": "0:02:13", "remaining_time": "2:13:33", "throughput": 5314.7, "total_tokens": 709376} {"current_steps": 700, "total_steps": 42420, "loss": 2.226, "lr": 0.00016478076379066478, "epoch": 0.33003300330033003, "percentage": 1.65, "elapsed_time": "0:02:14", "remaining_time": "2:13:26", "throughput": 5313.88, "total_tokens": 713824} {"current_steps": 705, "total_steps": 42420, "loss": 0.8771, "lr": 0.00016595945308816595, "epoch": 0.3323903818953324, "percentage": 1.66, "elapsed_time": "0:02:15", "remaining_time": "2:13:26", "throughput": 5315.52, "total_tokens": 719264} {"current_steps": 710, "total_steps": 42420, "loss": 0.7023, "lr": 0.00016713814238566714, "epoch": 0.3347477604903347, "percentage": 1.67, "elapsed_time": "0:02:16", "remaining_time": "2:13:24", "throughput": 5317.13, "total_tokens": 724512} {"current_steps": 715, "total_steps": 42420, "loss": 0.5576, "lr": 0.00016831683168316833, "epoch": 0.3371051390853371, "percentage": 1.69, "elapsed_time": "0:02:17", "remaining_time": "2:13:20", "throughput": 5318.28, "total_tokens": 729504} {"current_steps": 720, "total_steps": 42420, "loss": 0.4639, "lr": 0.0001694955209806695, "epoch": 0.33946251768033947, "percentage": 1.7, "elapsed_time": "0:02:18", "remaining_time": "2:13:14", "throughput": 5317.37, "total_tokens": 733984} {"current_steps": 725, "total_steps": 42420, "loss": 0.3454, "lr": 0.0001706742102781707, "epoch": 0.3418198962753418, "percentage": 1.71, "elapsed_time": "0:02:18", "remaining_time": "2:13:13", "throughput": 5318.43, "total_tokens": 739232} {"current_steps": 730, "total_steps": 42420, "loss": 0.3328, "lr": 0.00017185289957567185, "epoch": 0.34417727487034416, "percentage": 1.72, "elapsed_time": "0:02:19", "remaining_time": "2:13:12", "throughput": 5319.19, "total_tokens": 744416} {"current_steps": 735, "total_steps": 42420, "loss": 0.3668, "lr": 0.00017303158887317302, "epoch": 0.3465346534653465, "percentage": 1.73, "elapsed_time": "0:02:20", "remaining_time": "2:13:12", "throughput": 5321.33, "total_tokens": 749952} {"current_steps": 740, "total_steps": 42420, "loss": 0.4491, "lr": 0.0001742102781706742, "epoch": 0.3488920320603489, "percentage": 1.74, "elapsed_time": "0:02:21", "remaining_time": "2:13:04", "throughput": 5320.13, "total_tokens": 754176} {"current_steps": 745, "total_steps": 42420, "loss": 0.6665, "lr": 0.0001753889674681754, "epoch": 0.35124941065535126, "percentage": 1.76, "elapsed_time": "0:02:22", "remaining_time": "2:12:59", "throughput": 5319.75, "total_tokens": 758848} {"current_steps": 750, "total_steps": 42420, "loss": 0.4134, "lr": 0.00017656765676567657, "epoch": 0.3536067892503536, "percentage": 1.77, "elapsed_time": "0:02:23", "remaining_time": "2:12:56", "throughput": 5320.81, "total_tokens": 763840} {"current_steps": 755, "total_steps": 42420, "loss": 0.5052, "lr": 0.00017774634606317774, "epoch": 0.35596416784535595, "percentage": 1.78, "elapsed_time": "0:02:24", "remaining_time": "2:12:58", "throughput": 5323.18, "total_tokens": 769568} {"current_steps": 760, "total_steps": 42420, "loss": 0.3307, "lr": 0.00017892503536067893, "epoch": 0.3583215464403583, "percentage": 1.79, "elapsed_time": "0:02:25", "remaining_time": "2:13:03", "throughput": 5326.74, "total_tokens": 775840} {"current_steps": 765, "total_steps": 42420, "loss": 0.4993, "lr": 0.00018010372465818012, "epoch": 0.3606789250353607, "percentage": 1.8, "elapsed_time": "0:02:26", "remaining_time": "2:12:55", "throughput": 5324.96, "total_tokens": 779936} {"current_steps": 770, "total_steps": 42420, "loss": 0.6488, "lr": 0.0001812824139556813, "epoch": 0.36303630363036304, "percentage": 1.82, "elapsed_time": "0:02:27", "remaining_time": "2:12:52", "throughput": 5324.73, "total_tokens": 784800} {"current_steps": 775, "total_steps": 42420, "loss": 0.9929, "lr": 0.00018246110325318245, "epoch": 0.3653936822253654, "percentage": 1.83, "elapsed_time": "0:02:28", "remaining_time": "2:12:56", "throughput": 5326.8, "total_tokens": 790688} {"current_steps": 780, "total_steps": 42420, "loss": 1.429, "lr": 0.00018363979255068365, "epoch": 0.36775106082036774, "percentage": 1.84, "elapsed_time": "0:02:29", "remaining_time": "2:13:11", "throughput": 5331.62, "total_tokens": 798144} {"current_steps": 785, "total_steps": 42420, "loss": 0.4839, "lr": 0.0001848184818481848, "epoch": 0.3701084394153701, "percentage": 1.85, "elapsed_time": "0:02:30", "remaining_time": "2:13:04", "throughput": 5330.07, "total_tokens": 802400} {"current_steps": 790, "total_steps": 42420, "loss": 0.5315, "lr": 0.000185997171145686, "epoch": 0.3724658180103725, "percentage": 1.86, "elapsed_time": "0:02:31", "remaining_time": "2:12:58", "throughput": 5330.14, "total_tokens": 807008} {"current_steps": 795, "total_steps": 42420, "loss": 0.3204, "lr": 0.0001871758604431872, "epoch": 0.37482319660537483, "percentage": 1.87, "elapsed_time": "0:02:32", "remaining_time": "2:12:53", "throughput": 5330.36, "total_tokens": 811712} {"current_steps": 800, "total_steps": 42420, "loss": 0.3983, "lr": 0.00018835454974068836, "epoch": 0.3771805752003772, "percentage": 1.89, "elapsed_time": "0:02:33", "remaining_time": "2:12:51", "throughput": 5330.42, "total_tokens": 816800} {"current_steps": 805, "total_steps": 42420, "loss": 0.3992, "lr": 0.00018953323903818953, "epoch": 0.3795379537953795, "percentage": 1.9, "elapsed_time": "0:02:34", "remaining_time": "2:12:53", "throughput": 5332.34, "total_tokens": 822464} {"current_steps": 810, "total_steps": 42420, "loss": 0.4352, "lr": 0.0001907119283356907, "epoch": 0.38189533239038187, "percentage": 1.91, "elapsed_time": "0:02:35", "remaining_time": "2:12:54", "throughput": 5333.22, "total_tokens": 827904} {"current_steps": 815, "total_steps": 42420, "loss": 0.5217, "lr": 0.0001918906176331919, "epoch": 0.38425271098538427, "percentage": 1.92, "elapsed_time": "0:02:36", "remaining_time": "2:12:51", "throughput": 5333.28, "total_tokens": 832768} {"current_steps": 820, "total_steps": 42420, "loss": 0.4634, "lr": 0.00019306930693069308, "epoch": 0.3866100895803866, "percentage": 1.93, "elapsed_time": "0:02:37", "remaining_time": "2:12:47", "throughput": 5332.83, "total_tokens": 837504} {"current_steps": 825, "total_steps": 42420, "loss": 0.3625, "lr": 0.00019424799622819424, "epoch": 0.38896746817538896, "percentage": 1.94, "elapsed_time": "0:02:37", "remaining_time": "2:12:38", "throughput": 5330.89, "total_tokens": 841504} {"current_steps": 830, "total_steps": 42420, "loss": 0.3958, "lr": 0.00019542668552569544, "epoch": 0.3913248467703913, "percentage": 1.96, "elapsed_time": "0:02:38", "remaining_time": "2:12:35", "throughput": 5331.34, "total_tokens": 846464} {"current_steps": 835, "total_steps": 42420, "loss": 0.3812, "lr": 0.0001966053748231966, "epoch": 0.39368222536539366, "percentage": 1.97, "elapsed_time": "0:02:39", "remaining_time": "2:12:34", "throughput": 5332.73, "total_tokens": 851776} {"current_steps": 840, "total_steps": 42420, "loss": 0.3434, "lr": 0.0001977840641206978, "epoch": 0.39603960396039606, "percentage": 1.98, "elapsed_time": "0:02:40", "remaining_time": "2:12:30", "throughput": 5332.54, "total_tokens": 856512} {"current_steps": 845, "total_steps": 42420, "loss": 0.2224, "lr": 0.00019896275341819896, "epoch": 0.3983969825553984, "percentage": 1.99, "elapsed_time": "0:02:41", "remaining_time": "2:12:31", "throughput": 5334.15, "total_tokens": 862080} {"current_steps": 850, "total_steps": 42420, "loss": 1.2703, "lr": 0.00020014144271570015, "epoch": 0.40075436115040075, "percentage": 2.0, "elapsed_time": "0:02:42", "remaining_time": "2:12:34", "throughput": 5335.85, "total_tokens": 867872} {"current_steps": 855, "total_steps": 42420, "loss": 2.386, "lr": 0.00020132013201320132, "epoch": 0.4031117397454031, "percentage": 2.02, "elapsed_time": "0:02:43", "remaining_time": "2:12:29", "throughput": 5335.24, "total_tokens": 872384} {"current_steps": 860, "total_steps": 42420, "loss": 0.5015, "lr": 0.00020249882131070248, "epoch": 0.40546911834040544, "percentage": 2.03, "elapsed_time": "0:02:44", "remaining_time": "2:12:22", "throughput": 5334.02, "total_tokens": 876640} {"current_steps": 865, "total_steps": 42420, "loss": 0.3249, "lr": 0.0002036775106082037, "epoch": 0.40782649693540785, "percentage": 2.04, "elapsed_time": "0:02:45", "remaining_time": "2:12:22", "throughput": 5336.16, "total_tokens": 882240} {"current_steps": 870, "total_steps": 42420, "loss": 0.3597, "lr": 0.00020485619990570487, "epoch": 0.4101838755304102, "percentage": 2.05, "elapsed_time": "0:02:46", "remaining_time": "2:12:21", "throughput": 5337.86, "total_tokens": 887616} {"current_steps": 875, "total_steps": 42420, "loss": 0.4537, "lr": 0.00020603488920320603, "epoch": 0.41254125412541254, "percentage": 2.06, "elapsed_time": "0:02:47", "remaining_time": "2:12:22", "throughput": 5338.69, "total_tokens": 893024} {"current_steps": 880, "total_steps": 42420, "loss": 0.4624, "lr": 0.0002072135785007072, "epoch": 0.4148986327204149, "percentage": 2.07, "elapsed_time": "0:02:48", "remaining_time": "2:12:22", "throughput": 5339.98, "total_tokens": 898464} {"current_steps": 885, "total_steps": 42420, "loss": 0.3331, "lr": 0.0002083922677982084, "epoch": 0.41725601131541723, "percentage": 2.09, "elapsed_time": "0:02:49", "remaining_time": "2:12:16", "throughput": 5338.12, "total_tokens": 902688} {"current_steps": 890, "total_steps": 42420, "loss": 0.4819, "lr": 0.00020957095709570959, "epoch": 0.41961338991041963, "percentage": 2.1, "elapsed_time": "0:02:50", "remaining_time": "2:12:14", "throughput": 5338.44, "total_tokens": 907744} {"current_steps": 895, "total_steps": 42420, "loss": 0.4812, "lr": 0.00021074964639321075, "epoch": 0.421970768505422, "percentage": 2.11, "elapsed_time": "0:02:50", "remaining_time": "2:12:11", "throughput": 5338.2, "total_tokens": 912608} {"current_steps": 900, "total_steps": 42420, "loss": 0.3775, "lr": 0.00021192833569071194, "epoch": 0.4243281471004243, "percentage": 2.12, "elapsed_time": "0:02:51", "remaining_time": "2:12:10", "throughput": 5338.36, "total_tokens": 917632} {"current_steps": 905, "total_steps": 42420, "loss": 0.2862, "lr": 0.0002131070249882131, "epoch": 0.42668552569542667, "percentage": 2.13, "elapsed_time": "0:02:52", "remaining_time": "2:12:04", "throughput": 5337.03, "total_tokens": 921920} {"current_steps": 910, "total_steps": 42420, "loss": 1.3819, "lr": 0.00021428571428571427, "epoch": 0.429042904290429, "percentage": 2.15, "elapsed_time": "0:02:53", "remaining_time": "2:12:01", "throughput": 5337.8, "total_tokens": 926976} {"current_steps": 915, "total_steps": 42420, "loss": 1.5002, "lr": 0.00021546440358321547, "epoch": 0.4314002828854314, "percentage": 2.16, "elapsed_time": "0:02:54", "remaining_time": "2:11:57", "throughput": 5337.9, "total_tokens": 931680} {"current_steps": 920, "total_steps": 42420, "loss": 0.422, "lr": 0.00021664309288071666, "epoch": 0.43375766148043376, "percentage": 2.17, "elapsed_time": "0:02:55", "remaining_time": "2:11:56", "throughput": 5338.76, "total_tokens": 936992} {"current_steps": 925, "total_steps": 42420, "loss": 0.5411, "lr": 0.00021782178217821783, "epoch": 0.4361150400754361, "percentage": 2.18, "elapsed_time": "0:02:56", "remaining_time": "2:11:51", "throughput": 5337.96, "total_tokens": 941408} {"current_steps": 930, "total_steps": 42420, "loss": 0.4124, "lr": 0.000219000471475719, "epoch": 0.43847241867043846, "percentage": 2.19, "elapsed_time": "0:02:57", "remaining_time": "2:11:49", "throughput": 5338.34, "total_tokens": 946464} {"current_steps": 935, "total_steps": 42420, "loss": 0.3652, "lr": 0.00022017916077322018, "epoch": 0.4408297972654408, "percentage": 2.2, "elapsed_time": "0:02:58", "remaining_time": "2:11:47", "throughput": 5338.87, "total_tokens": 951488} {"current_steps": 940, "total_steps": 42420, "loss": 0.3534, "lr": 0.00022135785007072138, "epoch": 0.4431871758604432, "percentage": 2.22, "elapsed_time": "0:02:59", "remaining_time": "2:11:44", "throughput": 5339.01, "total_tokens": 956416} {"current_steps": 945, "total_steps": 42420, "loss": 0.457, "lr": 0.00022253653936822254, "epoch": 0.44554455445544555, "percentage": 2.23, "elapsed_time": "0:03:00", "remaining_time": "2:11:45", "throughput": 5340.69, "total_tokens": 961984} {"current_steps": 950, "total_steps": 42420, "loss": 0.6354, "lr": 0.0002237152286657237, "epoch": 0.4479019330504479, "percentage": 2.24, "elapsed_time": "0:03:01", "remaining_time": "2:11:58", "throughput": 5344.65, "total_tokens": 969504} {"current_steps": 955, "total_steps": 42420, "loss": 0.3561, "lr": 0.0002248939179632249, "epoch": 0.45025931164545024, "percentage": 2.25, "elapsed_time": "0:03:02", "remaining_time": "2:11:55", "throughput": 5344.85, "total_tokens": 974368} {"current_steps": 960, "total_steps": 42420, "loss": 0.3256, "lr": 0.00022607260726072607, "epoch": 0.4526166902404526, "percentage": 2.26, "elapsed_time": "0:03:03", "remaining_time": "2:11:51", "throughput": 5345.23, "total_tokens": 979168} {"current_steps": 965, "total_steps": 42420, "loss": 0.3572, "lr": 0.00022725129655822726, "epoch": 0.454974068835455, "percentage": 2.27, "elapsed_time": "0:03:04", "remaining_time": "2:11:46", "throughput": 5344.5, "total_tokens": 983616} {"current_steps": 970, "total_steps": 42420, "loss": 0.2827, "lr": 0.00022842998585572845, "epoch": 0.45733144743045734, "percentage": 2.29, "elapsed_time": "0:03:04", "remaining_time": "2:11:42", "throughput": 5344.02, "total_tokens": 988224} {"current_steps": 975, "total_steps": 42420, "loss": 0.4548, "lr": 0.00022960867515322962, "epoch": 0.4596888260254597, "percentage": 2.3, "elapsed_time": "0:03:05", "remaining_time": "2:11:37", "throughput": 5343.36, "total_tokens": 992704} {"current_steps": 980, "total_steps": 42420, "loss": 0.4035, "lr": 0.00023078736445073078, "epoch": 0.46204620462046203, "percentage": 2.31, "elapsed_time": "0:03:06", "remaining_time": "2:11:35", "throughput": 5343.43, "total_tokens": 997664} {"current_steps": 985, "total_steps": 42420, "loss": 0.3721, "lr": 0.00023196605374823195, "epoch": 0.4644035832154644, "percentage": 2.32, "elapsed_time": "0:03:07", "remaining_time": "2:11:32", "throughput": 5343.99, "total_tokens": 1002688} {"current_steps": 990, "total_steps": 42420, "loss": 0.3156, "lr": 0.00023314474304573317, "epoch": 0.4667609618104668, "percentage": 2.33, "elapsed_time": "0:03:08", "remaining_time": "2:11:38", "throughput": 5346.99, "total_tokens": 1009216} {"current_steps": 995, "total_steps": 42420, "loss": 0.3762, "lr": 0.00023432343234323433, "epoch": 0.4691183404054691, "percentage": 2.35, "elapsed_time": "0:03:09", "remaining_time": "2:11:34", "throughput": 5346.49, "total_tokens": 1013856} {"current_steps": 1000, "total_steps": 42420, "loss": 0.3464, "lr": 0.0002355021216407355, "epoch": 0.47147571900047147, "percentage": 2.36, "elapsed_time": "0:03:10", "remaining_time": "2:11:31", "throughput": 5345.88, "total_tokens": 1018560} {"current_steps": 1005, "total_steps": 42420, "loss": 0.5504, "lr": 0.0002366808109382367, "epoch": 0.4738330975954738, "percentage": 2.37, "elapsed_time": "0:03:11", "remaining_time": "2:11:28", "throughput": 5345.68, "total_tokens": 1023264} {"current_steps": 1010, "total_steps": 42420, "loss": 0.4903, "lr": 0.00023785950023573786, "epoch": 0.47619047619047616, "percentage": 2.38, "elapsed_time": "0:03:12", "remaining_time": "2:11:27", "throughput": 5346.38, "total_tokens": 1028480} {"current_steps": 1015, "total_steps": 42420, "loss": 0.6117, "lr": 0.00023903818953323905, "epoch": 0.47854785478547857, "percentage": 2.39, "elapsed_time": "0:03:13", "remaining_time": "2:11:26", "throughput": 5347.11, "total_tokens": 1033696} {"current_steps": 1020, "total_steps": 42420, "loss": 0.6776, "lr": 0.00024021687883074021, "epoch": 0.4809052333804809, "percentage": 2.4, "elapsed_time": "0:03:14", "remaining_time": "2:11:25", "throughput": 5347.82, "total_tokens": 1038944} {"current_steps": 1025, "total_steps": 42420, "loss": 0.3072, "lr": 0.0002413955681282414, "epoch": 0.48326261197548326, "percentage": 2.42, "elapsed_time": "0:03:15", "remaining_time": "2:11:19", "throughput": 5346.83, "total_tokens": 1043168} {"current_steps": 1030, "total_steps": 42420, "loss": 0.303, "lr": 0.00024257425742574257, "epoch": 0.4856199905704856, "percentage": 2.43, "elapsed_time": "0:03:16", "remaining_time": "2:11:19", "throughput": 5347.76, "total_tokens": 1048544} {"current_steps": 1035, "total_steps": 42420, "loss": 0.5621, "lr": 0.00024375294672324374, "epoch": 0.48797736916548795, "percentage": 2.44, "elapsed_time": "0:03:16", "remaining_time": "2:11:16", "throughput": 5348.13, "total_tokens": 1053472} {"current_steps": 1040, "total_steps": 42420, "loss": 0.3903, "lr": 0.00024493163602074496, "epoch": 0.49033474776049035, "percentage": 2.45, "elapsed_time": "0:03:17", "remaining_time": "2:11:14", "throughput": 5349.26, "total_tokens": 1058688} {"current_steps": 1045, "total_steps": 42420, "loss": 0.3808, "lr": 0.0002461103253182461, "epoch": 0.4926921263554927, "percentage": 2.46, "elapsed_time": "0:03:18", "remaining_time": "2:11:08", "throughput": 5348.14, "total_tokens": 1062912} {"current_steps": 1050, "total_steps": 42420, "loss": 0.3494, "lr": 0.0002472890146157473, "epoch": 0.49504950495049505, "percentage": 2.48, "elapsed_time": "0:03:19", "remaining_time": "2:11:02", "throughput": 5347.51, "total_tokens": 1067168} {"current_steps": 1055, "total_steps": 42420, "loss": 0.4335, "lr": 0.0002484677039132485, "epoch": 0.4974068835454974, "percentage": 2.49, "elapsed_time": "0:03:20", "remaining_time": "2:11:07", "throughput": 5349.94, "total_tokens": 1073536} {"current_steps": 1060, "total_steps": 42420, "loss": 0.4044, "lr": 0.0002496463932107496, "epoch": 0.49976426214049974, "percentage": 2.5, "elapsed_time": "0:03:21", "remaining_time": "2:11:06", "throughput": 5350.73, "total_tokens": 1078784} {"current_steps": 1065, "total_steps": 42420, "loss": 0.3578, "lr": 0.0002508250825082508, "epoch": 0.5021216407355021, "percentage": 2.51, "elapsed_time": "0:03:22", "remaining_time": "2:11:04", "throughput": 5351.02, "total_tokens": 1083808} {"current_steps": 1070, "total_steps": 42420, "loss": 0.5309, "lr": 0.000252003771805752, "epoch": 0.5044790193305044, "percentage": 2.52, "elapsed_time": "0:03:23", "remaining_time": "2:11:00", "throughput": 5350.67, "total_tokens": 1088384} {"current_steps": 1075, "total_steps": 42420, "loss": 0.3229, "lr": 0.0002531824611032532, "epoch": 0.5068363979255068, "percentage": 2.53, "elapsed_time": "0:03:24", "remaining_time": "2:10:59", "throughput": 5350.44, "total_tokens": 1093408} {"current_steps": 1080, "total_steps": 42420, "loss": 0.2852, "lr": 0.00025436115040075434, "epoch": 0.5091937765205092, "percentage": 2.55, "elapsed_time": "0:03:25", "remaining_time": "2:10:57", "throughput": 5350.66, "total_tokens": 1098400} {"current_steps": 1085, "total_steps": 42420, "loss": 0.4302, "lr": 0.0002555398396982556, "epoch": 0.5115511551155115, "percentage": 2.56, "elapsed_time": "0:03:26", "remaining_time": "2:10:53", "throughput": 5350.13, "total_tokens": 1102912} {"current_steps": 1090, "total_steps": 42420, "loss": 1.1511, "lr": 0.0002567185289957567, "epoch": 0.5139085337105139, "percentage": 2.57, "elapsed_time": "0:03:26", "remaining_time": "2:10:48", "throughput": 5348.72, "total_tokens": 1107104} {"current_steps": 1095, "total_steps": 42420, "loss": 0.5706, "lr": 0.0002578972182932579, "epoch": 0.5162659123055162, "percentage": 2.58, "elapsed_time": "0:03:27", "remaining_time": "2:10:43", "throughput": 5347.92, "total_tokens": 1111488} {"current_steps": 1100, "total_steps": 42420, "loss": 0.3648, "lr": 0.0002590759075907591, "epoch": 0.5186232909005186, "percentage": 2.59, "elapsed_time": "0:03:28", "remaining_time": "2:10:45", "throughput": 5349.55, "total_tokens": 1117280} {"current_steps": 1105, "total_steps": 42420, "loss": 0.3716, "lr": 0.00026025459688826025, "epoch": 0.520980669495521, "percentage": 2.6, "elapsed_time": "0:03:29", "remaining_time": "2:10:43", "throughput": 5349.63, "total_tokens": 1122208} {"current_steps": 1110, "total_steps": 42420, "loss": 0.3691, "lr": 0.00026143328618576144, "epoch": 0.5233380480905233, "percentage": 2.62, "elapsed_time": "0:03:30", "remaining_time": "2:10:41", "throughput": 5349.98, "total_tokens": 1127232} {"current_steps": 1115, "total_steps": 42420, "loss": 0.385, "lr": 0.0002626119754832626, "epoch": 0.5256954266855257, "percentage": 2.63, "elapsed_time": "0:03:31", "remaining_time": "2:10:40", "throughput": 5350.39, "total_tokens": 1132416} {"current_steps": 1120, "total_steps": 42420, "loss": 0.3231, "lr": 0.0002637906647807638, "epoch": 0.528052805280528, "percentage": 2.64, "elapsed_time": "0:03:32", "remaining_time": "2:10:39", "throughput": 5350.65, "total_tokens": 1137472} {"current_steps": 1125, "total_steps": 42420, "loss": 0.3214, "lr": 0.000264969354078265, "epoch": 0.5304101838755304, "percentage": 2.65, "elapsed_time": "0:03:33", "remaining_time": "2:10:39", "throughput": 5351.45, "total_tokens": 1142848} {"current_steps": 1130, "total_steps": 42420, "loss": 0.9129, "lr": 0.00026614804337576615, "epoch": 0.5327675624705328, "percentage": 2.66, "elapsed_time": "0:03:34", "remaining_time": "2:10:35", "throughput": 5351.1, "total_tokens": 1147488} {"current_steps": 1135, "total_steps": 42420, "loss": 0.5154, "lr": 0.00026732673267326735, "epoch": 0.5351249410655351, "percentage": 2.68, "elapsed_time": "0:03:35", "remaining_time": "2:10:29", "throughput": 5350.39, "total_tokens": 1151712} {"current_steps": 1140, "total_steps": 42420, "loss": 0.336, "lr": 0.0002685054219707685, "epoch": 0.5374823196605375, "percentage": 2.69, "elapsed_time": "0:03:36", "remaining_time": "2:10:28", "throughput": 5349.77, "total_tokens": 1156608} {"current_steps": 1145, "total_steps": 42420, "loss": 0.2949, "lr": 0.0002696841112682697, "epoch": 0.5398396982555398, "percentage": 2.7, "elapsed_time": "0:03:37", "remaining_time": "2:10:24", "throughput": 5349.09, "total_tokens": 1160992} {"current_steps": 1150, "total_steps": 42420, "loss": 0.3457, "lr": 0.00027086280056577087, "epoch": 0.5421970768505422, "percentage": 2.71, "elapsed_time": "0:03:37", "remaining_time": "2:10:18", "throughput": 5347.8, "total_tokens": 1165152} {"current_steps": 1155, "total_steps": 42420, "loss": 0.3598, "lr": 0.00027204148986327206, "epoch": 0.5445544554455446, "percentage": 2.72, "elapsed_time": "0:03:38", "remaining_time": "2:10:14", "throughput": 5347.21, "total_tokens": 1169600} {"current_steps": 1160, "total_steps": 42420, "loss": 0.3796, "lr": 0.00027322017916077326, "epoch": 0.5469118340405469, "percentage": 2.73, "elapsed_time": "0:03:39", "remaining_time": "2:10:16", "throughput": 5348.42, "total_tokens": 1175392} {"current_steps": 1165, "total_steps": 42420, "loss": 0.373, "lr": 0.0002743988684582744, "epoch": 0.5492692126355493, "percentage": 2.75, "elapsed_time": "0:03:40", "remaining_time": "2:10:18", "throughput": 5349.86, "total_tokens": 1181216} {"current_steps": 1170, "total_steps": 42420, "loss": 0.3233, "lr": 0.0002755775577557756, "epoch": 0.5516265912305516, "percentage": 2.76, "elapsed_time": "0:03:41", "remaining_time": "2:10:15", "throughput": 5349.21, "total_tokens": 1185856} {"current_steps": 1175, "total_steps": 42420, "loss": 0.2561, "lr": 0.0002767562470532768, "epoch": 0.553983969825554, "percentage": 2.77, "elapsed_time": "0:03:42", "remaining_time": "2:10:13", "throughput": 5349.13, "total_tokens": 1190720} {"current_steps": 1180, "total_steps": 42420, "loss": 0.4202, "lr": 0.0002779349363507779, "epoch": 0.5563413484205564, "percentage": 2.78, "elapsed_time": "0:03:43", "remaining_time": "2:10:09", "throughput": 5348.33, "total_tokens": 1195072} {"current_steps": 1185, "total_steps": 42420, "loss": 0.3718, "lr": 0.0002791136256482791, "epoch": 0.5586987270155587, "percentage": 2.79, "elapsed_time": "0:03:44", "remaining_time": "2:10:03", "throughput": 5346.92, "total_tokens": 1199136} {"current_steps": 1190, "total_steps": 42420, "loss": 0.3885, "lr": 0.0002802923149457803, "epoch": 0.5610561056105611, "percentage": 2.81, "elapsed_time": "0:03:45", "remaining_time": "2:10:04", "throughput": 5347.74, "total_tokens": 1204608} {"current_steps": 1195, "total_steps": 42420, "loss": 0.3319, "lr": 0.0002814710042432815, "epoch": 0.5634134842055634, "percentage": 2.82, "elapsed_time": "0:03:46", "remaining_time": "2:10:08", "throughput": 5350.19, "total_tokens": 1211008} {"current_steps": 1200, "total_steps": 42420, "loss": 0.3784, "lr": 0.0002826496935407827, "epoch": 0.5657708628005658, "percentage": 2.83, "elapsed_time": "0:03:47", "remaining_time": "2:10:03", "throughput": 5349.34, "total_tokens": 1215296} {"current_steps": 1205, "total_steps": 42420, "loss": 0.4083, "lr": 0.00028382838283828383, "epoch": 0.5681282413955682, "percentage": 2.84, "elapsed_time": "0:03:48", "remaining_time": "2:09:59", "throughput": 5348.4, "total_tokens": 1219584} {"current_steps": 1210, "total_steps": 42420, "loss": 0.3702, "lr": 0.000285007072135785, "epoch": 0.5704856199905705, "percentage": 2.85, "elapsed_time": "0:03:48", "remaining_time": "2:09:54", "throughput": 5347.62, "total_tokens": 1223808} {"current_steps": 1215, "total_steps": 42420, "loss": 0.3854, "lr": 0.00028618576143328616, "epoch": 0.5728429985855729, "percentage": 2.86, "elapsed_time": "0:03:49", "remaining_time": "2:09:49", "throughput": 5346.66, "total_tokens": 1228000} {"current_steps": 1220, "total_steps": 42420, "loss": 0.3384, "lr": 0.00028736445073078735, "epoch": 0.5752003771805752, "percentage": 2.88, "elapsed_time": "0:03:50", "remaining_time": "2:09:44", "throughput": 5345.98, "total_tokens": 1232256} {"current_steps": 1225, "total_steps": 42420, "loss": 0.3739, "lr": 0.0002885431400282886, "epoch": 0.5775577557755776, "percentage": 2.89, "elapsed_time": "0:03:51", "remaining_time": "2:09:45", "throughput": 5347.55, "total_tokens": 1237984} {"current_steps": 1230, "total_steps": 42420, "loss": 0.3584, "lr": 0.00028972182932578974, "epoch": 0.57991513437058, "percentage": 2.9, "elapsed_time": "0:03:52", "remaining_time": "2:09:42", "throughput": 5347.61, "total_tokens": 1242720} {"current_steps": 1235, "total_steps": 42420, "loss": 0.3524, "lr": 0.00029090051862329093, "epoch": 0.5822725129655822, "percentage": 2.91, "elapsed_time": "0:03:53", "remaining_time": "2:09:36", "throughput": 5346.53, "total_tokens": 1246816} {"current_steps": 1240, "total_steps": 42420, "loss": 0.3312, "lr": 0.00029207920792079207, "epoch": 0.5846298915605846, "percentage": 2.92, "elapsed_time": "0:03:54", "remaining_time": "2:09:33", "throughput": 5346.65, "total_tokens": 1251552} {"current_steps": 1245, "total_steps": 42420, "loss": 0.3108, "lr": 0.00029325789721829326, "epoch": 0.5869872701555869, "percentage": 2.93, "elapsed_time": "0:03:55", "remaining_time": "2:09:32", "throughput": 5347.03, "total_tokens": 1256576} {"current_steps": 1250, "total_steps": 42420, "loss": 0.3519, "lr": 0.00029443658651579445, "epoch": 0.5893446487505893, "percentage": 2.95, "elapsed_time": "0:03:55", "remaining_time": "2:09:28", "throughput": 5346.36, "total_tokens": 1261088} {"current_steps": 1255, "total_steps": 42420, "loss": 0.338, "lr": 0.0002956152758132956, "epoch": 0.5917020273455917, "percentage": 2.96, "elapsed_time": "0:03:56", "remaining_time": "2:09:26", "throughput": 5345.96, "total_tokens": 1265728} {"current_steps": 1260, "total_steps": 42420, "loss": 0.3166, "lr": 0.00029679396511079684, "epoch": 0.594059405940594, "percentage": 2.97, "elapsed_time": "0:03:57", "remaining_time": "2:09:23", "throughput": 5346.11, "total_tokens": 1270496} {"current_steps": 1265, "total_steps": 42420, "loss": 0.3244, "lr": 0.000297972654408298, "epoch": 0.5964167845355964, "percentage": 2.98, "elapsed_time": "0:03:58", "remaining_time": "2:09:18", "throughput": 5345.55, "total_tokens": 1274720} {"current_steps": 1270, "total_steps": 42420, "loss": 0.3507, "lr": 0.00029915134370579917, "epoch": 0.5987741631305987, "percentage": 2.99, "elapsed_time": "0:03:59", "remaining_time": "2:09:15", "throughput": 5345.18, "total_tokens": 1279360} {"current_steps": 1275, "total_steps": 42420, "loss": 0.3497, "lr": 0.00030033003300330036, "epoch": 0.6011315417256011, "percentage": 3.01, "elapsed_time": "0:04:00", "remaining_time": "2:09:13", "throughput": 5345.98, "total_tokens": 1284512} {"current_steps": 1280, "total_steps": 42420, "loss": 0.3899, "lr": 0.0003015087223008015, "epoch": 0.6034889203206035, "percentage": 3.02, "elapsed_time": "0:04:01", "remaining_time": "2:09:16", "throughput": 5347.66, "total_tokens": 1290560} {"current_steps": 1285, "total_steps": 42420, "loss": 0.3635, "lr": 0.0003026874115983027, "epoch": 0.6058462989156058, "percentage": 3.03, "elapsed_time": "0:04:02", "remaining_time": "2:09:14", "throughput": 5347.31, "total_tokens": 1295296} {"current_steps": 1290, "total_steps": 42420, "loss": 0.3473, "lr": 0.00030386610089580383, "epoch": 0.6082036775106082, "percentage": 3.04, "elapsed_time": "0:04:03", "remaining_time": "2:09:15", "throughput": 5348.28, "total_tokens": 1300928} {"current_steps": 1295, "total_steps": 42420, "loss": 0.3651, "lr": 0.0003050447901933051, "epoch": 0.6105610561056105, "percentage": 3.05, "elapsed_time": "0:04:04", "remaining_time": "2:09:12", "throughput": 5347.95, "total_tokens": 1305472} {"current_steps": 1300, "total_steps": 42420, "loss": 0.3623, "lr": 0.00030622347949080627, "epoch": 0.6129184347006129, "percentage": 3.06, "elapsed_time": "0:04:05", "remaining_time": "2:09:09", "throughput": 5347.65, "total_tokens": 1310176} {"current_steps": 1305, "total_steps": 42420, "loss": 0.3046, "lr": 0.0003074021687883074, "epoch": 0.6152758132956153, "percentage": 3.08, "elapsed_time": "0:04:05", "remaining_time": "2:09:07", "throughput": 5348.0, "total_tokens": 1315104} {"current_steps": 1310, "total_steps": 42420, "loss": 0.3569, "lr": 0.0003085808580858086, "epoch": 0.6176331918906176, "percentage": 3.09, "elapsed_time": "0:04:06", "remaining_time": "2:09:09", "throughput": 5349.38, "total_tokens": 1320928} {"current_steps": 1315, "total_steps": 42420, "loss": 0.3808, "lr": 0.00030975954738330974, "epoch": 0.61999057048562, "percentage": 3.1, "elapsed_time": "0:04:08", "remaining_time": "2:09:13", "throughput": 5351.67, "total_tokens": 1327520} {"current_steps": 1320, "total_steps": 42420, "loss": 0.3457, "lr": 0.00031093823668081093, "epoch": 0.6223479490806223, "percentage": 3.11, "elapsed_time": "0:04:08", "remaining_time": "2:09:11", "throughput": 5352.02, "total_tokens": 1332448} {"current_steps": 1325, "total_steps": 42420, "loss": 0.3702, "lr": 0.0003121169259783121, "epoch": 0.6247053276756247, "percentage": 3.12, "elapsed_time": "0:04:09", "remaining_time": "2:09:09", "throughput": 5352.42, "total_tokens": 1337440} {"current_steps": 1330, "total_steps": 42420, "loss": 0.3777, "lr": 0.0003132956152758133, "epoch": 0.6270627062706271, "percentage": 3.14, "elapsed_time": "0:04:10", "remaining_time": "2:09:12", "throughput": 5354.23, "total_tokens": 1343584} {"current_steps": 1335, "total_steps": 42420, "loss": 0.3508, "lr": 0.0003144743045733145, "epoch": 0.6294200848656294, "percentage": 3.15, "elapsed_time": "0:04:11", "remaining_time": "2:09:09", "throughput": 5353.59, "total_tokens": 1348032} {"current_steps": 1340, "total_steps": 42420, "loss": 0.3789, "lr": 0.00031565299387081565, "epoch": 0.6317774634606318, "percentage": 3.16, "elapsed_time": "0:04:13", "remaining_time": "2:09:18", "throughput": 5356.6, "total_tokens": 1355680} {"current_steps": 1345, "total_steps": 42420, "loss": 0.3067, "lr": 0.00031683168316831684, "epoch": 0.6341348420556341, "percentage": 3.17, "elapsed_time": "0:04:14", "remaining_time": "2:09:18", "throughput": 5357.09, "total_tokens": 1360896} {"current_steps": 1350, "total_steps": 42420, "loss": 0.4223, "lr": 0.00031801037246581803, "epoch": 0.6364922206506365, "percentage": 3.18, "elapsed_time": "0:04:14", "remaining_time": "2:09:15", "throughput": 5356.73, "total_tokens": 1365568} {"current_steps": 1355, "total_steps": 42420, "loss": 0.3345, "lr": 0.0003191890617633192, "epoch": 0.6388495992456389, "percentage": 3.19, "elapsed_time": "0:04:15", "remaining_time": "2:09:10", "throughput": 5356.39, "total_tokens": 1369888} {"current_steps": 1360, "total_steps": 42420, "loss": 0.3444, "lr": 0.00032036775106082037, "epoch": 0.6412069778406412, "percentage": 3.21, "elapsed_time": "0:04:16", "remaining_time": "2:09:09", "throughput": 5356.72, "total_tokens": 1375040} {"current_steps": 1365, "total_steps": 42420, "loss": 0.3332, "lr": 0.00032154644035832156, "epoch": 0.6435643564356436, "percentage": 3.22, "elapsed_time": "0:04:17", "remaining_time": "2:09:15", "throughput": 5359.41, "total_tokens": 1381952} {"current_steps": 1370, "total_steps": 42420, "loss": 0.3551, "lr": 0.00032272512965582275, "epoch": 0.6459217350306459, "percentage": 3.23, "elapsed_time": "0:04:18", "remaining_time": "2:09:12", "throughput": 5359.48, "total_tokens": 1386656} {"current_steps": 1375, "total_steps": 42420, "loss": 0.3484, "lr": 0.00032390381895332394, "epoch": 0.6482791136256483, "percentage": 3.24, "elapsed_time": "0:04:19", "remaining_time": "2:09:10", "throughput": 5359.99, "total_tokens": 1391616} {"current_steps": 1380, "total_steps": 42420, "loss": 0.3189, "lr": 0.0003250825082508251, "epoch": 0.6506364922206507, "percentage": 3.25, "elapsed_time": "0:04:20", "remaining_time": "2:09:10", "throughput": 5360.6, "total_tokens": 1397088} {"current_steps": 1385, "total_steps": 42420, "loss": 0.3058, "lr": 0.0003262611975483263, "epoch": 0.652993870815653, "percentage": 3.26, "elapsed_time": "0:04:21", "remaining_time": "2:09:07", "throughput": 5360.33, "total_tokens": 1401632} {"current_steps": 1390, "total_steps": 42420, "loss": 0.4481, "lr": 0.0003274398868458274, "epoch": 0.6553512494106554, "percentage": 3.28, "elapsed_time": "0:04:22", "remaining_time": "2:09:02", "throughput": 5359.48, "total_tokens": 1405728} {"current_steps": 1395, "total_steps": 42420, "loss": 0.3389, "lr": 0.0003286185761433286, "epoch": 0.6577086280056577, "percentage": 3.29, "elapsed_time": "0:04:23", "remaining_time": "2:09:00", "throughput": 5359.87, "total_tokens": 1410688} {"current_steps": 1400, "total_steps": 42420, "loss": 0.3737, "lr": 0.00032979726544082985, "epoch": 0.6600660066006601, "percentage": 3.3, "elapsed_time": "0:04:24", "remaining_time": "2:09:01", "throughput": 5361.19, "total_tokens": 1416416} {"current_steps": 1405, "total_steps": 42420, "loss": 0.3508, "lr": 0.000330975954738331, "epoch": 0.6624233851956625, "percentage": 3.31, "elapsed_time": "0:04:25", "remaining_time": "2:08:59", "throughput": 5361.75, "total_tokens": 1421600} {"current_steps": 1410, "total_steps": 42420, "loss": 0.3675, "lr": 0.0003321546440358322, "epoch": 0.6647807637906648, "percentage": 3.32, "elapsed_time": "0:04:26", "remaining_time": "2:08:58", "throughput": 5361.74, "total_tokens": 1426624} {"current_steps": 1415, "total_steps": 42420, "loss": 0.3676, "lr": 0.0003333333333333333, "epoch": 0.6671381423856672, "percentage": 3.34, "elapsed_time": "0:04:27", "remaining_time": "2:09:10", "throughput": 5365.55, "total_tokens": 1435008} {"current_steps": 1420, "total_steps": 42420, "loss": 0.3292, "lr": 0.0003345120226308345, "epoch": 0.6694955209806694, "percentage": 3.35, "elapsed_time": "0:04:28", "remaining_time": "2:09:07", "throughput": 5365.67, "total_tokens": 1439776} {"current_steps": 1425, "total_steps": 42420, "loss": 0.3262, "lr": 0.0003356907119283357, "epoch": 0.6718528995756718, "percentage": 3.36, "elapsed_time": "0:04:29", "remaining_time": "2:09:04", "throughput": 5365.0, "total_tokens": 1444352} {"current_steps": 1430, "total_steps": 42420, "loss": 0.3828, "lr": 0.00033686940122583685, "epoch": 0.6742102781706742, "percentage": 3.37, "elapsed_time": "0:04:30", "remaining_time": "2:09:04", "throughput": 5365.66, "total_tokens": 1449632} {"current_steps": 1435, "total_steps": 42420, "loss": 0.2976, "lr": 0.0003380480905233381, "epoch": 0.6765676567656765, "percentage": 3.38, "elapsed_time": "0:04:31", "remaining_time": "2:09:09", "throughput": 5368.41, "total_tokens": 1456576} {"current_steps": 1440, "total_steps": 42420, "loss": 0.329, "lr": 0.00033922677982083923, "epoch": 0.6789250353606789, "percentage": 3.39, "elapsed_time": "0:04:32", "remaining_time": "2:09:05", "throughput": 5367.89, "total_tokens": 1460928} {"current_steps": 1445, "total_steps": 42420, "loss": 0.2935, "lr": 0.0003404054691183404, "epoch": 0.6812824139556812, "percentage": 3.41, "elapsed_time": "0:04:33", "remaining_time": "2:09:04", "throughput": 5368.6, "total_tokens": 1466272} {"current_steps": 1450, "total_steps": 42420, "loss": 0.4656, "lr": 0.0003415841584158416, "epoch": 0.6836397925506836, "percentage": 3.42, "elapsed_time": "0:04:33", "remaining_time": "2:09:01", "throughput": 5367.95, "total_tokens": 1470784} {"current_steps": 1455, "total_steps": 42420, "loss": 0.3241, "lr": 0.00034276284771334275, "epoch": 0.685997171145686, "percentage": 3.43, "elapsed_time": "0:04:34", "remaining_time": "2:09:00", "throughput": 5368.38, "total_tokens": 1475904} {"current_steps": 1460, "total_steps": 42420, "loss": 0.3795, "lr": 0.00034394153701084395, "epoch": 0.6883545497406883, "percentage": 3.44, "elapsed_time": "0:04:35", "remaining_time": "2:08:57", "throughput": 5368.01, "total_tokens": 1480416} {"current_steps": 1465, "total_steps": 42420, "loss": 0.357, "lr": 0.0003451202263083451, "epoch": 0.6907119283356907, "percentage": 3.45, "elapsed_time": "0:04:36", "remaining_time": "2:08:53", "throughput": 5368.05, "total_tokens": 1484896} {"current_steps": 1470, "total_steps": 42420, "loss": 0.3484, "lr": 0.00034629891560584633, "epoch": 0.693069306930693, "percentage": 3.47, "elapsed_time": "0:04:37", "remaining_time": "2:08:51", "throughput": 5368.47, "total_tokens": 1489984} {"current_steps": 1475, "total_steps": 42420, "loss": 0.3644, "lr": 0.0003474776049033475, "epoch": 0.6954266855256954, "percentage": 3.48, "elapsed_time": "0:04:38", "remaining_time": "2:08:49", "throughput": 5368.74, "total_tokens": 1494912} {"current_steps": 1480, "total_steps": 42420, "loss": 0.3543, "lr": 0.00034865629420084866, "epoch": 0.6977840641206978, "percentage": 3.49, "elapsed_time": "0:04:39", "remaining_time": "2:08:48", "throughput": 5368.97, "total_tokens": 1499968} {"current_steps": 1485, "total_steps": 42420, "loss": 0.3366, "lr": 0.00034983498349834986, "epoch": 0.7001414427157001, "percentage": 3.5, "elapsed_time": "0:04:40", "remaining_time": "2:08:44", "throughput": 5368.17, "total_tokens": 1504288} {"current_steps": 1490, "total_steps": 42420, "loss": 0.3292, "lr": 0.000351013672795851, "epoch": 0.7024988213107025, "percentage": 3.51, "elapsed_time": "0:04:41", "remaining_time": "2:08:43", "throughput": 5368.9, "total_tokens": 1509632} {"current_steps": 1495, "total_steps": 42420, "loss": 0.3294, "lr": 0.0003521923620933522, "epoch": 0.7048561999057048, "percentage": 3.52, "elapsed_time": "0:04:42", "remaining_time": "2:08:43", "throughput": 5368.83, "total_tokens": 1514784} {"current_steps": 1500, "total_steps": 42420, "loss": 0.349, "lr": 0.0003533710513908534, "epoch": 0.7072135785007072, "percentage": 3.54, "elapsed_time": "0:04:43", "remaining_time": "2:08:41", "throughput": 5368.86, "total_tokens": 1519648} {"current_steps": 1505, "total_steps": 42420, "loss": 0.3793, "lr": 0.0003545497406883546, "epoch": 0.7095709570957096, "percentage": 3.55, "elapsed_time": "0:04:43", "remaining_time": "2:08:40", "throughput": 5369.13, "total_tokens": 1524768} {"current_steps": 1510, "total_steps": 42420, "loss": 0.3445, "lr": 0.00035572842998585577, "epoch": 0.7119283356907119, "percentage": 3.56, "elapsed_time": "0:04:44", "remaining_time": "2:08:38", "throughput": 5368.96, "total_tokens": 1529600} {"current_steps": 1515, "total_steps": 42420, "loss": 0.3247, "lr": 0.0003569071192833569, "epoch": 0.7142857142857143, "percentage": 3.57, "elapsed_time": "0:04:45", "remaining_time": "2:08:39", "throughput": 5369.75, "total_tokens": 1535232} {"current_steps": 1520, "total_steps": 42420, "loss": 0.3718, "lr": 0.0003580858085808581, "epoch": 0.7166430928807166, "percentage": 3.58, "elapsed_time": "0:04:46", "remaining_time": "2:08:35", "throughput": 5369.32, "total_tokens": 1539648} {"current_steps": 1525, "total_steps": 42420, "loss": 0.354, "lr": 0.0003592644978783593, "epoch": 0.719000471475719, "percentage": 3.6, "elapsed_time": "0:04:47", "remaining_time": "2:08:30", "throughput": 5368.04, "total_tokens": 1543488} {"current_steps": 1530, "total_steps": 42420, "loss": 0.3317, "lr": 0.00036044318717586043, "epoch": 0.7213578500707214, "percentage": 3.61, "elapsed_time": "0:04:48", "remaining_time": "2:08:32", "throughput": 5369.31, "total_tokens": 1549440} {"current_steps": 1535, "total_steps": 42420, "loss": 0.3229, "lr": 0.0003616218764733616, "epoch": 0.7237152286657237, "percentage": 3.62, "elapsed_time": "0:04:49", "remaining_time": "2:08:32", "throughput": 5370.22, "total_tokens": 1554976} {"current_steps": 1540, "total_steps": 42420, "loss": 0.3774, "lr": 0.0003628005657708628, "epoch": 0.7260726072607261, "percentage": 3.63, "elapsed_time": "0:04:50", "remaining_time": "2:08:29", "throughput": 5369.85, "total_tokens": 1559552} {"current_steps": 1545, "total_steps": 42420, "loss": 0.3369, "lr": 0.000363979255068364, "epoch": 0.7284299858557284, "percentage": 3.64, "elapsed_time": "0:04:51", "remaining_time": "2:08:29", "throughput": 5370.31, "total_tokens": 1564960} {"current_steps": 1550, "total_steps": 42420, "loss": 0.3177, "lr": 0.0003651579443658652, "epoch": 0.7307873644507308, "percentage": 3.65, "elapsed_time": "0:04:52", "remaining_time": "2:08:27", "throughput": 5370.0, "total_tokens": 1569632} {"current_steps": 1555, "total_steps": 42420, "loss": 0.3585, "lr": 0.00036633663366336634, "epoch": 0.7331447430457332, "percentage": 3.67, "elapsed_time": "0:04:53", "remaining_time": "2:08:24", "throughput": 5369.83, "total_tokens": 1574272} {"current_steps": 1560, "total_steps": 42420, "loss": 0.29, "lr": 0.00036751532296086753, "epoch": 0.7355021216407355, "percentage": 3.68, "elapsed_time": "0:04:54", "remaining_time": "2:08:23", "throughput": 5370.19, "total_tokens": 1579392} {"current_steps": 1565, "total_steps": 42420, "loss": 0.286, "lr": 0.00036869401225836867, "epoch": 0.7378595002357379, "percentage": 3.69, "elapsed_time": "0:04:55", "remaining_time": "2:08:24", "throughput": 5370.79, "total_tokens": 1585056} {"current_steps": 1570, "total_steps": 42420, "loss": 0.2527, "lr": 0.00036987270155586986, "epoch": 0.7402168788307402, "percentage": 3.7, "elapsed_time": "0:04:55", "remaining_time": "2:08:20", "throughput": 5370.62, "total_tokens": 1589472} {"current_steps": 1575, "total_steps": 42420, "loss": 0.3773, "lr": 0.0003710513908533711, "epoch": 0.7425742574257426, "percentage": 3.71, "elapsed_time": "0:04:56", "remaining_time": "2:08:19", "throughput": 5371.04, "total_tokens": 1594624} {"current_steps": 1580, "total_steps": 42420, "loss": 0.421, "lr": 0.00037223008015087225, "epoch": 0.744931636020745, "percentage": 3.72, "elapsed_time": "0:04:57", "remaining_time": "2:08:17", "throughput": 5371.46, "total_tokens": 1599648} {"current_steps": 1585, "total_steps": 42420, "loss": 0.3473, "lr": 0.00037340876944837344, "epoch": 0.7472890146157473, "percentage": 3.74, "elapsed_time": "0:04:58", "remaining_time": "2:08:16", "throughput": 5371.68, "total_tokens": 1604672} {"current_steps": 1590, "total_steps": 42420, "loss": 0.2975, "lr": 0.0003745874587458746, "epoch": 0.7496463932107497, "percentage": 3.75, "elapsed_time": "0:04:59", "remaining_time": "2:08:16", "throughput": 5372.26, "total_tokens": 1610144} {"current_steps": 1595, "total_steps": 42420, "loss": 0.4531, "lr": 0.00037576614804337577, "epoch": 0.752003771805752, "percentage": 3.76, "elapsed_time": "0:05:00", "remaining_time": "2:08:20", "throughput": 5373.87, "total_tokens": 1616640} {"current_steps": 1600, "total_steps": 42420, "loss": 0.3638, "lr": 0.00037694483734087696, "epoch": 0.7543611504007544, "percentage": 3.77, "elapsed_time": "0:05:01", "remaining_time": "2:08:16", "throughput": 5373.55, "total_tokens": 1621056} {"current_steps": 1605, "total_steps": 42420, "loss": 0.3567, "lr": 0.0003781235266383781, "epoch": 0.7567185289957568, "percentage": 3.78, "elapsed_time": "0:05:02", "remaining_time": "2:08:12", "throughput": 5373.25, "total_tokens": 1625408} {"current_steps": 1610, "total_steps": 42420, "loss": 0.3505, "lr": 0.00037930221593587935, "epoch": 0.759075907590759, "percentage": 3.8, "elapsed_time": "0:05:03", "remaining_time": "2:08:13", "throughput": 5374.43, "total_tokens": 1631296} {"current_steps": 1615, "total_steps": 42420, "loss": 0.3419, "lr": 0.0003804809052333805, "epoch": 0.7614332861857614, "percentage": 3.81, "elapsed_time": "0:05:04", "remaining_time": "2:08:11", "throughput": 5374.66, "total_tokens": 1636224} {"current_steps": 1620, "total_steps": 42420, "loss": 0.3578, "lr": 0.0003816595945308817, "epoch": 0.7637906647807637, "percentage": 3.82, "elapsed_time": "0:05:05", "remaining_time": "2:08:12", "throughput": 5376.2, "total_tokens": 1642144} {"current_steps": 1625, "total_steps": 42420, "loss": 0.346, "lr": 0.00038283828382838287, "epoch": 0.7661480433757661, "percentage": 3.83, "elapsed_time": "0:05:06", "remaining_time": "2:08:14", "throughput": 5377.89, "total_tokens": 1648256} {"current_steps": 1630, "total_steps": 42420, "loss": 0.3596, "lr": 0.000384016973125884, "epoch": 0.7685054219707685, "percentage": 3.84, "elapsed_time": "0:05:07", "remaining_time": "2:08:14", "throughput": 5378.31, "total_tokens": 1653696} {"current_steps": 1635, "total_steps": 42420, "loss": 0.3195, "lr": 0.0003851956624233852, "epoch": 0.7708628005657708, "percentage": 3.85, "elapsed_time": "0:05:08", "remaining_time": "2:08:15", "throughput": 5379.45, "total_tokens": 1659584} {"current_steps": 1640, "total_steps": 42420, "loss": 0.3413, "lr": 0.00038637435172088634, "epoch": 0.7732201791607732, "percentage": 3.87, "elapsed_time": "0:05:09", "remaining_time": "2:08:15", "throughput": 5380.12, "total_tokens": 1664992} {"current_steps": 1645, "total_steps": 42420, "loss": 0.258, "lr": 0.0003875530410183876, "epoch": 0.7755775577557755, "percentage": 3.88, "elapsed_time": "0:05:10", "remaining_time": "2:08:12", "throughput": 5379.96, "total_tokens": 1669632} {"current_steps": 1650, "total_steps": 42420, "loss": 0.3939, "lr": 0.0003887317303158888, "epoch": 0.7779349363507779, "percentage": 3.89, "elapsed_time": "0:05:11", "remaining_time": "2:08:13", "throughput": 5380.45, "total_tokens": 1675232} {"current_steps": 1655, "total_steps": 42420, "loss": 0.3876, "lr": 0.0003899104196133899, "epoch": 0.7802923149457803, "percentage": 3.9, "elapsed_time": "0:05:12", "remaining_time": "2:08:10", "throughput": 5380.24, "total_tokens": 1679776} {"current_steps": 1660, "total_steps": 42420, "loss": 0.2632, "lr": 0.0003910891089108911, "epoch": 0.7826496935407826, "percentage": 3.91, "elapsed_time": "0:05:13", "remaining_time": "2:08:07", "throughput": 5380.18, "total_tokens": 1684512} {"current_steps": 1665, "total_steps": 42420, "loss": 0.3931, "lr": 0.00039226779820839225, "epoch": 0.785007072135785, "percentage": 3.93, "elapsed_time": "0:05:13", "remaining_time": "2:08:05", "throughput": 5379.49, "total_tokens": 1688992} {"current_steps": 1670, "total_steps": 42420, "loss": 0.3471, "lr": 0.00039344648750589344, "epoch": 0.7873644507307873, "percentage": 3.94, "elapsed_time": "0:05:14", "remaining_time": "2:08:02", "throughput": 5379.21, "total_tokens": 1693504} {"current_steps": 1675, "total_steps": 42420, "loss": 0.3218, "lr": 0.00039462517680339463, "epoch": 0.7897218293257897, "percentage": 3.95, "elapsed_time": "0:05:15", "remaining_time": "2:08:01", "throughput": 5379.73, "total_tokens": 1698784} {"current_steps": 1680, "total_steps": 42420, "loss": 0.2853, "lr": 0.00039580386610089583, "epoch": 0.7920792079207921, "percentage": 3.96, "elapsed_time": "0:05:16", "remaining_time": "2:08:03", "throughput": 5380.88, "total_tokens": 1704800} {"current_steps": 1685, "total_steps": 42420, "loss": 0.4272, "lr": 0.000396982555398397, "epoch": 0.7944365865157944, "percentage": 3.97, "elapsed_time": "0:05:17", "remaining_time": "2:08:02", "throughput": 5381.03, "total_tokens": 1709984} {"current_steps": 1690, "total_steps": 42420, "loss": 0.3592, "lr": 0.00039816124469589816, "epoch": 0.7967939651107968, "percentage": 3.98, "elapsed_time": "0:05:18", "remaining_time": "2:08:01", "throughput": 5380.97, "total_tokens": 1715040} {"current_steps": 1695, "total_steps": 42420, "loss": 0.368, "lr": 0.00039933993399339935, "epoch": 0.7991513437057991, "percentage": 4.0, "elapsed_time": "0:05:19", "remaining_time": "2:08:01", "throughput": 5382.16, "total_tokens": 1720832} {"current_steps": 1700, "total_steps": 42420, "loss": 0.3526, "lr": 0.00040051862329090054, "epoch": 0.8015087223008015, "percentage": 4.01, "elapsed_time": "0:05:20", "remaining_time": "2:08:01", "throughput": 5382.66, "total_tokens": 1726080} {"current_steps": 1705, "total_steps": 42420, "loss": 0.3149, "lr": 0.0004016973125884017, "epoch": 0.8038661008958039, "percentage": 4.02, "elapsed_time": "0:05:21", "remaining_time": "2:07:57", "throughput": 5382.45, "total_tokens": 1730592} {"current_steps": 1710, "total_steps": 42420, "loss": 0.3028, "lr": 0.0004028760018859029, "epoch": 0.8062234794908062, "percentage": 4.03, "elapsed_time": "0:05:22", "remaining_time": "2:07:57", "throughput": 5382.98, "total_tokens": 1735968} {"current_steps": 1715, "total_steps": 42420, "loss": 0.364, "lr": 0.00040405469118340407, "epoch": 0.8085808580858086, "percentage": 4.04, "elapsed_time": "0:05:23", "remaining_time": "2:07:57", "throughput": 5383.12, "total_tokens": 1741216} {"current_steps": 1720, "total_steps": 42420, "loss": 0.3697, "lr": 0.00040523338048090526, "epoch": 0.8109382366808109, "percentage": 4.05, "elapsed_time": "0:05:24", "remaining_time": "2:07:52", "throughput": 5382.08, "total_tokens": 1745024} {"current_steps": 1725, "total_steps": 42420, "loss": 0.3427, "lr": 0.00040641206977840645, "epoch": 0.8132956152758133, "percentage": 4.07, "elapsed_time": "0:05:25", "remaining_time": "2:07:51", "throughput": 5382.3, "total_tokens": 1750272} {"current_steps": 1730, "total_steps": 42420, "loss": 0.338, "lr": 0.0004075907590759076, "epoch": 0.8156529938708157, "percentage": 4.08, "elapsed_time": "0:05:26", "remaining_time": "2:07:52", "throughput": 5383.07, "total_tokens": 1755968} {"current_steps": 1735, "total_steps": 42420, "loss": 0.3394, "lr": 0.0004087694483734088, "epoch": 0.818010372465818, "percentage": 4.09, "elapsed_time": "0:05:27", "remaining_time": "2:07:52", "throughput": 5383.8, "total_tokens": 1761600} {"current_steps": 1740, "total_steps": 42420, "loss": 0.3625, "lr": 0.0004099481376709099, "epoch": 0.8203677510608204, "percentage": 4.1, "elapsed_time": "0:05:28", "remaining_time": "2:07:51", "throughput": 5383.64, "total_tokens": 1766592} {"current_steps": 1745, "total_steps": 42420, "loss": 0.2932, "lr": 0.0004111268269684111, "epoch": 0.8227251296558227, "percentage": 4.11, "elapsed_time": "0:05:28", "remaining_time": "2:07:47", "throughput": 5382.54, "total_tokens": 1770624} {"current_steps": 1750, "total_steps": 42420, "loss": 0.3663, "lr": 0.00041230551626591236, "epoch": 0.8250825082508251, "percentage": 4.13, "elapsed_time": "0:05:29", "remaining_time": "2:07:45", "throughput": 5382.08, "total_tokens": 1775168} {"current_steps": 1755, "total_steps": 42420, "loss": 0.341, "lr": 0.0004134842055634135, "epoch": 0.8274398868458275, "percentage": 4.14, "elapsed_time": "0:05:30", "remaining_time": "2:07:41", "throughput": 5381.05, "total_tokens": 1779264} {"current_steps": 1760, "total_steps": 42420, "loss": 0.3075, "lr": 0.0004146628948609147, "epoch": 0.8297972654408298, "percentage": 4.15, "elapsed_time": "0:05:31", "remaining_time": "2:07:37", "throughput": 5380.17, "total_tokens": 1783232} {"current_steps": 1765, "total_steps": 42420, "loss": 0.2663, "lr": 0.00041584158415841583, "epoch": 0.8321546440358322, "percentage": 4.16, "elapsed_time": "0:05:32", "remaining_time": "2:07:35", "throughput": 5380.33, "total_tokens": 1788288} {"current_steps": 1770, "total_steps": 42420, "loss": 0.3118, "lr": 0.000417020273455917, "epoch": 0.8345120226308345, "percentage": 4.17, "elapsed_time": "0:05:33", "remaining_time": "2:07:31", "throughput": 5379.63, "total_tokens": 1792384} {"current_steps": 1775, "total_steps": 42420, "loss": 0.3946, "lr": 0.0004181989627534182, "epoch": 0.8368694012258369, "percentage": 4.18, "elapsed_time": "0:05:34", "remaining_time": "2:07:28", "throughput": 5379.36, "total_tokens": 1796800} {"current_steps": 1780, "total_steps": 42420, "loss": 0.3865, "lr": 0.00041937765205091936, "epoch": 0.8392267798208393, "percentage": 4.2, "elapsed_time": "0:05:34", "remaining_time": "2:07:27", "throughput": 5379.78, "total_tokens": 1802016} {"current_steps": 1785, "total_steps": 42420, "loss": 0.3468, "lr": 0.0004205563413484206, "epoch": 0.8415841584158416, "percentage": 4.21, "elapsed_time": "0:05:35", "remaining_time": "2:07:28", "throughput": 5380.64, "total_tokens": 1807680} {"current_steps": 1790, "total_steps": 42420, "loss": 0.35, "lr": 0.00042173503064592174, "epoch": 0.843941537010844, "percentage": 4.22, "elapsed_time": "0:05:36", "remaining_time": "2:07:26", "throughput": 5380.45, "total_tokens": 1812512} {"current_steps": 1795, "total_steps": 42420, "loss": 0.3435, "lr": 0.00042291371994342293, "epoch": 0.8462989156058462, "percentage": 4.23, "elapsed_time": "0:05:37", "remaining_time": "2:07:24", "throughput": 5380.58, "total_tokens": 1817408} {"current_steps": 1800, "total_steps": 42420, "loss": 0.3286, "lr": 0.0004240924092409241, "epoch": 0.8486562942008486, "percentage": 4.24, "elapsed_time": "0:05:38", "remaining_time": "2:07:20", "throughput": 5379.53, "total_tokens": 1821376} {"current_steps": 1805, "total_steps": 42420, "loss": 0.357, "lr": 0.00042527109853842526, "epoch": 0.851013672795851, "percentage": 4.26, "elapsed_time": "0:05:39", "remaining_time": "2:07:17", "throughput": 5379.55, "total_tokens": 1826048} {"current_steps": 1810, "total_steps": 42420, "loss": 0.3898, "lr": 0.00042644978783592646, "epoch": 0.8533710513908533, "percentage": 4.27, "elapsed_time": "0:05:40", "remaining_time": "2:07:15", "throughput": 5379.4, "total_tokens": 1830720} {"current_steps": 1815, "total_steps": 42420, "loss": 0.3548, "lr": 0.0004276284771334276, "epoch": 0.8557284299858557, "percentage": 4.28, "elapsed_time": "0:05:41", "remaining_time": "2:07:14", "throughput": 5379.43, "total_tokens": 1835648} {"current_steps": 1820, "total_steps": 42420, "loss": 0.3378, "lr": 0.00042880716643092884, "epoch": 0.858085808580858, "percentage": 4.29, "elapsed_time": "0:05:42", "remaining_time": "2:07:10", "throughput": 5378.67, "total_tokens": 1839808} {"current_steps": 1825, "total_steps": 42420, "loss": 0.3066, "lr": 0.00042998585572843003, "epoch": 0.8604431871758604, "percentage": 4.3, "elapsed_time": "0:05:42", "remaining_time": "2:07:08", "throughput": 5378.73, "total_tokens": 1844512} {"current_steps": 1830, "total_steps": 42420, "loss": 0.2676, "lr": 0.0004311645450259312, "epoch": 0.8628005657708628, "percentage": 4.31, "elapsed_time": "0:05:43", "remaining_time": "2:07:05", "throughput": 5378.37, "total_tokens": 1849056} {"current_steps": 1835, "total_steps": 42420, "loss": 0.3935, "lr": 0.00043234323432343237, "epoch": 0.8651579443658651, "percentage": 4.33, "elapsed_time": "0:05:44", "remaining_time": "2:07:02", "throughput": 5377.87, "total_tokens": 1853472} {"current_steps": 1840, "total_steps": 42420, "loss": 0.2622, "lr": 0.0004335219236209335, "epoch": 0.8675153229608675, "percentage": 4.34, "elapsed_time": "0:05:45", "remaining_time": "2:06:59", "throughput": 5377.36, "total_tokens": 1857920} {"current_steps": 1845, "total_steps": 42420, "loss": 0.3765, "lr": 0.0004347006129184347, "epoch": 0.8698727015558698, "percentage": 4.35, "elapsed_time": "0:05:46", "remaining_time": "2:06:59", "throughput": 5377.74, "total_tokens": 1863232} {"current_steps": 1850, "total_steps": 42420, "loss": 0.2918, "lr": 0.0004358793022159359, "epoch": 0.8722300801508722, "percentage": 4.36, "elapsed_time": "0:05:47", "remaining_time": "2:06:56", "throughput": 5377.26, "total_tokens": 1867616} {"current_steps": 1855, "total_steps": 42420, "loss": 0.3571, "lr": 0.0004370579915134371, "epoch": 0.8745874587458746, "percentage": 4.37, "elapsed_time": "0:05:48", "remaining_time": "2:06:53", "throughput": 5377.07, "total_tokens": 1872160} {"current_steps": 1860, "total_steps": 42420, "loss": 0.3486, "lr": 0.0004382366808109383, "epoch": 0.8769448373408769, "percentage": 4.38, "elapsed_time": "0:05:49", "remaining_time": "2:06:51", "throughput": 5376.97, "total_tokens": 1876832} {"current_steps": 1865, "total_steps": 42420, "loss": 0.3249, "lr": 0.0004394153701084394, "epoch": 0.8793022159358793, "percentage": 4.4, "elapsed_time": "0:05:49", "remaining_time": "2:06:48", "throughput": 5376.42, "total_tokens": 1881056} {"current_steps": 1870, "total_steps": 42420, "loss": 0.3298, "lr": 0.0004405940594059406, "epoch": 0.8816595945308816, "percentage": 4.41, "elapsed_time": "0:05:50", "remaining_time": "2:06:46", "throughput": 5376.3, "total_tokens": 1885952} {"current_steps": 1875, "total_steps": 42420, "loss": 0.3082, "lr": 0.0004417727487034418, "epoch": 0.884016973125884, "percentage": 4.42, "elapsed_time": "0:05:51", "remaining_time": "2:06:44", "throughput": 5376.06, "total_tokens": 1890592} {"current_steps": 1880, "total_steps": 42420, "loss": 0.362, "lr": 0.00044295143800094294, "epoch": 0.8863743517208864, "percentage": 4.43, "elapsed_time": "0:05:52", "remaining_time": "2:06:41", "throughput": 5375.59, "total_tokens": 1895072} {"current_steps": 1885, "total_steps": 42420, "loss": 0.3138, "lr": 0.00044413012729844413, "epoch": 0.8887317303158887, "percentage": 4.44, "elapsed_time": "0:05:53", "remaining_time": "2:06:38", "throughput": 5375.11, "total_tokens": 1899424} {"current_steps": 1890, "total_steps": 42420, "loss": 0.3805, "lr": 0.0004453088165959453, "epoch": 0.8910891089108911, "percentage": 4.46, "elapsed_time": "0:05:54", "remaining_time": "2:06:40", "throughput": 5376.41, "total_tokens": 1905632} {"current_steps": 1895, "total_steps": 42420, "loss": 0.3706, "lr": 0.0004464875058934465, "epoch": 0.8934464875058934, "percentage": 4.47, "elapsed_time": "0:05:55", "remaining_time": "2:06:38", "throughput": 5375.94, "total_tokens": 1910176} {"current_steps": 1900, "total_steps": 42420, "loss": 0.311, "lr": 0.0004476661951909477, "epoch": 0.8958038661008958, "percentage": 4.48, "elapsed_time": "0:05:56", "remaining_time": "2:06:39", "throughput": 5376.63, "total_tokens": 1915808} {"current_steps": 1905, "total_steps": 42420, "loss": 0.3297, "lr": 0.00044884488448844885, "epoch": 0.8981612446958982, "percentage": 4.49, "elapsed_time": "0:05:57", "remaining_time": "2:06:38", "throughput": 5376.59, "total_tokens": 1920896} {"current_steps": 1910, "total_steps": 42420, "loss": 0.3903, "lr": 0.00045002357378595004, "epoch": 0.9005186232909005, "percentage": 4.5, "elapsed_time": "0:05:58", "remaining_time": "2:06:33", "throughput": 5375.24, "total_tokens": 1924576} {"current_steps": 1915, "total_steps": 42420, "loss": 0.3361, "lr": 0.0004512022630834512, "epoch": 0.9028760018859029, "percentage": 4.51, "elapsed_time": "0:05:59", "remaining_time": "2:06:36", "throughput": 5376.07, "total_tokens": 1930752} {"current_steps": 1920, "total_steps": 42420, "loss": 0.3571, "lr": 0.00045238095238095237, "epoch": 0.9052333804809052, "percentage": 4.53, "elapsed_time": "0:05:59", "remaining_time": "2:06:31", "throughput": 5375.29, "total_tokens": 1934560} {"current_steps": 1925, "total_steps": 42420, "loss": 0.3569, "lr": 0.0004535596416784536, "epoch": 0.9075907590759076, "percentage": 4.54, "elapsed_time": "0:06:00", "remaining_time": "2:06:27", "throughput": 5374.39, "total_tokens": 1938368} {"current_steps": 1930, "total_steps": 42420, "loss": 0.3239, "lr": 0.00045473833097595476, "epoch": 0.90994813767091, "percentage": 4.55, "elapsed_time": "0:06:01", "remaining_time": "2:06:24", "throughput": 5373.8, "total_tokens": 1942720} {"current_steps": 1935, "total_steps": 42420, "loss": 0.3762, "lr": 0.00045591702027345595, "epoch": 0.9123055162659123, "percentage": 4.56, "elapsed_time": "0:06:02", "remaining_time": "2:06:25", "throughput": 5374.97, "total_tokens": 1948768} {"current_steps": 1940, "total_steps": 42420, "loss": 0.3422, "lr": 0.0004570957095709571, "epoch": 0.9146628948609147, "percentage": 4.57, "elapsed_time": "0:06:03", "remaining_time": "2:06:23", "throughput": 5375.0, "total_tokens": 1953376} {"current_steps": 1945, "total_steps": 42420, "loss": 0.3228, "lr": 0.0004582743988684583, "epoch": 0.917020273455917, "percentage": 4.59, "elapsed_time": "0:06:04", "remaining_time": "2:06:20", "throughput": 5374.34, "total_tokens": 1957632} {"current_steps": 1950, "total_steps": 42420, "loss": 0.3021, "lr": 0.00045945308816595947, "epoch": 0.9193776520509194, "percentage": 4.6, "elapsed_time": "0:06:05", "remaining_time": "2:06:17", "throughput": 5373.8, "total_tokens": 1961920} {"current_steps": 1955, "total_steps": 42420, "loss": 1.7603, "lr": 0.0004606317774634606, "epoch": 0.9217350306459218, "percentage": 4.61, "elapsed_time": "0:06:06", "remaining_time": "2:06:19", "throughput": 5375.24, "total_tokens": 1968480} {"current_steps": 1960, "total_steps": 42420, "loss": 0.4208, "lr": 0.00046181046676096186, "epoch": 0.9240924092409241, "percentage": 4.62, "elapsed_time": "0:06:07", "remaining_time": "2:06:18", "throughput": 5375.44, "total_tokens": 1973440} {"current_steps": 1965, "total_steps": 42420, "loss": 0.3508, "lr": 0.000462989156058463, "epoch": 0.9264497878359265, "percentage": 4.63, "elapsed_time": "0:06:07", "remaining_time": "2:06:16", "throughput": 5375.19, "total_tokens": 1978016} {"current_steps": 1970, "total_steps": 42420, "loss": 0.3594, "lr": 0.0004641678453559642, "epoch": 0.9288071664309288, "percentage": 4.64, "elapsed_time": "0:06:09", "remaining_time": "2:06:16", "throughput": 5376.15, "total_tokens": 1983808} {"current_steps": 1975, "total_steps": 42420, "loss": 0.3424, "lr": 0.0004653465346534654, "epoch": 0.9311645450259312, "percentage": 4.66, "elapsed_time": "0:06:09", "remaining_time": "2:06:16", "throughput": 5376.77, "total_tokens": 1989248} {"current_steps": 1980, "total_steps": 42420, "loss": 0.3315, "lr": 0.0004665252239509665, "epoch": 0.9335219236209336, "percentage": 4.67, "elapsed_time": "0:06:11", "remaining_time": "2:06:18", "throughput": 5377.95, "total_tokens": 1995392} {"current_steps": 1985, "total_steps": 42420, "loss": 0.3466, "lr": 0.0004677039132484677, "epoch": 0.9358793022159358, "percentage": 4.68, "elapsed_time": "0:06:11", "remaining_time": "2:06:16", "throughput": 5377.93, "total_tokens": 2000256} {"current_steps": 1990, "total_steps": 42420, "loss": 0.3233, "lr": 0.00046888260254596885, "epoch": 0.9382366808109383, "percentage": 4.69, "elapsed_time": "0:06:12", "remaining_time": "2:06:14", "throughput": 5377.53, "total_tokens": 2004800} {"current_steps": 1995, "total_steps": 42420, "loss": 0.3414, "lr": 0.0004700612918434701, "epoch": 0.9405940594059405, "percentage": 4.7, "elapsed_time": "0:06:13", "remaining_time": "2:06:12", "throughput": 5377.73, "total_tokens": 2009760} {"current_steps": 2000, "total_steps": 42420, "loss": 0.2682, "lr": 0.0004712399811409713, "epoch": 0.9429514380009429, "percentage": 4.71, "elapsed_time": "0:06:14", "remaining_time": "2:06:12", "throughput": 5378.07, "total_tokens": 2015072} {"current_steps": 2005, "total_steps": 42420, "loss": 0.3925, "lr": 0.00047241867043847243, "epoch": 0.9453088165959453, "percentage": 4.73, "elapsed_time": "0:06:15", "remaining_time": "2:06:07", "throughput": 5377.22, "total_tokens": 2018848} {"current_steps": 2010, "total_steps": 42420, "loss": 0.3972, "lr": 0.0004735973597359736, "epoch": 0.9476661951909476, "percentage": 4.74, "elapsed_time": "0:06:16", "remaining_time": "2:06:08", "throughput": 5377.96, "total_tokens": 2024704} {"current_steps": 2015, "total_steps": 42420, "loss": 0.3648, "lr": 0.00047477604903347476, "epoch": 0.95002357378595, "percentage": 4.75, "elapsed_time": "0:06:17", "remaining_time": "2:06:07", "throughput": 5378.06, "total_tokens": 2029632} {"current_steps": 2020, "total_steps": 42420, "loss": 0.3507, "lr": 0.00047595473833097595, "epoch": 0.9523809523809523, "percentage": 4.76, "elapsed_time": "0:06:18", "remaining_time": "2:06:06", "throughput": 5378.17, "total_tokens": 2034720} {"current_steps": 2025, "total_steps": 42420, "loss": 0.3571, "lr": 0.00047713342762847714, "epoch": 0.9547383309759547, "percentage": 4.77, "elapsed_time": "0:06:19", "remaining_time": "2:06:04", "throughput": 5377.68, "total_tokens": 2039168} {"current_steps": 2030, "total_steps": 42420, "loss": 0.3557, "lr": 0.00047831211692597834, "epoch": 0.9570957095709571, "percentage": 4.79, "elapsed_time": "0:06:20", "remaining_time": "2:06:02", "throughput": 5377.66, "total_tokens": 2044032} {"current_steps": 2035, "total_steps": 42420, "loss": 0.3591, "lr": 0.00047949080622347953, "epoch": 0.9594530881659594, "percentage": 4.8, "elapsed_time": "0:06:21", "remaining_time": "2:06:04", "throughput": 5378.64, "total_tokens": 2050080} {"current_steps": 2040, "total_steps": 42420, "loss": 0.3638, "lr": 0.00048066949552098067, "epoch": 0.9618104667609618, "percentage": 4.81, "elapsed_time": "0:06:22", "remaining_time": "2:06:01", "throughput": 5378.65, "total_tokens": 2054752} {"current_steps": 2045, "total_steps": 42420, "loss": 0.3518, "lr": 0.00048184818481848186, "epoch": 0.9641678453559641, "percentage": 4.82, "elapsed_time": "0:06:22", "remaining_time": "2:05:58", "throughput": 5378.18, "total_tokens": 2059008} {"current_steps": 2050, "total_steps": 42420, "loss": 0.2934, "lr": 0.00048302687411598305, "epoch": 0.9665252239509665, "percentage": 4.83, "elapsed_time": "0:06:23", "remaining_time": "2:05:58", "throughput": 5379.13, "total_tokens": 2064576} {"current_steps": 2055, "total_steps": 42420, "loss": 0.3758, "lr": 0.0004842055634134842, "epoch": 0.9688826025459689, "percentage": 4.84, "elapsed_time": "0:06:24", "remaining_time": "2:05:56", "throughput": 5379.23, "total_tokens": 2069408} {"current_steps": 2060, "total_steps": 42420, "loss": 0.3317, "lr": 0.0004853842527109854, "epoch": 0.9712399811409712, "percentage": 4.86, "elapsed_time": "0:06:25", "remaining_time": "2:05:54", "throughput": 5379.42, "total_tokens": 2074304} {"current_steps": 2065, "total_steps": 42420, "loss": 0.2824, "lr": 0.0004865629420084866, "epoch": 0.9735973597359736, "percentage": 4.87, "elapsed_time": "0:06:26", "remaining_time": "2:05:51", "throughput": 5379.04, "total_tokens": 2078688} {"current_steps": 2070, "total_steps": 42420, "loss": 0.4135, "lr": 0.00048774163130598777, "epoch": 0.9759547383309759, "percentage": 4.88, "elapsed_time": "0:06:27", "remaining_time": "2:05:48", "throughput": 5378.45, "total_tokens": 2082848} {"current_steps": 2075, "total_steps": 42420, "loss": 0.331, "lr": 0.000488920320603489, "epoch": 0.9783121169259783, "percentage": 4.89, "elapsed_time": "0:06:28", "remaining_time": "2:05:50", "throughput": 5379.82, "total_tokens": 2089184} {"current_steps": 2080, "total_steps": 42420, "loss": 0.3312, "lr": 0.0004900990099009901, "epoch": 0.9806694955209807, "percentage": 4.9, "elapsed_time": "0:06:29", "remaining_time": "2:05:49", "throughput": 5379.87, "total_tokens": 2094176} {"current_steps": 2085, "total_steps": 42420, "loss": 0.3536, "lr": 0.0004912776991984913, "epoch": 0.983026874115983, "percentage": 4.92, "elapsed_time": "0:06:30", "remaining_time": "2:05:47", "throughput": 5379.95, "total_tokens": 2098976} {"current_steps": 2090, "total_steps": 42420, "loss": 0.2999, "lr": 0.0004924563884959925, "epoch": 0.9853842527109854, "percentage": 4.93, "elapsed_time": "0:06:31", "remaining_time": "2:05:46", "throughput": 5380.25, "total_tokens": 2104224} {"current_steps": 2095, "total_steps": 42420, "loss": 0.3507, "lr": 0.0004936350777934936, "epoch": 0.9877416313059877, "percentage": 4.94, "elapsed_time": "0:06:31", "remaining_time": "2:05:44", "throughput": 5379.55, "total_tokens": 2108480} {"current_steps": 2100, "total_steps": 42420, "loss": 0.3545, "lr": 0.0004948137670909949, "epoch": 0.9900990099009901, "percentage": 4.95, "elapsed_time": "0:06:32", "remaining_time": "2:05:42", "throughput": 5379.84, "total_tokens": 2113440} {"current_steps": 2105, "total_steps": 42420, "loss": 0.3136, "lr": 0.000495992456388496, "epoch": 0.9924563884959925, "percentage": 4.96, "elapsed_time": "0:06:33", "remaining_time": "2:05:40", "throughput": 5380.01, "total_tokens": 2118112} {"current_steps": 2110, "total_steps": 42420, "loss": 0.3551, "lr": 0.0004971711456859971, "epoch": 0.9948137670909948, "percentage": 4.97, "elapsed_time": "0:06:34", "remaining_time": "2:05:41", "throughput": 5380.36, "total_tokens": 2123872} {"current_steps": 2115, "total_steps": 42420, "loss": 0.3729, "lr": 0.0004983498349834983, "epoch": 0.9971711456859972, "percentage": 4.99, "elapsed_time": "0:06:35", "remaining_time": "2:05:39", "throughput": 5380.22, "total_tokens": 2128544} {"current_steps": 2120, "total_steps": 42420, "loss": 0.3414, "lr": 0.0004995285242809995, "epoch": 0.9995285242809995, "percentage": 5.0, "elapsed_time": "0:06:36", "remaining_time": "2:05:37", "throughput": 5380.29, "total_tokens": 2133408} {"current_steps": 2121, "total_steps": 42420, "eval_loss": 0.3479316532611847, "epoch": 1.0, "percentage": 5.0, "elapsed_time": "0:06:55", "remaining_time": "2:11:37", "throughput": 5134.34, "total_tokens": 2134176} {"current_steps": 2125, "total_steps": 42420, "loss": 0.3419, "lr": 0.0005007072135785007, "epoch": 1.0018859028760019, "percentage": 5.01, "elapsed_time": "0:06:58", "remaining_time": "2:12:19", "throughput": 5106.46, "total_tokens": 2138112} {"current_steps": 2130, "total_steps": 42420, "loss": 0.3339, "lr": 0.0005018859028760019, "epoch": 1.0042432814710043, "percentage": 5.02, "elapsed_time": "0:06:59", "remaining_time": "2:12:18", "throughput": 5107.49, "total_tokens": 2143584} {"current_steps": 2135, "total_steps": 42420, "loss": 0.3532, "lr": 0.0005030645921735031, "epoch": 1.0066006600660067, "percentage": 5.03, "elapsed_time": "0:07:00", "remaining_time": "2:12:15", "throughput": 5107.94, "total_tokens": 2148192} {"current_steps": 2140, "total_steps": 42420, "loss": 0.3161, "lr": 0.0005042432814710043, "epoch": 1.0089580386610089, "percentage": 5.04, "elapsed_time": "0:07:01", "remaining_time": "2:12:14", "throughput": 5108.83, "total_tokens": 2153568} {"current_steps": 2145, "total_steps": 42420, "loss": 0.3598, "lr": 0.0005054219707685053, "epoch": 1.0113154172560113, "percentage": 5.06, "elapsed_time": "0:07:02", "remaining_time": "2:12:12", "throughput": 5109.68, "total_tokens": 2158752} {"current_steps": 2150, "total_steps": 42420, "loss": 0.2967, "lr": 0.0005066006600660066, "epoch": 1.0136727958510137, "percentage": 5.07, "elapsed_time": "0:07:03", "remaining_time": "2:12:11", "throughput": 5110.73, "total_tokens": 2164320} {"current_steps": 2155, "total_steps": 42420, "loss": 0.3607, "lr": 0.0005077793493635078, "epoch": 1.016030174446016, "percentage": 5.08, "elapsed_time": "0:07:04", "remaining_time": "2:12:09", "throughput": 5110.98, "total_tokens": 2169120} {"current_steps": 2160, "total_steps": 42420, "loss": 0.3301, "lr": 0.000508958038661009, "epoch": 1.0183875530410185, "percentage": 5.09, "elapsed_time": "0:07:05", "remaining_time": "2:12:07", "throughput": 5111.14, "total_tokens": 2173952} {"current_steps": 2165, "total_steps": 42420, "loss": 0.3521, "lr": 0.0005101367279585102, "epoch": 1.0207449316360206, "percentage": 5.1, "elapsed_time": "0:07:06", "remaining_time": "2:12:03", "throughput": 5111.33, "total_tokens": 2178240} {"current_steps": 2170, "total_steps": 42420, "loss": 0.3555, "lr": 0.0005113154172560112, "epoch": 1.023102310231023, "percentage": 5.12, "elapsed_time": "0:07:07", "remaining_time": "2:12:02", "throughput": 5112.41, "total_tokens": 2183744} {"current_steps": 2175, "total_steps": 42420, "loss": 0.3665, "lr": 0.0005124941065535125, "epoch": 1.0254596888260255, "percentage": 5.13, "elapsed_time": "0:07:08", "remaining_time": "2:12:00", "throughput": 5113.0, "total_tokens": 2188576} {"current_steps": 2180, "total_steps": 42420, "loss": 0.3666, "lr": 0.0005136727958510136, "epoch": 1.0278170674210279, "percentage": 5.14, "elapsed_time": "0:07:08", "remaining_time": "2:11:56", "throughput": 5113.01, "total_tokens": 2192896} {"current_steps": 2185, "total_steps": 42420, "loss": 0.359, "lr": 0.0005148514851485149, "epoch": 1.0301744460160303, "percentage": 5.15, "elapsed_time": "0:07:09", "remaining_time": "2:11:55", "throughput": 5114.16, "total_tokens": 2198368} {"current_steps": 2190, "total_steps": 42420, "loss": 0.3371, "lr": 0.0005160301744460161, "epoch": 1.0325318246110324, "percentage": 5.16, "elapsed_time": "0:07:10", "remaining_time": "2:11:51", "throughput": 5114.24, "total_tokens": 2202624} {"current_steps": 2195, "total_steps": 42420, "loss": 0.3081, "lr": 0.0005172088637435172, "epoch": 1.0348892032060348, "percentage": 5.17, "elapsed_time": "0:07:11", "remaining_time": "2:11:47", "throughput": 5114.26, "total_tokens": 2206848} {"current_steps": 2200, "total_steps": 42420, "loss": 0.3217, "lr": 0.0005183875530410184, "epoch": 1.0372465818010372, "percentage": 5.19, "elapsed_time": "0:07:12", "remaining_time": "2:11:47", "throughput": 5115.47, "total_tokens": 2212736} {"current_steps": 2205, "total_steps": 42420, "loss": 0.2524, "lr": 0.0005195662423385195, "epoch": 1.0396039603960396, "percentage": 5.2, "elapsed_time": "0:07:13", "remaining_time": "2:11:44", "throughput": 5115.54, "total_tokens": 2217152} {"current_steps": 2210, "total_steps": 42420, "loss": 0.3785, "lr": 0.0005207449316360208, "epoch": 1.041961338991042, "percentage": 5.21, "elapsed_time": "0:07:14", "remaining_time": "2:11:44", "throughput": 5116.65, "total_tokens": 2222816} {"current_steps": 2215, "total_steps": 42420, "loss": 0.3531, "lr": 0.0005219236209335219, "epoch": 1.0443187175860442, "percentage": 5.22, "elapsed_time": "0:07:15", "remaining_time": "2:11:41", "throughput": 5116.99, "total_tokens": 2227552} {"current_steps": 2220, "total_steps": 42420, "loss": 0.3169, "lr": 0.0005231023102310231, "epoch": 1.0466760961810466, "percentage": 5.23, "elapsed_time": "0:07:16", "remaining_time": "2:11:39", "throughput": 5117.96, "total_tokens": 2232704} {"current_steps": 2225, "total_steps": 42420, "loss": 0.3431, "lr": 0.0005242809995285243, "epoch": 1.049033474776049, "percentage": 5.25, "elapsed_time": "0:07:17", "remaining_time": "2:11:38", "throughput": 5118.41, "total_tokens": 2237792} {"current_steps": 2230, "total_steps": 42420, "loss": 0.3201, "lr": 0.0005254596888260254, "epoch": 1.0513908533710514, "percentage": 5.26, "elapsed_time": "0:07:18", "remaining_time": "2:11:34", "throughput": 5118.64, "total_tokens": 2242240} {"current_steps": 2235, "total_steps": 42420, "loss": 0.3431, "lr": 0.0005266383781235267, "epoch": 1.0537482319660538, "percentage": 5.27, "elapsed_time": "0:07:19", "remaining_time": "2:11:34", "throughput": 5119.71, "total_tokens": 2247936} {"current_steps": 2240, "total_steps": 42420, "loss": 0.309, "lr": 0.0005278170674210278, "epoch": 1.056105610561056, "percentage": 5.28, "elapsed_time": "0:07:20", "remaining_time": "2:11:36", "throughput": 5121.31, "total_tokens": 2254496} {"current_steps": 2245, "total_steps": 42420, "loss": 0.3629, "lr": 0.000528995756718529, "epoch": 1.0584629891560584, "percentage": 5.29, "elapsed_time": "0:07:21", "remaining_time": "2:11:31", "throughput": 5120.97, "total_tokens": 2258368} {"current_steps": 2250, "total_steps": 42420, "loss": 0.3748, "lr": 0.0005301744460160301, "epoch": 1.0608203677510608, "percentage": 5.3, "elapsed_time": "0:07:21", "remaining_time": "2:11:29", "throughput": 5121.4, "total_tokens": 2263168} {"current_steps": 2255, "total_steps": 42420, "loss": 0.329, "lr": 0.0005313531353135314, "epoch": 1.0631777463460632, "percentage": 5.32, "elapsed_time": "0:07:22", "remaining_time": "2:11:26", "throughput": 5121.69, "total_tokens": 2267648} {"current_steps": 2260, "total_steps": 42420, "loss": 0.3382, "lr": 0.0005325318246110326, "epoch": 1.0655351249410656, "percentage": 5.33, "elapsed_time": "0:07:23", "remaining_time": "2:11:24", "throughput": 5122.48, "total_tokens": 2272864} {"current_steps": 2265, "total_steps": 42420, "loss": 0.3489, "lr": 0.0005337105139085337, "epoch": 1.0678925035360678, "percentage": 5.34, "elapsed_time": "0:07:24", "remaining_time": "2:11:21", "throughput": 5123.07, "total_tokens": 2277632} {"current_steps": 2270, "total_steps": 42420, "loss": 0.3477, "lr": 0.0005348892032060349, "epoch": 1.0702498821310702, "percentage": 5.35, "elapsed_time": "0:07:25", "remaining_time": "2:11:19", "throughput": 5123.66, "total_tokens": 2282464} {"current_steps": 2275, "total_steps": 42420, "loss": 0.339, "lr": 0.000536067892503536, "epoch": 1.0726072607260726, "percentage": 5.36, "elapsed_time": "0:07:26", "remaining_time": "2:11:16", "throughput": 5124.0, "total_tokens": 2287264} {"current_steps": 2280, "total_steps": 42420, "loss": 0.3635, "lr": 0.0005372465818010373, "epoch": 1.074964639321075, "percentage": 5.37, "elapsed_time": "0:07:27", "remaining_time": "2:11:14", "throughput": 5124.46, "total_tokens": 2292000} {"current_steps": 2285, "total_steps": 42420, "loss": 0.359, "lr": 0.0005384252710985384, "epoch": 1.0773220179160774, "percentage": 5.39, "elapsed_time": "0:07:28", "remaining_time": "2:11:13", "throughput": 5125.81, "total_tokens": 2297696} {"current_steps": 2290, "total_steps": 42420, "loss": 0.3518, "lr": 0.0005396039603960396, "epoch": 1.0796793965110796, "percentage": 5.4, "elapsed_time": "0:07:29", "remaining_time": "2:11:11", "throughput": 5126.47, "total_tokens": 2302624} {"current_steps": 2295, "total_steps": 42420, "loss": 0.3419, "lr": 0.0005407826496935408, "epoch": 1.082036775106082, "percentage": 5.41, "elapsed_time": "0:07:30", "remaining_time": "2:11:09", "throughput": 5127.66, "total_tokens": 2308128} {"current_steps": 2300, "total_steps": 42420, "loss": 0.3284, "lr": 0.0005419613389910419, "epoch": 1.0843941537010844, "percentage": 5.42, "elapsed_time": "0:07:31", "remaining_time": "2:11:10", "throughput": 5129.15, "total_tokens": 2314272} {"current_steps": 2305, "total_steps": 42420, "loss": 0.3001, "lr": 0.0005431400282885432, "epoch": 1.0867515322960868, "percentage": 5.43, "elapsed_time": "0:07:32", "remaining_time": "2:11:07", "throughput": 5129.45, "total_tokens": 2318784} {"current_steps": 2310, "total_steps": 42420, "loss": 0.3402, "lr": 0.0005443187175860443, "epoch": 1.0891089108910892, "percentage": 5.45, "elapsed_time": "0:07:32", "remaining_time": "2:11:04", "throughput": 5129.92, "total_tokens": 2323392} {"current_steps": 2315, "total_steps": 42420, "loss": 0.3111, "lr": 0.0005454974068835456, "epoch": 1.0914662894860914, "percentage": 5.46, "elapsed_time": "0:07:33", "remaining_time": "2:11:03", "throughput": 5130.65, "total_tokens": 2328704} {"current_steps": 2320, "total_steps": 42420, "loss": 0.4087, "lr": 0.0005466760961810466, "epoch": 1.0938236680810938, "percentage": 5.47, "elapsed_time": "0:07:34", "remaining_time": "2:11:02", "throughput": 5131.69, "total_tokens": 2334368} {"current_steps": 2325, "total_steps": 42420, "loss": 0.3512, "lr": 0.0005478547854785478, "epoch": 1.0961810466760962, "percentage": 5.48, "elapsed_time": "0:07:35", "remaining_time": "2:11:02", "throughput": 5132.94, "total_tokens": 2340288} {"current_steps": 2330, "total_steps": 42420, "loss": 0.3594, "lr": 0.0005490334747760491, "epoch": 1.0985384252710986, "percentage": 5.49, "elapsed_time": "0:07:36", "remaining_time": "2:11:00", "throughput": 5133.02, "total_tokens": 2344960} {"current_steps": 2335, "total_steps": 42420, "loss": 0.3534, "lr": 0.0005502121640735502, "epoch": 1.100895803866101, "percentage": 5.5, "elapsed_time": "0:07:37", "remaining_time": "2:10:56", "throughput": 5132.78, "total_tokens": 2349024} {"current_steps": 2340, "total_steps": 42420, "loss": 0.3597, "lr": 0.0005513908533710515, "epoch": 1.1032531824611032, "percentage": 5.52, "elapsed_time": "0:07:38", "remaining_time": "2:10:53", "throughput": 5133.23, "total_tokens": 2353792} {"current_steps": 2345, "total_steps": 42420, "loss": 0.3492, "lr": 0.0005525695426685525, "epoch": 1.1056105610561056, "percentage": 5.53, "elapsed_time": "0:07:39", "remaining_time": "2:10:54", "throughput": 5134.27, "total_tokens": 2359744} {"current_steps": 2350, "total_steps": 42420, "loss": 0.3551, "lr": 0.0005537482319660537, "epoch": 1.107967939651108, "percentage": 5.54, "elapsed_time": "0:07:40", "remaining_time": "2:10:51", "throughput": 5134.23, "total_tokens": 2364288} {"current_steps": 2355, "total_steps": 42420, "loss": 0.3607, "lr": 0.0005549269212635549, "epoch": 1.1103253182461104, "percentage": 5.55, "elapsed_time": "0:07:41", "remaining_time": "2:10:52", "throughput": 5135.34, "total_tokens": 2370240} {"current_steps": 2360, "total_steps": 42420, "loss": 0.3579, "lr": 0.0005561056105610561, "epoch": 1.1126826968411128, "percentage": 5.56, "elapsed_time": "0:07:42", "remaining_time": "2:10:49", "throughput": 5135.85, "total_tokens": 2375040} {"current_steps": 2365, "total_steps": 42420, "loss": 0.306, "lr": 0.0005572842998585574, "epoch": 1.115040075436115, "percentage": 5.58, "elapsed_time": "0:07:43", "remaining_time": "2:10:47", "throughput": 5136.33, "total_tokens": 2379936} {"current_steps": 2370, "total_steps": 42420, "loss": 0.3726, "lr": 0.0005584629891560584, "epoch": 1.1173974540311173, "percentage": 5.59, "elapsed_time": "0:07:44", "remaining_time": "2:10:46", "throughput": 5137.17, "total_tokens": 2385216} {"current_steps": 2375, "total_steps": 42420, "loss": 0.3512, "lr": 0.0005596416784535597, "epoch": 1.1197548326261197, "percentage": 5.6, "elapsed_time": "0:07:45", "remaining_time": "2:10:46", "throughput": 5138.7, "total_tokens": 2391360} {"current_steps": 2380, "total_steps": 42420, "loss": 0.3229, "lr": 0.0005608203677510608, "epoch": 1.1221122112211221, "percentage": 5.61, "elapsed_time": "0:07:46", "remaining_time": "2:10:44", "throughput": 5139.39, "total_tokens": 2396544} {"current_steps": 2385, "total_steps": 42420, "loss": 0.3292, "lr": 0.000561999057048562, "epoch": 1.1244695898161245, "percentage": 5.62, "elapsed_time": "0:07:47", "remaining_time": "2:10:44", "throughput": 5140.31, "total_tokens": 2402048} {"current_steps": 2390, "total_steps": 42420, "loss": 0.2735, "lr": 0.0005631777463460632, "epoch": 1.1268269684111267, "percentage": 5.63, "elapsed_time": "0:07:48", "remaining_time": "2:10:42", "throughput": 5141.11, "total_tokens": 2407264} {"current_steps": 2395, "total_steps": 42420, "loss": 0.3214, "lr": 0.0005643564356435643, "epoch": 1.1291843470061291, "percentage": 5.65, "elapsed_time": "0:07:49", "remaining_time": "2:10:40", "throughput": 5141.57, "total_tokens": 2412320} {"current_steps": 2400, "total_steps": 42420, "loss": 0.3732, "lr": 0.0005655351249410656, "epoch": 1.1315417256011315, "percentage": 5.66, "elapsed_time": "0:07:50", "remaining_time": "2:10:38", "throughput": 5142.08, "total_tokens": 2417280} {"current_steps": 2405, "total_steps": 42420, "loss": 0.3451, "lr": 0.0005667138142385667, "epoch": 1.133899104196134, "percentage": 5.67, "elapsed_time": "0:07:51", "remaining_time": "2:10:37", "throughput": 5142.78, "total_tokens": 2422368} {"current_steps": 2410, "total_steps": 42420, "loss": 0.2808, "lr": 0.000567892503536068, "epoch": 1.1362564827911363, "percentage": 5.68, "elapsed_time": "0:07:51", "remaining_time": "2:10:34", "throughput": 5143.06, "total_tokens": 2427040} {"current_steps": 2415, "total_steps": 42420, "loss": 0.2933, "lr": 0.0005690711928335691, "epoch": 1.1386138613861387, "percentage": 5.69, "elapsed_time": "0:07:53", "remaining_time": "2:10:36", "throughput": 5144.21, "total_tokens": 2433472} {"current_steps": 2420, "total_steps": 42420, "loss": 0.3708, "lr": 0.0005702498821310702, "epoch": 1.140971239981141, "percentage": 5.7, "elapsed_time": "0:07:54", "remaining_time": "2:10:36", "throughput": 5145.45, "total_tokens": 2439360} {"current_steps": 2425, "total_steps": 42420, "loss": 0.3081, "lr": 0.0005714285714285714, "epoch": 1.1433286185761433, "percentage": 5.72, "elapsed_time": "0:07:54", "remaining_time": "2:10:31", "throughput": 5144.67, "total_tokens": 2442912} {"current_steps": 2430, "total_steps": 42420, "loss": 0.3444, "lr": 0.0005726072607260726, "epoch": 1.1456859971711457, "percentage": 5.73, "elapsed_time": "0:07:55", "remaining_time": "2:10:30", "throughput": 5145.72, "total_tokens": 2448512} {"current_steps": 2435, "total_steps": 42420, "loss": 0.3665, "lr": 0.0005737859500235739, "epoch": 1.1480433757661481, "percentage": 5.74, "elapsed_time": "0:07:56", "remaining_time": "2:10:28", "throughput": 5145.92, "total_tokens": 2453152} {"current_steps": 2440, "total_steps": 42420, "loss": 0.3047, "lr": 0.000574964639321075, "epoch": 1.1504007543611503, "percentage": 5.75, "elapsed_time": "0:07:57", "remaining_time": "2:10:30", "throughput": 5147.77, "total_tokens": 2460128} {"current_steps": 2445, "total_steps": 42420, "loss": 0.3173, "lr": 0.0005761433286185761, "epoch": 1.1527581329561527, "percentage": 5.76, "elapsed_time": "0:07:58", "remaining_time": "2:10:28", "throughput": 5147.74, "total_tokens": 2464736} {"current_steps": 2450, "total_steps": 42420, "loss": 0.3092, "lr": 0.0005773220179160773, "epoch": 1.155115511551155, "percentage": 5.78, "elapsed_time": "0:07:59", "remaining_time": "2:10:26", "throughput": 5148.29, "total_tokens": 2469920} {"current_steps": 2455, "total_steps": 42420, "loss": 0.3852, "lr": 0.0005785007072135785, "epoch": 1.1574728901461575, "percentage": 5.79, "elapsed_time": "0:08:00", "remaining_time": "2:10:25", "throughput": 5149.31, "total_tokens": 2475392} {"current_steps": 2460, "total_steps": 42420, "loss": 0.3671, "lr": 0.0005796793965110797, "epoch": 1.15983026874116, "percentage": 5.8, "elapsed_time": "0:08:01", "remaining_time": "2:10:24", "throughput": 5149.71, "total_tokens": 2480416} {"current_steps": 2465, "total_steps": 42420, "loss": 0.3078, "lr": 0.0005808580858085809, "epoch": 1.1621876473361623, "percentage": 5.81, "elapsed_time": "0:08:02", "remaining_time": "2:10:25", "throughput": 5151.13, "total_tokens": 2486784} {"current_steps": 2470, "total_steps": 42420, "loss": 0.3907, "lr": 0.000582036775106082, "epoch": 1.1645450259311645, "percentage": 5.82, "elapsed_time": "0:08:03", "remaining_time": "2:10:23", "throughput": 5152.11, "total_tokens": 2492224} {"current_steps": 2475, "total_steps": 42420, "loss": 0.3406, "lr": 0.0005832154644035832, "epoch": 1.166902404526167, "percentage": 5.83, "elapsed_time": "0:08:04", "remaining_time": "2:10:19", "throughput": 5151.45, "total_tokens": 2496000} {"current_steps": 2480, "total_steps": 42420, "loss": 0.3505, "lr": 0.0005843941537010844, "epoch": 1.1692597831211693, "percentage": 5.85, "elapsed_time": "0:08:05", "remaining_time": "2:10:15", "throughput": 5151.07, "total_tokens": 2499712} {"current_steps": 2485, "total_steps": 42420, "loss": 0.3768, "lr": 0.0005855728429985856, "epoch": 1.1716171617161717, "percentage": 5.86, "elapsed_time": "0:08:06", "remaining_time": "2:10:15", "throughput": 5152.34, "total_tokens": 2505696} {"current_steps": 2490, "total_steps": 42420, "loss": 0.3311, "lr": 0.0005867515322960868, "epoch": 1.1739745403111739, "percentage": 5.87, "elapsed_time": "0:08:07", "remaining_time": "2:10:14", "throughput": 5153.16, "total_tokens": 2511168} {"current_steps": 2495, "total_steps": 42420, "loss": 0.3248, "lr": 0.0005879302215935878, "epoch": 1.1763319189061763, "percentage": 5.88, "elapsed_time": "0:08:08", "remaining_time": "2:10:13", "throughput": 5153.76, "total_tokens": 2516416} {"current_steps": 2500, "total_steps": 42420, "loss": 0.3526, "lr": 0.0005891089108910891, "epoch": 1.1786892975011787, "percentage": 5.89, "elapsed_time": "0:08:09", "remaining_time": "2:10:12", "throughput": 5154.73, "total_tokens": 2521888} {"current_steps": 2505, "total_steps": 42420, "loss": 0.3769, "lr": 0.0005902876001885903, "epoch": 1.181046676096181, "percentage": 5.91, "elapsed_time": "0:08:10", "remaining_time": "2:10:08", "throughput": 5154.66, "total_tokens": 2526048} {"current_steps": 2510, "total_steps": 42420, "loss": 0.2982, "lr": 0.0005914662894860915, "epoch": 1.1834040546911835, "percentage": 5.92, "elapsed_time": "0:08:11", "remaining_time": "2:10:07", "throughput": 5155.6, "total_tokens": 2531616} {"current_steps": 2515, "total_steps": 42420, "loss": 0.3473, "lr": 0.0005926449787835927, "epoch": 1.1857614332861859, "percentage": 5.93, "elapsed_time": "0:08:11", "remaining_time": "2:10:05", "throughput": 5156.41, "total_tokens": 2536736} {"current_steps": 2520, "total_steps": 42420, "loss": 0.2936, "lr": 0.0005938236680810938, "epoch": 1.188118811881188, "percentage": 5.94, "elapsed_time": "0:08:13", "remaining_time": "2:10:05", "throughput": 5157.63, "total_tokens": 2542752} {"current_steps": 2525, "total_steps": 42420, "loss": 0.3895, "lr": 0.000595002357378595, "epoch": 1.1904761904761905, "percentage": 5.95, "elapsed_time": "0:08:13", "remaining_time": "2:10:02", "throughput": 5157.72, "total_tokens": 2547168} {"current_steps": 2530, "total_steps": 42420, "loss": 0.2977, "lr": 0.0005961810466760961, "epoch": 1.1928335690711929, "percentage": 5.96, "elapsed_time": "0:08:14", "remaining_time": "2:10:01", "throughput": 5158.33, "total_tokens": 2552352} {"current_steps": 2535, "total_steps": 42420, "loss": 0.291, "lr": 0.0005973597359735974, "epoch": 1.1951909476661953, "percentage": 5.98, "elapsed_time": "0:08:15", "remaining_time": "2:09:59", "throughput": 5158.84, "total_tokens": 2557376} {"current_steps": 2540, "total_steps": 42420, "loss": 0.2825, "lr": 0.0005985384252710986, "epoch": 1.1975483262611974, "percentage": 5.99, "elapsed_time": "0:08:16", "remaining_time": "2:09:56", "throughput": 5158.99, "total_tokens": 2561792} {"current_steps": 2545, "total_steps": 42420, "loss": 0.2804, "lr": 0.0005997171145685997, "epoch": 1.1999057048561999, "percentage": 6.0, "elapsed_time": "0:08:17", "remaining_time": "2:09:53", "throughput": 5158.8, "total_tokens": 2565952} {"current_steps": 2550, "total_steps": 42420, "loss": 0.3493, "lr": 0.0006008958038661009, "epoch": 1.2022630834512023, "percentage": 6.01, "elapsed_time": "0:08:18", "remaining_time": "2:09:52", "throughput": 5159.37, "total_tokens": 2571232} {"current_steps": 2555, "total_steps": 42420, "loss": 0.2784, "lr": 0.000602074493163602, "epoch": 1.2046204620462047, "percentage": 6.02, "elapsed_time": "0:08:19", "remaining_time": "2:09:48", "throughput": 5159.21, "total_tokens": 2575488} {"current_steps": 2560, "total_steps": 42420, "loss": 0.3179, "lr": 0.0006032531824611033, "epoch": 1.206977840641207, "percentage": 6.03, "elapsed_time": "0:08:20", "remaining_time": "2:09:49", "throughput": 5160.31, "total_tokens": 2581536} {"current_steps": 2565, "total_steps": 42420, "loss": 0.2552, "lr": 0.0006044318717586044, "epoch": 1.2093352192362095, "percentage": 6.05, "elapsed_time": "0:08:21", "remaining_time": "2:09:46", "throughput": 5160.39, "total_tokens": 2585984} {"current_steps": 2570, "total_steps": 42420, "loss": 0.3601, "lr": 0.0006056105610561056, "epoch": 1.2116925978312116, "percentage": 6.06, "elapsed_time": "0:08:21", "remaining_time": "2:09:43", "throughput": 5160.53, "total_tokens": 2590336} {"current_steps": 2575, "total_steps": 42420, "loss": 0.3103, "lr": 0.0006067892503536068, "epoch": 1.214049976426214, "percentage": 6.07, "elapsed_time": "0:08:22", "remaining_time": "2:09:42", "throughput": 5161.54, "total_tokens": 2595968} {"current_steps": 2580, "total_steps": 42420, "loss": 0.3807, "lr": 0.000607967939651108, "epoch": 1.2164073550212164, "percentage": 6.08, "elapsed_time": "0:08:23", "remaining_time": "2:09:39", "throughput": 5161.56, "total_tokens": 2600480} {"current_steps": 2585, "total_steps": 42420, "loss": 0.3115, "lr": 0.0006091466289486092, "epoch": 1.2187647336162188, "percentage": 6.09, "elapsed_time": "0:08:24", "remaining_time": "2:09:36", "throughput": 5161.63, "total_tokens": 2604704} {"current_steps": 2590, "total_steps": 42420, "loss": 0.3238, "lr": 0.0006103253182461103, "epoch": 1.221122112211221, "percentage": 6.11, "elapsed_time": "0:08:25", "remaining_time": "2:09:35", "throughput": 5162.28, "total_tokens": 2609984} {"current_steps": 2595, "total_steps": 42420, "loss": 0.3196, "lr": 0.0006115040075436115, "epoch": 1.2234794908062234, "percentage": 6.12, "elapsed_time": "0:08:26", "remaining_time": "2:09:31", "throughput": 5161.99, "total_tokens": 2613984} {"current_steps": 2600, "total_steps": 42420, "loss": 0.3534, "lr": 0.0006126826968411126, "epoch": 1.2258368694012258, "percentage": 6.13, "elapsed_time": "0:08:27", "remaining_time": "2:09:29", "throughput": 5162.44, "total_tokens": 2618912} {"current_steps": 2605, "total_steps": 42420, "loss": 0.3886, "lr": 0.0006138613861386139, "epoch": 1.2281942479962282, "percentage": 6.14, "elapsed_time": "0:08:28", "remaining_time": "2:09:32", "throughput": 5164.15, "total_tokens": 2626304} {"current_steps": 2610, "total_steps": 42420, "loss": 0.3801, "lr": 0.0006150400754361151, "epoch": 1.2305516265912306, "percentage": 6.15, "elapsed_time": "0:08:29", "remaining_time": "2:09:30", "throughput": 5164.55, "total_tokens": 2630944} {"current_steps": 2615, "total_steps": 42420, "loss": 0.3595, "lr": 0.0006162187647336162, "epoch": 1.232909005186233, "percentage": 6.16, "elapsed_time": "0:08:30", "remaining_time": "2:09:26", "throughput": 5164.28, "total_tokens": 2635008} {"current_steps": 2620, "total_steps": 42420, "loss": 0.3278, "lr": 0.0006173974540311174, "epoch": 1.2352663837812352, "percentage": 6.18, "elapsed_time": "0:08:31", "remaining_time": "2:09:23", "throughput": 5164.27, "total_tokens": 2639264} {"current_steps": 2625, "total_steps": 42420, "loss": 0.3044, "lr": 0.0006185761433286185, "epoch": 1.2376237623762376, "percentage": 6.19, "elapsed_time": "0:08:32", "remaining_time": "2:09:23", "throughput": 5165.03, "total_tokens": 2644928} {"current_steps": 2630, "total_steps": 42420, "loss": 0.3501, "lr": 0.0006197548326261198, "epoch": 1.23998114097124, "percentage": 6.2, "elapsed_time": "0:08:33", "remaining_time": "2:09:21", "throughput": 5165.5, "total_tokens": 2649952} {"current_steps": 2635, "total_steps": 42420, "loss": 0.3028, "lr": 0.0006209335219236209, "epoch": 1.2423385195662424, "percentage": 6.21, "elapsed_time": "0:08:33", "remaining_time": "2:09:19", "throughput": 5165.86, "total_tokens": 2654688} {"current_steps": 2640, "total_steps": 42420, "loss": 0.2892, "lr": 0.0006221122112211222, "epoch": 1.2446958981612446, "percentage": 6.22, "elapsed_time": "0:08:34", "remaining_time": "2:09:17", "throughput": 5166.45, "total_tokens": 2659680} {"current_steps": 2645, "total_steps": 42420, "loss": 0.3508, "lr": 0.0006232909005186233, "epoch": 1.247053276756247, "percentage": 6.24, "elapsed_time": "0:08:35", "remaining_time": "2:09:15", "throughput": 5167.16, "total_tokens": 2664960} {"current_steps": 2650, "total_steps": 42420, "loss": 0.4181, "lr": 0.0006244695898161244, "epoch": 1.2494106553512494, "percentage": 6.25, "elapsed_time": "0:08:36", "remaining_time": "2:09:12", "throughput": 5167.09, "total_tokens": 2669120} {"current_steps": 2655, "total_steps": 42420, "loss": 0.3298, "lr": 0.0006256482791136257, "epoch": 1.2517680339462518, "percentage": 6.26, "elapsed_time": "0:08:37", "remaining_time": "2:09:09", "throughput": 5167.27, "total_tokens": 2673728} {"current_steps": 2660, "total_steps": 42420, "loss": 0.3305, "lr": 0.0006268269684111268, "epoch": 1.2541254125412542, "percentage": 6.27, "elapsed_time": "0:08:38", "remaining_time": "2:09:08", "throughput": 5167.81, "total_tokens": 2678880} {"current_steps": 2665, "total_steps": 42420, "loss": 0.3441, "lr": 0.0006280056577086281, "epoch": 1.2564827911362566, "percentage": 6.28, "elapsed_time": "0:08:39", "remaining_time": "2:09:03", "throughput": 5167.13, "total_tokens": 2682208} {"current_steps": 2670, "total_steps": 42420, "loss": 0.299, "lr": 0.0006291843470061291, "epoch": 1.2588401697312588, "percentage": 6.29, "elapsed_time": "0:08:39", "remaining_time": "2:09:00", "throughput": 5166.88, "total_tokens": 2686400} {"current_steps": 2675, "total_steps": 42420, "loss": 0.3363, "lr": 0.0006303630363036303, "epoch": 1.2611975483262612, "percentage": 6.31, "elapsed_time": "0:08:40", "remaining_time": "2:08:58", "throughput": 5167.04, "total_tokens": 2691008} {"current_steps": 2680, "total_steps": 42420, "loss": 0.3518, "lr": 0.0006315417256011316, "epoch": 1.2635549269212636, "percentage": 6.32, "elapsed_time": "0:08:41", "remaining_time": "2:08:58", "throughput": 5168.34, "total_tokens": 2697184} {"current_steps": 2685, "total_steps": 42420, "loss": 0.3165, "lr": 0.0006327204148986327, "epoch": 1.265912305516266, "percentage": 6.33, "elapsed_time": "0:08:42", "remaining_time": "2:08:56", "throughput": 5168.51, "total_tokens": 2701856} {"current_steps": 2690, "total_steps": 42420, "loss": 0.3082, "lr": 0.000633899104196134, "epoch": 1.2682696841112682, "percentage": 6.34, "elapsed_time": "0:08:43", "remaining_time": "2:08:54", "throughput": 5169.21, "total_tokens": 2707072} {"current_steps": 2695, "total_steps": 42420, "loss": 0.2728, "lr": 0.000635077793493635, "epoch": 1.2706270627062706, "percentage": 6.35, "elapsed_time": "0:08:44", "remaining_time": "2:08:51", "throughput": 5169.16, "total_tokens": 2711456} {"current_steps": 2700, "total_steps": 42420, "loss": 0.2836, "lr": 0.0006362564827911363, "epoch": 1.272984441301273, "percentage": 6.36, "elapsed_time": "0:08:45", "remaining_time": "2:08:49", "throughput": 5169.33, "total_tokens": 2716128} {"current_steps": 2705, "total_steps": 42420, "loss": 0.2961, "lr": 0.0006374351720886374, "epoch": 1.2753418198962754, "percentage": 6.38, "elapsed_time": "0:08:46", "remaining_time": "2:08:46", "throughput": 5169.45, "total_tokens": 2720544} {"current_steps": 2710, "total_steps": 42420, "loss": 0.2236, "lr": 0.0006386138613861386, "epoch": 1.2776991984912778, "percentage": 6.39, "elapsed_time": "0:08:47", "remaining_time": "2:08:45", "throughput": 5170.27, "total_tokens": 2725888} {"current_steps": 2715, "total_steps": 42420, "loss": 0.3901, "lr": 0.0006397925506836399, "epoch": 1.2800565770862802, "percentage": 6.4, "elapsed_time": "0:08:48", "remaining_time": "2:08:44", "throughput": 5170.92, "total_tokens": 2731104} {"current_steps": 2720, "total_steps": 42420, "loss": 0.3481, "lr": 0.0006409712399811409, "epoch": 1.2824139556812824, "percentage": 6.41, "elapsed_time": "0:08:49", "remaining_time": "2:08:41", "throughput": 5171.4, "total_tokens": 2735936} {"current_steps": 2725, "total_steps": 42420, "loss": 0.2563, "lr": 0.0006421499292786422, "epoch": 1.2847713342762848, "percentage": 6.42, "elapsed_time": "0:08:49", "remaining_time": "2:08:39", "throughput": 5171.89, "total_tokens": 2740896} {"current_steps": 2730, "total_steps": 42420, "loss": 0.3802, "lr": 0.0006433286185761433, "epoch": 1.2871287128712872, "percentage": 6.44, "elapsed_time": "0:08:50", "remaining_time": "2:08:37", "throughput": 5172.15, "total_tokens": 2745696} {"current_steps": 2735, "total_steps": 42420, "loss": 0.3159, "lr": 0.0006445073078736445, "epoch": 1.2894860914662896, "percentage": 6.45, "elapsed_time": "0:08:51", "remaining_time": "2:08:36", "throughput": 5172.16, "total_tokens": 2750400} {"current_steps": 2740, "total_steps": 42420, "loss": 0.3449, "lr": 0.0006456859971711457, "epoch": 1.2918434700612917, "percentage": 6.46, "elapsed_time": "0:08:52", "remaining_time": "2:08:34", "throughput": 5172.69, "total_tokens": 2755360} {"current_steps": 2745, "total_steps": 42420, "loss": 0.3115, "lr": 0.0006468646864686468, "epoch": 1.2942008486562941, "percentage": 6.47, "elapsed_time": "0:08:53", "remaining_time": "2:08:35", "throughput": 5174.19, "total_tokens": 2762112} {"current_steps": 2750, "total_steps": 42420, "loss": 0.3516, "lr": 0.0006480433757661481, "epoch": 1.2965582272512965, "percentage": 6.48, "elapsed_time": "0:08:54", "remaining_time": "2:08:33", "throughput": 5174.39, "total_tokens": 2766784} {"current_steps": 2755, "total_steps": 42420, "loss": 0.3387, "lr": 0.0006492220650636492, "epoch": 1.298915605846299, "percentage": 6.49, "elapsed_time": "0:08:55", "remaining_time": "2:08:32", "throughput": 5175.16, "total_tokens": 2772160} {"current_steps": 2760, "total_steps": 42420, "loss": 0.3345, "lr": 0.0006504007543611505, "epoch": 1.3012729844413014, "percentage": 6.51, "elapsed_time": "0:08:56", "remaining_time": "2:08:32", "throughput": 5176.24, "total_tokens": 2778112} {"current_steps": 2765, "total_steps": 42420, "loss": 0.3762, "lr": 0.0006515794436586516, "epoch": 1.3036303630363038, "percentage": 6.52, "elapsed_time": "0:08:57", "remaining_time": "2:08:29", "throughput": 5176.53, "total_tokens": 2782848} {"current_steps": 2770, "total_steps": 42420, "loss": 0.3378, "lr": 0.0006527581329561527, "epoch": 1.305987741631306, "percentage": 6.53, "elapsed_time": "0:08:58", "remaining_time": "2:08:29", "throughput": 5177.43, "total_tokens": 2788416} {"current_steps": 2775, "total_steps": 42420, "loss": 0.3485, "lr": 0.0006539368222536539, "epoch": 1.3083451202263083, "percentage": 6.54, "elapsed_time": "0:08:59", "remaining_time": "2:08:27", "throughput": 5177.7, "total_tokens": 2793440} {"current_steps": 2780, "total_steps": 42420, "loss": 0.3583, "lr": 0.0006551155115511551, "epoch": 1.3107024988213107, "percentage": 6.55, "elapsed_time": "0:09:00", "remaining_time": "2:08:25", "throughput": 5177.85, "total_tokens": 2797984} {"current_steps": 2785, "total_steps": 42420, "loss": 0.3407, "lr": 0.0006562942008486564, "epoch": 1.3130598774163131, "percentage": 6.57, "elapsed_time": "0:09:01", "remaining_time": "2:08:23", "throughput": 5178.34, "total_tokens": 2803136} {"current_steps": 2790, "total_steps": 42420, "loss": 0.3597, "lr": 0.0006574728901461575, "epoch": 1.3154172560113153, "percentage": 6.58, "elapsed_time": "0:09:02", "remaining_time": "2:08:24", "throughput": 5179.49, "total_tokens": 2809344} {"current_steps": 2795, "total_steps": 42420, "loss": 0.3331, "lr": 0.0006586515794436586, "epoch": 1.3177746346063177, "percentage": 6.59, "elapsed_time": "0:09:03", "remaining_time": "2:08:22", "throughput": 5179.76, "total_tokens": 2814176} {"current_steps": 2800, "total_steps": 42420, "loss": 0.2891, "lr": 0.0006598302687411598, "epoch": 1.3201320132013201, "percentage": 6.6, "elapsed_time": "0:09:04", "remaining_time": "2:08:20", "throughput": 5180.31, "total_tokens": 2819296} {"current_steps": 2805, "total_steps": 42420, "loss": 0.3554, "lr": 0.000661008958038661, "epoch": 1.3224893917963225, "percentage": 6.61, "elapsed_time": "0:09:05", "remaining_time": "2:08:18", "throughput": 5180.63, "total_tokens": 2823936} {"current_steps": 2810, "total_steps": 42420, "loss": 0.3701, "lr": 0.0006621876473361622, "epoch": 1.324846770391325, "percentage": 6.62, "elapsed_time": "0:09:06", "remaining_time": "2:08:17", "throughput": 5181.28, "total_tokens": 2829248} {"current_steps": 2815, "total_steps": 42420, "loss": 0.3693, "lr": 0.0006633663366336634, "epoch": 1.3272041489863273, "percentage": 6.64, "elapsed_time": "0:09:06", "remaining_time": "2:08:15", "throughput": 5181.9, "total_tokens": 2834304} {"current_steps": 2820, "total_steps": 42420, "loss": 0.3408, "lr": 0.0006645450259311646, "epoch": 1.3295615275813295, "percentage": 6.65, "elapsed_time": "0:09:07", "remaining_time": "2:08:12", "throughput": 5182.15, "total_tokens": 2838944} {"current_steps": 2825, "total_steps": 42420, "loss": 0.3421, "lr": 0.0006657237152286657, "epoch": 1.331918906176332, "percentage": 6.66, "elapsed_time": "0:09:08", "remaining_time": "2:08:11", "throughput": 5182.53, "total_tokens": 2843872} {"current_steps": 2830, "total_steps": 42420, "loss": 0.3249, "lr": 0.0006669024045261669, "epoch": 1.3342762847713343, "percentage": 6.67, "elapsed_time": "0:09:09", "remaining_time": "2:08:11", "throughput": 5183.63, "total_tokens": 2850176} {"current_steps": 2835, "total_steps": 42420, "loss": 0.3823, "lr": 0.0006680810938236681, "epoch": 1.3366336633663367, "percentage": 6.68, "elapsed_time": "0:09:10", "remaining_time": "2:08:09", "throughput": 5183.82, "total_tokens": 2854880} {"current_steps": 2840, "total_steps": 42420, "loss": 0.3298, "lr": 0.0006692597831211693, "epoch": 1.338991041961339, "percentage": 6.69, "elapsed_time": "0:09:11", "remaining_time": "2:08:08", "throughput": 5184.65, "total_tokens": 2860320} {"current_steps": 2845, "total_steps": 42420, "loss": 0.3241, "lr": 0.0006704384724186704, "epoch": 1.3413484205563413, "percentage": 6.71, "elapsed_time": "0:09:12", "remaining_time": "2:08:05", "throughput": 5184.75, "total_tokens": 2864640} {"current_steps": 2850, "total_steps": 42420, "loss": 0.351, "lr": 0.0006716171617161716, "epoch": 1.3437057991513437, "percentage": 6.72, "elapsed_time": "0:09:13", "remaining_time": "2:08:04", "throughput": 5185.23, "total_tokens": 2869824} {"current_steps": 2855, "total_steps": 42420, "loss": 0.3295, "lr": 0.0006727958510136728, "epoch": 1.346063177746346, "percentage": 6.73, "elapsed_time": "0:09:14", "remaining_time": "2:08:05", "throughput": 5186.49, "total_tokens": 2876416} {"current_steps": 2860, "total_steps": 42420, "loss": 0.3459, "lr": 0.000673974540311174, "epoch": 1.3484205563413485, "percentage": 6.74, "elapsed_time": "0:09:15", "remaining_time": "2:08:03", "throughput": 5186.65, "total_tokens": 2880928} {"current_steps": 2865, "total_steps": 42420, "loss": 0.3321, "lr": 0.0006751532296086752, "epoch": 1.350777934936351, "percentage": 6.75, "elapsed_time": "0:09:16", "remaining_time": "2:08:01", "throughput": 5186.95, "total_tokens": 2885792} {"current_steps": 2870, "total_steps": 42420, "loss": 0.358, "lr": 0.0006763319189061763, "epoch": 1.353135313531353, "percentage": 6.77, "elapsed_time": "0:09:17", "remaining_time": "2:07:57", "throughput": 5186.52, "total_tokens": 2889600} {"current_steps": 2875, "total_steps": 42420, "loss": 0.3433, "lr": 0.0006775106082036775, "epoch": 1.3554926921263555, "percentage": 6.78, "elapsed_time": "0:09:18", "remaining_time": "2:07:56", "throughput": 5187.5, "total_tokens": 2895168} {"current_steps": 2880, "total_steps": 42420, "loss": 0.3328, "lr": 0.0006786892975011786, "epoch": 1.3578500707213579, "percentage": 6.79, "elapsed_time": "0:09:19", "remaining_time": "2:07:55", "throughput": 5187.93, "total_tokens": 2900256} {"current_steps": 2885, "total_steps": 42420, "loss": 0.3452, "lr": 0.0006798679867986799, "epoch": 1.3602074493163603, "percentage": 6.8, "elapsed_time": "0:09:19", "remaining_time": "2:07:52", "throughput": 5187.97, "total_tokens": 2904640} {"current_steps": 2890, "total_steps": 42420, "loss": 0.3389, "lr": 0.0006810466760961811, "epoch": 1.3625648279113625, "percentage": 6.81, "elapsed_time": "0:09:20", "remaining_time": "2:07:49", "throughput": 5187.77, "total_tokens": 2908896} {"current_steps": 2895, "total_steps": 42420, "loss": 0.3613, "lr": 0.0006822253653936822, "epoch": 1.3649222065063649, "percentage": 6.82, "elapsed_time": "0:09:21", "remaining_time": "2:07:49", "throughput": 5188.77, "total_tokens": 2914880} {"current_steps": 2900, "total_steps": 42420, "loss": 0.3106, "lr": 0.0006834040546911834, "epoch": 1.3672795851013673, "percentage": 6.84, "elapsed_time": "0:09:22", "remaining_time": "2:07:47", "throughput": 5188.94, "total_tokens": 2919456} {"current_steps": 2905, "total_steps": 42420, "loss": 0.3865, "lr": 0.0006845827439886846, "epoch": 1.3696369636963697, "percentage": 6.85, "elapsed_time": "0:09:23", "remaining_time": "2:07:45", "throughput": 5189.29, "total_tokens": 2924384} {"current_steps": 2910, "total_steps": 42420, "loss": 0.3478, "lr": 0.0006857614332861858, "epoch": 1.371994342291372, "percentage": 6.86, "elapsed_time": "0:09:24", "remaining_time": "2:07:44", "throughput": 5189.66, "total_tokens": 2929440} {"current_steps": 2915, "total_steps": 42420, "loss": 0.322, "lr": 0.0006869401225836869, "epoch": 1.3743517208863745, "percentage": 6.87, "elapsed_time": "0:09:25", "remaining_time": "2:07:42", "throughput": 5189.65, "total_tokens": 2934176} {"current_steps": 2920, "total_steps": 42420, "loss": 0.3117, "lr": 0.0006881188118811881, "epoch": 1.3767090994813767, "percentage": 6.88, "elapsed_time": "0:09:26", "remaining_time": "2:07:39", "throughput": 5189.48, "total_tokens": 2938208} {"current_steps": 2925, "total_steps": 42420, "loss": 0.344, "lr": 0.0006892975011786893, "epoch": 1.379066478076379, "percentage": 6.9, "elapsed_time": "0:09:27", "remaining_time": "2:07:36", "throughput": 5189.74, "total_tokens": 2942976} {"current_steps": 2930, "total_steps": 42420, "loss": 0.3081, "lr": 0.0006904761904761905, "epoch": 1.3814238566713815, "percentage": 6.91, "elapsed_time": "0:09:28", "remaining_time": "2:07:35", "throughput": 5190.23, "total_tokens": 2948160} {"current_steps": 2935, "total_steps": 42420, "loss": 0.3321, "lr": 0.0006916548797736917, "epoch": 1.3837812352663839, "percentage": 6.92, "elapsed_time": "0:09:28", "remaining_time": "2:07:34", "throughput": 5190.81, "total_tokens": 2953568} {"current_steps": 2940, "total_steps": 42420, "loss": 0.3714, "lr": 0.0006928335690711929, "epoch": 1.386138613861386, "percentage": 6.93, "elapsed_time": "0:09:29", "remaining_time": "2:07:34", "throughput": 5191.42, "total_tokens": 2959008} {"current_steps": 2945, "total_steps": 42420, "loss": 0.3308, "lr": 0.000694012258368694, "epoch": 1.3884959924563884, "percentage": 6.94, "elapsed_time": "0:09:31", "remaining_time": "2:07:34", "throughput": 5192.54, "total_tokens": 2965088} {"current_steps": 2950, "total_steps": 42420, "loss": 0.3471, "lr": 0.0006951909476661951, "epoch": 1.3908533710513908, "percentage": 6.95, "elapsed_time": "0:09:31", "remaining_time": "2:07:32", "throughput": 5192.89, "total_tokens": 2969888} {"current_steps": 2955, "total_steps": 42420, "loss": 0.3436, "lr": 0.0006963696369636964, "epoch": 1.3932107496463932, "percentage": 6.97, "elapsed_time": "0:09:32", "remaining_time": "2:07:31", "throughput": 5193.81, "total_tokens": 2975680} {"current_steps": 2960, "total_steps": 42420, "loss": 0.3075, "lr": 0.0006975483262611976, "epoch": 1.3955681282413956, "percentage": 6.98, "elapsed_time": "0:09:33", "remaining_time": "2:07:28", "throughput": 5193.67, "total_tokens": 2979712} {"current_steps": 2965, "total_steps": 42420, "loss": 0.2832, "lr": 0.0006987270155586988, "epoch": 1.397925506836398, "percentage": 6.99, "elapsed_time": "0:09:34", "remaining_time": "2:07:27", "throughput": 5194.39, "total_tokens": 2985216} {"current_steps": 2970, "total_steps": 42420, "loss": 0.2936, "lr": 0.0006999057048561999, "epoch": 1.4002828854314002, "percentage": 7.0, "elapsed_time": "0:09:35", "remaining_time": "2:07:25", "throughput": 5194.72, "total_tokens": 2990048} {"current_steps": 2975, "total_steps": 42420, "loss": 0.4255, "lr": 0.000701084394153701, "epoch": 1.4026402640264026, "percentage": 7.01, "elapsed_time": "0:09:36", "remaining_time": "2:07:24", "throughput": 5195.37, "total_tokens": 2995328} {"current_steps": 2980, "total_steps": 42420, "loss": 0.3718, "lr": 0.0007022630834512023, "epoch": 1.404997642621405, "percentage": 7.02, "elapsed_time": "0:09:37", "remaining_time": "2:07:21", "throughput": 5195.4, "total_tokens": 2999616} {"current_steps": 2985, "total_steps": 42420, "loss": 0.3636, "lr": 0.0007034417727487034, "epoch": 1.4073550212164074, "percentage": 7.04, "elapsed_time": "0:09:38", "remaining_time": "2:07:19", "throughput": 5195.8, "total_tokens": 3004704} {"current_steps": 2990, "total_steps": 42420, "loss": 0.3538, "lr": 0.0007046204620462047, "epoch": 1.4097123998114096, "percentage": 7.05, "elapsed_time": "0:09:39", "remaining_time": "2:07:19", "throughput": 5196.42, "total_tokens": 3010144} {"current_steps": 2995, "total_steps": 42420, "loss": 0.3522, "lr": 0.0007057991513437058, "epoch": 1.412069778406412, "percentage": 7.06, "elapsed_time": "0:09:40", "remaining_time": "2:07:18", "throughput": 5197.0, "total_tokens": 3015616} {"current_steps": 3000, "total_steps": 42420, "loss": 0.3547, "lr": 0.0007069778406412069, "epoch": 1.4144271570014144, "percentage": 7.07, "elapsed_time": "0:09:41", "remaining_time": "2:07:16", "throughput": 5197.65, "total_tokens": 3020832} {"current_steps": 3005, "total_steps": 42420, "loss": 0.3387, "lr": 0.0007081565299387082, "epoch": 1.4167845355964168, "percentage": 7.08, "elapsed_time": "0:09:42", "remaining_time": "2:07:14", "throughput": 5197.72, "total_tokens": 3025408} {"current_steps": 3010, "total_steps": 42420, "loss": 0.3325, "lr": 0.0007093352192362093, "epoch": 1.4191419141914192, "percentage": 7.1, "elapsed_time": "0:09:42", "remaining_time": "2:07:11", "throughput": 5197.42, "total_tokens": 3029216} {"current_steps": 3015, "total_steps": 42420, "loss": 0.3557, "lr": 0.0007105139085337106, "epoch": 1.4214992927864216, "percentage": 7.11, "elapsed_time": "0:09:43", "remaining_time": "2:07:09", "throughput": 5197.8, "total_tokens": 3034080} {"current_steps": 3020, "total_steps": 42420, "loss": 0.3509, "lr": 0.0007116925978312116, "epoch": 1.4238566713814238, "percentage": 7.12, "elapsed_time": "0:09:44", "remaining_time": "2:07:07", "throughput": 5198.28, "total_tokens": 3039232} {"current_steps": 3025, "total_steps": 42420, "loss": 0.2983, "lr": 0.0007128712871287129, "epoch": 1.4262140499764262, "percentage": 7.13, "elapsed_time": "0:09:45", "remaining_time": "2:07:05", "throughput": 5198.21, "total_tokens": 3043552} {"current_steps": 3030, "total_steps": 42420, "loss": 0.2946, "lr": 0.0007140499764262141, "epoch": 1.4285714285714286, "percentage": 7.14, "elapsed_time": "0:09:46", "remaining_time": "2:07:04", "throughput": 5199.17, "total_tokens": 3049440} {"current_steps": 3035, "total_steps": 42420, "loss": 0.3071, "lr": 0.0007152286657237152, "epoch": 1.430928807166431, "percentage": 7.15, "elapsed_time": "0:09:47", "remaining_time": "2:07:03", "throughput": 5199.79, "total_tokens": 3054752} {"current_steps": 3040, "total_steps": 42420, "loss": 0.362, "lr": 0.0007164073550212165, "epoch": 1.4332861857614332, "percentage": 7.17, "elapsed_time": "0:09:48", "remaining_time": "2:07:02", "throughput": 5200.28, "total_tokens": 3059904} {"current_steps": 3045, "total_steps": 42420, "loss": 0.3121, "lr": 0.0007175860443187175, "epoch": 1.4356435643564356, "percentage": 7.18, "elapsed_time": "0:09:49", "remaining_time": "2:06:58", "throughput": 5199.88, "total_tokens": 3063648} {"current_steps": 3050, "total_steps": 42420, "loss": 0.3775, "lr": 0.0007187647336162188, "epoch": 1.438000942951438, "percentage": 7.19, "elapsed_time": "0:09:50", "remaining_time": "2:06:56", "throughput": 5199.98, "total_tokens": 3068320} {"current_steps": 3055, "total_steps": 42420, "loss": 0.3592, "lr": 0.0007199434229137199, "epoch": 1.4403583215464404, "percentage": 7.2, "elapsed_time": "0:09:51", "remaining_time": "2:06:56", "throughput": 5200.96, "total_tokens": 3074080} {"current_steps": 3060, "total_steps": 42420, "loss": 0.5095, "lr": 0.0007211221122112211, "epoch": 1.4427157001414428, "percentage": 7.21, "elapsed_time": "0:09:52", "remaining_time": "2:06:55", "throughput": 5201.6, "total_tokens": 3079584} {"current_steps": 3065, "total_steps": 42420, "loss": 0.3632, "lr": 0.0007223008015087224, "epoch": 1.4450730787364452, "percentage": 7.23, "elapsed_time": "0:09:52", "remaining_time": "2:06:53", "throughput": 5201.81, "total_tokens": 3084256} {"current_steps": 3070, "total_steps": 42420, "loss": 0.3334, "lr": 0.0007234794908062234, "epoch": 1.4474304573314474, "percentage": 7.24, "elapsed_time": "0:09:53", "remaining_time": "2:06:51", "throughput": 5202.13, "total_tokens": 3089248} {"current_steps": 3075, "total_steps": 42420, "loss": 0.344, "lr": 0.0007246581801037247, "epoch": 1.4497878359264498, "percentage": 7.25, "elapsed_time": "0:09:54", "remaining_time": "2:06:49", "throughput": 5201.89, "total_tokens": 3093504} {"current_steps": 3080, "total_steps": 42420, "loss": 0.3324, "lr": 0.0007258368694012258, "epoch": 1.4521452145214522, "percentage": 7.26, "elapsed_time": "0:09:55", "remaining_time": "2:06:50", "throughput": 5203.36, "total_tokens": 3100480} {"current_steps": 3085, "total_steps": 42420, "loss": 0.3541, "lr": 0.0007270155586987271, "epoch": 1.4545025931164546, "percentage": 7.27, "elapsed_time": "0:09:56", "remaining_time": "2:06:49", "throughput": 5204.12, "total_tokens": 3105888} {"current_steps": 3090, "total_steps": 42420, "loss": 0.319, "lr": 0.0007281942479962282, "epoch": 1.4568599717114568, "percentage": 7.28, "elapsed_time": "0:09:57", "remaining_time": "2:06:48", "throughput": 5204.65, "total_tokens": 3111200} {"current_steps": 3095, "total_steps": 42420, "loss": 0.3519, "lr": 0.0007293729372937293, "epoch": 1.4592173503064592, "percentage": 7.3, "elapsed_time": "0:09:58", "remaining_time": "2:06:46", "throughput": 5204.93, "total_tokens": 3116000} {"current_steps": 3100, "total_steps": 42420, "loss": 0.3399, "lr": 0.0007305516265912306, "epoch": 1.4615747289014616, "percentage": 7.31, "elapsed_time": "0:09:59", "remaining_time": "2:06:45", "throughput": 5205.38, "total_tokens": 3121152} {"current_steps": 3105, "total_steps": 42420, "loss": 0.3883, "lr": 0.0007317303158887317, "epoch": 1.463932107496464, "percentage": 7.32, "elapsed_time": "0:10:00", "remaining_time": "2:06:45", "throughput": 5206.22, "total_tokens": 3127264} {"current_steps": 3110, "total_steps": 42420, "loss": 0.3144, "lr": 0.000732909005186233, "epoch": 1.4662894860914664, "percentage": 7.33, "elapsed_time": "0:10:01", "remaining_time": "2:06:43", "throughput": 5206.29, "total_tokens": 3131744} {"current_steps": 3115, "total_steps": 42420, "loss": 0.3396, "lr": 0.0007340876944837341, "epoch": 1.4686468646864688, "percentage": 7.34, "elapsed_time": "0:10:02", "remaining_time": "2:06:41", "throughput": 5206.52, "total_tokens": 3136480} {"current_steps": 3120, "total_steps": 42420, "loss": 0.289, "lr": 0.0007352663837812352, "epoch": 1.471004243281471, "percentage": 7.36, "elapsed_time": "0:10:03", "remaining_time": "2:06:39", "throughput": 5206.66, "total_tokens": 3141088} {"current_steps": 3125, "total_steps": 42420, "loss": 0.3594, "lr": 0.0007364450730787364, "epoch": 1.4733616218764733, "percentage": 7.37, "elapsed_time": "0:10:04", "remaining_time": "2:06:38", "throughput": 5207.19, "total_tokens": 3146464} {"current_steps": 3130, "total_steps": 42420, "loss": 0.3746, "lr": 0.0007376237623762376, "epoch": 1.4757190004714758, "percentage": 7.38, "elapsed_time": "0:10:05", "remaining_time": "2:06:36", "throughput": 5207.73, "total_tokens": 3151520} {"current_steps": 3135, "total_steps": 42420, "loss": 0.3467, "lr": 0.0007388024516737389, "epoch": 1.4780763790664782, "percentage": 7.39, "elapsed_time": "0:10:06", "remaining_time": "2:06:34", "throughput": 5207.96, "total_tokens": 3156192} {"current_steps": 3140, "total_steps": 42420, "loss": 0.3478, "lr": 0.00073998114097124, "epoch": 1.4804337576614803, "percentage": 7.4, "elapsed_time": "0:10:06", "remaining_time": "2:06:31", "throughput": 5207.98, "total_tokens": 3160672} {"current_steps": 3145, "total_steps": 42420, "loss": 0.3768, "lr": 0.0007411598302687412, "epoch": 1.4827911362564827, "percentage": 7.41, "elapsed_time": "0:10:07", "remaining_time": "2:06:30", "throughput": 5208.15, "total_tokens": 3165632} {"current_steps": 3150, "total_steps": 42420, "loss": 0.3797, "lr": 0.0007423385195662423, "epoch": 1.4851485148514851, "percentage": 7.43, "elapsed_time": "0:10:08", "remaining_time": "2:06:29", "throughput": 5208.73, "total_tokens": 3170944} {"current_steps": 3155, "total_steps": 42420, "loss": 0.3549, "lr": 0.0007435172088637435, "epoch": 1.4875058934464875, "percentage": 7.44, "elapsed_time": "0:10:09", "remaining_time": "2:06:27", "throughput": 5208.75, "total_tokens": 3175392} {"current_steps": 3160, "total_steps": 42420, "loss": 0.3603, "lr": 0.0007446958981612447, "epoch": 1.48986327204149, "percentage": 7.45, "elapsed_time": "0:10:10", "remaining_time": "2:06:26", "throughput": 5209.76, "total_tokens": 3181440} {"current_steps": 3165, "total_steps": 42420, "loss": 0.3552, "lr": 0.0007458745874587459, "epoch": 1.4922206506364923, "percentage": 7.46, "elapsed_time": "0:10:11", "remaining_time": "2:06:24", "throughput": 5209.51, "total_tokens": 3185632} {"current_steps": 3170, "total_steps": 42420, "loss": 0.3401, "lr": 0.0007470532767562471, "epoch": 1.4945780292314945, "percentage": 7.47, "elapsed_time": "0:10:12", "remaining_time": "2:06:22", "throughput": 5209.6, "total_tokens": 3190240} {"current_steps": 3175, "total_steps": 42420, "loss": 0.3378, "lr": 0.0007482319660537482, "epoch": 1.496935407826497, "percentage": 7.48, "elapsed_time": "0:10:13", "remaining_time": "2:06:21", "throughput": 5210.1, "total_tokens": 3195488} {"current_steps": 3180, "total_steps": 42420, "loss": 0.3192, "lr": 0.0007494106553512494, "epoch": 1.4992927864214993, "percentage": 7.5, "elapsed_time": "0:10:14", "remaining_time": "2:06:20", "throughput": 5210.74, "total_tokens": 3201024} {"current_steps": 3185, "total_steps": 42420, "loss": 0.3345, "lr": 0.0007505893446487506, "epoch": 1.5016501650165015, "percentage": 7.51, "elapsed_time": "0:10:15", "remaining_time": "2:06:18", "throughput": 5211.01, "total_tokens": 3205888} {"current_steps": 3190, "total_steps": 42420, "loss": 0.317, "lr": 0.0007517680339462518, "epoch": 1.504007543611504, "percentage": 7.52, "elapsed_time": "0:10:15", "remaining_time": "2:06:15", "throughput": 5210.77, "total_tokens": 3209824} {"current_steps": 3195, "total_steps": 42420, "loss": 0.3859, "lr": 0.000752946723243753, "epoch": 1.5063649222065063, "percentage": 7.53, "elapsed_time": "0:10:16", "remaining_time": "2:06:12", "throughput": 5210.66, "total_tokens": 3214016} {"current_steps": 3200, "total_steps": 42420, "loss": 0.362, "lr": 0.0007541254125412541, "epoch": 1.5087223008015087, "percentage": 7.54, "elapsed_time": "0:10:17", "remaining_time": "2:06:10", "throughput": 5210.47, "total_tokens": 3218336} {"current_steps": 3205, "total_steps": 42420, "loss": 0.3547, "lr": 0.0007553041018387554, "epoch": 1.511079679396511, "percentage": 7.56, "elapsed_time": "0:10:18", "remaining_time": "2:06:09", "throughput": 5210.88, "total_tokens": 3223616} {"current_steps": 3210, "total_steps": 42420, "loss": 0.3363, "lr": 0.0007564827911362565, "epoch": 1.5134370579915135, "percentage": 7.57, "elapsed_time": "0:10:19", "remaining_time": "2:06:07", "throughput": 5211.22, "total_tokens": 3228544} {"current_steps": 3215, "total_steps": 42420, "loss": 0.3341, "lr": 0.0007576614804337577, "epoch": 1.515794436586516, "percentage": 7.58, "elapsed_time": "0:10:20", "remaining_time": "2:06:05", "throughput": 5211.11, "total_tokens": 3232832} {"current_steps": 3220, "total_steps": 42420, "loss": 0.3535, "lr": 0.0007588401697312589, "epoch": 1.5181518151815183, "percentage": 7.59, "elapsed_time": "0:10:21", "remaining_time": "2:06:02", "throughput": 5211.0, "total_tokens": 3237152} {"current_steps": 3225, "total_steps": 42420, "loss": 0.3071, "lr": 0.00076001885902876, "epoch": 1.5205091937765205, "percentage": 7.6, "elapsed_time": "0:10:22", "remaining_time": "2:06:00", "throughput": 5211.07, "total_tokens": 3241824} {"current_steps": 3230, "total_steps": 42420, "loss": 0.3054, "lr": 0.0007611975483262612, "epoch": 1.522866572371523, "percentage": 7.61, "elapsed_time": "0:10:22", "remaining_time": "2:05:58", "throughput": 5211.02, "total_tokens": 3246112} {"current_steps": 3235, "total_steps": 42420, "loss": 0.2998, "lr": 0.0007623762376237624, "epoch": 1.525223950966525, "percentage": 7.63, "elapsed_time": "0:10:23", "remaining_time": "2:05:55", "throughput": 5211.13, "total_tokens": 3250592} {"current_steps": 3240, "total_steps": 42420, "loss": 0.4226, "lr": 0.0007635549269212636, "epoch": 1.5275813295615275, "percentage": 7.64, "elapsed_time": "0:10:24", "remaining_time": "2:05:52", "throughput": 5210.84, "total_tokens": 3254528} {"current_steps": 3245, "total_steps": 42420, "loss": 0.3863, "lr": 0.0007647336162187648, "epoch": 1.5299387081565299, "percentage": 7.65, "elapsed_time": "0:10:25", "remaining_time": "2:05:52", "throughput": 5211.5, "total_tokens": 3260160} {"current_steps": 3250, "total_steps": 42420, "loss": 0.3494, "lr": 0.0007659123055162659, "epoch": 1.5322960867515323, "percentage": 7.66, "elapsed_time": "0:10:26", "remaining_time": "2:05:49", "throughput": 5211.7, "total_tokens": 3264672} {"current_steps": 3255, "total_steps": 42420, "loss": 0.3571, "lr": 0.0007670909948137671, "epoch": 1.5346534653465347, "percentage": 7.67, "elapsed_time": "0:10:27", "remaining_time": "2:05:47", "throughput": 5211.74, "total_tokens": 3269344} {"current_steps": 3260, "total_steps": 42420, "loss": 0.3169, "lr": 0.0007682696841112683, "epoch": 1.537010843941537, "percentage": 7.69, "elapsed_time": "0:10:28", "remaining_time": "2:05:45", "throughput": 5211.69, "total_tokens": 3273536} {"current_steps": 3265, "total_steps": 42420, "loss": 0.2972, "lr": 0.0007694483734087695, "epoch": 1.5393682225365395, "percentage": 7.7, "elapsed_time": "0:10:28", "remaining_time": "2:05:42", "throughput": 5211.79, "total_tokens": 3278016} {"current_steps": 3270, "total_steps": 42420, "loss": 0.3531, "lr": 0.0007706270627062707, "epoch": 1.541725601131542, "percentage": 7.71, "elapsed_time": "0:10:29", "remaining_time": "2:05:41", "throughput": 5212.18, "total_tokens": 3282976} {"current_steps": 3275, "total_steps": 42420, "loss": 0.2917, "lr": 0.0007718057520037718, "epoch": 1.544082979726544, "percentage": 7.72, "elapsed_time": "0:10:30", "remaining_time": "2:05:41", "throughput": 5213.18, "total_tokens": 3289184} {"current_steps": 3280, "total_steps": 42420, "loss": 0.3114, "lr": 0.000772984441301273, "epoch": 1.5464403583215465, "percentage": 7.73, "elapsed_time": "0:10:31", "remaining_time": "2:05:39", "throughput": 5213.33, "total_tokens": 3293952} {"current_steps": 3285, "total_steps": 42420, "loss": 0.395, "lr": 0.0007741631305987742, "epoch": 1.5487977369165487, "percentage": 7.74, "elapsed_time": "0:10:32", "remaining_time": "2:05:37", "throughput": 5213.24, "total_tokens": 3298432} {"current_steps": 3290, "total_steps": 42420, "loss": 0.3324, "lr": 0.0007753418198962754, "epoch": 1.551155115511551, "percentage": 7.76, "elapsed_time": "0:10:33", "remaining_time": "2:05:35", "throughput": 5213.39, "total_tokens": 3303168} {"current_steps": 3295, "total_steps": 42420, "loss": 0.3504, "lr": 0.0007765205091937766, "epoch": 1.5535124941065535, "percentage": 7.77, "elapsed_time": "0:10:34", "remaining_time": "2:05:35", "throughput": 5214.26, "total_tokens": 3309184} {"current_steps": 3300, "total_steps": 42420, "loss": 0.3444, "lr": 0.0007776991984912776, "epoch": 1.5558698727015559, "percentage": 7.78, "elapsed_time": "0:10:35", "remaining_time": "2:05:35", "throughput": 5215.02, "total_tokens": 3314976} {"current_steps": 3305, "total_steps": 42420, "loss": 0.349, "lr": 0.0007788778877887789, "epoch": 1.5582272512965583, "percentage": 7.79, "elapsed_time": "0:10:36", "remaining_time": "2:05:32", "throughput": 5214.78, "total_tokens": 3318976} {"current_steps": 3310, "total_steps": 42420, "loss": 0.3562, "lr": 0.0007800565770862801, "epoch": 1.5605846298915607, "percentage": 7.8, "elapsed_time": "0:10:37", "remaining_time": "2:05:30", "throughput": 5215.05, "total_tokens": 3323904} {"current_steps": 3315, "total_steps": 42420, "loss": 0.3582, "lr": 0.0007812352663837813, "epoch": 1.562942008486563, "percentage": 7.81, "elapsed_time": "0:10:38", "remaining_time": "2:05:29", "throughput": 5215.45, "total_tokens": 3328896} {"current_steps": 3320, "total_steps": 42420, "loss": 0.3489, "lr": 0.0007824139556812825, "epoch": 1.5652993870815655, "percentage": 7.83, "elapsed_time": "0:10:39", "remaining_time": "2:05:26", "throughput": 5215.39, "total_tokens": 3333184} {"current_steps": 3325, "total_steps": 42420, "loss": 0.3369, "lr": 0.0007835926449787835, "epoch": 1.5676567656765676, "percentage": 7.84, "elapsed_time": "0:10:40", "remaining_time": "2:05:25", "throughput": 5215.84, "total_tokens": 3338336} {"current_steps": 3330, "total_steps": 42420, "loss": 0.2943, "lr": 0.0007847713342762848, "epoch": 1.57001414427157, "percentage": 7.85, "elapsed_time": "0:10:41", "remaining_time": "2:05:24", "throughput": 5216.51, "total_tokens": 3343808} {"current_steps": 3335, "total_steps": 42420, "loss": 0.3647, "lr": 0.0007859500235737859, "epoch": 1.5723715228665722, "percentage": 7.86, "elapsed_time": "0:10:41", "remaining_time": "2:05:22", "throughput": 5216.61, "total_tokens": 3348512} {"current_steps": 3340, "total_steps": 42420, "loss": 0.2974, "lr": 0.0007871287128712872, "epoch": 1.5747289014615746, "percentage": 7.87, "elapsed_time": "0:10:42", "remaining_time": "2:05:20", "throughput": 5216.67, "total_tokens": 3352992} {"current_steps": 3345, "total_steps": 42420, "loss": 0.3529, "lr": 0.0007883074021687884, "epoch": 1.577086280056577, "percentage": 7.89, "elapsed_time": "0:10:43", "remaining_time": "2:05:19", "throughput": 5217.06, "total_tokens": 3358080} {"current_steps": 3350, "total_steps": 42420, "loss": 0.2835, "lr": 0.0007894860914662895, "epoch": 1.5794436586515794, "percentage": 7.9, "elapsed_time": "0:10:44", "remaining_time": "2:05:18", "throughput": 5217.3, "total_tokens": 3363200} {"current_steps": 3355, "total_steps": 42420, "loss": 0.3447, "lr": 0.0007906647807637907, "epoch": 1.5818010372465818, "percentage": 7.91, "elapsed_time": "0:10:45", "remaining_time": "2:05:15", "throughput": 5217.14, "total_tokens": 3367616} {"current_steps": 3360, "total_steps": 42420, "loss": 0.3116, "lr": 0.0007918434700612918, "epoch": 1.5841584158415842, "percentage": 7.92, "elapsed_time": "0:10:46", "remaining_time": "2:05:13", "throughput": 5216.93, "total_tokens": 3371968} {"current_steps": 3365, "total_steps": 42420, "loss": 0.3345, "lr": 0.0007930221593587931, "epoch": 1.5865157944365866, "percentage": 7.93, "elapsed_time": "0:10:47", "remaining_time": "2:05:13", "throughput": 5218.0, "total_tokens": 3378048} {"current_steps": 3370, "total_steps": 42420, "loss": 0.2996, "lr": 0.0007942008486562942, "epoch": 1.588873173031589, "percentage": 7.94, "elapsed_time": "0:10:48", "remaining_time": "2:05:11", "throughput": 5217.88, "total_tokens": 3382272} {"current_steps": 3375, "total_steps": 42420, "loss": 0.3453, "lr": 0.0007953795379537954, "epoch": 1.5912305516265912, "percentage": 7.96, "elapsed_time": "0:10:49", "remaining_time": "2:05:10", "throughput": 5218.32, "total_tokens": 3387520} {"current_steps": 3380, "total_steps": 42420, "loss": 0.4298, "lr": 0.0007965582272512966, "epoch": 1.5935879302215936, "percentage": 7.97, "elapsed_time": "0:10:50", "remaining_time": "2:05:08", "throughput": 5218.73, "total_tokens": 3392640} {"current_steps": 3385, "total_steps": 42420, "loss": 0.3466, "lr": 0.0007977369165487977, "epoch": 1.5959453088165958, "percentage": 7.98, "elapsed_time": "0:10:51", "remaining_time": "2:05:07", "throughput": 5219.26, "total_tokens": 3397824} {"current_steps": 3390, "total_steps": 42420, "loss": 0.3444, "lr": 0.000798915605846299, "epoch": 1.5983026874115982, "percentage": 7.99, "elapsed_time": "0:10:51", "remaining_time": "2:05:06", "throughput": 5219.88, "total_tokens": 3403360} {"current_steps": 3395, "total_steps": 42420, "loss": 0.3449, "lr": 0.0008000942951438001, "epoch": 1.6006600660066006, "percentage": 8.0, "elapsed_time": "0:10:52", "remaining_time": "2:05:05", "throughput": 5220.17, "total_tokens": 3408320} {"current_steps": 3400, "total_steps": 42420, "loss": 0.3688, "lr": 0.0008012729844413013, "epoch": 1.603017444601603, "percentage": 8.02, "elapsed_time": "0:10:53", "remaining_time": "2:05:02", "throughput": 5220.07, "total_tokens": 3412576} {"current_steps": 3405, "total_steps": 42420, "loss": 0.3242, "lr": 0.0008024516737388024, "epoch": 1.6053748231966054, "percentage": 8.03, "elapsed_time": "0:10:54", "remaining_time": "2:05:00", "throughput": 5219.95, "total_tokens": 3416960} {"current_steps": 3410, "total_steps": 42420, "loss": 0.3496, "lr": 0.0008036303630363037, "epoch": 1.6077322017916078, "percentage": 8.04, "elapsed_time": "0:10:55", "remaining_time": "2:04:59", "throughput": 5220.26, "total_tokens": 3421984} {"current_steps": 3415, "total_steps": 42420, "loss": 0.3425, "lr": 0.0008048090523338049, "epoch": 1.6100895803866102, "percentage": 8.05, "elapsed_time": "0:10:56", "remaining_time": "2:04:56", "throughput": 5220.09, "total_tokens": 3426112} {"current_steps": 3420, "total_steps": 42420, "loss": 0.3074, "lr": 0.000805987741631306, "epoch": 1.6124469589816126, "percentage": 8.06, "elapsed_time": "0:10:57", "remaining_time": "2:04:54", "throughput": 5220.0, "total_tokens": 3430528} {"current_steps": 3425, "total_steps": 42420, "loss": 0.3107, "lr": 0.0008071664309288072, "epoch": 1.6148043375766148, "percentage": 8.07, "elapsed_time": "0:10:58", "remaining_time": "2:04:53", "throughput": 5220.48, "total_tokens": 3435808} {"current_steps": 3430, "total_steps": 42420, "loss": 0.3829, "lr": 0.0008083451202263083, "epoch": 1.6171617161716172, "percentage": 8.09, "elapsed_time": "0:10:59", "remaining_time": "2:04:52", "throughput": 5221.03, "total_tokens": 3441152} {"current_steps": 3435, "total_steps": 42420, "loss": 0.2858, "lr": 0.0008095238095238096, "epoch": 1.6195190947666194, "percentage": 8.1, "elapsed_time": "0:10:59", "remaining_time": "2:04:49", "throughput": 5221.06, "total_tokens": 3445440} {"current_steps": 3440, "total_steps": 42420, "loss": 0.2984, "lr": 0.0008107024988213107, "epoch": 1.6218764733616218, "percentage": 8.11, "elapsed_time": "0:11:01", "remaining_time": "2:04:50", "throughput": 5221.8, "total_tokens": 3451648} {"current_steps": 3445, "total_steps": 42420, "loss": 0.2912, "lr": 0.000811881188118812, "epoch": 1.6242338519566242, "percentage": 8.12, "elapsed_time": "0:11:01", "remaining_time": "2:04:48", "throughput": 5222.15, "total_tokens": 3456672} {"current_steps": 3450, "total_steps": 42420, "loss": 0.3858, "lr": 0.0008130598774163131, "epoch": 1.6265912305516266, "percentage": 8.13, "elapsed_time": "0:11:02", "remaining_time": "2:04:48", "throughput": 5222.81, "total_tokens": 3462368} {"current_steps": 3455, "total_steps": 42420, "loss": 0.3706, "lr": 0.0008142385667138142, "epoch": 1.628948609146629, "percentage": 8.14, "elapsed_time": "0:11:04", "remaining_time": "2:04:49", "throughput": 5223.97, "total_tokens": 3469120} {"current_steps": 3460, "total_steps": 42420, "loss": 0.8939, "lr": 0.0008154172560113155, "epoch": 1.6313059877416314, "percentage": 8.16, "elapsed_time": "0:11:05", "remaining_time": "2:04:48", "throughput": 5224.52, "total_tokens": 3474560} {"current_steps": 3465, "total_steps": 42420, "loss": 4.798, "lr": 0.0008165959453088166, "epoch": 1.6336633663366338, "percentage": 8.17, "elapsed_time": "0:11:05", "remaining_time": "2:04:46", "throughput": 5224.88, "total_tokens": 3479424} {"current_steps": 3470, "total_steps": 42420, "loss": 2.7323, "lr": 0.0008177746346063179, "epoch": 1.6360207449316362, "percentage": 8.18, "elapsed_time": "0:11:06", "remaining_time": "2:04:44", "throughput": 5225.07, "total_tokens": 3484064} {"current_steps": 3475, "total_steps": 42420, "loss": 0.9376, "lr": 0.0008189533239038189, "epoch": 1.6383781235266384, "percentage": 8.19, "elapsed_time": "0:11:07", "remaining_time": "2:04:42", "throughput": 5225.11, "total_tokens": 3488608} {"current_steps": 3480, "total_steps": 42420, "loss": 0.5443, "lr": 0.0008201320132013201, "epoch": 1.6407355021216408, "percentage": 8.2, "elapsed_time": "0:11:08", "remaining_time": "2:04:39", "throughput": 5224.93, "total_tokens": 3492640} {"current_steps": 3485, "total_steps": 42420, "loss": 0.9043, "lr": 0.0008213107024988214, "epoch": 1.643092880716643, "percentage": 8.22, "elapsed_time": "0:11:09", "remaining_time": "2:04:39", "throughput": 5225.75, "total_tokens": 3498400} {"current_steps": 3490, "total_steps": 42420, "loss": 0.8193, "lr": 0.0008224893917963225, "epoch": 1.6454502593116453, "percentage": 8.23, "elapsed_time": "0:11:10", "remaining_time": "2:04:37", "throughput": 5226.11, "total_tokens": 3503488} {"current_steps": 3495, "total_steps": 42420, "loss": 0.6015, "lr": 0.0008236680810938238, "epoch": 1.6478076379066477, "percentage": 8.24, "elapsed_time": "0:11:11", "remaining_time": "2:04:35", "throughput": 5226.25, "total_tokens": 3508096} {"current_steps": 3500, "total_steps": 42420, "loss": 0.4676, "lr": 0.0008248467703913248, "epoch": 1.6501650165016502, "percentage": 8.25, "elapsed_time": "0:11:12", "remaining_time": "2:04:34", "throughput": 5226.49, "total_tokens": 3512992} {"current_steps": 3505, "total_steps": 42420, "loss": 0.4078, "lr": 0.000826025459688826, "epoch": 1.6525223950966526, "percentage": 8.26, "elapsed_time": "0:11:13", "remaining_time": "2:04:33", "throughput": 5227.06, "total_tokens": 3518688} {"current_steps": 3510, "total_steps": 42420, "loss": 0.5027, "lr": 0.0008272041489863272, "epoch": 1.654879773691655, "percentage": 8.27, "elapsed_time": "0:11:14", "remaining_time": "2:04:33", "throughput": 5227.45, "total_tokens": 3524160} {"current_steps": 3515, "total_steps": 42420, "loss": 0.4324, "lr": 0.0008283828382838284, "epoch": 1.6572371522866574, "percentage": 8.29, "elapsed_time": "0:11:15", "remaining_time": "2:04:31", "throughput": 5227.54, "total_tokens": 3528896} {"current_steps": 3520, "total_steps": 42420, "loss": 0.6192, "lr": 0.0008295615275813297, "epoch": 1.6595945308816598, "percentage": 8.3, "elapsed_time": "0:11:15", "remaining_time": "2:04:30", "throughput": 5227.72, "total_tokens": 3533760} {"current_steps": 3525, "total_steps": 42420, "loss": 0.6153, "lr": 0.0008307402168788307, "epoch": 1.661951909476662, "percentage": 8.31, "elapsed_time": "0:11:16", "remaining_time": "2:04:27", "throughput": 5227.71, "total_tokens": 3537952} {"current_steps": 3530, "total_steps": 42420, "loss": 0.6056, "lr": 0.000831918906176332, "epoch": 1.6643092880716643, "percentage": 8.32, "elapsed_time": "0:11:17", "remaining_time": "2:04:25", "throughput": 5227.88, "total_tokens": 3542464} {"current_steps": 3535, "total_steps": 42420, "loss": 0.3251, "lr": 0.0008330975954738331, "epoch": 1.6666666666666665, "percentage": 8.33, "elapsed_time": "0:11:18", "remaining_time": "2:04:22", "throughput": 5227.77, "total_tokens": 3546656} {"current_steps": 3540, "total_steps": 42420, "loss": 0.5946, "lr": 0.0008342762847713343, "epoch": 1.669024045261669, "percentage": 8.35, "elapsed_time": "0:11:19", "remaining_time": "2:04:22", "throughput": 5228.4, "total_tokens": 3552288} {"current_steps": 3545, "total_steps": 42420, "loss": 0.3858, "lr": 0.0008354549740688355, "epoch": 1.6713814238566713, "percentage": 8.36, "elapsed_time": "0:11:20", "remaining_time": "2:04:19", "throughput": 5228.35, "total_tokens": 3556544} {"current_steps": 3550, "total_steps": 42420, "loss": 0.3648, "lr": 0.0008366336633663366, "epoch": 1.6737388024516737, "percentage": 8.37, "elapsed_time": "0:11:21", "remaining_time": "2:04:18", "throughput": 5228.78, "total_tokens": 3561984} {"current_steps": 3555, "total_steps": 42420, "loss": 0.3396, "lr": 0.0008378123526638379, "epoch": 1.6760961810466761, "percentage": 8.38, "elapsed_time": "0:11:22", "remaining_time": "2:04:16", "throughput": 5229.0, "total_tokens": 3566656} {"current_steps": 3560, "total_steps": 42420, "loss": 0.6981, "lr": 0.000838991041961339, "epoch": 1.6784535596416785, "percentage": 8.39, "elapsed_time": "0:11:22", "remaining_time": "2:04:14", "throughput": 5228.87, "total_tokens": 3571072} {"current_steps": 3565, "total_steps": 42420, "loss": 0.5329, "lr": 0.0008401697312588402, "epoch": 1.680810938236681, "percentage": 8.4, "elapsed_time": "0:11:24", "remaining_time": "2:04:15", "throughput": 5229.67, "total_tokens": 3577184} {"current_steps": 3570, "total_steps": 42420, "loss": 0.3921, "lr": 0.0008413484205563414, "epoch": 1.6831683168316833, "percentage": 8.42, "elapsed_time": "0:11:24", "remaining_time": "2:04:14", "throughput": 5230.0, "total_tokens": 3582496} {"current_steps": 3575, "total_steps": 42420, "loss": 0.3694, "lr": 0.0008425271098538425, "epoch": 1.6855256954266855, "percentage": 8.43, "elapsed_time": "0:11:25", "remaining_time": "2:04:12", "throughput": 5230.01, "total_tokens": 3587232} {"current_steps": 3580, "total_steps": 42420, "loss": 0.5184, "lr": 0.0008437057991513437, "epoch": 1.687883074021688, "percentage": 8.44, "elapsed_time": "0:11:26", "remaining_time": "2:04:11", "throughput": 5230.41, "total_tokens": 3592256} {"current_steps": 3585, "total_steps": 42420, "loss": 0.4626, "lr": 0.0008448844884488449, "epoch": 1.69024045261669, "percentage": 8.45, "elapsed_time": "0:11:27", "remaining_time": "2:04:09", "throughput": 5230.74, "total_tokens": 3597184} {"current_steps": 3590, "total_steps": 42420, "loss": 0.599, "lr": 0.0008460631777463462, "epoch": 1.6925978312116925, "percentage": 8.46, "elapsed_time": "0:11:28", "remaining_time": "2:04:07", "throughput": 5230.86, "total_tokens": 3601760} {"current_steps": 3595, "total_steps": 42420, "loss": 0.5399, "lr": 0.0008472418670438473, "epoch": 1.694955209806695, "percentage": 8.47, "elapsed_time": "0:11:29", "remaining_time": "2:04:05", "throughput": 5231.0, "total_tokens": 3606432} {"current_steps": 3600, "total_steps": 42420, "loss": 0.9764, "lr": 0.0008484205563413484, "epoch": 1.6973125884016973, "percentage": 8.49, "elapsed_time": "0:11:30", "remaining_time": "2:04:04", "throughput": 5231.35, "total_tokens": 3611424} {"current_steps": 3605, "total_steps": 42420, "loss": 0.6582, "lr": 0.0008495992456388496, "epoch": 1.6996699669966997, "percentage": 8.5, "elapsed_time": "0:11:31", "remaining_time": "2:04:02", "throughput": 5231.4, "total_tokens": 3616064} {"current_steps": 3610, "total_steps": 42420, "loss": 0.6603, "lr": 0.0008507779349363508, "epoch": 1.702027345591702, "percentage": 8.51, "elapsed_time": "0:11:32", "remaining_time": "2:04:00", "throughput": 5231.4, "total_tokens": 3620704} {"current_steps": 3615, "total_steps": 42420, "loss": 0.5496, "lr": 0.000851956624233852, "epoch": 1.7043847241867045, "percentage": 8.52, "elapsed_time": "0:11:33", "remaining_time": "2:04:00", "throughput": 5232.07, "total_tokens": 3626624} {"current_steps": 3620, "total_steps": 42420, "loss": 0.4669, "lr": 0.0008531353135313532, "epoch": 1.706742102781707, "percentage": 8.53, "elapsed_time": "0:11:33", "remaining_time": "2:03:58", "throughput": 5231.81, "total_tokens": 3630752} {"current_steps": 3625, "total_steps": 42420, "loss": 0.4278, "lr": 0.0008543140028288543, "epoch": 1.709099481376709, "percentage": 8.55, "elapsed_time": "0:11:34", "remaining_time": "2:03:55", "throughput": 5231.72, "total_tokens": 3634976} {"current_steps": 3630, "total_steps": 42420, "loss": 0.3459, "lr": 0.0008554926921263555, "epoch": 1.7114568599717115, "percentage": 8.56, "elapsed_time": "0:11:35", "remaining_time": "2:03:54", "throughput": 5232.02, "total_tokens": 3639904} {"current_steps": 3635, "total_steps": 42420, "loss": 0.4661, "lr": 0.0008566713814238567, "epoch": 1.7138142385667137, "percentage": 8.57, "elapsed_time": "0:11:36", "remaining_time": "2:03:52", "throughput": 5232.07, "total_tokens": 3644512} {"current_steps": 3640, "total_steps": 42420, "loss": 0.3505, "lr": 0.0008578500707213579, "epoch": 1.716171617161716, "percentage": 8.58, "elapsed_time": "0:11:37", "remaining_time": "2:03:50", "throughput": 5232.26, "total_tokens": 3649088} {"current_steps": 3645, "total_steps": 42420, "loss": 0.3731, "lr": 0.0008590287600188591, "epoch": 1.7185289957567185, "percentage": 8.59, "elapsed_time": "0:11:38", "remaining_time": "2:03:47", "throughput": 5232.23, "total_tokens": 3653440} {"current_steps": 3650, "total_steps": 42420, "loss": 0.4737, "lr": 0.0008602074493163601, "epoch": 1.7208863743517209, "percentage": 8.6, "elapsed_time": "0:11:39", "remaining_time": "2:03:46", "throughput": 5232.37, "total_tokens": 3658272} {"current_steps": 3655, "total_steps": 42420, "loss": 0.415, "lr": 0.0008613861386138614, "epoch": 1.7232437529467233, "percentage": 8.62, "elapsed_time": "0:11:40", "remaining_time": "2:03:44", "throughput": 5232.52, "total_tokens": 3662944} {"current_steps": 3660, "total_steps": 42420, "loss": 0.4464, "lr": 0.0008625648279113626, "epoch": 1.7256011315417257, "percentage": 8.63, "elapsed_time": "0:11:40", "remaining_time": "2:03:43", "throughput": 5232.82, "total_tokens": 3668128} {"current_steps": 3665, "total_steps": 42420, "loss": 0.4702, "lr": 0.0008637435172088638, "epoch": 1.727958510136728, "percentage": 8.64, "elapsed_time": "0:11:41", "remaining_time": "2:03:42", "throughput": 5233.26, "total_tokens": 3673440} {"current_steps": 3670, "total_steps": 42420, "loss": 0.4037, "lr": 0.000864922206506365, "epoch": 1.7303158887317305, "percentage": 8.65, "elapsed_time": "0:11:42", "remaining_time": "2:03:41", "throughput": 5233.77, "total_tokens": 3678624} {"current_steps": 3675, "total_steps": 42420, "loss": 0.3955, "lr": 0.0008661008958038661, "epoch": 1.7326732673267327, "percentage": 8.66, "elapsed_time": "0:11:43", "remaining_time": "2:03:39", "throughput": 5233.97, "total_tokens": 3683264} {"current_steps": 3680, "total_steps": 42420, "loss": 0.3887, "lr": 0.0008672795851013673, "epoch": 1.735030645921735, "percentage": 8.68, "elapsed_time": "0:11:44", "remaining_time": "2:03:39", "throughput": 5234.87, "total_tokens": 3689600} {"current_steps": 3685, "total_steps": 42420, "loss": 0.3747, "lr": 0.0008684582743988684, "epoch": 1.7373880245167372, "percentage": 8.69, "elapsed_time": "0:11:45", "remaining_time": "2:03:37", "throughput": 5234.88, "total_tokens": 3694080} {"current_steps": 3690, "total_steps": 42420, "loss": 0.4193, "lr": 0.0008696369636963697, "epoch": 1.7397454031117396, "percentage": 8.7, "elapsed_time": "0:11:46", "remaining_time": "2:03:35", "throughput": 5235.03, "total_tokens": 3698688} {"current_steps": 3695, "total_steps": 42420, "loss": 0.3581, "lr": 0.0008708156529938709, "epoch": 1.742102781706742, "percentage": 8.71, "elapsed_time": "0:11:47", "remaining_time": "2:03:32", "throughput": 5234.83, "total_tokens": 3702656} {"current_steps": 3700, "total_steps": 42420, "loss": 0.4255, "lr": 0.000871994342291372, "epoch": 1.7444601603017444, "percentage": 8.72, "elapsed_time": "0:11:48", "remaining_time": "2:03:31", "throughput": 5235.13, "total_tokens": 3707616} {"current_steps": 3705, "total_steps": 42420, "loss": 0.4085, "lr": 0.0008731730315888732, "epoch": 1.7468175388967468, "percentage": 8.73, "elapsed_time": "0:11:49", "remaining_time": "2:03:30", "throughput": 5235.93, "total_tokens": 3713408} {"current_steps": 3710, "total_steps": 42420, "loss": 0.3664, "lr": 0.0008743517208863743, "epoch": 1.7491749174917492, "percentage": 8.75, "elapsed_time": "0:11:50", "remaining_time": "2:03:30", "throughput": 5236.56, "total_tokens": 3719200} {"current_steps": 3715, "total_steps": 42420, "loss": 0.4249, "lr": 0.0008755304101838756, "epoch": 1.7515322960867516, "percentage": 8.76, "elapsed_time": "0:11:51", "remaining_time": "2:03:32", "throughput": 5237.92, "total_tokens": 3726624} {"current_steps": 3720, "total_steps": 42420, "loss": 0.3471, "lr": 0.0008767090994813767, "epoch": 1.753889674681754, "percentage": 8.77, "elapsed_time": "0:11:52", "remaining_time": "2:03:30", "throughput": 5238.12, "total_tokens": 3731488} {"current_steps": 3725, "total_steps": 42420, "loss": 0.3341, "lr": 0.0008778877887788779, "epoch": 1.7562470532767562, "percentage": 8.78, "elapsed_time": "0:11:53", "remaining_time": "2:03:29", "throughput": 5238.35, "total_tokens": 3736352} {"current_steps": 3730, "total_steps": 42420, "loss": 0.3563, "lr": 0.0008790664780763791, "epoch": 1.7586044318717586, "percentage": 8.79, "elapsed_time": "0:11:54", "remaining_time": "2:03:28", "throughput": 5238.51, "total_tokens": 3741376} {"current_steps": 3735, "total_steps": 42420, "loss": 0.3652, "lr": 0.0008802451673738803, "epoch": 1.7609618104667608, "percentage": 8.8, "elapsed_time": "0:11:55", "remaining_time": "2:03:25", "throughput": 5238.35, "total_tokens": 3745472} {"current_steps": 3740, "total_steps": 42420, "loss": 0.361, "lr": 0.0008814238566713815, "epoch": 1.7633191890617632, "percentage": 8.82, "elapsed_time": "0:11:55", "remaining_time": "2:03:24", "throughput": 5238.63, "total_tokens": 3750496} {"current_steps": 3745, "total_steps": 42420, "loss": 0.2464, "lr": 0.0008826025459688826, "epoch": 1.7656765676567656, "percentage": 8.83, "elapsed_time": "0:11:56", "remaining_time": "2:03:23", "throughput": 5239.22, "total_tokens": 3755808} {"current_steps": 3750, "total_steps": 42420, "loss": 0.6312, "lr": 0.0008837812352663838, "epoch": 1.768033946251768, "percentage": 8.84, "elapsed_time": "0:11:57", "remaining_time": "2:03:23", "throughput": 5239.81, "total_tokens": 3761728} {"current_steps": 3755, "total_steps": 42420, "loss": 0.5257, "lr": 0.0008849599245638849, "epoch": 1.7703913248467704, "percentage": 8.85, "elapsed_time": "0:11:58", "remaining_time": "2:03:22", "throughput": 5240.14, "total_tokens": 3766912} {"current_steps": 3760, "total_steps": 42420, "loss": 0.3358, "lr": 0.0008861386138613862, "epoch": 1.7727487034417728, "percentage": 8.86, "elapsed_time": "0:11:59", "remaining_time": "2:03:21", "throughput": 5240.59, "total_tokens": 3772512} {"current_steps": 3765, "total_steps": 42420, "loss": 0.5263, "lr": 0.0008873173031588874, "epoch": 1.7751060820367752, "percentage": 8.88, "elapsed_time": "0:12:00", "remaining_time": "2:03:21", "throughput": 5241.22, "total_tokens": 3778176} {"current_steps": 3770, "total_steps": 42420, "loss": 0.383, "lr": 0.0008884959924563886, "epoch": 1.7774634606317776, "percentage": 8.89, "elapsed_time": "0:12:01", "remaining_time": "2:03:19", "throughput": 5241.51, "total_tokens": 3783296} {"current_steps": 3775, "total_steps": 42420, "loss": 0.4335, "lr": 0.0008896746817538897, "epoch": 1.7798208392267798, "percentage": 8.9, "elapsed_time": "0:12:02", "remaining_time": "2:03:16", "throughput": 5241.21, "total_tokens": 3787104} {"current_steps": 3780, "total_steps": 42420, "loss": 0.3618, "lr": 0.0008908533710513908, "epoch": 1.7821782178217822, "percentage": 8.91, "elapsed_time": "0:12:03", "remaining_time": "2:03:17", "throughput": 5242.15, "total_tokens": 3793760} {"current_steps": 3785, "total_steps": 42420, "loss": 0.3448, "lr": 0.0008920320603488921, "epoch": 1.7845355964167844, "percentage": 8.92, "elapsed_time": "0:12:04", "remaining_time": "2:03:16", "throughput": 5242.57, "total_tokens": 3798976} {"current_steps": 3790, "total_steps": 42420, "loss": 0.43, "lr": 0.0008932107496463932, "epoch": 1.7868929750117868, "percentage": 8.93, "elapsed_time": "0:12:05", "remaining_time": "2:03:15", "throughput": 5243.13, "total_tokens": 3804416} {"current_steps": 3795, "total_steps": 42420, "loss": 0.3931, "lr": 0.0008943894389438945, "epoch": 1.7892503536067892, "percentage": 8.95, "elapsed_time": "0:12:06", "remaining_time": "2:03:14", "throughput": 5243.41, "total_tokens": 3809376} {"current_steps": 3800, "total_steps": 42420, "loss": 0.4523, "lr": 0.0008955681282413956, "epoch": 1.7916077322017916, "percentage": 8.96, "elapsed_time": "0:12:07", "remaining_time": "2:03:13", "throughput": 5243.77, "total_tokens": 3814880} {"current_steps": 3805, "total_steps": 42420, "loss": 0.4363, "lr": 0.0008967468175388967, "epoch": 1.793965110796794, "percentage": 8.97, "elapsed_time": "0:12:08", "remaining_time": "2:03:14", "throughput": 5244.7, "total_tokens": 3821248} {"current_steps": 3810, "total_steps": 42420, "loss": 0.3433, "lr": 0.000897925506836398, "epoch": 1.7963224893917964, "percentage": 8.98, "elapsed_time": "0:12:09", "remaining_time": "2:03:11", "throughput": 5244.57, "total_tokens": 3825440} {"current_steps": 3815, "total_steps": 42420, "loss": 0.2555, "lr": 0.0008991041961338991, "epoch": 1.7986798679867988, "percentage": 8.99, "elapsed_time": "0:12:10", "remaining_time": "2:03:11", "throughput": 5245.1, "total_tokens": 3831328} {"current_steps": 3820, "total_steps": 42420, "loss": 0.5159, "lr": 0.0009002828854314004, "epoch": 1.8010372465818012, "percentage": 9.01, "elapsed_time": "0:12:11", "remaining_time": "2:03:10", "throughput": 5245.46, "total_tokens": 3836384} {"current_steps": 3825, "total_steps": 42420, "loss": 0.4142, "lr": 0.0009014615747289014, "epoch": 1.8033946251768034, "percentage": 9.02, "elapsed_time": "0:12:12", "remaining_time": "2:03:08", "throughput": 5245.86, "total_tokens": 3841408} {"current_steps": 3830, "total_steps": 42420, "loss": 0.3761, "lr": 0.0009026402640264026, "epoch": 1.8057520037718058, "percentage": 9.03, "elapsed_time": "0:12:13", "remaining_time": "2:03:07", "throughput": 5246.17, "total_tokens": 3846336} {"current_steps": 3835, "total_steps": 42420, "loss": 0.4117, "lr": 0.0009038189533239039, "epoch": 1.808109382366808, "percentage": 9.04, "elapsed_time": "0:12:14", "remaining_time": "2:03:05", "throughput": 5246.11, "total_tokens": 3850752} {"current_steps": 3840, "total_steps": 42420, "loss": 0.3866, "lr": 0.000904997642621405, "epoch": 1.8104667609618104, "percentage": 9.05, "elapsed_time": "0:12:14", "remaining_time": "2:03:03", "throughput": 5245.9, "total_tokens": 3855136} {"current_steps": 3845, "total_steps": 42420, "loss": 0.3893, "lr": 0.0009061763319189063, "epoch": 1.8128241395568128, "percentage": 9.06, "elapsed_time": "0:12:15", "remaining_time": "2:03:00", "throughput": 5245.4, "total_tokens": 3858688} {"current_steps": 3850, "total_steps": 42420, "loss": 0.3717, "lr": 0.0009073550212164073, "epoch": 1.8151815181518152, "percentage": 9.08, "elapsed_time": "0:12:16", "remaining_time": "2:02:58", "throughput": 5245.61, "total_tokens": 3863424} {"current_steps": 3855, "total_steps": 42420, "loss": 0.3039, "lr": 0.0009085337105139086, "epoch": 1.8175388967468176, "percentage": 9.09, "elapsed_time": "0:12:17", "remaining_time": "2:02:57", "throughput": 5246.09, "total_tokens": 3868832} {"current_steps": 3860, "total_steps": 42420, "loss": 0.4214, "lr": 0.0009097123998114097, "epoch": 1.81989627534182, "percentage": 9.1, "elapsed_time": "0:12:18", "remaining_time": "2:02:55", "throughput": 5246.06, "total_tokens": 3873376} {"current_steps": 3865, "total_steps": 42420, "loss": 0.4241, "lr": 0.0009108910891089109, "epoch": 1.8222536539368224, "percentage": 9.11, "elapsed_time": "0:12:19", "remaining_time": "2:02:54", "throughput": 5246.35, "total_tokens": 3878272} {"current_steps": 3870, "total_steps": 42420, "loss": 0.3898, "lr": 0.0009120697784064122, "epoch": 1.8246110325318248, "percentage": 9.12, "elapsed_time": "0:12:20", "remaining_time": "2:02:53", "throughput": 5246.62, "total_tokens": 3883488} {"current_steps": 3875, "total_steps": 42420, "loss": 0.3896, "lr": 0.0009132484677039132, "epoch": 1.826968411126827, "percentage": 9.13, "elapsed_time": "0:12:21", "remaining_time": "2:02:52", "throughput": 5247.05, "total_tokens": 3888864} {"current_steps": 3880, "total_steps": 42420, "loss": 0.3339, "lr": 0.0009144271570014145, "epoch": 1.8293257897218294, "percentage": 9.15, "elapsed_time": "0:12:22", "remaining_time": "2:02:52", "throughput": 5247.4, "total_tokens": 3894496} {"current_steps": 3885, "total_steps": 42420, "loss": 0.3918, "lr": 0.0009156058462989156, "epoch": 1.8316831683168315, "percentage": 9.16, "elapsed_time": "0:12:23", "remaining_time": "2:02:49", "throughput": 5247.28, "total_tokens": 3898784} {"current_steps": 3890, "total_steps": 42420, "loss": 0.3573, "lr": 0.0009167845355964169, "epoch": 1.834040546911834, "percentage": 9.17, "elapsed_time": "0:12:23", "remaining_time": "2:02:49", "throughput": 5247.65, "total_tokens": 3904160} {"current_steps": 3895, "total_steps": 42420, "loss": 0.3239, "lr": 0.000917963224893918, "epoch": 1.8363979255068363, "percentage": 9.18, "elapsed_time": "0:12:24", "remaining_time": "2:02:46", "throughput": 5247.7, "total_tokens": 3908608} {"current_steps": 3900, "total_steps": 42420, "loss": 0.7588, "lr": 0.0009191419141914191, "epoch": 1.8387553041018387, "percentage": 9.19, "elapsed_time": "0:12:25", "remaining_time": "2:02:44", "throughput": 5247.6, "total_tokens": 3912896} {"current_steps": 3905, "total_steps": 42420, "loss": 0.4815, "lr": 0.0009203206034889204, "epoch": 1.8411126826968411, "percentage": 9.21, "elapsed_time": "0:12:26", "remaining_time": "2:02:44", "throughput": 5248.17, "total_tokens": 3918496} {"current_steps": 3910, "total_steps": 42420, "loss": 0.596, "lr": 0.0009214992927864215, "epoch": 1.8434700612918435, "percentage": 9.22, "elapsed_time": "0:12:27", "remaining_time": "2:02:42", "throughput": 5248.12, "total_tokens": 3922976} {"current_steps": 3915, "total_steps": 42420, "loss": 0.3567, "lr": 0.0009226779820839228, "epoch": 1.845827439886846, "percentage": 9.23, "elapsed_time": "0:12:28", "remaining_time": "2:02:40", "throughput": 5248.13, "total_tokens": 3927360} {"current_steps": 3920, "total_steps": 42420, "loss": 0.4033, "lr": 0.0009238566713814239, "epoch": 1.8481848184818483, "percentage": 9.24, "elapsed_time": "0:12:29", "remaining_time": "2:02:37", "throughput": 5248.26, "total_tokens": 3931872} {"current_steps": 3925, "total_steps": 42420, "loss": 0.3639, "lr": 0.000925035360678925, "epoch": 1.8505421970768505, "percentage": 9.25, "elapsed_time": "0:12:30", "remaining_time": "2:02:35", "throughput": 5248.03, "total_tokens": 3936032} {"current_steps": 3930, "total_steps": 42420, "loss": 0.4129, "lr": 0.0009262140499764262, "epoch": 1.852899575671853, "percentage": 9.26, "elapsed_time": "0:12:31", "remaining_time": "2:02:35", "throughput": 5248.54, "total_tokens": 3941728} {"current_steps": 3935, "total_steps": 42420, "loss": 0.3848, "lr": 0.0009273927392739274, "epoch": 1.855256954266855, "percentage": 9.28, "elapsed_time": "0:12:31", "remaining_time": "2:02:33", "throughput": 5248.44, "total_tokens": 3946176} {"current_steps": 3940, "total_steps": 42420, "loss": 0.3632, "lr": 0.0009285714285714287, "epoch": 1.8576143328618575, "percentage": 9.29, "elapsed_time": "0:12:32", "remaining_time": "2:02:32", "throughput": 5248.52, "total_tokens": 3950976} {"current_steps": 3945, "total_steps": 42420, "loss": 0.3622, "lr": 0.0009297501178689298, "epoch": 1.85997171145686, "percentage": 9.3, "elapsed_time": "0:12:33", "remaining_time": "2:02:29", "throughput": 5248.39, "total_tokens": 3955264} {"current_steps": 3950, "total_steps": 42420, "loss": 0.3943, "lr": 0.0009309288071664309, "epoch": 1.8623290900518623, "percentage": 9.31, "elapsed_time": "0:12:34", "remaining_time": "2:02:28", "throughput": 5248.85, "total_tokens": 3960512} {"current_steps": 3955, "total_steps": 42420, "loss": 0.3494, "lr": 0.0009321074964639321, "epoch": 1.8646864686468647, "percentage": 9.32, "elapsed_time": "0:12:35", "remaining_time": "2:02:27", "throughput": 5249.11, "total_tokens": 3965408} {"current_steps": 3960, "total_steps": 42420, "loss": 0.3943, "lr": 0.0009332861857614333, "epoch": 1.8670438472418671, "percentage": 9.34, "elapsed_time": "0:12:36", "remaining_time": "2:02:26", "throughput": 5249.64, "total_tokens": 3971168} {"current_steps": 3965, "total_steps": 42420, "loss": 0.3555, "lr": 0.0009344648750589345, "epoch": 1.8694012258368695, "percentage": 9.35, "elapsed_time": "0:12:37", "remaining_time": "2:02:24", "throughput": 5249.48, "total_tokens": 3975456} {"current_steps": 3970, "total_steps": 42420, "loss": 0.4356, "lr": 0.0009356435643564357, "epoch": 1.871758604431872, "percentage": 9.36, "elapsed_time": "0:12:38", "remaining_time": "2:02:23", "throughput": 5249.92, "total_tokens": 3980864} {"current_steps": 3975, "total_steps": 42420, "loss": 0.2582, "lr": 0.0009368222536539369, "epoch": 1.874115983026874, "percentage": 9.37, "elapsed_time": "0:12:39", "remaining_time": "2:02:24", "throughput": 5250.74, "total_tokens": 3987392} {"current_steps": 3980, "total_steps": 42420, "loss": 0.5689, "lr": 0.000938000942951438, "epoch": 1.8764733616218765, "percentage": 9.38, "elapsed_time": "0:12:40", "remaining_time": "2:02:22", "throughput": 5250.88, "total_tokens": 3992064} {"current_steps": 3985, "total_steps": 42420, "loss": 0.3072, "lr": 0.0009391796322489392, "epoch": 1.8788307402168787, "percentage": 9.39, "elapsed_time": "0:12:41", "remaining_time": "2:02:20", "throughput": 5250.89, "total_tokens": 3996512} {"current_steps": 3990, "total_steps": 42420, "loss": 0.5118, "lr": 0.0009403583215464404, "epoch": 1.881188118811881, "percentage": 9.41, "elapsed_time": "0:12:42", "remaining_time": "2:02:20", "throughput": 5251.26, "total_tokens": 4001984} {"current_steps": 3995, "total_steps": 42420, "loss": 0.4153, "lr": 0.0009415370108439416, "epoch": 1.8835454974068835, "percentage": 9.42, "elapsed_time": "0:12:42", "remaining_time": "2:02:17", "throughput": 5251.27, "total_tokens": 4006304} {"current_steps": 4000, "total_steps": 42420, "loss": 0.3934, "lr": 0.0009427157001414427, "epoch": 1.8859028760018859, "percentage": 9.43, "elapsed_time": "0:12:43", "remaining_time": "2:02:17", "throughput": 5251.85, "total_tokens": 4012064} {"current_steps": 4005, "total_steps": 42420, "loss": 0.3921, "lr": 0.0009438943894389439, "epoch": 1.8882602545968883, "percentage": 9.44, "elapsed_time": "0:12:44", "remaining_time": "2:02:17", "throughput": 5252.58, "total_tokens": 4018048} {"current_steps": 4010, "total_steps": 42420, "loss": 0.32, "lr": 0.0009450730787364451, "epoch": 1.8906176331918907, "percentage": 9.45, "elapsed_time": "0:12:45", "remaining_time": "2:02:16", "throughput": 5252.82, "total_tokens": 4023104} {"current_steps": 4015, "total_steps": 42420, "loss": 0.3152, "lr": 0.0009462517680339463, "epoch": 1.892975011786893, "percentage": 9.46, "elapsed_time": "0:12:46", "remaining_time": "2:02:15", "throughput": 5253.34, "total_tokens": 4028800} {"current_steps": 4020, "total_steps": 42420, "loss": 0.3961, "lr": 0.0009474304573314475, "epoch": 1.8953323903818955, "percentage": 9.48, "elapsed_time": "0:12:47", "remaining_time": "2:02:13", "throughput": 5253.25, "total_tokens": 4033152} {"current_steps": 4025, "total_steps": 42420, "loss": 0.313, "lr": 0.0009486091466289486, "epoch": 1.8976897689768977, "percentage": 9.49, "elapsed_time": "0:12:48", "remaining_time": "2:02:12", "throughput": 5253.39, "total_tokens": 4038048} {"current_steps": 4030, "total_steps": 42420, "loss": 0.3795, "lr": 0.0009497878359264498, "epoch": 1.9000471475719, "percentage": 9.5, "elapsed_time": "0:12:49", "remaining_time": "2:02:11", "throughput": 5253.95, "total_tokens": 4043680} {"current_steps": 4035, "total_steps": 42420, "loss": 0.3698, "lr": 0.000950966525223951, "epoch": 1.9024045261669023, "percentage": 9.51, "elapsed_time": "0:12:50", "remaining_time": "2:02:10", "throughput": 5254.18, "total_tokens": 4048544} {"current_steps": 4040, "total_steps": 42420, "loss": 0.3046, "lr": 0.0009521452145214522, "epoch": 1.9047619047619047, "percentage": 9.52, "elapsed_time": "0:12:51", "remaining_time": "2:02:07", "throughput": 5253.82, "total_tokens": 4052448} {"current_steps": 4045, "total_steps": 42420, "loss": 0.2899, "lr": 0.0009533239038189534, "epoch": 1.907119283356907, "percentage": 9.54, "elapsed_time": "0:12:52", "remaining_time": "2:02:05", "throughput": 5253.81, "total_tokens": 4056992} {"current_steps": 4050, "total_steps": 42420, "loss": 0.4447, "lr": 0.0009545025931164545, "epoch": 1.9094766619519095, "percentage": 9.55, "elapsed_time": "0:12:53", "remaining_time": "2:02:05", "throughput": 5254.57, "total_tokens": 4062848} {"current_steps": 4055, "total_steps": 42420, "loss": 0.3754, "lr": 0.0009556812824139557, "epoch": 1.9118340405469119, "percentage": 9.56, "elapsed_time": "0:12:54", "remaining_time": "2:02:03", "throughput": 5254.86, "total_tokens": 4067840} {"current_steps": 4060, "total_steps": 42420, "loss": 0.3917, "lr": 0.0009568599717114569, "epoch": 1.9141914191419143, "percentage": 9.57, "elapsed_time": "0:12:54", "remaining_time": "2:02:02", "throughput": 5254.92, "total_tokens": 4072416} {"current_steps": 4065, "total_steps": 42420, "loss": 0.3264, "lr": 0.0009580386610089581, "epoch": 1.9165487977369167, "percentage": 9.58, "elapsed_time": "0:12:55", "remaining_time": "2:02:00", "throughput": 5255.08, "total_tokens": 4077152} {"current_steps": 4070, "total_steps": 42420, "loss": 0.2651, "lr": 0.0009592173503064592, "epoch": 1.918906176331919, "percentage": 9.59, "elapsed_time": "0:12:56", "remaining_time": "2:01:58", "throughput": 5255.15, "total_tokens": 4081632} {"current_steps": 4075, "total_steps": 42420, "loss": 0.33, "lr": 0.0009603960396039604, "epoch": 1.9212635549269212, "percentage": 9.61, "elapsed_time": "0:12:57", "remaining_time": "2:01:56", "throughput": 5255.23, "total_tokens": 4086336} {"current_steps": 4080, "total_steps": 42420, "loss": 0.3247, "lr": 0.0009615747289014616, "epoch": 1.9236209335219236, "percentage": 9.62, "elapsed_time": "0:12:58", "remaining_time": "2:01:55", "throughput": 5255.45, "total_tokens": 4091296} {"current_steps": 4085, "total_steps": 42420, "loss": 0.4034, "lr": 0.0009627534181989628, "epoch": 1.9259783121169258, "percentage": 9.63, "elapsed_time": "0:12:59", "remaining_time": "2:01:55", "throughput": 5256.08, "total_tokens": 4097216} {"current_steps": 4090, "total_steps": 42420, "loss": 0.3809, "lr": 0.000963932107496464, "epoch": 1.9283356907119282, "percentage": 9.64, "elapsed_time": "0:13:00", "remaining_time": "2:01:54", "throughput": 5256.43, "total_tokens": 4102432} {"current_steps": 4095, "total_steps": 42420, "loss": 0.3914, "lr": 0.0009651107967939652, "epoch": 1.9306930693069306, "percentage": 9.65, "elapsed_time": "0:13:01", "remaining_time": "2:01:54", "throughput": 5257.26, "total_tokens": 4108608} {"current_steps": 4100, "total_steps": 42420, "loss": 0.3438, "lr": 0.0009662894860914663, "epoch": 1.933050447901933, "percentage": 9.67, "elapsed_time": "0:13:02", "remaining_time": "2:01:52", "throughput": 5257.69, "total_tokens": 4113760} {"current_steps": 4105, "total_steps": 42420, "loss": 0.3232, "lr": 0.0009674681753889674, "epoch": 1.9354078264969354, "percentage": 9.68, "elapsed_time": "0:13:03", "remaining_time": "2:01:50", "throughput": 5257.49, "total_tokens": 4117888} {"current_steps": 4110, "total_steps": 42420, "loss": 0.3254, "lr": 0.0009686468646864687, "epoch": 1.9377652050919378, "percentage": 9.69, "elapsed_time": "0:13:04", "remaining_time": "2:01:49", "throughput": 5257.76, "total_tokens": 4123072} {"current_steps": 4115, "total_steps": 42420, "loss": 0.3532, "lr": 0.0009698255539839699, "epoch": 1.9401225836869402, "percentage": 9.7, "elapsed_time": "0:13:05", "remaining_time": "2:01:51", "throughput": 5259.0, "total_tokens": 4130624} {"current_steps": 4120, "total_steps": 42420, "loss": 0.3828, "lr": 0.0009710042432814711, "epoch": 1.9424799622819426, "percentage": 9.71, "elapsed_time": "0:13:06", "remaining_time": "2:01:49", "throughput": 5259.0, "total_tokens": 4135232} {"current_steps": 4125, "total_steps": 42420, "loss": 0.4693, "lr": 0.0009721829325789722, "epoch": 1.9448373408769448, "percentage": 9.72, "elapsed_time": "0:13:07", "remaining_time": "2:01:48", "throughput": 5258.98, "total_tokens": 4139840} {"current_steps": 4130, "total_steps": 42420, "loss": 0.3459, "lr": 0.0009733616218764733, "epoch": 1.9471947194719472, "percentage": 9.74, "elapsed_time": "0:13:08", "remaining_time": "2:01:46", "throughput": 5259.17, "total_tokens": 4144640} {"current_steps": 4135, "total_steps": 42420, "loss": 0.5717, "lr": 0.0009745403111739746, "epoch": 1.9495520980669494, "percentage": 9.75, "elapsed_time": "0:13:09", "remaining_time": "2:01:47", "throughput": 5260.07, "total_tokens": 4151456} {"current_steps": 4140, "total_steps": 42420, "loss": 0.2401, "lr": 0.0009757190004714757, "epoch": 1.9519094766619518, "percentage": 9.76, "elapsed_time": "0:13:10", "remaining_time": "2:01:45", "throughput": 5260.11, "total_tokens": 4156032} {"current_steps": 4145, "total_steps": 42420, "loss": 0.3634, "lr": 0.000976897689768977, "epoch": 1.9542668552569542, "percentage": 9.77, "elapsed_time": "0:13:10", "remaining_time": "2:01:43", "throughput": 5260.0, "total_tokens": 4160512} {"current_steps": 4150, "total_steps": 42420, "loss": 0.408, "lr": 0.0009780763790664781, "epoch": 1.9566242338519566, "percentage": 9.78, "elapsed_time": "0:13:11", "remaining_time": "2:01:41", "throughput": 5260.03, "total_tokens": 4164992} {"current_steps": 4155, "total_steps": 42420, "loss": 0.3618, "lr": 0.0009792550683639792, "epoch": 1.958981612446959, "percentage": 9.79, "elapsed_time": "0:13:12", "remaining_time": "2:01:39", "throughput": 5259.59, "total_tokens": 4168768} {"current_steps": 4160, "total_steps": 42420, "loss": 0.3306, "lr": 0.0009804337576614804, "epoch": 1.9613389910419614, "percentage": 9.81, "elapsed_time": "0:13:13", "remaining_time": "2:01:38", "throughput": 5259.95, "total_tokens": 4174048} {"current_steps": 4165, "total_steps": 42420, "loss": 0.2632, "lr": 0.0009816124469589817, "epoch": 1.9636963696369638, "percentage": 9.82, "elapsed_time": "0:13:14", "remaining_time": "2:01:38", "throughput": 5260.45, "total_tokens": 4179808} {"current_steps": 4170, "total_steps": 42420, "loss": 0.3713, "lr": 0.0009827911362564829, "epoch": 1.9660537482319662, "percentage": 9.83, "elapsed_time": "0:13:15", "remaining_time": "2:01:38", "throughput": 5261.12, "total_tokens": 4185952} {"current_steps": 4175, "total_steps": 42420, "loss": 0.3697, "lr": 0.000983969825553984, "epoch": 1.9684111268269684, "percentage": 9.84, "elapsed_time": "0:13:16", "remaining_time": "2:01:37", "throughput": 5261.53, "total_tokens": 4191200} {"current_steps": 4180, "total_steps": 42420, "loss": 0.3076, "lr": 0.0009851485148514852, "epoch": 1.9707685054219708, "percentage": 9.85, "elapsed_time": "0:13:17", "remaining_time": "2:01:36", "throughput": 5261.98, "total_tokens": 4196736} {"current_steps": 4185, "total_steps": 42420, "loss": 0.4158, "lr": 0.0009863272041489863, "epoch": 1.973125884016973, "percentage": 9.87, "elapsed_time": "0:13:18", "remaining_time": "2:01:36", "throughput": 5262.52, "total_tokens": 4202688} {"current_steps": 4190, "total_steps": 42420, "loss": 0.38, "lr": 0.0009875058934464876, "epoch": 1.9754832626119754, "percentage": 9.88, "elapsed_time": "0:13:19", "remaining_time": "2:01:35", "throughput": 5262.79, "total_tokens": 4207904} {"current_steps": 4195, "total_steps": 42420, "loss": 0.3521, "lr": 0.0009886845827439888, "epoch": 1.9778406412069778, "percentage": 9.89, "elapsed_time": "0:13:20", "remaining_time": "2:01:33", "throughput": 5262.75, "total_tokens": 4212288} {"current_steps": 4200, "total_steps": 42420, "loss": 0.3586, "lr": 0.00098986327204149, "epoch": 1.9801980198019802, "percentage": 9.9, "elapsed_time": "0:13:21", "remaining_time": "2:01:31", "throughput": 5262.84, "total_tokens": 4217120} {"current_steps": 4205, "total_steps": 42420, "loss": 0.359, "lr": 0.000991041961338991, "epoch": 1.9825553983969826, "percentage": 9.91, "elapsed_time": "0:13:22", "remaining_time": "2:01:31", "throughput": 5263.43, "total_tokens": 4222880} {"current_steps": 4210, "total_steps": 42420, "loss": 0.3322, "lr": 0.0009922206506364922, "epoch": 1.984912776991985, "percentage": 9.92, "elapsed_time": "0:13:23", "remaining_time": "2:01:30", "throughput": 5263.81, "total_tokens": 4228256} {"current_steps": 4215, "total_steps": 42420, "loss": 0.3314, "lr": 0.0009933993399339936, "epoch": 1.9872701555869874, "percentage": 9.94, "elapsed_time": "0:13:24", "remaining_time": "2:01:28", "throughput": 5263.85, "total_tokens": 4232928} {"current_steps": 4220, "total_steps": 42420, "loss": 0.4007, "lr": 0.0009945780292314947, "epoch": 1.9896275341819898, "percentage": 9.95, "elapsed_time": "0:13:24", "remaining_time": "2:01:26", "throughput": 5263.66, "total_tokens": 4237152} {"current_steps": 4225, "total_steps": 42420, "loss": 0.3255, "lr": 0.0009957567185289958, "epoch": 1.991984912776992, "percentage": 9.96, "elapsed_time": "0:13:25", "remaining_time": "2:01:25", "throughput": 5263.88, "total_tokens": 4242208} {"current_steps": 4230, "total_steps": 42420, "loss": 0.4168, "lr": 0.000996935407826497, "epoch": 1.9943422913719944, "percentage": 9.97, "elapsed_time": "0:13:26", "remaining_time": "2:01:24", "throughput": 5264.16, "total_tokens": 4247456} {"current_steps": 4235, "total_steps": 42420, "loss": 0.3225, "lr": 0.0009981140971239981, "epoch": 1.9966996699669965, "percentage": 9.98, "elapsed_time": "0:13:28", "remaining_time": "2:01:27", "throughput": 5265.45, "total_tokens": 4255552} {"current_steps": 4240, "total_steps": 42420, "loss": 0.5563, "lr": 0.0009992927864214993, "epoch": 1.999057048561999, "percentage": 10.0, "elapsed_time": "0:13:29", "remaining_time": "2:01:26", "throughput": 5265.84, "total_tokens": 4260960} {"current_steps": 4242, "total_steps": 42420, "eval_loss": 0.3527408838272095, "epoch": 2.0, "percentage": 10.0, "elapsed_time": "0:13:48", "remaining_time": "2:04:17", "throughput": 5144.88, "total_tokens": 4262848} {"current_steps": 4245, "total_steps": 42420, "loss": 0.3795, "lr": 0.000999999993228679, "epoch": 2.0014144271570014, "percentage": 10.01, "elapsed_time": "0:13:51", "remaining_time": "2:04:35", "throughput": 5130.94, "total_tokens": 4265280} {"current_steps": 4250, "total_steps": 42420, "loss": 0.4732, "lr": 0.0009999999170513208, "epoch": 2.0037718057520038, "percentage": 10.02, "elapsed_time": "0:13:52", "remaining_time": "2:04:33", "throughput": 5130.85, "total_tokens": 4269472} {"current_steps": 4255, "total_steps": 42420, "loss": 0.435, "lr": 0.0009999997562324663, "epoch": 2.006129184347006, "percentage": 10.03, "elapsed_time": "0:13:53", "remaining_time": "2:04:31", "throughput": 5131.32, "total_tokens": 4274624} {"current_steps": 4260, "total_steps": 42420, "loss": 0.4132, "lr": 0.0009999995107721421, "epoch": 2.0084865629420086, "percentage": 10.04, "elapsed_time": "0:13:53", "remaining_time": "2:04:30", "throughput": 5131.72, "total_tokens": 4279744} {"current_steps": 4265, "total_steps": 42420, "loss": 0.3711, "lr": 0.0009999991806703907, "epoch": 2.010843941537011, "percentage": 10.05, "elapsed_time": "0:13:54", "remaining_time": "2:04:29", "throughput": 5132.04, "total_tokens": 4284736} {"current_steps": 4270, "total_steps": 42420, "loss": 0.3329, "lr": 0.0009999987659272673, "epoch": 2.0132013201320134, "percentage": 10.07, "elapsed_time": "0:13:55", "remaining_time": "2:04:27", "throughput": 5131.96, "total_tokens": 4289088} {"current_steps": 4275, "total_steps": 42420, "loss": 0.3446, "lr": 0.0009999982665428425, "epoch": 2.0155586987270158, "percentage": 10.08, "elapsed_time": "0:13:56", "remaining_time": "2:04:24", "throughput": 5131.93, "total_tokens": 4293184} {"current_steps": 4280, "total_steps": 42420, "loss": 0.3579, "lr": 0.0009999976825172006, "epoch": 2.0179160773220177, "percentage": 10.09, "elapsed_time": "0:13:57", "remaining_time": "2:04:22", "throughput": 5131.95, "total_tokens": 4297440} {"current_steps": 4285, "total_steps": 42420, "loss": 0.3843, "lr": 0.0009999970138504404, "epoch": 2.02027345591702, "percentage": 10.1, "elapsed_time": "0:13:58", "remaining_time": "2:04:21", "throughput": 5132.57, "total_tokens": 4303136} {"current_steps": 4290, "total_steps": 42420, "loss": 0.3611, "lr": 0.0009999962605426753, "epoch": 2.0226308345120225, "percentage": 10.11, "elapsed_time": "0:13:59", "remaining_time": "2:04:19", "throughput": 5132.77, "total_tokens": 4307872} {"current_steps": 4295, "total_steps": 42420, "loss": 0.3819, "lr": 0.0009999954225940327, "epoch": 2.024988213107025, "percentage": 10.12, "elapsed_time": "0:14:00", "remaining_time": "2:04:17", "throughput": 5132.89, "total_tokens": 4312448} {"current_steps": 4300, "total_steps": 42420, "loss": 0.4071, "lr": 0.0009999945000046546, "epoch": 2.0273455917020273, "percentage": 10.14, "elapsed_time": "0:14:01", "remaining_time": "2:04:15", "throughput": 5132.98, "total_tokens": 4316992} {"current_steps": 4305, "total_steps": 42420, "loss": 0.4096, "lr": 0.000999993492774697, "epoch": 2.0297029702970297, "percentage": 10.15, "elapsed_time": "0:14:02", "remaining_time": "2:04:15", "throughput": 5133.86, "total_tokens": 4323200} {"current_steps": 4310, "total_steps": 42420, "loss": 0.3728, "lr": 0.0009999924009043305, "epoch": 2.032060348892032, "percentage": 10.16, "elapsed_time": "0:14:03", "remaining_time": "2:04:15", "throughput": 5134.7, "total_tokens": 4329344} {"current_steps": 4315, "total_steps": 42420, "loss": 0.3855, "lr": 0.00099999122439374, "epoch": 2.0344177274870345, "percentage": 10.17, "elapsed_time": "0:14:04", "remaining_time": "2:04:13", "throughput": 5134.93, "total_tokens": 4334304} {"current_steps": 4320, "total_steps": 42420, "loss": 0.3761, "lr": 0.0009999899632431243, "epoch": 2.036775106082037, "percentage": 10.18, "elapsed_time": "0:14:05", "remaining_time": "2:04:13", "throughput": 5135.5, "total_tokens": 4339968} {"current_steps": 4325, "total_steps": 42420, "loss": 0.3967, "lr": 0.0009999886174526973, "epoch": 2.0391324846770393, "percentage": 10.2, "elapsed_time": "0:14:06", "remaining_time": "2:04:12", "throughput": 5135.99, "total_tokens": 4345568} {"current_steps": 4330, "total_steps": 42420, "loss": 0.3677, "lr": 0.000999987187022687, "epoch": 2.0414898632720413, "percentage": 10.21, "elapsed_time": "0:14:07", "remaining_time": "2:04:11", "throughput": 5136.31, "total_tokens": 4350624} {"current_steps": 4335, "total_steps": 42420, "loss": 0.3832, "lr": 0.0009999856719533349, "epoch": 2.0438472418670437, "percentage": 10.22, "elapsed_time": "0:14:07", "remaining_time": "2:04:08", "throughput": 5136.32, "total_tokens": 4354720} {"current_steps": 4340, "total_steps": 42420, "loss": 0.3969, "lr": 0.0009999840722448978, "epoch": 2.046204620462046, "percentage": 10.23, "elapsed_time": "0:14:09", "remaining_time": "2:04:10", "throughput": 5137.51, "total_tokens": 4362432} {"current_steps": 4345, "total_steps": 42420, "loss": 0.3595, "lr": 0.0009999823878976467, "epoch": 2.0485619990570485, "percentage": 10.24, "elapsed_time": "0:14:10", "remaining_time": "2:04:09", "throughput": 5138.39, "total_tokens": 4368448} {"current_steps": 4350, "total_steps": 42420, "loss": 0.3566, "lr": 0.0009999806189118663, "epoch": 2.050919377652051, "percentage": 10.25, "elapsed_time": "0:14:11", "remaining_time": "2:04:08", "throughput": 5138.85, "total_tokens": 4373696} {"current_steps": 4355, "total_steps": 42420, "loss": 0.3784, "lr": 0.0009999787652878564, "epoch": 2.0532767562470533, "percentage": 10.27, "elapsed_time": "0:14:11", "remaining_time": "2:04:06", "throughput": 5138.8, "total_tokens": 4378144} {"current_steps": 4360, "total_steps": 42420, "loss": 0.395, "lr": 0.0009999768270259309, "epoch": 2.0556341348420557, "percentage": 10.28, "elapsed_time": "0:14:12", "remaining_time": "2:04:05", "throughput": 5139.31, "total_tokens": 4383648} {"current_steps": 4365, "total_steps": 42420, "loss": 0.3554, "lr": 0.0009999748041264174, "epoch": 2.057991513437058, "percentage": 10.29, "elapsed_time": "0:14:13", "remaining_time": "2:04:05", "throughput": 5140.03, "total_tokens": 4389440} {"current_steps": 4370, "total_steps": 42420, "loss": 0.4362, "lr": 0.0009999726965896588, "epoch": 2.0603488920320605, "percentage": 10.3, "elapsed_time": "0:14:15", "remaining_time": "2:04:04", "throughput": 5140.62, "total_tokens": 4395456} {"current_steps": 4375, "total_steps": 42420, "loss": 0.361, "lr": 0.0009999705044160118, "epoch": 2.062706270627063, "percentage": 10.31, "elapsed_time": "0:14:15", "remaining_time": "2:04:03", "throughput": 5140.7, "total_tokens": 4400064} {"current_steps": 4380, "total_steps": 42420, "loss": 0.3709, "lr": 0.0009999682276058473, "epoch": 2.065063649222065, "percentage": 10.33, "elapsed_time": "0:14:16", "remaining_time": "2:04:01", "throughput": 5140.78, "total_tokens": 4404640} {"current_steps": 4385, "total_steps": 42420, "loss": 0.3395, "lr": 0.0009999658661595508, "epoch": 2.0674210278170673, "percentage": 10.34, "elapsed_time": "0:14:17", "remaining_time": "2:03:59", "throughput": 5140.86, "total_tokens": 4409152} {"current_steps": 4390, "total_steps": 42420, "loss": 0.3724, "lr": 0.0009999634200775222, "epoch": 2.0697784064120697, "percentage": 10.35, "elapsed_time": "0:14:18", "remaining_time": "2:03:58", "throughput": 5141.65, "total_tokens": 4415232} {"current_steps": 4395, "total_steps": 42420, "loss": 0.3753, "lr": 0.0009999608893601754, "epoch": 2.072135785007072, "percentage": 10.36, "elapsed_time": "0:14:19", "remaining_time": "2:03:57", "throughput": 5141.78, "total_tokens": 4419904} {"current_steps": 4400, "total_steps": 42420, "loss": 0.3362, "lr": 0.0009999582740079392, "epoch": 2.0744931636020745, "percentage": 10.37, "elapsed_time": "0:14:20", "remaining_time": "2:03:54", "throughput": 5141.58, "total_tokens": 4423552} {"current_steps": 4405, "total_steps": 42420, "loss": 0.3914, "lr": 0.0009999555740212555, "epoch": 2.076850542197077, "percentage": 10.38, "elapsed_time": "0:14:21", "remaining_time": "2:03:54", "throughput": 5142.51, "total_tokens": 4430272} {"current_steps": 4410, "total_steps": 42420, "loss": 0.444, "lr": 0.0009999527894005821, "epoch": 2.0792079207920793, "percentage": 10.4, "elapsed_time": "0:14:22", "remaining_time": "2:03:52", "throughput": 5142.65, "total_tokens": 4434976} {"current_steps": 4415, "total_steps": 42420, "loss": 0.6722, "lr": 0.0009999499201463905, "epoch": 2.0815652993870817, "percentage": 10.41, "elapsed_time": "0:14:23", "remaining_time": "2:03:53", "throughput": 5143.6, "total_tokens": 4441536} {"current_steps": 4420, "total_steps": 42420, "loss": 0.4308, "lr": 0.0009999469662591656, "epoch": 2.083922677982084, "percentage": 10.42, "elapsed_time": "0:14:24", "remaining_time": "2:03:51", "throughput": 5143.92, "total_tokens": 4446336} {"current_steps": 4425, "total_steps": 42420, "loss": 0.3729, "lr": 0.0009999439277394082, "epoch": 2.0862800565770865, "percentage": 10.43, "elapsed_time": "0:14:25", "remaining_time": "2:03:49", "throughput": 5143.68, "total_tokens": 4450400} {"current_steps": 4430, "total_steps": 42420, "loss": 0.4302, "lr": 0.0009999408045876324, "epoch": 2.0886374351720884, "percentage": 10.44, "elapsed_time": "0:14:26", "remaining_time": "2:03:47", "throughput": 5144.09, "total_tokens": 4455488} {"current_steps": 4435, "total_steps": 42420, "loss": 0.3509, "lr": 0.0009999375968043668, "epoch": 2.090994813767091, "percentage": 10.45, "elapsed_time": "0:14:27", "remaining_time": "2:03:46", "throughput": 5144.57, "total_tokens": 4460928} {"current_steps": 4440, "total_steps": 42420, "loss": 0.365, "lr": 0.0009999343043901546, "epoch": 2.0933521923620932, "percentage": 10.47, "elapsed_time": "0:14:27", "remaining_time": "2:03:44", "throughput": 5144.69, "total_tokens": 4465440} {"current_steps": 4445, "total_steps": 42420, "loss": 0.379, "lr": 0.0009999309273455528, "epoch": 2.0957095709570956, "percentage": 10.48, "elapsed_time": "0:14:28", "remaining_time": "2:03:41", "throughput": 5144.46, "total_tokens": 4469216} {"current_steps": 4450, "total_steps": 42420, "loss": 0.3811, "lr": 0.0009999274656711337, "epoch": 2.098066949552098, "percentage": 10.49, "elapsed_time": "0:14:29", "remaining_time": "2:03:40", "throughput": 5144.84, "total_tokens": 4474336} {"current_steps": 4455, "total_steps": 42420, "loss": 0.3476, "lr": 0.0009999239193674828, "epoch": 2.1004243281471005, "percentage": 10.5, "elapsed_time": "0:14:30", "remaining_time": "2:03:38", "throughput": 5144.83, "total_tokens": 4478528} {"current_steps": 4460, "total_steps": 42420, "loss": 0.5833, "lr": 0.0009999202884352005, "epoch": 2.102781706742103, "percentage": 10.51, "elapsed_time": "0:14:31", "remaining_time": "2:03:36", "throughput": 5145.13, "total_tokens": 4483328} {"current_steps": 4465, "total_steps": 42420, "loss": 0.3421, "lr": 0.0009999165728749016, "epoch": 2.1051390853371053, "percentage": 10.53, "elapsed_time": "0:14:32", "remaining_time": "2:03:34", "throughput": 5145.34, "total_tokens": 4487936} {"current_steps": 4470, "total_steps": 42420, "loss": 0.392, "lr": 0.000999912772687215, "epoch": 2.1074964639321077, "percentage": 10.54, "elapsed_time": "0:14:33", "remaining_time": "2:03:32", "throughput": 5145.31, "total_tokens": 4492128} {"current_steps": 4475, "total_steps": 42420, "loss": 0.3478, "lr": 0.000999908887872784, "epoch": 2.10985384252711, "percentage": 10.55, "elapsed_time": "0:14:34", "remaining_time": "2:03:31", "throughput": 5146.06, "total_tokens": 4497920} {"current_steps": 4480, "total_steps": 42420, "loss": 0.3915, "lr": 0.0009999049184322665, "epoch": 2.112211221122112, "percentage": 10.56, "elapsed_time": "0:14:34", "remaining_time": "2:03:29", "throughput": 5146.19, "total_tokens": 4502688} {"current_steps": 4485, "total_steps": 42420, "loss": 0.4569, "lr": 0.000999900864366334, "epoch": 2.1145685997171144, "percentage": 10.57, "elapsed_time": "0:14:35", "remaining_time": "2:03:27", "throughput": 5146.18, "total_tokens": 4506752} {"current_steps": 4490, "total_steps": 42420, "loss": 0.3968, "lr": 0.0009998967256756729, "epoch": 2.116925978312117, "percentage": 10.58, "elapsed_time": "0:14:36", "remaining_time": "2:03:25", "throughput": 5146.49, "total_tokens": 4511648} {"current_steps": 4495, "total_steps": 42420, "loss": 0.363, "lr": 0.000999892502360984, "epoch": 2.119283356907119, "percentage": 10.6, "elapsed_time": "0:14:37", "remaining_time": "2:03:24", "throughput": 5147.05, "total_tokens": 4516992} {"current_steps": 4500, "total_steps": 42420, "loss": 0.2815, "lr": 0.000999888194422982, "epoch": 2.1216407355021216, "percentage": 10.61, "elapsed_time": "0:14:38", "remaining_time": "2:03:23", "throughput": 5147.63, "total_tokens": 4522560} {"current_steps": 4505, "total_steps": 42420, "loss": 0.2983, "lr": 0.0009998838018623966, "epoch": 2.123998114097124, "percentage": 10.62, "elapsed_time": "0:14:39", "remaining_time": "2:03:22", "throughput": 5147.9, "total_tokens": 4527584} {"current_steps": 4510, "total_steps": 42420, "loss": 0.4211, "lr": 0.0009998793246799708, "epoch": 2.1263554926921264, "percentage": 10.63, "elapsed_time": "0:14:40", "remaining_time": "2:03:20", "throughput": 5148.32, "total_tokens": 4532768} {"current_steps": 4515, "total_steps": 42420, "loss": 0.3392, "lr": 0.0009998747628764632, "epoch": 2.128712871287129, "percentage": 10.64, "elapsed_time": "0:14:41", "remaining_time": "2:03:19", "throughput": 5148.55, "total_tokens": 4537536} {"current_steps": 4520, "total_steps": 42420, "loss": 0.3646, "lr": 0.0009998701164526452, "epoch": 2.1310702498821312, "percentage": 10.66, "elapsed_time": "0:14:42", "remaining_time": "2:03:16", "throughput": 5148.44, "total_tokens": 4541600} {"current_steps": 4525, "total_steps": 42420, "loss": 0.3896, "lr": 0.000999865385409304, "epoch": 2.1334276284771336, "percentage": 10.67, "elapsed_time": "0:14:43", "remaining_time": "2:03:15", "throughput": 5149.06, "total_tokens": 4547168} {"current_steps": 4530, "total_steps": 42420, "loss": 0.3367, "lr": 0.0009998605697472403, "epoch": 2.1357850070721356, "percentage": 10.68, "elapsed_time": "0:14:43", "remaining_time": "2:03:13", "throughput": 5149.06, "total_tokens": 4551616} {"current_steps": 4535, "total_steps": 42420, "loss": 0.2875, "lr": 0.0009998556694672692, "epoch": 2.138142385667138, "percentage": 10.69, "elapsed_time": "0:14:44", "remaining_time": "2:03:12", "throughput": 5149.48, "total_tokens": 4556736} {"current_steps": 4540, "total_steps": 42420, "loss": 0.4143, "lr": 0.0009998506845702202, "epoch": 2.1404997642621404, "percentage": 10.7, "elapsed_time": "0:14:45", "remaining_time": "2:03:10", "throughput": 5149.59, "total_tokens": 4561088} {"current_steps": 4545, "total_steps": 42420, "loss": 0.4621, "lr": 0.0009998456150569375, "epoch": 2.142857142857143, "percentage": 10.71, "elapsed_time": "0:14:46", "remaining_time": "2:03:07", "throughput": 5149.42, "total_tokens": 4565056} {"current_steps": 4550, "total_steps": 42420, "loss": 0.39, "lr": 0.000999840460928279, "epoch": 2.145214521452145, "percentage": 10.73, "elapsed_time": "0:14:47", "remaining_time": "2:03:05", "throughput": 5149.58, "total_tokens": 4569664} {"current_steps": 4555, "total_steps": 42420, "loss": 0.4048, "lr": 0.0009998352221851172, "epoch": 2.1475719000471476, "percentage": 10.74, "elapsed_time": "0:14:48", "remaining_time": "2:03:04", "throughput": 5149.67, "total_tokens": 4574368} {"current_steps": 4560, "total_steps": 42420, "loss": 0.3723, "lr": 0.0009998298988283393, "epoch": 2.14992927864215, "percentage": 10.75, "elapsed_time": "0:14:49", "remaining_time": "2:03:02", "throughput": 5149.9, "total_tokens": 4579072} {"current_steps": 4565, "total_steps": 42420, "loss": 0.3625, "lr": 0.0009998244908588458, "epoch": 2.1522866572371524, "percentage": 10.76, "elapsed_time": "0:14:50", "remaining_time": "2:03:01", "throughput": 5150.3, "total_tokens": 4584320} {"current_steps": 4570, "total_steps": 42420, "loss": 0.371, "lr": 0.0009998189982775528, "epoch": 2.154644035832155, "percentage": 10.77, "elapsed_time": "0:14:50", "remaining_time": "2:02:59", "throughput": 5150.29, "total_tokens": 4588608} {"current_steps": 4575, "total_steps": 42420, "loss": 0.3494, "lr": 0.0009998134210853897, "epoch": 2.157001414427157, "percentage": 10.79, "elapsed_time": "0:14:51", "remaining_time": "2:02:56", "throughput": 5150.26, "total_tokens": 4592928} {"current_steps": 4580, "total_steps": 42420, "loss": 0.3514, "lr": 0.0009998077592833006, "epoch": 2.159358793022159, "percentage": 10.8, "elapsed_time": "0:14:52", "remaining_time": "2:02:56", "throughput": 5151.07, "total_tokens": 4599072} {"current_steps": 4585, "total_steps": 42420, "loss": 0.3805, "lr": 0.0009998020128722444, "epoch": 2.1617161716171616, "percentage": 10.81, "elapsed_time": "0:14:53", "remaining_time": "2:02:54", "throughput": 5151.36, "total_tokens": 4603808} {"current_steps": 4590, "total_steps": 42420, "loss": 0.1962, "lr": 0.0009997961818531934, "epoch": 2.164073550212164, "percentage": 10.82, "elapsed_time": "0:14:54", "remaining_time": "2:02:53", "throughput": 5151.56, "total_tokens": 4608512} {"current_steps": 4595, "total_steps": 42420, "loss": 0.3126, "lr": 0.000999790266227135, "epoch": 2.1664309288071664, "percentage": 10.83, "elapsed_time": "0:14:55", "remaining_time": "2:02:53", "throughput": 5152.35, "total_tokens": 4615200} {"current_steps": 4600, "total_steps": 42420, "loss": 0.4648, "lr": 0.0009997842659950705, "epoch": 2.1687883074021688, "percentage": 10.84, "elapsed_time": "0:14:56", "remaining_time": "2:02:53", "throughput": 5152.92, "total_tokens": 4620992} {"current_steps": 4605, "total_steps": 42420, "loss": 0.3599, "lr": 0.0009997781811580154, "epoch": 2.171145685997171, "percentage": 10.86, "elapsed_time": "0:14:57", "remaining_time": "2:02:50", "throughput": 5152.88, "total_tokens": 4625248} {"current_steps": 4610, "total_steps": 42420, "loss": 0.3241, "lr": 0.000999772011717, "epoch": 2.1735030645921736, "percentage": 10.87, "elapsed_time": "0:14:58", "remaining_time": "2:02:49", "throughput": 5153.05, "total_tokens": 4629856} {"current_steps": 4615, "total_steps": 42420, "loss": 0.5135, "lr": 0.000999765757673069, "epoch": 2.175860443187176, "percentage": 10.88, "elapsed_time": "0:14:59", "remaining_time": "2:02:46", "throughput": 5152.94, "total_tokens": 4634048} {"current_steps": 4620, "total_steps": 42420, "loss": 0.3132, "lr": 0.0009997594190272806, "epoch": 2.1782178217821784, "percentage": 10.89, "elapsed_time": "0:15:00", "remaining_time": "2:02:46", "throughput": 5153.48, "total_tokens": 4639744} {"current_steps": 4625, "total_steps": 42420, "loss": 0.3517, "lr": 0.0009997529957807077, "epoch": 2.1805752003771808, "percentage": 10.9, "elapsed_time": "0:15:01", "remaining_time": "2:02:45", "throughput": 5154.16, "total_tokens": 4645472} {"current_steps": 4630, "total_steps": 42420, "loss": 0.3339, "lr": 0.0009997464879344382, "epoch": 2.1829325789721827, "percentage": 10.91, "elapsed_time": "0:15:02", "remaining_time": "2:02:44", "throughput": 5154.73, "total_tokens": 4651200} {"current_steps": 4635, "total_steps": 42420, "loss": 0.3773, "lr": 0.0009997398954895735, "epoch": 2.185289957567185, "percentage": 10.93, "elapsed_time": "0:15:03", "remaining_time": "2:02:43", "throughput": 5154.99, "total_tokens": 4656288} {"current_steps": 4640, "total_steps": 42420, "loss": 0.2953, "lr": 0.0009997332184472295, "epoch": 2.1876473361621875, "percentage": 10.94, "elapsed_time": "0:15:04", "remaining_time": "2:02:41", "throughput": 5155.08, "total_tokens": 4660832} {"current_steps": 4645, "total_steps": 42420, "loss": 0.3478, "lr": 0.0009997264568085368, "epoch": 2.19000471475719, "percentage": 10.95, "elapsed_time": "0:15:05", "remaining_time": "2:02:40", "throughput": 5155.59, "total_tokens": 4666304} {"current_steps": 4650, "total_steps": 42420, "loss": 0.2856, "lr": 0.0009997196105746395, "epoch": 2.1923620933521923, "percentage": 10.96, "elapsed_time": "0:15:05", "remaining_time": "2:02:38", "throughput": 5155.61, "total_tokens": 4670592} {"current_steps": 4655, "total_steps": 42420, "loss": 0.3962, "lr": 0.0009997126797466971, "epoch": 2.1947194719471947, "percentage": 10.97, "elapsed_time": "0:15:06", "remaining_time": "2:02:36", "throughput": 5155.59, "total_tokens": 4674848} {"current_steps": 4660, "total_steps": 42420, "loss": 0.4261, "lr": 0.0009997056643258828, "epoch": 2.197076850542197, "percentage": 10.99, "elapsed_time": "0:15:07", "remaining_time": "2:02:34", "throughput": 5155.7, "total_tokens": 4679200} {"current_steps": 4665, "total_steps": 42420, "loss": 0.2899, "lr": 0.0009996985643133838, "epoch": 2.1994342291371995, "percentage": 11.0, "elapsed_time": "0:15:08", "remaining_time": "2:02:32", "throughput": 5155.96, "total_tokens": 4683936} {"current_steps": 4670, "total_steps": 42420, "loss": 0.6573, "lr": 0.0009996913797104023, "epoch": 2.201791607732202, "percentage": 11.01, "elapsed_time": "0:15:09", "remaining_time": "2:02:30", "throughput": 5156.21, "total_tokens": 4688864} {"current_steps": 4675, "total_steps": 42420, "loss": 0.5003, "lr": 0.0009996841105181548, "epoch": 2.2041489863272044, "percentage": 11.02, "elapsed_time": "0:15:10", "remaining_time": "2:02:28", "throughput": 5156.53, "total_tokens": 4693600} {"current_steps": 4680, "total_steps": 42420, "loss": 0.5135, "lr": 0.0009996767567378713, "epoch": 2.2065063649222063, "percentage": 11.03, "elapsed_time": "0:15:11", "remaining_time": "2:02:26", "throughput": 5156.5, "total_tokens": 4697920} {"current_steps": 4685, "total_steps": 42420, "loss": 0.3924, "lr": 0.0009996693183707969, "epoch": 2.2088637435172087, "percentage": 11.04, "elapsed_time": "0:15:11", "remaining_time": "2:02:25", "throughput": 5156.57, "total_tokens": 4702432} {"current_steps": 4690, "total_steps": 42420, "loss": 0.3491, "lr": 0.0009996617954181905, "epoch": 2.211221122112211, "percentage": 11.06, "elapsed_time": "0:15:12", "remaining_time": "2:02:24", "throughput": 5157.05, "total_tokens": 4707936} {"current_steps": 4695, "total_steps": 42420, "loss": 0.3901, "lr": 0.0009996541878813264, "epoch": 2.2135785007072135, "percentage": 11.07, "elapsed_time": "0:15:13", "remaining_time": "2:02:22", "throughput": 5157.31, "total_tokens": 4712864} {"current_steps": 4700, "total_steps": 42420, "loss": 0.389, "lr": 0.0009996464957614916, "epoch": 2.215935879302216, "percentage": 11.08, "elapsed_time": "0:15:14", "remaining_time": "2:02:21", "throughput": 5157.51, "total_tokens": 4717728} {"current_steps": 4705, "total_steps": 42420, "loss": 0.3767, "lr": 0.000999638719059989, "epoch": 2.2182932578972183, "percentage": 11.09, "elapsed_time": "0:15:15", "remaining_time": "2:02:19", "throughput": 5157.54, "total_tokens": 4722176} {"current_steps": 4710, "total_steps": 42420, "loss": 0.3659, "lr": 0.0009996308577781342, "epoch": 2.2206506364922207, "percentage": 11.1, "elapsed_time": "0:15:16", "remaining_time": "2:02:17", "throughput": 5157.78, "total_tokens": 4726912} {"current_steps": 4715, "total_steps": 42420, "loss": 0.3607, "lr": 0.0009996229119172586, "epoch": 2.223008015087223, "percentage": 11.12, "elapsed_time": "0:15:17", "remaining_time": "2:02:16", "throughput": 5158.31, "total_tokens": 4732640} {"current_steps": 4720, "total_steps": 42420, "loss": 0.3663, "lr": 0.000999614881478707, "epoch": 2.2253653936822255, "percentage": 11.13, "elapsed_time": "0:15:18", "remaining_time": "2:02:15", "throughput": 5158.76, "total_tokens": 4737920} {"current_steps": 4725, "total_steps": 42420, "loss": 0.3111, "lr": 0.0009996067664638393, "epoch": 2.227722772277228, "percentage": 11.14, "elapsed_time": "0:15:19", "remaining_time": "2:02:15", "throughput": 5159.34, "total_tokens": 4743680} {"current_steps": 4730, "total_steps": 42420, "loss": 0.4588, "lr": 0.0009995985668740286, "epoch": 2.23008015087223, "percentage": 11.15, "elapsed_time": "0:15:20", "remaining_time": "2:02:15", "throughput": 5160.15, "total_tokens": 4750112} {"current_steps": 4735, "total_steps": 42420, "loss": 0.3439, "lr": 0.0009995902827106633, "epoch": 2.2324375294672323, "percentage": 11.16, "elapsed_time": "0:15:21", "remaining_time": "2:02:14", "throughput": 5160.79, "total_tokens": 4756032} {"current_steps": 4740, "total_steps": 42420, "loss": 0.3608, "lr": 0.0009995819139751456, "epoch": 2.2347949080622347, "percentage": 11.17, "elapsed_time": "0:15:22", "remaining_time": "2:02:13", "throughput": 5161.14, "total_tokens": 4761120} {"current_steps": 4745, "total_steps": 42420, "loss": 0.4285, "lr": 0.0009995734606688925, "epoch": 2.237152286657237, "percentage": 11.19, "elapsed_time": "0:15:23", "remaining_time": "2:02:11", "throughput": 5161.48, "total_tokens": 4766112} {"current_steps": 4750, "total_steps": 42420, "loss": 0.3147, "lr": 0.0009995649227933345, "epoch": 2.2395096652522395, "percentage": 11.2, "elapsed_time": "0:15:24", "remaining_time": "2:02:10", "throughput": 5161.9, "total_tokens": 4771552} {"current_steps": 4755, "total_steps": 42420, "loss": 0.509, "lr": 0.0009995563003499174, "epoch": 2.241867043847242, "percentage": 11.21, "elapsed_time": "0:15:25", "remaining_time": "2:02:09", "throughput": 5162.43, "total_tokens": 4777056} {"current_steps": 4760, "total_steps": 42420, "loss": 0.4503, "lr": 0.0009995475933401005, "epoch": 2.2442244224422443, "percentage": 11.22, "elapsed_time": "0:15:26", "remaining_time": "2:02:09", "throughput": 5162.8, "total_tokens": 4782624} {"current_steps": 4765, "total_steps": 42420, "loss": 0.3971, "lr": 0.000999538801765358, "epoch": 2.2465818010372467, "percentage": 11.23, "elapsed_time": "0:15:27", "remaining_time": "2:02:06", "throughput": 5162.53, "total_tokens": 4786368} {"current_steps": 4770, "total_steps": 42420, "loss": 0.3327, "lr": 0.000999529925627178, "epoch": 2.248939179632249, "percentage": 11.24, "elapsed_time": "0:15:28", "remaining_time": "2:02:04", "throughput": 5162.71, "total_tokens": 4791072} {"current_steps": 4775, "total_steps": 42420, "loss": 0.3935, "lr": 0.0009995209649270632, "epoch": 2.251296558227251, "percentage": 11.26, "elapsed_time": "0:15:28", "remaining_time": "2:02:02", "throughput": 5162.8, "total_tokens": 4795456} {"current_steps": 4780, "total_steps": 42420, "loss": 0.2614, "lr": 0.0009995119196665302, "epoch": 2.2536539368222535, "percentage": 11.27, "elapsed_time": "0:15:29", "remaining_time": "2:02:01", "throughput": 5162.87, "total_tokens": 4800224} {"current_steps": 4785, "total_steps": 42420, "loss": 0.3405, "lr": 0.0009995027898471107, "epoch": 2.256011315417256, "percentage": 11.28, "elapsed_time": "0:15:30", "remaining_time": "2:01:59", "throughput": 5163.01, "total_tokens": 4804832} {"current_steps": 4790, "total_steps": 42420, "loss": 0.3172, "lr": 0.00099949357547035, "epoch": 2.2583686940122583, "percentage": 11.29, "elapsed_time": "0:15:31", "remaining_time": "2:01:57", "throughput": 5163.07, "total_tokens": 4809216} {"current_steps": 4795, "total_steps": 42420, "loss": 0.3354, "lr": 0.0009994842765378075, "epoch": 2.2607260726072607, "percentage": 11.3, "elapsed_time": "0:15:32", "remaining_time": "2:01:56", "throughput": 5163.33, "total_tokens": 4814208} {"current_steps": 4800, "total_steps": 42420, "loss": 0.3162, "lr": 0.000999474893051058, "epoch": 2.263083451202263, "percentage": 11.32, "elapsed_time": "0:15:33", "remaining_time": "2:01:54", "throughput": 5163.3, "total_tokens": 4818592} {"current_steps": 4805, "total_steps": 42420, "loss": 0.3659, "lr": 0.0009994654250116898, "epoch": 2.2654408297972655, "percentage": 11.33, "elapsed_time": "0:15:34", "remaining_time": "2:01:54", "throughput": 5164.15, "total_tokens": 4825312} {"current_steps": 4810, "total_steps": 42420, "loss": 0.3155, "lr": 0.0009994558724213053, "epoch": 2.267798208392268, "percentage": 11.34, "elapsed_time": "0:15:35", "remaining_time": "2:01:53", "throughput": 5164.56, "total_tokens": 4830624} {"current_steps": 4815, "total_steps": 42420, "loss": 0.2872, "lr": 0.000999446235281522, "epoch": 2.2701555869872703, "percentage": 11.35, "elapsed_time": "0:15:36", "remaining_time": "2:01:51", "throughput": 5164.5, "total_tokens": 4834944} {"current_steps": 4820, "total_steps": 42420, "loss": 0.451, "lr": 0.0009994365135939714, "epoch": 2.2725129655822727, "percentage": 11.36, "elapsed_time": "0:15:37", "remaining_time": "2:01:49", "throughput": 5164.61, "total_tokens": 4839552} {"current_steps": 4825, "total_steps": 42420, "loss": 0.3584, "lr": 0.000999426707360299, "epoch": 2.274870344177275, "percentage": 11.37, "elapsed_time": "0:15:37", "remaining_time": "2:01:48", "throughput": 5164.85, "total_tokens": 4844480} {"current_steps": 4830, "total_steps": 42420, "loss": 0.3885, "lr": 0.0009994168165821644, "epoch": 2.2772277227722775, "percentage": 11.39, "elapsed_time": "0:15:38", "remaining_time": "2:01:47", "throughput": 5165.2, "total_tokens": 4849888} {"current_steps": 4835, "total_steps": 42420, "loss": 0.3165, "lr": 0.0009994068412612427, "epoch": 2.2795851013672794, "percentage": 11.4, "elapsed_time": "0:15:39", "remaining_time": "2:01:46", "throughput": 5165.35, "total_tokens": 4854880} {"current_steps": 4840, "total_steps": 42420, "loss": 0.3674, "lr": 0.0009993967813992223, "epoch": 2.281942479962282, "percentage": 11.41, "elapsed_time": "0:15:41", "remaining_time": "2:01:47", "throughput": 5166.32, "total_tokens": 4862432} {"current_steps": 4845, "total_steps": 42420, "loss": 0.3062, "lr": 0.0009993866369978057, "epoch": 2.2842998585572842, "percentage": 11.42, "elapsed_time": "0:15:42", "remaining_time": "2:01:48", "throughput": 5167.39, "total_tokens": 4869376} {"current_steps": 4850, "total_steps": 42420, "loss": 0.4502, "lr": 0.000999376408058711, "epoch": 2.2866572371522866, "percentage": 11.43, "elapsed_time": "0:15:43", "remaining_time": "2:01:48", "throughput": 5167.95, "total_tokens": 4875616} {"current_steps": 4855, "total_steps": 42420, "loss": 0.388, "lr": 0.0009993660945836688, "epoch": 2.289014615747289, "percentage": 11.45, "elapsed_time": "0:15:44", "remaining_time": "2:01:46", "throughput": 5168.21, "total_tokens": 4880544} {"current_steps": 4860, "total_steps": 42420, "loss": 0.3443, "lr": 0.0009993556965744261, "epoch": 2.2913719943422914, "percentage": 11.46, "elapsed_time": "0:15:45", "remaining_time": "2:01:45", "throughput": 5168.76, "total_tokens": 4886176} {"current_steps": 4865, "total_steps": 42420, "loss": 0.4661, "lr": 0.0009993452140327423, "epoch": 2.293729372937294, "percentage": 11.47, "elapsed_time": "0:15:46", "remaining_time": "2:01:43", "throughput": 5168.78, "total_tokens": 4890496} {"current_steps": 4870, "total_steps": 42420, "loss": 0.3957, "lr": 0.000999334646960392, "epoch": 2.2960867515322962, "percentage": 11.48, "elapsed_time": "0:15:47", "remaining_time": "2:01:42", "throughput": 5168.97, "total_tokens": 4895712} {"current_steps": 4875, "total_steps": 42420, "loss": 0.3396, "lr": 0.0009993239953591645, "epoch": 2.298444130127298, "percentage": 11.49, "elapsed_time": "0:15:48", "remaining_time": "2:01:41", "throughput": 5169.16, "total_tokens": 4900640} {"current_steps": 4880, "total_steps": 42420, "loss": 0.3523, "lr": 0.0009993132592308624, "epoch": 2.3008015087223006, "percentage": 11.5, "elapsed_time": "0:15:48", "remaining_time": "2:01:39", "throughput": 5169.24, "total_tokens": 4905344} {"current_steps": 4885, "total_steps": 42420, "loss": 0.3474, "lr": 0.0009993024385773035, "epoch": 2.303158887317303, "percentage": 11.52, "elapsed_time": "0:15:49", "remaining_time": "2:01:38", "throughput": 5169.52, "total_tokens": 4910240} {"current_steps": 4890, "total_steps": 42420, "loss": 0.2599, "lr": 0.0009992915334003194, "epoch": 2.3055162659123054, "percentage": 11.53, "elapsed_time": "0:15:50", "remaining_time": "2:01:36", "throughput": 5169.69, "total_tokens": 4915072} {"current_steps": 4895, "total_steps": 42420, "loss": 0.4502, "lr": 0.0009992805437017563, "epoch": 2.307873644507308, "percentage": 11.54, "elapsed_time": "0:15:51", "remaining_time": "2:01:35", "throughput": 5170.02, "total_tokens": 4920096} {"current_steps": 4900, "total_steps": 42420, "loss": 0.3395, "lr": 0.0009992694694834741, "epoch": 2.31023102310231, "percentage": 11.55, "elapsed_time": "0:15:52", "remaining_time": "2:01:34", "throughput": 5170.35, "total_tokens": 4925280} {"current_steps": 4905, "total_steps": 42420, "loss": 0.3181, "lr": 0.000999258310747348, "epoch": 2.3125884016973126, "percentage": 11.56, "elapsed_time": "0:15:54", "remaining_time": "2:01:36", "throughput": 5171.8, "total_tokens": 4933920} {"current_steps": 4910, "total_steps": 42420, "loss": 0.3906, "lr": 0.000999247067495267, "epoch": 2.314945780292315, "percentage": 11.57, "elapsed_time": "0:15:54", "remaining_time": "2:01:35", "throughput": 5171.91, "total_tokens": 4938720} {"current_steps": 4915, "total_steps": 42420, "loss": 0.3772, "lr": 0.000999235739729134, "epoch": 2.3173031588873174, "percentage": 11.59, "elapsed_time": "0:15:55", "remaining_time": "2:01:34", "throughput": 5172.52, "total_tokens": 4944672} {"current_steps": 4920, "total_steps": 42420, "loss": 0.358, "lr": 0.000999224327450867, "epoch": 2.31966053748232, "percentage": 11.6, "elapsed_time": "0:15:56", "remaining_time": "2:01:33", "throughput": 5172.93, "total_tokens": 4949856} {"current_steps": 4925, "total_steps": 42420, "loss": 0.3931, "lr": 0.0009992128306623976, "epoch": 2.322017916077322, "percentage": 11.61, "elapsed_time": "0:15:57", "remaining_time": "2:01:32", "throughput": 5173.39, "total_tokens": 4955168} {"current_steps": 4930, "total_steps": 42420, "loss": 0.3018, "lr": 0.0009992012493656721, "epoch": 2.3243752946723246, "percentage": 11.62, "elapsed_time": "0:15:58", "remaining_time": "2:01:30", "throughput": 5173.47, "total_tokens": 4959744} {"current_steps": 4935, "total_steps": 42420, "loss": 0.2874, "lr": 0.000999189583562651, "epoch": 2.3267326732673266, "percentage": 11.63, "elapsed_time": "0:15:59", "remaining_time": "2:01:28", "throughput": 5173.63, "total_tokens": 4964448} {"current_steps": 4940, "total_steps": 42420, "loss": 0.3585, "lr": 0.0009991778332553091, "epoch": 2.329090051862329, "percentage": 11.65, "elapsed_time": "0:16:00", "remaining_time": "2:01:27", "throughput": 5174.12, "total_tokens": 4969792} {"current_steps": 4945, "total_steps": 42420, "loss": 0.3499, "lr": 0.0009991659984456356, "epoch": 2.3314474304573314, "percentage": 11.66, "elapsed_time": "0:16:01", "remaining_time": "2:01:26", "throughput": 5174.55, "total_tokens": 4975136} {"current_steps": 4950, "total_steps": 42420, "loss": 0.308, "lr": 0.0009991540791356342, "epoch": 2.333804809052334, "percentage": 11.67, "elapsed_time": "0:16:02", "remaining_time": "2:01:24", "throughput": 5174.77, "total_tokens": 4979872} {"current_steps": 4955, "total_steps": 42420, "loss": 0.4224, "lr": 0.0009991420753273222, "epoch": 2.336162187647336, "percentage": 11.68, "elapsed_time": "0:16:03", "remaining_time": "2:01:23", "throughput": 5175.1, "total_tokens": 4984896} {"current_steps": 4960, "total_steps": 42420, "loss": 0.3626, "lr": 0.0009991299870227316, "epoch": 2.3385195662423386, "percentage": 11.69, "elapsed_time": "0:16:04", "remaining_time": "2:01:22", "throughput": 5175.62, "total_tokens": 4990464} {"current_steps": 4965, "total_steps": 42420, "loss": 0.4334, "lr": 0.000999117814223909, "epoch": 2.340876944837341, "percentage": 11.7, "elapsed_time": "0:16:05", "remaining_time": "2:01:20", "throughput": 5175.94, "total_tokens": 4995520} {"current_steps": 4970, "total_steps": 42420, "loss": 0.3843, "lr": 0.0009991055569329153, "epoch": 2.3432343234323434, "percentage": 11.72, "elapsed_time": "0:16:05", "remaining_time": "2:01:18", "throughput": 5175.99, "total_tokens": 4999936} {"current_steps": 4975, "total_steps": 42420, "loss": 0.4012, "lr": 0.000999093215151825, "epoch": 2.3455917020273453, "percentage": 11.73, "elapsed_time": "0:16:07", "remaining_time": "2:01:18", "throughput": 5176.71, "total_tokens": 5006176} {"current_steps": 4980, "total_steps": 42420, "loss": 0.3975, "lr": 0.0009990807888827272, "epoch": 2.3479490806223478, "percentage": 11.74, "elapsed_time": "0:16:07", "remaining_time": "2:01:17", "throughput": 5176.89, "total_tokens": 5011136} {"current_steps": 4985, "total_steps": 42420, "loss": 0.3815, "lr": 0.000999068278127726, "epoch": 2.35030645921735, "percentage": 11.75, "elapsed_time": "0:16:08", "remaining_time": "2:01:15", "throughput": 5177.09, "total_tokens": 5015776} {"current_steps": 4990, "total_steps": 42420, "loss": 0.3538, "lr": 0.000999055682888939, "epoch": 2.3526638378123526, "percentage": 11.76, "elapsed_time": "0:16:09", "remaining_time": "2:01:13", "throughput": 5177.14, "total_tokens": 5020064} {"current_steps": 4995, "total_steps": 42420, "loss": 0.3733, "lr": 0.0009990430031684985, "epoch": 2.355021216407355, "percentage": 11.78, "elapsed_time": "0:16:10", "remaining_time": "2:01:12", "throughput": 5177.58, "total_tokens": 5025600} {"current_steps": 5000, "total_steps": 42420, "loss": 0.3484, "lr": 0.0009990302389685506, "epoch": 2.3573785950023574, "percentage": 11.79, "elapsed_time": "0:16:11", "remaining_time": "2:01:11", "throughput": 5178.05, "total_tokens": 5031200} {"current_steps": 5005, "total_steps": 42420, "loss": 0.3403, "lr": 0.0009990173902912562, "epoch": 2.3597359735973598, "percentage": 11.8, "elapsed_time": "0:16:12", "remaining_time": "2:01:09", "throughput": 5178.07, "total_tokens": 5035360} {"current_steps": 5010, "total_steps": 42420, "loss": 0.41, "lr": 0.0009990044571387907, "epoch": 2.362093352192362, "percentage": 11.81, "elapsed_time": "0:16:13", "remaining_time": "2:01:07", "throughput": 5178.41, "total_tokens": 5040288} {"current_steps": 5015, "total_steps": 42420, "loss": 0.3405, "lr": 0.0009989914395133433, "epoch": 2.3644507307873646, "percentage": 11.82, "elapsed_time": "0:16:14", "remaining_time": "2:01:07", "throughput": 5178.96, "total_tokens": 5045952} {"current_steps": 5020, "total_steps": 42420, "loss": 0.3572, "lr": 0.0009989783374171172, "epoch": 2.366808109382367, "percentage": 11.83, "elapsed_time": "0:16:15", "remaining_time": "2:01:06", "throughput": 5179.52, "total_tokens": 5051584} {"current_steps": 5025, "total_steps": 42420, "loss": 0.3613, "lr": 0.0009989651508523312, "epoch": 2.3691654879773694, "percentage": 11.85, "elapsed_time": "0:16:16", "remaining_time": "2:01:04", "throughput": 5179.6, "total_tokens": 5056128} {"current_steps": 5030, "total_steps": 42420, "loss": 0.2774, "lr": 0.0009989518798212168, "epoch": 2.3715228665723718, "percentage": 11.86, "elapsed_time": "0:16:17", "remaining_time": "2:01:03", "throughput": 5179.99, "total_tokens": 5061440} {"current_steps": 5035, "total_steps": 42420, "loss": 0.4311, "lr": 0.000998938524326021, "epoch": 2.3738802451673737, "percentage": 11.87, "elapsed_time": "0:16:18", "remaining_time": "2:01:02", "throughput": 5180.37, "total_tokens": 5066848} {"current_steps": 5040, "total_steps": 42420, "loss": 0.4225, "lr": 0.0009989250843690045, "epoch": 2.376237623762376, "percentage": 11.88, "elapsed_time": "0:16:19", "remaining_time": "2:01:01", "throughput": 5180.68, "total_tokens": 5072000} {"current_steps": 5045, "total_steps": 42420, "loss": 0.4055, "lr": 0.0009989115599524426, "epoch": 2.3785950023573785, "percentage": 11.89, "elapsed_time": "0:16:19", "remaining_time": "2:00:59", "throughput": 5180.86, "total_tokens": 5076832} {"current_steps": 5050, "total_steps": 42420, "loss": 0.3433, "lr": 0.0009988979510786244, "epoch": 2.380952380952381, "percentage": 11.9, "elapsed_time": "0:16:20", "remaining_time": "2:00:58", "throughput": 5181.2, "total_tokens": 5082176} {"current_steps": 5055, "total_steps": 42420, "loss": 0.3959, "lr": 0.000998884257749854, "epoch": 2.3833097595473833, "percentage": 11.92, "elapsed_time": "0:16:21", "remaining_time": "2:00:57", "throughput": 5181.3, "total_tokens": 5086944} {"current_steps": 5060, "total_steps": 42420, "loss": 0.3493, "lr": 0.0009988704799684493, "epoch": 2.3856671381423857, "percentage": 11.93, "elapsed_time": "0:16:22", "remaining_time": "2:00:55", "throughput": 5181.46, "total_tokens": 5091712} {"current_steps": 5065, "total_steps": 42420, "loss": 0.3292, "lr": 0.0009988566177367426, "epoch": 2.388024516737388, "percentage": 11.94, "elapsed_time": "0:16:23", "remaining_time": "2:00:54", "throughput": 5181.93, "total_tokens": 5097440} {"current_steps": 5070, "total_steps": 42420, "loss": 0.3922, "lr": 0.0009988426710570808, "epoch": 2.3903818953323905, "percentage": 11.95, "elapsed_time": "0:16:24", "remaining_time": "2:00:53", "throughput": 5182.19, "total_tokens": 5102656} {"current_steps": 5075, "total_steps": 42420, "loss": 0.3453, "lr": 0.0009988286399318245, "epoch": 2.3927392739273925, "percentage": 11.96, "elapsed_time": "0:16:25", "remaining_time": "2:00:53", "throughput": 5182.77, "total_tokens": 5108448} {"current_steps": 5080, "total_steps": 42420, "loss": 0.3495, "lr": 0.0009988145243633494, "epoch": 2.395096652522395, "percentage": 11.98, "elapsed_time": "0:16:26", "remaining_time": "2:00:52", "throughput": 5183.21, "total_tokens": 5113888} {"current_steps": 5085, "total_steps": 42420, "loss": 0.3572, "lr": 0.0009988003243540443, "epoch": 2.3974540311173973, "percentage": 11.99, "elapsed_time": "0:16:27", "remaining_time": "2:00:49", "throughput": 5183.11, "total_tokens": 5117824} {"current_steps": 5090, "total_steps": 42420, "loss": 0.3781, "lr": 0.0009987860399063137, "epoch": 2.3998114097123997, "percentage": 12.0, "elapsed_time": "0:16:28", "remaining_time": "2:00:48", "throughput": 5183.42, "total_tokens": 5122976} {"current_steps": 5095, "total_steps": 42420, "loss": 0.2512, "lr": 0.0009987716710225755, "epoch": 2.402168788307402, "percentage": 12.01, "elapsed_time": "0:16:29", "remaining_time": "2:00:46", "throughput": 5183.32, "total_tokens": 5127296} {"current_steps": 5100, "total_steps": 42420, "loss": 0.5497, "lr": 0.0009987572177052617, "epoch": 2.4045261669024045, "percentage": 12.02, "elapsed_time": "0:16:30", "remaining_time": "2:00:45", "throughput": 5183.48, "total_tokens": 5132064} {"current_steps": 5105, "total_steps": 42420, "loss": 0.3995, "lr": 0.0009987426799568195, "epoch": 2.406883545497407, "percentage": 12.03, "elapsed_time": "0:16:30", "remaining_time": "2:00:43", "throughput": 5183.6, "total_tokens": 5136576} {"current_steps": 5110, "total_steps": 42420, "loss": 0.3612, "lr": 0.00099872805777971, "epoch": 2.4092409240924093, "percentage": 12.05, "elapsed_time": "0:16:31", "remaining_time": "2:00:42", "throughput": 5183.87, "total_tokens": 5141888} {"current_steps": 5115, "total_steps": 42420, "loss": 0.3532, "lr": 0.000998713351176408, "epoch": 2.4115983026874117, "percentage": 12.06, "elapsed_time": "0:16:32", "remaining_time": "2:00:40", "throughput": 5184.29, "total_tokens": 5147104} {"current_steps": 5120, "total_steps": 42420, "loss": 0.2842, "lr": 0.0009986985601494033, "epoch": 2.413955681282414, "percentage": 12.07, "elapsed_time": "0:16:33", "remaining_time": "2:00:40", "throughput": 5184.87, "total_tokens": 5152832} {"current_steps": 5125, "total_steps": 42420, "loss": 0.3122, "lr": 0.0009986836847012, "epoch": 2.4163130598774165, "percentage": 12.08, "elapsed_time": "0:16:34", "remaining_time": "2:00:38", "throughput": 5185.29, "total_tokens": 5158016} {"current_steps": 5130, "total_steps": 42420, "loss": 0.3515, "lr": 0.000998668724834316, "epoch": 2.418670438472419, "percentage": 12.09, "elapsed_time": "0:16:35", "remaining_time": "2:00:37", "throughput": 5185.53, "total_tokens": 5162976} {"current_steps": 5135, "total_steps": 42420, "loss": 0.316, "lr": 0.0009986536805512837, "epoch": 2.421027817067421, "percentage": 12.11, "elapsed_time": "0:16:36", "remaining_time": "2:00:35", "throughput": 5185.52, "total_tokens": 5167296} {"current_steps": 5140, "total_steps": 42420, "loss": 0.3589, "lr": 0.00099863855185465, "epoch": 2.4233851956624233, "percentage": 12.12, "elapsed_time": "0:16:37", "remaining_time": "2:00:36", "throughput": 5186.77, "total_tokens": 5175104} {"current_steps": 5145, "total_steps": 42420, "loss": 0.3868, "lr": 0.0009986233387469757, "epoch": 2.4257425742574257, "percentage": 12.13, "elapsed_time": "0:16:38", "remaining_time": "2:00:35", "throughput": 5187.12, "total_tokens": 5180416} {"current_steps": 5150, "total_steps": 42420, "loss": 0.3695, "lr": 0.0009986080412308365, "epoch": 2.428099952852428, "percentage": 12.14, "elapsed_time": "0:16:39", "remaining_time": "2:00:34", "throughput": 5187.74, "total_tokens": 5186272} {"current_steps": 5155, "total_steps": 42420, "loss": 0.3437, "lr": 0.0009985926593088217, "epoch": 2.4304573314474305, "percentage": 12.15, "elapsed_time": "0:16:40", "remaining_time": "2:00:33", "throughput": 5187.82, "total_tokens": 5190784} {"current_steps": 5160, "total_steps": 42420, "loss": 0.425, "lr": 0.0009985771929835352, "epoch": 2.432814710042433, "percentage": 12.16, "elapsed_time": "0:16:41", "remaining_time": "2:00:30", "throughput": 5187.83, "total_tokens": 5195040} {"current_steps": 5165, "total_steps": 42420, "loss": 0.3975, "lr": 0.0009985616422575954, "epoch": 2.4351720886374353, "percentage": 12.18, "elapsed_time": "0:16:42", "remaining_time": "2:00:29", "throughput": 5187.78, "total_tokens": 5199424} {"current_steps": 5170, "total_steps": 42420, "loss": 0.3704, "lr": 0.0009985460071336347, "epoch": 2.4375294672324377, "percentage": 12.19, "elapsed_time": "0:16:43", "remaining_time": "2:00:27", "throughput": 5188.16, "total_tokens": 5204576} {"current_steps": 5175, "total_steps": 42420, "loss": 0.3613, "lr": 0.0009985302876142999, "epoch": 2.4398868458274396, "percentage": 12.2, "elapsed_time": "0:16:44", "remaining_time": "2:00:26", "throughput": 5188.52, "total_tokens": 5209920} {"current_steps": 5180, "total_steps": 42420, "loss": 0.3602, "lr": 0.000998514483702252, "epoch": 2.442244224422442, "percentage": 12.21, "elapsed_time": "0:16:45", "remaining_time": "2:00:25", "throughput": 5188.91, "total_tokens": 5215200} {"current_steps": 5185, "total_steps": 42420, "loss": 0.3479, "lr": 0.000998498595400166, "epoch": 2.4446016030174444, "percentage": 12.22, "elapsed_time": "0:16:45", "remaining_time": "2:00:23", "throughput": 5188.89, "total_tokens": 5219296} {"current_steps": 5190, "total_steps": 42420, "loss": 0.3427, "lr": 0.000998482622710732, "epoch": 2.446958981612447, "percentage": 12.23, "elapsed_time": "0:16:46", "remaining_time": "2:00:21", "throughput": 5189.08, "total_tokens": 5224128} {"current_steps": 5195, "total_steps": 42420, "loss": 0.3681, "lr": 0.0009984665656366538, "epoch": 2.4493163602074493, "percentage": 12.25, "elapsed_time": "0:16:47", "remaining_time": "2:00:20", "throughput": 5189.45, "total_tokens": 5229344} {"current_steps": 5200, "total_steps": 42420, "loss": 0.3087, "lr": 0.0009984504241806493, "epoch": 2.4516737388024517, "percentage": 12.26, "elapsed_time": "0:16:48", "remaining_time": "2:00:20", "throughput": 5189.91, "total_tokens": 5235200} {"current_steps": 5205, "total_steps": 42420, "loss": 0.3721, "lr": 0.0009984341983454515, "epoch": 2.454031117397454, "percentage": 12.27, "elapsed_time": "0:16:49", "remaining_time": "2:00:19", "throughput": 5190.54, "total_tokens": 5241184} {"current_steps": 5210, "total_steps": 42420, "loss": 0.3639, "lr": 0.0009984178881338067, "epoch": 2.4563884959924565, "percentage": 12.28, "elapsed_time": "0:16:50", "remaining_time": "2:00:17", "throughput": 5190.63, "total_tokens": 5245664} {"current_steps": 5215, "total_steps": 42420, "loss": 0.4588, "lr": 0.000998401493548476, "epoch": 2.458745874587459, "percentage": 12.29, "elapsed_time": "0:16:51", "remaining_time": "2:00:16", "throughput": 5190.72, "total_tokens": 5250400} {"current_steps": 5220, "total_steps": 42420, "loss": 0.3802, "lr": 0.0009983850145922348, "epoch": 2.4611032531824613, "percentage": 12.31, "elapsed_time": "0:16:52", "remaining_time": "2:00:14", "throughput": 5190.9, "total_tokens": 5255008} {"current_steps": 5225, "total_steps": 42420, "loss": 0.3861, "lr": 0.0009983684512678728, "epoch": 2.4634606317774637, "percentage": 12.32, "elapsed_time": "0:16:53", "remaining_time": "2:00:12", "throughput": 5190.93, "total_tokens": 5259392} {"current_steps": 5230, "total_steps": 42420, "loss": 0.3022, "lr": 0.000998351803578194, "epoch": 2.465818010372466, "percentage": 12.33, "elapsed_time": "0:16:54", "remaining_time": "2:00:10", "throughput": 5191.03, "total_tokens": 5263712} {"current_steps": 5235, "total_steps": 42420, "loss": 0.4213, "lr": 0.0009983350715260162, "epoch": 2.468175388967468, "percentage": 12.34, "elapsed_time": "0:16:54", "remaining_time": "2:00:08", "throughput": 5190.94, "total_tokens": 5267936} {"current_steps": 5240, "total_steps": 42420, "loss": 0.4757, "lr": 0.000998318255114172, "epoch": 2.4705327675624704, "percentage": 12.35, "elapsed_time": "0:16:55", "remaining_time": "2:00:07", "throughput": 5191.23, "total_tokens": 5272928} {"current_steps": 5245, "total_steps": 42420, "loss": 0.4436, "lr": 0.0009983013543455083, "epoch": 2.472890146157473, "percentage": 12.36, "elapsed_time": "0:16:56", "remaining_time": "2:00:05", "throughput": 5191.54, "total_tokens": 5277920} {"current_steps": 5250, "total_steps": 42420, "loss": 0.3143, "lr": 0.0009982843692228859, "epoch": 2.4752475247524752, "percentage": 12.38, "elapsed_time": "0:16:57", "remaining_time": "2:00:03", "throughput": 5191.74, "total_tokens": 5282656} {"current_steps": 5255, "total_steps": 42420, "loss": 0.4461, "lr": 0.0009982672997491803, "epoch": 2.4776049033474776, "percentage": 12.39, "elapsed_time": "0:16:58", "remaining_time": "2:00:02", "throughput": 5191.94, "total_tokens": 5287424} {"current_steps": 5260, "total_steps": 42420, "loss": 0.3363, "lr": 0.0009982501459272807, "epoch": 2.47996228194248, "percentage": 12.4, "elapsed_time": "0:16:59", "remaining_time": "2:00:01", "throughput": 5192.31, "total_tokens": 5292832} {"current_steps": 5265, "total_steps": 42420, "loss": 0.3388, "lr": 0.0009982329077600914, "epoch": 2.4823196605374824, "percentage": 12.41, "elapsed_time": "0:17:00", "remaining_time": "1:59:59", "throughput": 5192.24, "total_tokens": 5296992} {"current_steps": 5270, "total_steps": 42420, "loss": 0.3743, "lr": 0.0009982155852505301, "epoch": 2.484677039132485, "percentage": 12.42, "elapsed_time": "0:17:01", "remaining_time": "1:59:57", "throughput": 5192.36, "total_tokens": 5301760} {"current_steps": 5275, "total_steps": 42420, "loss": 0.2692, "lr": 0.0009981981784015297, "epoch": 2.487034417727487, "percentage": 12.44, "elapsed_time": "0:17:01", "remaining_time": "1:59:56", "throughput": 5192.51, "total_tokens": 5306464} {"current_steps": 5280, "total_steps": 42420, "loss": 0.3728, "lr": 0.0009981806872160365, "epoch": 2.489391796322489, "percentage": 12.45, "elapsed_time": "0:17:02", "remaining_time": "1:59:54", "throughput": 5192.68, "total_tokens": 5311360} {"current_steps": 5285, "total_steps": 42420, "loss": 0.3552, "lr": 0.0009981631116970115, "epoch": 2.4917491749174916, "percentage": 12.46, "elapsed_time": "0:17:03", "remaining_time": "1:59:53", "throughput": 5192.89, "total_tokens": 5316224} {"current_steps": 5290, "total_steps": 42420, "loss": 0.349, "lr": 0.0009981454518474303, "epoch": 2.494106553512494, "percentage": 12.47, "elapsed_time": "0:17:04", "remaining_time": "1:59:51", "throughput": 5192.78, "total_tokens": 5320288} {"current_steps": 5295, "total_steps": 42420, "loss": 0.4067, "lr": 0.0009981277076702819, "epoch": 2.4964639321074964, "percentage": 12.48, "elapsed_time": "0:17:05", "remaining_time": "1:59:50", "throughput": 5192.96, "total_tokens": 5325440} {"current_steps": 5300, "total_steps": 42420, "loss": 0.3188, "lr": 0.0009981098791685703, "epoch": 2.498821310702499, "percentage": 12.49, "elapsed_time": "0:17:06", "remaining_time": "1:59:49", "throughput": 5193.36, "total_tokens": 5331072} {"current_steps": 5305, "total_steps": 42420, "loss": 0.3854, "lr": 0.0009980919663453137, "epoch": 2.501178689297501, "percentage": 12.51, "elapsed_time": "0:17:07", "remaining_time": "1:59:47", "throughput": 5193.41, "total_tokens": 5335488} {"current_steps": 5310, "total_steps": 42420, "loss": 0.3228, "lr": 0.0009980739692035444, "epoch": 2.5035360678925036, "percentage": 12.52, "elapsed_time": "0:17:08", "remaining_time": "1:59:46", "throughput": 5193.57, "total_tokens": 5340576} {"current_steps": 5315, "total_steps": 42420, "loss": 0.3532, "lr": 0.0009980558877463086, "epoch": 2.505893446487506, "percentage": 12.53, "elapsed_time": "0:17:09", "remaining_time": "1:59:45", "throughput": 5193.88, "total_tokens": 5345856} {"current_steps": 5320, "total_steps": 42420, "loss": 0.3503, "lr": 0.000998037721976668, "epoch": 2.5082508250825084, "percentage": 12.54, "elapsed_time": "0:17:10", "remaining_time": "1:59:44", "throughput": 5194.35, "total_tokens": 5351456} {"current_steps": 5325, "total_steps": 42420, "loss": 0.3599, "lr": 0.0009980194718976966, "epoch": 2.510608203677511, "percentage": 12.55, "elapsed_time": "0:17:11", "remaining_time": "1:59:42", "throughput": 5194.48, "total_tokens": 5356096} {"current_steps": 5330, "total_steps": 42420, "loss": 0.368, "lr": 0.0009980011375124852, "epoch": 2.512965582272513, "percentage": 12.56, "elapsed_time": "0:17:11", "remaining_time": "1:59:40", "throughput": 5194.41, "total_tokens": 5360096} {"current_steps": 5335, "total_steps": 42420, "loss": 0.3239, "lr": 0.0009979827188241364, "epoch": 2.515322960867515, "percentage": 12.58, "elapsed_time": "0:17:12", "remaining_time": "1:59:39", "throughput": 5194.54, "total_tokens": 5364768} {"current_steps": 5340, "total_steps": 42420, "loss": 0.481, "lr": 0.0009979642158357686, "epoch": 2.5176803394625176, "percentage": 12.59, "elapsed_time": "0:17:13", "remaining_time": "1:59:37", "throughput": 5194.51, "total_tokens": 5368992} {"current_steps": 5345, "total_steps": 42420, "loss": 0.3427, "lr": 0.000997945628550514, "epoch": 2.52003771805752, "percentage": 12.6, "elapsed_time": "0:17:14", "remaining_time": "1:59:34", "throughput": 5194.42, "total_tokens": 5373024} {"current_steps": 5350, "total_steps": 42420, "loss": 0.3648, "lr": 0.0009979269569715193, "epoch": 2.5223950966525224, "percentage": 12.61, "elapsed_time": "0:17:15", "remaining_time": "1:59:34", "throughput": 5194.92, "total_tokens": 5378688} {"current_steps": 5355, "total_steps": 42420, "loss": 0.3685, "lr": 0.0009979082011019448, "epoch": 2.5247524752475248, "percentage": 12.62, "elapsed_time": "0:17:16", "remaining_time": "1:59:31", "throughput": 5194.67, "total_tokens": 5382336} {"current_steps": 5360, "total_steps": 42420, "loss": 0.3556, "lr": 0.0009978893609449663, "epoch": 2.527109853842527, "percentage": 12.64, "elapsed_time": "0:17:16", "remaining_time": "1:59:29", "throughput": 5194.57, "total_tokens": 5386464} {"current_steps": 5365, "total_steps": 42420, "loss": 0.3863, "lr": 0.0009978704365037722, "epoch": 2.5294672324375296, "percentage": 12.65, "elapsed_time": "0:17:17", "remaining_time": "1:59:28", "throughput": 5194.57, "total_tokens": 5391072} {"current_steps": 5370, "total_steps": 42420, "loss": 0.3428, "lr": 0.0009978514277815668, "epoch": 2.531824611032532, "percentage": 12.66, "elapsed_time": "0:17:18", "remaining_time": "1:59:26", "throughput": 5194.52, "total_tokens": 5395360} {"current_steps": 5375, "total_steps": 42420, "loss": 0.3026, "lr": 0.0009978323347815676, "epoch": 2.534181989627534, "percentage": 12.67, "elapsed_time": "0:17:19", "remaining_time": "1:59:25", "throughput": 5195.01, "total_tokens": 5400800} {"current_steps": 5380, "total_steps": 42420, "loss": 0.2172, "lr": 0.000997813157507007, "epoch": 2.5365393682225363, "percentage": 12.68, "elapsed_time": "0:17:20", "remaining_time": "1:59:23", "throughput": 5195.14, "total_tokens": 5405280} {"current_steps": 5385, "total_steps": 42420, "loss": 0.2319, "lr": 0.000997793895961131, "epoch": 2.5388967468175387, "percentage": 12.69, "elapsed_time": "0:17:21", "remaining_time": "1:59:22", "throughput": 5195.62, "total_tokens": 5411072} {"current_steps": 5390, "total_steps": 42420, "loss": 0.3917, "lr": 0.0009977745501472007, "epoch": 2.541254125412541, "percentage": 12.71, "elapsed_time": "0:17:22", "remaining_time": "1:59:21", "throughput": 5196.0, "total_tokens": 5416480} {"current_steps": 5395, "total_steps": 42420, "loss": 0.4335, "lr": 0.0009977551200684906, "epoch": 2.5436115040075435, "percentage": 12.72, "elapsed_time": "0:17:23", "remaining_time": "1:59:20", "throughput": 5196.07, "total_tokens": 5421120} {"current_steps": 5400, "total_steps": 42420, "loss": 0.4388, "lr": 0.0009977356057282903, "epoch": 2.545968882602546, "percentage": 12.73, "elapsed_time": "0:17:24", "remaining_time": "1:59:18", "throughput": 5196.44, "total_tokens": 5426432} {"current_steps": 5405, "total_steps": 42420, "loss": 0.4461, "lr": 0.000997716007129903, "epoch": 2.5483262611975483, "percentage": 12.74, "elapsed_time": "0:17:25", "remaining_time": "1:59:17", "throughput": 5196.63, "total_tokens": 5431168} {"current_steps": 5410, "total_steps": 42420, "loss": 0.3312, "lr": 0.000997696324276646, "epoch": 2.5506836397925507, "percentage": 12.75, "elapsed_time": "0:17:26", "remaining_time": "1:59:16", "throughput": 5197.15, "total_tokens": 5436960} {"current_steps": 5415, "total_steps": 42420, "loss": 0.4506, "lr": 0.0009976765571718521, "epoch": 2.553041018387553, "percentage": 12.77, "elapsed_time": "0:17:27", "remaining_time": "1:59:16", "throughput": 5197.8, "total_tokens": 5443104} {"current_steps": 5420, "total_steps": 42420, "loss": 0.4078, "lr": 0.0009976567058188673, "epoch": 2.5553983969825556, "percentage": 12.78, "elapsed_time": "0:17:28", "remaining_time": "1:59:16", "throughput": 5198.43, "total_tokens": 5449312} {"current_steps": 5425, "total_steps": 42420, "loss": 0.3584, "lr": 0.0009976367702210515, "epoch": 2.557755775577558, "percentage": 12.79, "elapsed_time": "0:17:29", "remaining_time": "1:59:14", "throughput": 5198.73, "total_tokens": 5454304} {"current_steps": 5430, "total_steps": 42420, "loss": 0.3445, "lr": 0.0009976167503817801, "epoch": 2.5601131541725604, "percentage": 12.8, "elapsed_time": "0:17:29", "remaining_time": "1:59:12", "throughput": 5198.61, "total_tokens": 5458400} {"current_steps": 5435, "total_steps": 42420, "loss": 0.3369, "lr": 0.000997596646304442, "epoch": 2.5624705327675623, "percentage": 12.81, "elapsed_time": "0:17:30", "remaining_time": "1:59:10", "throughput": 5198.76, "total_tokens": 5463040} {"current_steps": 5440, "total_steps": 42420, "loss": 0.4166, "lr": 0.0009975764579924402, "epoch": 2.5648279113625647, "percentage": 12.82, "elapsed_time": "0:17:31", "remaining_time": "1:59:09", "throughput": 5198.81, "total_tokens": 5467616} {"current_steps": 5445, "total_steps": 42420, "loss": 0.2971, "lr": 0.0009975561854491926, "epoch": 2.567185289957567, "percentage": 12.84, "elapsed_time": "0:17:32", "remaining_time": "1:59:07", "throughput": 5198.98, "total_tokens": 5472320} {"current_steps": 5450, "total_steps": 42420, "loss": 0.3411, "lr": 0.0009975358286781306, "epoch": 2.5695426685525695, "percentage": 12.85, "elapsed_time": "0:17:33", "remaining_time": "1:59:05", "throughput": 5199.1, "total_tokens": 5476800} {"current_steps": 5455, "total_steps": 42420, "loss": 0.2874, "lr": 0.0009975153876827009, "epoch": 2.571900047147572, "percentage": 12.86, "elapsed_time": "0:17:34", "remaining_time": "1:59:04", "throughput": 5199.43, "total_tokens": 5481952} {"current_steps": 5460, "total_steps": 42420, "loss": 0.3403, "lr": 0.000997494862466363, "epoch": 2.5742574257425743, "percentage": 12.87, "elapsed_time": "0:17:35", "remaining_time": "1:59:02", "throughput": 5199.47, "total_tokens": 5486432} {"current_steps": 5465, "total_steps": 42420, "loss": 0.3318, "lr": 0.000997474253032592, "epoch": 2.5766148043375767, "percentage": 12.88, "elapsed_time": "0:17:36", "remaining_time": "1:59:01", "throughput": 5199.54, "total_tokens": 5490976} {"current_steps": 5470, "total_steps": 42420, "loss": 0.39, "lr": 0.0009974535593848767, "epoch": 2.578972182932579, "percentage": 12.89, "elapsed_time": "0:17:36", "remaining_time": "1:58:59", "throughput": 5199.48, "total_tokens": 5495232} {"current_steps": 5475, "total_steps": 42420, "loss": 0.3597, "lr": 0.0009974327815267198, "epoch": 2.581329561527581, "percentage": 12.91, "elapsed_time": "0:17:37", "remaining_time": "1:58:57", "throughput": 5199.63, "total_tokens": 5499936} {"current_steps": 5480, "total_steps": 42420, "loss": 0.3392, "lr": 0.0009974119194616391, "epoch": 2.5836869401225835, "percentage": 12.92, "elapsed_time": "0:17:38", "remaining_time": "1:58:56", "throughput": 5199.78, "total_tokens": 5504736} {"current_steps": 5485, "total_steps": 42420, "loss": 0.3525, "lr": 0.000997390973193166, "epoch": 2.586044318717586, "percentage": 12.93, "elapsed_time": "0:17:39", "remaining_time": "1:58:55", "throughput": 5200.12, "total_tokens": 5509952} {"current_steps": 5490, "total_steps": 42420, "loss": 0.3403, "lr": 0.0009973699427248462, "epoch": 2.5884016973125883, "percentage": 12.94, "elapsed_time": "0:17:40", "remaining_time": "1:58:53", "throughput": 5200.3, "total_tokens": 5514688} {"current_steps": 5495, "total_steps": 42420, "loss": 0.3352, "lr": 0.0009973488280602403, "epoch": 2.5907590759075907, "percentage": 12.95, "elapsed_time": "0:17:41", "remaining_time": "1:58:52", "throughput": 5200.46, "total_tokens": 5519520} {"current_steps": 5500, "total_steps": 42420, "loss": 0.3636, "lr": 0.000997327629202922, "epoch": 2.593116454502593, "percentage": 12.97, "elapsed_time": "0:17:42", "remaining_time": "1:58:49", "throughput": 5200.35, "total_tokens": 5523488} {"current_steps": 5505, "total_steps": 42420, "loss": 0.352, "lr": 0.0009973063461564803, "epoch": 2.5954738330975955, "percentage": 12.98, "elapsed_time": "0:17:43", "remaining_time": "1:58:48", "throughput": 5200.79, "total_tokens": 5528928} {"current_steps": 5510, "total_steps": 42420, "loss": 0.3596, "lr": 0.000997284978924518, "epoch": 2.597831211692598, "percentage": 12.99, "elapsed_time": "0:17:44", "remaining_time": "1:58:47", "throughput": 5201.09, "total_tokens": 5534144} {"current_steps": 5515, "total_steps": 42420, "loss": 0.3611, "lr": 0.000997263527510652, "epoch": 2.6001885902876003, "percentage": 13.0, "elapsed_time": "0:17:45", "remaining_time": "1:58:47", "throughput": 5201.51, "total_tokens": 5539872} {"current_steps": 5520, "total_steps": 42420, "loss": 0.457, "lr": 0.0009972419919185142, "epoch": 2.6025459688826027, "percentage": 13.01, "elapsed_time": "0:17:45", "remaining_time": "1:58:45", "throughput": 5201.66, "total_tokens": 5544896} {"current_steps": 5525, "total_steps": 42420, "loss": 0.3504, "lr": 0.0009972203721517496, "epoch": 2.604903347477605, "percentage": 13.02, "elapsed_time": "0:17:46", "remaining_time": "1:58:44", "throughput": 5201.76, "total_tokens": 5549440} {"current_steps": 5530, "total_steps": 42420, "loss": 0.4059, "lr": 0.0009971986682140185, "epoch": 2.6072607260726075, "percentage": 13.04, "elapsed_time": "0:17:47", "remaining_time": "1:58:42", "throughput": 5201.96, "total_tokens": 5554464} {"current_steps": 5535, "total_steps": 42420, "loss": 0.3657, "lr": 0.0009971768801089948, "epoch": 2.6096181046676095, "percentage": 13.05, "elapsed_time": "0:17:48", "remaining_time": "1:58:41", "throughput": 5201.92, "total_tokens": 5558688} {"current_steps": 5540, "total_steps": 42420, "loss": 0.3303, "lr": 0.0009971550078403668, "epoch": 2.611975483262612, "percentage": 13.06, "elapsed_time": "0:17:49", "remaining_time": "1:58:39", "throughput": 5202.05, "total_tokens": 5563456} {"current_steps": 5545, "total_steps": 42420, "loss": 0.3445, "lr": 0.0009971330514118371, "epoch": 2.6143328618576143, "percentage": 13.07, "elapsed_time": "0:17:50", "remaining_time": "1:58:38", "throughput": 5202.28, "total_tokens": 5568320} {"current_steps": 5550, "total_steps": 42420, "loss": 0.3364, "lr": 0.0009971110108271228, "epoch": 2.6166902404526167, "percentage": 13.08, "elapsed_time": "0:17:51", "remaining_time": "1:58:36", "throughput": 5202.54, "total_tokens": 5573472} {"current_steps": 5555, "total_steps": 42420, "loss": 0.2923, "lr": 0.0009970888860899548, "epoch": 2.619047619047619, "percentage": 13.1, "elapsed_time": "0:17:52", "remaining_time": "1:58:35", "throughput": 5202.66, "total_tokens": 5578080} {"current_steps": 5560, "total_steps": 42420, "loss": 0.3393, "lr": 0.0009970666772040784, "epoch": 2.6214049976426215, "percentage": 13.11, "elapsed_time": "0:17:52", "remaining_time": "1:58:33", "throughput": 5202.54, "total_tokens": 5582240} {"current_steps": 5565, "total_steps": 42420, "loss": 0.3333, "lr": 0.0009970443841732532, "epoch": 2.623762376237624, "percentage": 13.12, "elapsed_time": "0:17:53", "remaining_time": "1:58:31", "throughput": 5202.62, "total_tokens": 5586688} {"current_steps": 5570, "total_steps": 42420, "loss": 0.3077, "lr": 0.0009970220070012533, "epoch": 2.6261197548326263, "percentage": 13.13, "elapsed_time": "0:17:54", "remaining_time": "1:58:30", "throughput": 5202.69, "total_tokens": 5591584} {"current_steps": 5575, "total_steps": 42420, "loss": 0.3671, "lr": 0.0009969995456918664, "epoch": 2.6284771334276282, "percentage": 13.14, "elapsed_time": "0:17:55", "remaining_time": "1:58:30", "throughput": 5203.55, "total_tokens": 5598816} {"current_steps": 5580, "total_steps": 42420, "loss": 0.3145, "lr": 0.000996977000248895, "epoch": 2.6308345120226306, "percentage": 13.15, "elapsed_time": "0:17:57", "remaining_time": "1:58:31", "throughput": 5204.47, "total_tokens": 5605952} {"current_steps": 5585, "total_steps": 42420, "loss": 0.2925, "lr": 0.000996954370676156, "epoch": 2.633191890617633, "percentage": 13.17, "elapsed_time": "0:17:58", "remaining_time": "1:58:30", "throughput": 5204.82, "total_tokens": 5611328} {"current_steps": 5590, "total_steps": 42420, "loss": 0.3361, "lr": 0.0009969316569774793, "epoch": 2.6355492692126354, "percentage": 13.18, "elapsed_time": "0:17:58", "remaining_time": "1:58:28", "throughput": 5204.9, "total_tokens": 5616000} {"current_steps": 5595, "total_steps": 42420, "loss": 0.3263, "lr": 0.0009969088591567107, "epoch": 2.637906647807638, "percentage": 13.19, "elapsed_time": "0:17:59", "remaining_time": "1:58:27", "throughput": 5205.06, "total_tokens": 5620672} {"current_steps": 5600, "total_steps": 42420, "loss": 0.2964, "lr": 0.0009968859772177094, "epoch": 2.6402640264026402, "percentage": 13.2, "elapsed_time": "0:18:00", "remaining_time": "1:58:26", "throughput": 5205.24, "total_tokens": 5625696} {"current_steps": 5605, "total_steps": 42420, "loss": 0.3831, "lr": 0.0009968630111643487, "epoch": 2.6426214049976426, "percentage": 13.21, "elapsed_time": "0:18:01", "remaining_time": "1:58:24", "throughput": 5205.57, "total_tokens": 5630880} {"current_steps": 5610, "total_steps": 42420, "loss": 0.3245, "lr": 0.0009968399610005166, "epoch": 2.644978783592645, "percentage": 13.22, "elapsed_time": "0:18:02", "remaining_time": "1:58:23", "throughput": 5205.63, "total_tokens": 5635360} {"current_steps": 5615, "total_steps": 42420, "loss": 0.3354, "lr": 0.0009968168267301147, "epoch": 2.6473361621876474, "percentage": 13.24, "elapsed_time": "0:18:03", "remaining_time": "1:58:21", "throughput": 5205.7, "total_tokens": 5640160} {"current_steps": 5620, "total_steps": 42420, "loss": 0.3164, "lr": 0.0009967936083570598, "epoch": 2.64969354078265, "percentage": 13.25, "elapsed_time": "0:18:04", "remaining_time": "1:58:20", "throughput": 5205.93, "total_tokens": 5645184} {"current_steps": 5625, "total_steps": 42420, "loss": 0.3888, "lr": 0.0009967703058852819, "epoch": 2.6520509193776522, "percentage": 13.26, "elapsed_time": "0:18:05", "remaining_time": "1:58:19", "throughput": 5205.94, "total_tokens": 5649792} {"current_steps": 5630, "total_steps": 42420, "loss": 0.363, "lr": 0.000996746919318726, "epoch": 2.6544082979726547, "percentage": 13.27, "elapsed_time": "0:18:06", "remaining_time": "1:58:18", "throughput": 5206.38, "total_tokens": 5655648} {"current_steps": 5635, "total_steps": 42420, "loss": 0.3196, "lr": 0.0009967234486613508, "epoch": 2.6567656765676566, "percentage": 13.28, "elapsed_time": "0:18:07", "remaining_time": "1:58:17", "throughput": 5206.67, "total_tokens": 5660832} {"current_steps": 5640, "total_steps": 42420, "loss": 0.484, "lr": 0.0009966998939171298, "epoch": 2.659123055162659, "percentage": 13.3, "elapsed_time": "0:18:08", "remaining_time": "1:58:15", "throughput": 5206.77, "total_tokens": 5665504} {"current_steps": 5645, "total_steps": 42420, "loss": 0.2606, "lr": 0.00099667625509005, "epoch": 2.6614804337576614, "percentage": 13.31, "elapsed_time": "0:18:08", "remaining_time": "1:58:14", "throughput": 5206.82, "total_tokens": 5670176} {"current_steps": 5650, "total_steps": 42420, "loss": 0.4262, "lr": 0.0009966525321841134, "epoch": 2.663837812352664, "percentage": 13.32, "elapsed_time": "0:18:09", "remaining_time": "1:58:13", "throughput": 5207.01, "total_tokens": 5675296} {"current_steps": 5655, "total_steps": 42420, "loss": 0.5317, "lr": 0.0009966287252033356, "epoch": 2.666195190947666, "percentage": 13.33, "elapsed_time": "0:18:10", "remaining_time": "1:58:10", "throughput": 5206.74, "total_tokens": 5678816} {"current_steps": 5660, "total_steps": 42420, "loss": 0.3137, "lr": 0.0009966048341517474, "epoch": 2.6685525695426686, "percentage": 13.34, "elapsed_time": "0:18:11", "remaining_time": "1:58:09", "throughput": 5207.02, "total_tokens": 5683808} {"current_steps": 5665, "total_steps": 42420, "loss": 0.3561, "lr": 0.0009965808590333923, "epoch": 2.670909948137671, "percentage": 13.35, "elapsed_time": "0:18:12", "remaining_time": "1:58:07", "throughput": 5207.03, "total_tokens": 5688352} {"current_steps": 5670, "total_steps": 42420, "loss": 0.3455, "lr": 0.0009965567998523293, "epoch": 2.6732673267326734, "percentage": 13.37, "elapsed_time": "0:18:13", "remaining_time": "1:58:06", "throughput": 5207.39, "total_tokens": 5693632} {"current_steps": 5675, "total_steps": 42420, "loss": 0.3923, "lr": 0.000996532656612631, "epoch": 2.6756247053276754, "percentage": 13.38, "elapsed_time": "0:18:14", "remaining_time": "1:58:06", "throughput": 5207.93, "total_tokens": 5699616} {"current_steps": 5680, "total_steps": 42420, "loss": 0.3181, "lr": 0.0009965084293183848, "epoch": 2.677982083922678, "percentage": 13.39, "elapsed_time": "0:18:15", "remaining_time": "1:58:04", "throughput": 5207.73, "total_tokens": 5703680} {"current_steps": 5685, "total_steps": 42420, "loss": 0.2971, "lr": 0.0009964841179736918, "epoch": 2.68033946251768, "percentage": 13.4, "elapsed_time": "0:18:16", "remaining_time": "1:58:02", "throughput": 5207.85, "total_tokens": 5708288} {"current_steps": 5690, "total_steps": 42420, "loss": 0.3074, "lr": 0.000996459722582667, "epoch": 2.6826968411126826, "percentage": 13.41, "elapsed_time": "0:18:17", "remaining_time": "1:58:02", "throughput": 5208.39, "total_tokens": 5714432} {"current_steps": 5695, "total_steps": 42420, "loss": 0.3367, "lr": 0.000996435243149441, "epoch": 2.685054219707685, "percentage": 13.43, "elapsed_time": "0:18:18", "remaining_time": "1:58:01", "throughput": 5208.66, "total_tokens": 5719584} {"current_steps": 5700, "total_steps": 42420, "loss": 0.38, "lr": 0.0009964106796781572, "epoch": 2.6874115983026874, "percentage": 13.44, "elapsed_time": "0:18:18", "remaining_time": "1:57:59", "throughput": 5208.75, "total_tokens": 5724096} {"current_steps": 5705, "total_steps": 42420, "loss": 0.3609, "lr": 0.0009963860321729739, "epoch": 2.68976897689769, "percentage": 13.45, "elapsed_time": "0:18:19", "remaining_time": "1:57:59", "throughput": 5209.14, "total_tokens": 5729952} {"current_steps": 5710, "total_steps": 42420, "loss": 0.358, "lr": 0.0009963613006380636, "epoch": 2.692126355492692, "percentage": 13.46, "elapsed_time": "0:18:20", "remaining_time": "1:57:58", "throughput": 5209.5, "total_tokens": 5735360} {"current_steps": 5715, "total_steps": 42420, "loss": 0.2999, "lr": 0.0009963364850776126, "epoch": 2.6944837340876946, "percentage": 13.47, "elapsed_time": "0:18:22", "remaining_time": "1:57:58", "throughput": 5210.13, "total_tokens": 5742080} {"current_steps": 5720, "total_steps": 42420, "loss": 0.4262, "lr": 0.0009963115854958223, "epoch": 2.696841112682697, "percentage": 13.48, "elapsed_time": "0:18:23", "remaining_time": "1:57:57", "throughput": 5210.54, "total_tokens": 5747808} {"current_steps": 5725, "total_steps": 42420, "loss": 0.3018, "lr": 0.0009962866018969072, "epoch": 2.6991984912776994, "percentage": 13.5, "elapsed_time": "0:18:24", "remaining_time": "1:57:57", "throughput": 5211.01, "total_tokens": 5753696} {"current_steps": 5730, "total_steps": 42420, "loss": 0.2789, "lr": 0.0009962615342850969, "epoch": 2.701555869872702, "percentage": 13.51, "elapsed_time": "0:18:24", "remaining_time": "1:57:55", "throughput": 5211.08, "total_tokens": 5758176} {"current_steps": 5735, "total_steps": 42420, "loss": 0.2672, "lr": 0.000996236382664635, "epoch": 2.7039132484677038, "percentage": 13.52, "elapsed_time": "0:18:25", "remaining_time": "1:57:53", "throughput": 5211.17, "total_tokens": 5762688} {"current_steps": 5740, "total_steps": 42420, "loss": 0.3165, "lr": 0.0009962111470397791, "epoch": 2.706270627062706, "percentage": 13.53, "elapsed_time": "0:18:26", "remaining_time": "1:57:53", "throughput": 5211.67, "total_tokens": 5768960} {"current_steps": 5745, "total_steps": 42420, "loss": 0.4255, "lr": 0.0009961858274148013, "epoch": 2.7086280056577086, "percentage": 13.54, "elapsed_time": "0:18:27", "remaining_time": "1:57:52", "throughput": 5211.76, "total_tokens": 5773760} {"current_steps": 5750, "total_steps": 42420, "loss": 0.3225, "lr": 0.0009961604237939874, "epoch": 2.710985384252711, "percentage": 13.55, "elapsed_time": "0:18:28", "remaining_time": "1:57:50", "throughput": 5211.93, "total_tokens": 5778624} {"current_steps": 5755, "total_steps": 42420, "loss": 0.3599, "lr": 0.0009961349361816384, "epoch": 2.7133427628477134, "percentage": 13.57, "elapsed_time": "0:18:29", "remaining_time": "1:57:49", "throughput": 5212.21, "total_tokens": 5783648} {"current_steps": 5760, "total_steps": 42420, "loss": 0.3672, "lr": 0.0009961093645820683, "epoch": 2.7157001414427158, "percentage": 13.58, "elapsed_time": "0:18:30", "remaining_time": "1:57:47", "throughput": 5212.16, "total_tokens": 5787936} {"current_steps": 5765, "total_steps": 42420, "loss": 0.3617, "lr": 0.0009960837089996065, "epoch": 2.718057520037718, "percentage": 13.59, "elapsed_time": "0:18:31", "remaining_time": "1:57:46", "throughput": 5212.45, "total_tokens": 5793216} {"current_steps": 5770, "total_steps": 42420, "loss": 0.2672, "lr": 0.0009960579694385954, "epoch": 2.7204148986327206, "percentage": 13.6, "elapsed_time": "0:18:32", "remaining_time": "1:57:45", "throughput": 5212.72, "total_tokens": 5798240} {"current_steps": 5775, "total_steps": 42420, "loss": 0.5423, "lr": 0.0009960321459033929, "epoch": 2.7227722772277225, "percentage": 13.61, "elapsed_time": "0:18:33", "remaining_time": "1:57:44", "throughput": 5213.2, "total_tokens": 5803968} {"current_steps": 5780, "total_steps": 42420, "loss": 0.3607, "lr": 0.0009960062383983703, "epoch": 2.725129655822725, "percentage": 13.63, "elapsed_time": "0:18:34", "remaining_time": "1:57:42", "throughput": 5213.32, "total_tokens": 5808576} {"current_steps": 5785, "total_steps": 42420, "loss": 0.3319, "lr": 0.0009959802469279133, "epoch": 2.7274870344177273, "percentage": 13.64, "elapsed_time": "0:18:35", "remaining_time": "1:57:41", "throughput": 5213.58, "total_tokens": 5813760} {"current_steps": 5790, "total_steps": 42420, "loss": 0.3794, "lr": 0.0009959541714964217, "epoch": 2.7298444130127297, "percentage": 13.65, "elapsed_time": "0:18:36", "remaining_time": "1:57:40", "throughput": 5213.63, "total_tokens": 5818496} {"current_steps": 5795, "total_steps": 42420, "loss": 0.3561, "lr": 0.0009959280121083097, "epoch": 2.732201791607732, "percentage": 13.66, "elapsed_time": "0:18:36", "remaining_time": "1:57:38", "throughput": 5213.52, "total_tokens": 5822496} {"current_steps": 5800, "total_steps": 42420, "loss": 0.3468, "lr": 0.0009959017687680054, "epoch": 2.7345591702027345, "percentage": 13.67, "elapsed_time": "0:18:37", "remaining_time": "1:57:36", "throughput": 5213.68, "total_tokens": 5827360} {"current_steps": 5805, "total_steps": 42420, "loss": 0.3566, "lr": 0.0009958754414799518, "epoch": 2.736916548797737, "percentage": 13.68, "elapsed_time": "0:18:38", "remaining_time": "1:57:36", "throughput": 5214.03, "total_tokens": 5832992} {"current_steps": 5810, "total_steps": 42420, "loss": 0.3705, "lr": 0.0009958490302486053, "epoch": 2.7392739273927393, "percentage": 13.7, "elapsed_time": "0:18:39", "remaining_time": "1:57:35", "throughput": 5214.49, "total_tokens": 5838784} {"current_steps": 5815, "total_steps": 42420, "loss": 0.3269, "lr": 0.0009958225350784372, "epoch": 2.7416313059877417, "percentage": 13.71, "elapsed_time": "0:18:40", "remaining_time": "1:57:33", "throughput": 5214.49, "total_tokens": 5843040} {"current_steps": 5820, "total_steps": 42420, "loss": 0.3679, "lr": 0.0009957959559739325, "epoch": 2.743988684582744, "percentage": 13.72, "elapsed_time": "0:18:41", "remaining_time": "1:57:32", "throughput": 5214.95, "total_tokens": 5848736} {"current_steps": 5825, "total_steps": 42420, "loss": 0.3533, "lr": 0.0009957692929395904, "epoch": 2.7463460631777465, "percentage": 13.73, "elapsed_time": "0:18:42", "remaining_time": "1:57:32", "throughput": 5215.32, "total_tokens": 5854464} {"current_steps": 5830, "total_steps": 42420, "loss": 0.3432, "lr": 0.0009957425459799247, "epoch": 2.748703441772749, "percentage": 13.74, "elapsed_time": "0:18:43", "remaining_time": "1:57:31", "throughput": 5215.51, "total_tokens": 5859680} {"current_steps": 5835, "total_steps": 42420, "loss": 0.3659, "lr": 0.0009957157150994632, "epoch": 2.751060820367751, "percentage": 13.76, "elapsed_time": "0:18:44", "remaining_time": "1:57:29", "throughput": 5215.55, "total_tokens": 5864192} {"current_steps": 5840, "total_steps": 42420, "loss": 0.3753, "lr": 0.0009956888003027481, "epoch": 2.7534181989627533, "percentage": 13.77, "elapsed_time": "0:18:45", "remaining_time": "1:57:28", "throughput": 5215.92, "total_tokens": 5869696} {"current_steps": 5845, "total_steps": 42420, "loss": 0.3787, "lr": 0.0009956618015943355, "epoch": 2.7557755775577557, "percentage": 13.78, "elapsed_time": "0:18:46", "remaining_time": "1:57:26", "throughput": 5215.86, "total_tokens": 5873920} {"current_steps": 5850, "total_steps": 42420, "loss": 0.331, "lr": 0.0009956347189787956, "epoch": 2.758132956152758, "percentage": 13.79, "elapsed_time": "0:18:47", "remaining_time": "1:57:25", "throughput": 5216.14, "total_tokens": 5878976} {"current_steps": 5855, "total_steps": 42420, "loss": 0.3846, "lr": 0.0009956075524607132, "epoch": 2.7604903347477605, "percentage": 13.8, "elapsed_time": "0:18:48", "remaining_time": "1:57:25", "throughput": 5216.59, "total_tokens": 5884992} {"current_steps": 5860, "total_steps": 42420, "loss": 0.3337, "lr": 0.000995580302044687, "epoch": 2.762847713342763, "percentage": 13.81, "elapsed_time": "0:18:48", "remaining_time": "1:57:23", "throughput": 5216.67, "total_tokens": 5889504} {"current_steps": 5865, "total_steps": 42420, "loss": 0.3485, "lr": 0.00099555296773533, "epoch": 2.7652050919377653, "percentage": 13.83, "elapsed_time": "0:18:49", "remaining_time": "1:57:22", "throughput": 5216.89, "total_tokens": 5894752} {"current_steps": 5870, "total_steps": 42420, "loss": 0.3103, "lr": 0.0009955255495372698, "epoch": 2.7675624705327677, "percentage": 13.84, "elapsed_time": "0:18:50", "remaining_time": "1:57:21", "throughput": 5217.11, "total_tokens": 5899872} {"current_steps": 5875, "total_steps": 42420, "loss": 0.3429, "lr": 0.0009954980474551477, "epoch": 2.7699198491277697, "percentage": 13.85, "elapsed_time": "0:18:51", "remaining_time": "1:57:19", "throughput": 5216.94, "total_tokens": 5903840} {"current_steps": 5880, "total_steps": 42420, "loss": 0.317, "lr": 0.000995470461493619, "epoch": 2.772277227722772, "percentage": 13.86, "elapsed_time": "0:18:52", "remaining_time": "1:57:18", "throughput": 5217.06, "total_tokens": 5908608} {"current_steps": 5885, "total_steps": 42420, "loss": 0.376, "lr": 0.0009954427916573538, "epoch": 2.7746346063177745, "percentage": 13.87, "elapsed_time": "0:18:53", "remaining_time": "1:57:16", "throughput": 5217.03, "total_tokens": 5912896} {"current_steps": 5890, "total_steps": 42420, "loss": 0.3472, "lr": 0.0009954150379510363, "epoch": 2.776991984912777, "percentage": 13.88, "elapsed_time": "0:18:54", "remaining_time": "1:57:15", "throughput": 5217.33, "total_tokens": 5918336} {"current_steps": 5895, "total_steps": 42420, "loss": 0.3382, "lr": 0.0009953872003793645, "epoch": 2.7793493635077793, "percentage": 13.9, "elapsed_time": "0:18:55", "remaining_time": "1:57:13", "throughput": 5217.36, "total_tokens": 5922944} {"current_steps": 5900, "total_steps": 42420, "loss": 0.4106, "lr": 0.0009953592789470507, "epoch": 2.7817067421027817, "percentage": 13.91, "elapsed_time": "0:18:56", "remaining_time": "1:57:12", "throughput": 5217.36, "total_tokens": 5927296} {"current_steps": 5905, "total_steps": 42420, "loss": 0.3516, "lr": 0.0009953312736588219, "epoch": 2.784064120697784, "percentage": 13.92, "elapsed_time": "0:18:56", "remaining_time": "1:57:10", "throughput": 5217.57, "total_tokens": 5932224} {"current_steps": 5910, "total_steps": 42420, "loss": 0.3067, "lr": 0.0009953031845194187, "epoch": 2.7864214992927865, "percentage": 13.93, "elapsed_time": "0:18:57", "remaining_time": "1:57:09", "throughput": 5217.61, "total_tokens": 5936928} {"current_steps": 5915, "total_steps": 42420, "loss": 0.3186, "lr": 0.000995275011533596, "epoch": 2.788778877887789, "percentage": 13.94, "elapsed_time": "0:18:58", "remaining_time": "1:57:08", "throughput": 5217.86, "total_tokens": 5942112} {"current_steps": 5920, "total_steps": 42420, "loss": 0.4263, "lr": 0.0009952467547061233, "epoch": 2.7911362564827913, "percentage": 13.96, "elapsed_time": "0:18:59", "remaining_time": "1:57:06", "throughput": 5217.84, "total_tokens": 5946400} {"current_steps": 5925, "total_steps": 42420, "loss": 0.3918, "lr": 0.0009952184140417837, "epoch": 2.7934936350777937, "percentage": 13.97, "elapsed_time": "0:19:00", "remaining_time": "1:57:04", "throughput": 5217.79, "total_tokens": 5950432} {"current_steps": 5930, "total_steps": 42420, "loss": 0.471, "lr": 0.000995189989545375, "epoch": 2.795851013672796, "percentage": 13.98, "elapsed_time": "0:19:01", "remaining_time": "1:57:03", "throughput": 5217.92, "total_tokens": 5955296} {"current_steps": 5935, "total_steps": 42420, "loss": 0.5553, "lr": 0.0009951614812217088, "epoch": 2.798208392267798, "percentage": 13.99, "elapsed_time": "0:19:02", "remaining_time": "1:57:01", "throughput": 5218.17, "total_tokens": 5960416} {"current_steps": 5940, "total_steps": 42420, "loss": 0.3345, "lr": 0.0009951328890756112, "epoch": 2.8005657708628005, "percentage": 14.0, "elapsed_time": "0:19:03", "remaining_time": "1:57:00", "throughput": 5218.36, "total_tokens": 5965280} {"current_steps": 5945, "total_steps": 42420, "loss": 0.3589, "lr": 0.0009951042131119226, "epoch": 2.802923149457803, "percentage": 14.01, "elapsed_time": "0:19:04", "remaining_time": "1:56:59", "throughput": 5218.54, "total_tokens": 5970304} {"current_steps": 5950, "total_steps": 42420, "loss": 0.4924, "lr": 0.000995075453335497, "epoch": 2.8052805280528053, "percentage": 14.03, "elapsed_time": "0:19:04", "remaining_time": "1:56:57", "throughput": 5218.69, "total_tokens": 5975104} {"current_steps": 5955, "total_steps": 42420, "loss": 0.3714, "lr": 0.000995046609751203, "epoch": 2.8076379066478077, "percentage": 14.04, "elapsed_time": "0:19:05", "remaining_time": "1:56:56", "throughput": 5218.71, "total_tokens": 5979808} {"current_steps": 5960, "total_steps": 42420, "loss": 0.3612, "lr": 0.0009950176823639231, "epoch": 2.80999528524281, "percentage": 14.05, "elapsed_time": "0:19:06", "remaining_time": "1:56:55", "throughput": 5218.93, "total_tokens": 5985024} {"current_steps": 5965, "total_steps": 42420, "loss": 0.3757, "lr": 0.000994988671178555, "epoch": 2.8123526638378125, "percentage": 14.06, "elapsed_time": "0:19:07", "remaining_time": "1:56:54", "throughput": 5219.05, "total_tokens": 5989920} {"current_steps": 5970, "total_steps": 42420, "loss": 0.3424, "lr": 0.000994959576200009, "epoch": 2.814710042432815, "percentage": 14.07, "elapsed_time": "0:19:08", "remaining_time": "1:56:52", "throughput": 5219.09, "total_tokens": 5994688} {"current_steps": 5975, "total_steps": 42420, "loss": 0.3641, "lr": 0.0009949303974332107, "epoch": 2.817067421027817, "percentage": 14.09, "elapsed_time": "0:19:09", "remaining_time": "1:56:50", "throughput": 5218.86, "total_tokens": 5998432} {"current_steps": 5980, "total_steps": 42420, "loss": 0.287, "lr": 0.0009949011348830996, "epoch": 2.8194247996228192, "percentage": 14.1, "elapsed_time": "0:19:10", "remaining_time": "1:56:49", "throughput": 5219.02, "total_tokens": 6003456} {"current_steps": 5985, "total_steps": 42420, "loss": 0.3819, "lr": 0.0009948717885546294, "epoch": 2.8217821782178216, "percentage": 14.11, "elapsed_time": "0:19:11", "remaining_time": "1:56:47", "throughput": 5218.91, "total_tokens": 6007648} {"current_steps": 5990, "total_steps": 42420, "loss": 0.3977, "lr": 0.0009948423584527678, "epoch": 2.824139556812824, "percentage": 14.12, "elapsed_time": "0:19:12", "remaining_time": "1:56:47", "throughput": 5219.2, "total_tokens": 6013184} {"current_steps": 5995, "total_steps": 42420, "loss": 0.3688, "lr": 0.000994812844582497, "epoch": 2.8264969354078264, "percentage": 14.13, "elapsed_time": "0:19:13", "remaining_time": "1:56:47", "throughput": 5219.87, "total_tokens": 6020256} {"current_steps": 6000, "total_steps": 42420, "loss": 0.3479, "lr": 0.0009947832469488127, "epoch": 2.828854314002829, "percentage": 14.14, "elapsed_time": "0:19:14", "remaining_time": "1:56:46", "throughput": 5220.05, "total_tokens": 6025088} {"current_steps": 6005, "total_steps": 42420, "loss": 0.3504, "lr": 0.0009947535655567257, "epoch": 2.8312116925978312, "percentage": 14.16, "elapsed_time": "0:19:14", "remaining_time": "1:56:43", "throughput": 5219.84, "total_tokens": 6028864} {"current_steps": 6010, "total_steps": 42420, "loss": 0.3788, "lr": 0.0009947238004112606, "epoch": 2.8335690711928336, "percentage": 14.17, "elapsed_time": "0:19:15", "remaining_time": "1:56:42", "throughput": 5220.09, "total_tokens": 6034080} {"current_steps": 6015, "total_steps": 42420, "loss": 0.3395, "lr": 0.0009946939515174562, "epoch": 2.835926449787836, "percentage": 14.18, "elapsed_time": "0:19:16", "remaining_time": "1:56:41", "throughput": 5220.15, "total_tokens": 6038656} {"current_steps": 6020, "total_steps": 42420, "loss": 0.3384, "lr": 0.000994664018880365, "epoch": 2.8382838283828384, "percentage": 14.19, "elapsed_time": "0:19:17", "remaining_time": "1:56:40", "throughput": 5220.22, "total_tokens": 6043552} {"current_steps": 6025, "total_steps": 42420, "loss": 0.3094, "lr": 0.0009946340025050544, "epoch": 2.840641206977841, "percentage": 14.2, "elapsed_time": "0:19:18", "remaining_time": "1:56:38", "throughput": 5220.1, "total_tokens": 6047712} {"current_steps": 6030, "total_steps": 42420, "loss": 0.2872, "lr": 0.0009946039023966055, "epoch": 2.8429985855728432, "percentage": 14.21, "elapsed_time": "0:19:19", "remaining_time": "1:56:37", "throughput": 5220.33, "total_tokens": 6052800} {"current_steps": 6035, "total_steps": 42420, "loss": 0.3761, "lr": 0.000994573718560114, "epoch": 2.845355964167845, "percentage": 14.23, "elapsed_time": "0:19:20", "remaining_time": "1:56:37", "throughput": 5220.92, "total_tokens": 6059200} {"current_steps": 6040, "total_steps": 42420, "loss": 0.4715, "lr": 0.0009945434510006891, "epoch": 2.8477133427628476, "percentage": 14.24, "elapsed_time": "0:19:21", "remaining_time": "1:56:35", "throughput": 5221.03, "total_tokens": 6064128} {"current_steps": 6045, "total_steps": 42420, "loss": 0.3813, "lr": 0.000994513099723455, "epoch": 2.85007072135785, "percentage": 14.25, "elapsed_time": "0:19:22", "remaining_time": "1:56:34", "throughput": 5221.14, "total_tokens": 6068992} {"current_steps": 6050, "total_steps": 42420, "loss": 0.5, "lr": 0.0009944826647335493, "epoch": 2.8524280999528524, "percentage": 14.26, "elapsed_time": "0:19:23", "remaining_time": "1:56:34", "throughput": 5221.59, "total_tokens": 6075232} {"current_steps": 6055, "total_steps": 42420, "loss": 0.384, "lr": 0.0009944521460361244, "epoch": 2.854785478547855, "percentage": 14.27, "elapsed_time": "0:19:24", "remaining_time": "1:56:33", "throughput": 5221.91, "total_tokens": 6080576} {"current_steps": 6060, "total_steps": 42420, "loss": 0.3779, "lr": 0.0009944215436363465, "epoch": 2.857142857142857, "percentage": 14.29, "elapsed_time": "0:19:25", "remaining_time": "1:56:31", "throughput": 5222.06, "total_tokens": 6085312} {"current_steps": 6065, "total_steps": 42420, "loss": 0.2713, "lr": 0.0009943908575393962, "epoch": 2.8595002357378596, "percentage": 14.3, "elapsed_time": "0:19:26", "remaining_time": "1:56:31", "throughput": 5222.38, "total_tokens": 6090944} {"current_steps": 6070, "total_steps": 42420, "loss": 0.3728, "lr": 0.0009943600877504678, "epoch": 2.861857614332862, "percentage": 14.31, "elapsed_time": "0:19:27", "remaining_time": "1:56:29", "throughput": 5222.26, "total_tokens": 6094912} {"current_steps": 6075, "total_steps": 42420, "loss": 0.3288, "lr": 0.0009943292342747703, "epoch": 2.864214992927864, "percentage": 14.32, "elapsed_time": "0:19:28", "remaining_time": "1:56:28", "throughput": 5222.51, "total_tokens": 6100416} {"current_steps": 6080, "total_steps": 42420, "loss": 0.429, "lr": 0.0009942982971175268, "epoch": 2.8665723715228664, "percentage": 14.33, "elapsed_time": "0:19:29", "remaining_time": "1:56:27", "throughput": 5222.73, "total_tokens": 6105536} {"current_steps": 6085, "total_steps": 42420, "loss": 0.4039, "lr": 0.0009942672762839743, "epoch": 2.8689297501178688, "percentage": 14.34, "elapsed_time": "0:19:29", "remaining_time": "1:56:26", "throughput": 5223.04, "total_tokens": 6110816} {"current_steps": 6090, "total_steps": 42420, "loss": 0.3058, "lr": 0.000994236171779364, "epoch": 2.871287128712871, "percentage": 14.36, "elapsed_time": "0:19:30", "remaining_time": "1:56:24", "throughput": 5223.05, "total_tokens": 6115136} {"current_steps": 6095, "total_steps": 42420, "loss": 0.42, "lr": 0.0009942049836089615, "epoch": 2.8736445073078736, "percentage": 14.37, "elapsed_time": "0:19:31", "remaining_time": "1:56:23", "throughput": 5223.42, "total_tokens": 6120864} {"current_steps": 6100, "total_steps": 42420, "loss": 0.333, "lr": 0.0009941737117780463, "epoch": 2.876001885902876, "percentage": 14.38, "elapsed_time": "0:19:32", "remaining_time": "1:56:22", "throughput": 5223.65, "total_tokens": 6126208} {"current_steps": 6105, "total_steps": 42420, "loss": 0.2929, "lr": 0.0009941423562919126, "epoch": 2.8783592644978784, "percentage": 14.39, "elapsed_time": "0:19:33", "remaining_time": "1:56:21", "throughput": 5223.67, "total_tokens": 6130656} {"current_steps": 6110, "total_steps": 42420, "loss": 0.269, "lr": 0.0009941109171558678, "epoch": 2.880716643092881, "percentage": 14.4, "elapsed_time": "0:19:34", "remaining_time": "1:56:19", "throughput": 5223.64, "total_tokens": 6135104} {"current_steps": 6115, "total_steps": 42420, "loss": 0.2919, "lr": 0.0009940793943752342, "epoch": 2.883074021687883, "percentage": 14.42, "elapsed_time": "0:19:35", "remaining_time": "1:56:18", "throughput": 5223.9, "total_tokens": 6140352} {"current_steps": 6120, "total_steps": 42420, "loss": 0.3572, "lr": 0.0009940477879553482, "epoch": 2.8854314002828856, "percentage": 14.43, "elapsed_time": "0:19:36", "remaining_time": "1:56:17", "throughput": 5224.02, "total_tokens": 6145376} {"current_steps": 6125, "total_steps": 42420, "loss": 0.3376, "lr": 0.0009940160979015602, "epoch": 2.887788778877888, "percentage": 14.44, "elapsed_time": "0:19:37", "remaining_time": "1:56:15", "throughput": 5223.89, "total_tokens": 6149376} {"current_steps": 6130, "total_steps": 42420, "loss": 0.3585, "lr": 0.0009939843242192346, "epoch": 2.8901461574728904, "percentage": 14.45, "elapsed_time": "0:19:38", "remaining_time": "1:56:14", "throughput": 5224.09, "total_tokens": 6154688} {"current_steps": 6135, "total_steps": 42420, "loss": 0.3546, "lr": 0.0009939524669137505, "epoch": 2.8925035360678923, "percentage": 14.46, "elapsed_time": "0:19:39", "remaining_time": "1:56:13", "throughput": 5224.06, "total_tokens": 6159232} {"current_steps": 6140, "total_steps": 42420, "loss": 0.3992, "lr": 0.0009939205259905004, "epoch": 2.8948609146628947, "percentage": 14.47, "elapsed_time": "0:19:39", "remaining_time": "1:56:11", "throughput": 5223.86, "total_tokens": 6162976} {"current_steps": 6145, "total_steps": 42420, "loss": 0.337, "lr": 0.0009938885014548918, "epoch": 2.897218293257897, "percentage": 14.49, "elapsed_time": "0:19:40", "remaining_time": "1:56:09", "throughput": 5223.84, "total_tokens": 6167328} {"current_steps": 6150, "total_steps": 42420, "loss": 0.4754, "lr": 0.0009938563933123454, "epoch": 2.8995756718528995, "percentage": 14.5, "elapsed_time": "0:19:41", "remaining_time": "1:56:08", "throughput": 5224.14, "total_tokens": 6172416} {"current_steps": 6155, "total_steps": 42420, "loss": 0.3217, "lr": 0.0009938242015682967, "epoch": 2.901933050447902, "percentage": 14.51, "elapsed_time": "0:19:42", "remaining_time": "1:56:07", "throughput": 5224.45, "total_tokens": 6177984} {"current_steps": 6160, "total_steps": 42420, "loss": 0.3322, "lr": 0.0009937919262281956, "epoch": 2.9042904290429044, "percentage": 14.52, "elapsed_time": "0:19:43", "remaining_time": "1:56:06", "throughput": 5224.64, "total_tokens": 6182944} {"current_steps": 6165, "total_steps": 42420, "loss": 0.3577, "lr": 0.0009937595672975055, "epoch": 2.9066478076379068, "percentage": 14.53, "elapsed_time": "0:19:44", "remaining_time": "1:56:04", "throughput": 5224.66, "total_tokens": 6187392} {"current_steps": 6170, "total_steps": 42420, "loss": 0.3359, "lr": 0.000993727124781704, "epoch": 2.909005186232909, "percentage": 14.55, "elapsed_time": "0:19:45", "remaining_time": "1:56:03", "throughput": 5224.83, "total_tokens": 6192320} {"current_steps": 6175, "total_steps": 42420, "loss": 0.4111, "lr": 0.0009936945986862835, "epoch": 2.911362564827911, "percentage": 14.56, "elapsed_time": "0:19:46", "remaining_time": "1:56:01", "throughput": 5224.83, "total_tokens": 6196736} {"current_steps": 6180, "total_steps": 42420, "loss": 0.3424, "lr": 0.0009936619890167499, "epoch": 2.9137199434229135, "percentage": 14.57, "elapsed_time": "0:19:46", "remaining_time": "1:55:59", "throughput": 5224.79, "total_tokens": 6201056} {"current_steps": 6185, "total_steps": 42420, "loss": 0.3492, "lr": 0.0009936292957786232, "epoch": 2.916077322017916, "percentage": 14.58, "elapsed_time": "0:19:48", "remaining_time": "1:56:00", "throughput": 5225.58, "total_tokens": 6208096} {"current_steps": 6190, "total_steps": 42420, "loss": 0.3904, "lr": 0.0009935965189774385, "epoch": 2.9184347006129183, "percentage": 14.59, "elapsed_time": "0:19:48", "remaining_time": "1:55:58", "throughput": 5225.53, "total_tokens": 6212832} {"current_steps": 6195, "total_steps": 42420, "loss": 0.3584, "lr": 0.0009935636586187436, "epoch": 2.9207920792079207, "percentage": 14.6, "elapsed_time": "0:19:49", "remaining_time": "1:55:57", "throughput": 5225.64, "total_tokens": 6217632} {"current_steps": 6200, "total_steps": 42420, "loss": 0.3675, "lr": 0.0009935307147081016, "epoch": 2.923149457802923, "percentage": 14.62, "elapsed_time": "0:19:50", "remaining_time": "1:55:56", "throughput": 5225.78, "total_tokens": 6222656} {"current_steps": 6205, "total_steps": 42420, "loss": 0.2328, "lr": 0.0009934976872510892, "epoch": 2.9255068363979255, "percentage": 14.63, "elapsed_time": "0:19:51", "remaining_time": "1:55:55", "throughput": 5226.03, "total_tokens": 6228160} {"current_steps": 6210, "total_steps": 42420, "loss": 0.4057, "lr": 0.0009934645762532976, "epoch": 2.927864214992928, "percentage": 14.64, "elapsed_time": "0:19:52", "remaining_time": "1:55:54", "throughput": 5226.38, "total_tokens": 6233504} {"current_steps": 6215, "total_steps": 42420, "loss": 0.3177, "lr": 0.0009934313817203317, "epoch": 2.9302215935879303, "percentage": 14.65, "elapsed_time": "0:19:53", "remaining_time": "1:55:53", "throughput": 5226.63, "total_tokens": 6238848} {"current_steps": 6220, "total_steps": 42420, "loss": 0.3239, "lr": 0.000993398103657811, "epoch": 2.9325789721829327, "percentage": 14.66, "elapsed_time": "0:19:54", "remaining_time": "1:55:52", "throughput": 5226.93, "total_tokens": 6244000} {"current_steps": 6225, "total_steps": 42420, "loss": 0.2842, "lr": 0.0009933647420713688, "epoch": 2.934936350777935, "percentage": 14.67, "elapsed_time": "0:19:55", "remaining_time": "1:55:50", "throughput": 5226.91, "total_tokens": 6248384} {"current_steps": 6230, "total_steps": 42420, "loss": 0.3455, "lr": 0.0009933312969666523, "epoch": 2.9372937293729375, "percentage": 14.69, "elapsed_time": "0:19:56", "remaining_time": "1:55:49", "throughput": 5227.06, "total_tokens": 6253440} {"current_steps": 6235, "total_steps": 42420, "loss": 0.405, "lr": 0.0009932977683493238, "epoch": 2.9396511079679395, "percentage": 14.7, "elapsed_time": "0:19:57", "remaining_time": "1:55:49", "throughput": 5227.36, "total_tokens": 6259232} {"current_steps": 6240, "total_steps": 42420, "loss": 0.3179, "lr": 0.0009932641562250587, "epoch": 2.942008486562942, "percentage": 14.71, "elapsed_time": "0:19:58", "remaining_time": "1:55:48", "throughput": 5227.71, "total_tokens": 6265216} {"current_steps": 6245, "total_steps": 42420, "loss": 0.3392, "lr": 0.0009932304605995473, "epoch": 2.9443658651579443, "percentage": 14.72, "elapsed_time": "0:19:59", "remaining_time": "1:55:47", "throughput": 5227.69, "total_tokens": 6269824} {"current_steps": 6250, "total_steps": 42420, "loss": 0.359, "lr": 0.0009931966814784932, "epoch": 2.9467232437529467, "percentage": 14.73, "elapsed_time": "0:20:00", "remaining_time": "1:55:47", "throughput": 5228.37, "total_tokens": 6276736} {"current_steps": 6255, "total_steps": 42420, "loss": 0.2993, "lr": 0.0009931628188676151, "epoch": 2.949080622347949, "percentage": 14.75, "elapsed_time": "0:20:01", "remaining_time": "1:55:47", "throughput": 5229.03, "total_tokens": 6283488} {"current_steps": 6260, "total_steps": 42420, "loss": 0.3918, "lr": 0.0009931288727726452, "epoch": 2.9514380009429515, "percentage": 14.76, "elapsed_time": "0:20:02", "remaining_time": "1:55:45", "throughput": 5228.91, "total_tokens": 6287360} {"current_steps": 6265, "total_steps": 42420, "loss": 0.3281, "lr": 0.00099309484319933, "epoch": 2.953795379537954, "percentage": 14.77, "elapsed_time": "0:20:03", "remaining_time": "1:55:44", "throughput": 5228.95, "total_tokens": 6292096} {"current_steps": 6270, "total_steps": 42420, "loss": 0.3015, "lr": 0.00099306073015343, "epoch": 2.9561527581329563, "percentage": 14.78, "elapsed_time": "0:20:04", "remaining_time": "1:55:43", "throughput": 5229.12, "total_tokens": 6297088} {"current_steps": 6275, "total_steps": 42420, "loss": 0.4382, "lr": 0.0009930265336407203, "epoch": 2.9585101367279583, "percentage": 14.79, "elapsed_time": "0:20:05", "remaining_time": "1:55:41", "throughput": 5229.15, "total_tokens": 6301760} {"current_steps": 6280, "total_steps": 42420, "loss": 0.3592, "lr": 0.0009929922536669894, "epoch": 2.9608675153229607, "percentage": 14.8, "elapsed_time": "0:20:06", "remaining_time": "1:55:40", "throughput": 5229.42, "total_tokens": 6307168} {"current_steps": 6285, "total_steps": 42420, "loss": 0.3673, "lr": 0.0009929578902380404, "epoch": 2.963224893917963, "percentage": 14.82, "elapsed_time": "0:20:06", "remaining_time": "1:55:39", "throughput": 5229.32, "total_tokens": 6311424} {"current_steps": 6290, "total_steps": 42420, "loss": 0.3992, "lr": 0.0009929234433596907, "epoch": 2.9655822725129655, "percentage": 14.83, "elapsed_time": "0:20:07", "remaining_time": "1:55:37", "throughput": 5229.2, "total_tokens": 6315520} {"current_steps": 6295, "total_steps": 42420, "loss": 0.3478, "lr": 0.0009928889130377714, "epoch": 2.967939651107968, "percentage": 14.84, "elapsed_time": "0:20:08", "remaining_time": "1:55:36", "throughput": 5229.67, "total_tokens": 6321568} {"current_steps": 6300, "total_steps": 42420, "loss": 0.4155, "lr": 0.0009928542992781277, "epoch": 2.9702970297029703, "percentage": 14.85, "elapsed_time": "0:20:09", "remaining_time": "1:55:36", "throughput": 5230.29, "total_tokens": 6327936} {"current_steps": 6305, "total_steps": 42420, "loss": 0.3323, "lr": 0.0009928196020866195, "epoch": 2.9726544082979727, "percentage": 14.86, "elapsed_time": "0:20:10", "remaining_time": "1:55:35", "throughput": 5230.37, "total_tokens": 6332736} {"current_steps": 6310, "total_steps": 42420, "loss": 0.3686, "lr": 0.0009927848214691204, "epoch": 2.975011786892975, "percentage": 14.88, "elapsed_time": "0:20:11", "remaining_time": "1:55:34", "throughput": 5230.53, "total_tokens": 6338112} {"current_steps": 6315, "total_steps": 42420, "loss": 0.3714, "lr": 0.0009927499574315178, "epoch": 2.9773691654879775, "percentage": 14.89, "elapsed_time": "0:20:12", "remaining_time": "1:55:32", "throughput": 5230.66, "total_tokens": 6342784} {"current_steps": 6320, "total_steps": 42420, "loss": 0.3474, "lr": 0.000992715009979714, "epoch": 2.97972654408298, "percentage": 14.9, "elapsed_time": "0:20:13", "remaining_time": "1:55:32", "throughput": 5230.99, "total_tokens": 6348448} {"current_steps": 6325, "total_steps": 42420, "loss": 0.4089, "lr": 0.0009926799791196246, "epoch": 2.9820839226779823, "percentage": 14.91, "elapsed_time": "0:20:14", "remaining_time": "1:55:31", "throughput": 5231.18, "total_tokens": 6353888} {"current_steps": 6330, "total_steps": 42420, "loss": 0.3579, "lr": 0.0009926448648571802, "epoch": 2.9844413012729847, "percentage": 14.92, "elapsed_time": "0:20:15", "remaining_time": "1:55:30", "throughput": 5231.5, "total_tokens": 6359584} {"current_steps": 6335, "total_steps": 42420, "loss": 0.3265, "lr": 0.0009926096671983248, "epoch": 2.9867986798679866, "percentage": 14.93, "elapsed_time": "0:20:16", "remaining_time": "1:55:30", "throughput": 5231.99, "total_tokens": 6365984} {"current_steps": 6340, "total_steps": 42420, "loss": 0.2858, "lr": 0.0009925743861490166, "epoch": 2.989156058462989, "percentage": 14.95, "elapsed_time": "0:20:17", "remaining_time": "1:55:29", "throughput": 5232.18, "total_tokens": 6371296} {"current_steps": 6345, "total_steps": 42420, "loss": 0.372, "lr": 0.0009925390217152283, "epoch": 2.9915134370579914, "percentage": 14.96, "elapsed_time": "0:20:18", "remaining_time": "1:55:28", "throughput": 5232.1, "total_tokens": 6375424} {"current_steps": 6350, "total_steps": 42420, "loss": 0.344, "lr": 0.0009925035739029465, "epoch": 2.993870815652994, "percentage": 14.97, "elapsed_time": "0:20:19", "remaining_time": "1:55:26", "throughput": 5232.29, "total_tokens": 6380384} {"current_steps": 6355, "total_steps": 42420, "loss": 0.3399, "lr": 0.000992468042718172, "epoch": 2.9962281942479962, "percentage": 14.98, "elapsed_time": "0:20:20", "remaining_time": "1:55:25", "throughput": 5232.44, "total_tokens": 6385632} {"current_steps": 6360, "total_steps": 42420, "loss": 0.3714, "lr": 0.0009924324281669194, "epoch": 2.9985855728429986, "percentage": 14.99, "elapsed_time": "0:20:21", "remaining_time": "1:55:24", "throughput": 5232.69, "total_tokens": 6390880} {"current_steps": 6363, "total_steps": 42420, "eval_loss": 0.36804601550102234, "epoch": 3.0, "percentage": 15.0, "elapsed_time": "0:20:40", "remaining_time": "1:57:11", "throughput": 5152.68, "total_tokens": 6393376} {"current_steps": 6365, "total_steps": 42420, "loss": 0.3897, "lr": 0.0009923967302552176, "epoch": 3.000942951438001, "percentage": 15.0, "elapsed_time": "0:20:43", "remaining_time": "1:57:24", "throughput": 5142.75, "total_tokens": 6395424} {"current_steps": 6370, "total_steps": 42420, "loss": 0.3378, "lr": 0.0009923609489891098, "epoch": 3.0033003300330035, "percentage": 15.02, "elapsed_time": "0:20:44", "remaining_time": "1:57:23", "throughput": 5143.04, "total_tokens": 6400896} {"current_steps": 6375, "total_steps": 42420, "loss": 0.4281, "lr": 0.0009923250843746534, "epoch": 3.005657708628006, "percentage": 15.03, "elapsed_time": "0:20:45", "remaining_time": "1:57:21", "throughput": 5143.1, "total_tokens": 6405216} {"current_steps": 6380, "total_steps": 42420, "loss": 0.3667, "lr": 0.0009922891364179194, "epoch": 3.008015087223008, "percentage": 15.04, "elapsed_time": "0:20:46", "remaining_time": "1:57:20", "throughput": 5143.47, "total_tokens": 6410624} {"current_steps": 6385, "total_steps": 42420, "loss": 0.358, "lr": 0.000992253105124993, "epoch": 3.01037246581801, "percentage": 15.05, "elapsed_time": "0:20:47", "remaining_time": "1:57:20", "throughput": 5144.22, "total_tokens": 6417600} {"current_steps": 6390, "total_steps": 42420, "loss": 0.3524, "lr": 0.000992216990501974, "epoch": 3.0127298444130126, "percentage": 15.06, "elapsed_time": "0:20:48", "remaining_time": "1:57:19", "throughput": 5144.3, "total_tokens": 6422304} {"current_steps": 6395, "total_steps": 42420, "loss": 0.3253, "lr": 0.000992180792554976, "epoch": 3.015087223008015, "percentage": 15.08, "elapsed_time": "0:20:49", "remaining_time": "1:57:17", "throughput": 5144.3, "total_tokens": 6426752} {"current_steps": 6400, "total_steps": 42420, "loss": 0.4586, "lr": 0.0009921445112901263, "epoch": 3.0174446016030174, "percentage": 15.09, "elapsed_time": "0:20:50", "remaining_time": "1:57:16", "throughput": 5144.61, "total_tokens": 6431936} {"current_steps": 6405, "total_steps": 42420, "loss": 0.3153, "lr": 0.000992108146713567, "epoch": 3.01980198019802, "percentage": 15.1, "elapsed_time": "0:20:51", "remaining_time": "1:57:15", "throughput": 5144.88, "total_tokens": 6437696} {"current_steps": 6410, "total_steps": 42420, "loss": 0.3023, "lr": 0.000992071698831454, "epoch": 3.022159358793022, "percentage": 15.11, "elapsed_time": "0:20:52", "remaining_time": "1:57:14", "throughput": 5144.95, "total_tokens": 6442304} {"current_steps": 6415, "total_steps": 42420, "loss": 0.2947, "lr": 0.0009920351676499576, "epoch": 3.0245167373880246, "percentage": 15.12, "elapsed_time": "0:20:52", "remaining_time": "1:57:12", "throughput": 5144.85, "total_tokens": 6446304} {"current_steps": 6420, "total_steps": 42420, "loss": 0.3388, "lr": 0.0009919985531752613, "epoch": 3.026874115983027, "percentage": 15.13, "elapsed_time": "0:20:53", "remaining_time": "1:57:10", "throughput": 5144.73, "total_tokens": 6450400} {"current_steps": 6425, "total_steps": 42420, "loss": 0.3849, "lr": 0.0009919618554135637, "epoch": 3.0292314945780294, "percentage": 15.15, "elapsed_time": "0:20:54", "remaining_time": "1:57:09", "throughput": 5144.89, "total_tokens": 6455328} {"current_steps": 6430, "total_steps": 42420, "loss": 0.3272, "lr": 0.0009919250743710768, "epoch": 3.0315888731730314, "percentage": 15.16, "elapsed_time": "0:20:55", "remaining_time": "1:57:07", "throughput": 5144.88, "total_tokens": 6459552} {"current_steps": 6435, "total_steps": 42420, "loss": 0.3994, "lr": 0.0009918882100540276, "epoch": 3.033946251768034, "percentage": 15.17, "elapsed_time": "0:20:56", "remaining_time": "1:57:05", "throughput": 5144.78, "total_tokens": 6463392} {"current_steps": 6440, "total_steps": 42420, "loss": 0.3528, "lr": 0.000991851262468656, "epoch": 3.036303630363036, "percentage": 15.18, "elapsed_time": "0:20:57", "remaining_time": "1:57:04", "throughput": 5145.17, "total_tokens": 6469088} {"current_steps": 6445, "total_steps": 42420, "loss": 0.3379, "lr": 0.000991814231621217, "epoch": 3.0386610089580386, "percentage": 15.19, "elapsed_time": "0:20:58", "remaining_time": "1:57:03", "throughput": 5145.43, "total_tokens": 6474560} {"current_steps": 6450, "total_steps": 42420, "loss": 0.3496, "lr": 0.0009917771175179787, "epoch": 3.041018387553041, "percentage": 15.21, "elapsed_time": "0:20:59", "remaining_time": "1:57:02", "throughput": 5145.44, "total_tokens": 6479104} {"current_steps": 6455, "total_steps": 42420, "loss": 0.3569, "lr": 0.0009917399201652246, "epoch": 3.0433757661480434, "percentage": 15.22, "elapsed_time": "0:21:00", "remaining_time": "1:57:00", "throughput": 5145.58, "total_tokens": 6484096} {"current_steps": 6460, "total_steps": 42420, "loss": 0.3544, "lr": 0.0009917026395692514, "epoch": 3.045733144743046, "percentage": 15.23, "elapsed_time": "0:21:01", "remaining_time": "1:56:59", "throughput": 5145.9, "total_tokens": 6489312} {"current_steps": 6465, "total_steps": 42420, "loss": 0.3264, "lr": 0.0009916652757363698, "epoch": 3.048090523338048, "percentage": 15.24, "elapsed_time": "0:21:01", "remaining_time": "1:56:58", "throughput": 5146.07, "total_tokens": 6494208} {"current_steps": 6470, "total_steps": 42420, "loss": 0.3546, "lr": 0.000991627828672905, "epoch": 3.0504479019330506, "percentage": 15.25, "elapsed_time": "0:21:02", "remaining_time": "1:56:56", "throughput": 5146.24, "total_tokens": 6498880} {"current_steps": 6475, "total_steps": 42420, "loss": 0.3349, "lr": 0.0009915902983851962, "epoch": 3.052805280528053, "percentage": 15.26, "elapsed_time": "0:21:03", "remaining_time": "1:56:55", "throughput": 5146.33, "total_tokens": 6503488} {"current_steps": 6480, "total_steps": 42420, "loss": 0.3335, "lr": 0.0009915526848795965, "epoch": 3.055162659123055, "percentage": 15.28, "elapsed_time": "0:21:04", "remaining_time": "1:56:54", "throughput": 5146.76, "total_tokens": 6509344} {"current_steps": 6485, "total_steps": 42420, "loss": 0.3711, "lr": 0.0009915149881624735, "epoch": 3.0575200377180574, "percentage": 15.29, "elapsed_time": "0:21:05", "remaining_time": "1:56:53", "throughput": 5147.0, "total_tokens": 6514784} {"current_steps": 6490, "total_steps": 42420, "loss": 0.3451, "lr": 0.0009914772082402084, "epoch": 3.0598774163130598, "percentage": 15.3, "elapsed_time": "0:21:06", "remaining_time": "1:56:52", "throughput": 5147.13, "total_tokens": 6519712} {"current_steps": 6495, "total_steps": 42420, "loss": 0.3726, "lr": 0.0009914393451191969, "epoch": 3.062234794908062, "percentage": 15.31, "elapsed_time": "0:21:07", "remaining_time": "1:56:51", "throughput": 5147.69, "total_tokens": 6525824} {"current_steps": 6500, "total_steps": 42420, "loss": 0.3535, "lr": 0.0009914013988058482, "epoch": 3.0645921735030646, "percentage": 15.32, "elapsed_time": "0:21:08", "remaining_time": "1:56:50", "throughput": 5147.98, "total_tokens": 6531136} {"current_steps": 6505, "total_steps": 42420, "loss": 0.3374, "lr": 0.0009913633693065863, "epoch": 3.066949552098067, "percentage": 15.33, "elapsed_time": "0:21:09", "remaining_time": "1:56:49", "throughput": 5148.07, "total_tokens": 6535872} {"current_steps": 6510, "total_steps": 42420, "loss": 0.3586, "lr": 0.000991325256627849, "epoch": 3.0693069306930694, "percentage": 15.35, "elapsed_time": "0:21:10", "remaining_time": "1:56:49", "throughput": 5148.45, "total_tokens": 6541824} {"current_steps": 6515, "total_steps": 42420, "loss": 0.3552, "lr": 0.0009912870607760877, "epoch": 3.0716643092880718, "percentage": 15.36, "elapsed_time": "0:21:11", "remaining_time": "1:56:47", "throughput": 5148.65, "total_tokens": 6546880} {"current_steps": 6520, "total_steps": 42420, "loss": 0.3432, "lr": 0.0009912487817577687, "epoch": 3.074021687883074, "percentage": 15.37, "elapsed_time": "0:21:12", "remaining_time": "1:56:46", "throughput": 5148.76, "total_tokens": 6551584} {"current_steps": 6525, "total_steps": 42420, "loss": 0.3629, "lr": 0.0009912104195793719, "epoch": 3.0763790664780766, "percentage": 15.38, "elapsed_time": "0:21:13", "remaining_time": "1:56:45", "throughput": 5149.0, "total_tokens": 6556672} {"current_steps": 6530, "total_steps": 42420, "loss": 0.3308, "lr": 0.0009911719742473914, "epoch": 3.0787364450730785, "percentage": 15.39, "elapsed_time": "0:21:14", "remaining_time": "1:56:44", "throughput": 5149.38, "total_tokens": 6562176} {"current_steps": 6535, "total_steps": 42420, "loss": 0.3679, "lr": 0.0009911334457683352, "epoch": 3.081093823668081, "percentage": 15.41, "elapsed_time": "0:21:15", "remaining_time": "1:56:42", "throughput": 5149.41, "total_tokens": 6566688} {"current_steps": 6540, "total_steps": 42420, "loss": 0.3587, "lr": 0.0009910948341487257, "epoch": 3.0834512022630833, "percentage": 15.42, "elapsed_time": "0:21:16", "remaining_time": "1:56:41", "throughput": 5149.82, "total_tokens": 6572416} {"current_steps": 6545, "total_steps": 42420, "loss": 0.3372, "lr": 0.000991056139395099, "epoch": 3.0858085808580857, "percentage": 15.43, "elapsed_time": "0:21:17", "remaining_time": "1:56:40", "throughput": 5149.77, "total_tokens": 6576832} {"current_steps": 6550, "total_steps": 42420, "loss": 0.2837, "lr": 0.0009910173615140056, "epoch": 3.088165959453088, "percentage": 15.44, "elapsed_time": "0:21:18", "remaining_time": "1:56:39", "throughput": 5150.27, "total_tokens": 6583008} {"current_steps": 6555, "total_steps": 42420, "loss": 0.4103, "lr": 0.00099097850051201, "epoch": 3.0905233380480905, "percentage": 15.45, "elapsed_time": "0:21:19", "remaining_time": "1:56:38", "throughput": 5150.58, "total_tokens": 6588544} {"current_steps": 6560, "total_steps": 42420, "loss": 0.3543, "lr": 0.0009909395563956905, "epoch": 3.092880716643093, "percentage": 15.46, "elapsed_time": "0:21:20", "remaining_time": "1:56:37", "throughput": 5150.78, "total_tokens": 6593632} {"current_steps": 6565, "total_steps": 42420, "loss": 0.3409, "lr": 0.0009909005291716397, "epoch": 3.0952380952380953, "percentage": 15.48, "elapsed_time": "0:21:20", "remaining_time": "1:56:35", "throughput": 5150.69, "total_tokens": 6597600} {"current_steps": 6570, "total_steps": 42420, "loss": 0.2645, "lr": 0.0009908614188464645, "epoch": 3.0975954738330977, "percentage": 15.49, "elapsed_time": "0:21:21", "remaining_time": "1:56:34", "throughput": 5150.94, "total_tokens": 6603040} {"current_steps": 6575, "total_steps": 42420, "loss": 0.3905, "lr": 0.0009908222254267852, "epoch": 3.0999528524281, "percentage": 15.5, "elapsed_time": "0:21:22", "remaining_time": "1:56:33", "throughput": 5151.14, "total_tokens": 6607872} {"current_steps": 6580, "total_steps": 42420, "loss": 0.4153, "lr": 0.000990782948919237, "epoch": 3.102310231023102, "percentage": 15.51, "elapsed_time": "0:21:23", "remaining_time": "1:56:32", "throughput": 5151.4, "total_tokens": 6613216} {"current_steps": 6585, "total_steps": 42420, "loss": 0.4029, "lr": 0.0009907435893304684, "epoch": 3.1046676096181045, "percentage": 15.52, "elapsed_time": "0:21:24", "remaining_time": "1:56:31", "throughput": 5151.96, "total_tokens": 6619360} {"current_steps": 6590, "total_steps": 42420, "loss": 0.3849, "lr": 0.0009907041466671426, "epoch": 3.107024988213107, "percentage": 15.54, "elapsed_time": "0:21:25", "remaining_time": "1:56:30", "throughput": 5152.15, "total_tokens": 6624160} {"current_steps": 6595, "total_steps": 42420, "loss": 0.3497, "lr": 0.0009906646209359366, "epoch": 3.1093823668081093, "percentage": 15.55, "elapsed_time": "0:21:26", "remaining_time": "1:56:28", "throughput": 5152.18, "total_tokens": 6628704} {"current_steps": 6600, "total_steps": 42420, "loss": 0.2769, "lr": 0.0009906250121435411, "epoch": 3.1117397454031117, "percentage": 15.56, "elapsed_time": "0:21:27", "remaining_time": "1:56:27", "throughput": 5152.34, "total_tokens": 6633504} {"current_steps": 6605, "total_steps": 42420, "loss": 0.4412, "lr": 0.0009905853202966616, "epoch": 3.114097123998114, "percentage": 15.57, "elapsed_time": "0:21:28", "remaining_time": "1:56:25", "throughput": 5152.11, "total_tokens": 6637312} {"current_steps": 6610, "total_steps": 42420, "loss": 0.3163, "lr": 0.0009905455454020167, "epoch": 3.1164545025931165, "percentage": 15.58, "elapsed_time": "0:21:29", "remaining_time": "1:56:23", "throughput": 5151.94, "total_tokens": 6641152} {"current_steps": 6615, "total_steps": 42420, "loss": 0.3484, "lr": 0.0009905056874663402, "epoch": 3.118811881188119, "percentage": 15.59, "elapsed_time": "0:21:30", "remaining_time": "1:56:23", "throughput": 5152.42, "total_tokens": 6647424} {"current_steps": 6620, "total_steps": 42420, "loss": 0.3269, "lr": 0.000990465746496379, "epoch": 3.1211692597831213, "percentage": 15.61, "elapsed_time": "0:21:31", "remaining_time": "1:56:21", "throughput": 5152.59, "total_tokens": 6652352} {"current_steps": 6625, "total_steps": 42420, "loss": 0.2565, "lr": 0.0009904257224988946, "epoch": 3.1235266383781237, "percentage": 15.62, "elapsed_time": "0:21:31", "remaining_time": "1:56:20", "throughput": 5152.78, "total_tokens": 6657248} {"current_steps": 6630, "total_steps": 42420, "loss": 0.3058, "lr": 0.0009903856154806624, "epoch": 3.1258840169731257, "percentage": 15.63, "elapsed_time": "0:21:32", "remaining_time": "1:56:18", "throughput": 5152.91, "total_tokens": 6661888} {"current_steps": 6635, "total_steps": 42420, "loss": 0.375, "lr": 0.0009903454254484717, "epoch": 3.128241395568128, "percentage": 15.64, "elapsed_time": "0:21:33", "remaining_time": "1:56:18", "throughput": 5153.21, "total_tokens": 6667360} {"current_steps": 6640, "total_steps": 42420, "loss": 0.4486, "lr": 0.000990305152409126, "epoch": 3.1305987741631305, "percentage": 15.65, "elapsed_time": "0:21:34", "remaining_time": "1:56:16", "throughput": 5153.37, "total_tokens": 6672224} {"current_steps": 6645, "total_steps": 42420, "loss": 0.3597, "lr": 0.000990264796369443, "epoch": 3.132956152758133, "percentage": 15.66, "elapsed_time": "0:21:35", "remaining_time": "1:56:15", "throughput": 5153.82, "total_tokens": 6677920} {"current_steps": 6650, "total_steps": 42420, "loss": 0.3804, "lr": 0.0009902243573362542, "epoch": 3.1353135313531353, "percentage": 15.68, "elapsed_time": "0:21:36", "remaining_time": "1:56:14", "throughput": 5154.15, "total_tokens": 6683328} {"current_steps": 6655, "total_steps": 42420, "loss": 0.3936, "lr": 0.0009901838353164052, "epoch": 3.1376709099481377, "percentage": 15.69, "elapsed_time": "0:21:37", "remaining_time": "1:56:13", "throughput": 5154.26, "total_tokens": 6688192} {"current_steps": 6660, "total_steps": 42420, "loss": 0.3596, "lr": 0.0009901432303167557, "epoch": 3.14002828854314, "percentage": 15.7, "elapsed_time": "0:21:38", "remaining_time": "1:56:12", "throughput": 5154.46, "total_tokens": 6693024} {"current_steps": 6665, "total_steps": 42420, "loss": 0.3522, "lr": 0.0009901025423441794, "epoch": 3.1423856671381425, "percentage": 15.71, "elapsed_time": "0:21:39", "remaining_time": "1:56:11", "throughput": 5154.9, "total_tokens": 6698944} {"current_steps": 6670, "total_steps": 42420, "loss": 0.3628, "lr": 0.0009900617714055644, "epoch": 3.144743045733145, "percentage": 15.72, "elapsed_time": "0:21:40", "remaining_time": "1:56:10", "throughput": 5155.2, "total_tokens": 6704352} {"current_steps": 6675, "total_steps": 42420, "loss": 0.3366, "lr": 0.000990020917507812, "epoch": 3.1471004243281473, "percentage": 15.74, "elapsed_time": "0:21:41", "remaining_time": "1:56:09", "throughput": 5155.52, "total_tokens": 6710048} {"current_steps": 6680, "total_steps": 42420, "loss": 0.3897, "lr": 0.0009899799806578386, "epoch": 3.1494578029231493, "percentage": 15.75, "elapsed_time": "0:21:42", "remaining_time": "1:56:08", "throughput": 5155.76, "total_tokens": 6715392} {"current_steps": 6685, "total_steps": 42420, "loss": 0.3409, "lr": 0.0009899389608625736, "epoch": 3.1518151815181517, "percentage": 15.76, "elapsed_time": "0:21:43", "remaining_time": "1:56:07", "throughput": 5156.08, "total_tokens": 6720992} {"current_steps": 6690, "total_steps": 42420, "loss": 0.3838, "lr": 0.0009898978581289616, "epoch": 3.154172560113154, "percentage": 15.77, "elapsed_time": "0:21:44", "remaining_time": "1:56:06", "throughput": 5156.2, "total_tokens": 6725536} {"current_steps": 6695, "total_steps": 42420, "loss": 0.3632, "lr": 0.0009898566724639598, "epoch": 3.1565299387081565, "percentage": 15.78, "elapsed_time": "0:21:45", "remaining_time": "1:56:05", "throughput": 5156.41, "total_tokens": 6730528} {"current_steps": 6700, "total_steps": 42420, "loss": 0.3746, "lr": 0.0009898154038745408, "epoch": 3.158887317303159, "percentage": 15.79, "elapsed_time": "0:21:46", "remaining_time": "1:56:03", "throughput": 5156.62, "total_tokens": 6735456} {"current_steps": 6705, "total_steps": 42420, "loss": 0.3371, "lr": 0.0009897740523676906, "epoch": 3.1612446958981613, "percentage": 15.81, "elapsed_time": "0:21:47", "remaining_time": "1:56:01", "throughput": 5156.59, "total_tokens": 6739744} {"current_steps": 6710, "total_steps": 42420, "loss": 0.3516, "lr": 0.000989732617950409, "epoch": 3.1636020744931637, "percentage": 15.82, "elapsed_time": "0:21:47", "remaining_time": "1:56:00", "throughput": 5156.95, "total_tokens": 6745120} {"current_steps": 6715, "total_steps": 42420, "loss": 0.387, "lr": 0.0009896911006297107, "epoch": 3.165959453088166, "percentage": 15.83, "elapsed_time": "0:21:48", "remaining_time": "1:55:59", "throughput": 5157.14, "total_tokens": 6750016} {"current_steps": 6720, "total_steps": 42420, "loss": 0.3334, "lr": 0.0009896495004126234, "epoch": 3.1683168316831685, "percentage": 15.84, "elapsed_time": "0:21:49", "remaining_time": "1:55:58", "throughput": 5157.45, "total_tokens": 6755616} {"current_steps": 6725, "total_steps": 42420, "loss": 0.3541, "lr": 0.0009896078173061893, "epoch": 3.170674210278171, "percentage": 15.85, "elapsed_time": "0:21:50", "remaining_time": "1:55:57", "throughput": 5157.75, "total_tokens": 6761088} {"current_steps": 6730, "total_steps": 42420, "loss": 0.3638, "lr": 0.000989566051317465, "epoch": 3.173031588873173, "percentage": 15.87, "elapsed_time": "0:21:51", "remaining_time": "1:55:56", "throughput": 5157.91, "total_tokens": 6766080} {"current_steps": 6735, "total_steps": 42420, "loss": 0.3598, "lr": 0.0009895242024535204, "epoch": 3.1753889674681752, "percentage": 15.88, "elapsed_time": "0:21:52", "remaining_time": "1:55:55", "throughput": 5158.01, "total_tokens": 6770784} {"current_steps": 6740, "total_steps": 42420, "loss": 0.3143, "lr": 0.0009894822707214401, "epoch": 3.1777463460631776, "percentage": 15.89, "elapsed_time": "0:21:53", "remaining_time": "1:55:54", "throughput": 5158.28, "total_tokens": 6776160} {"current_steps": 6745, "total_steps": 42420, "loss": 0.3468, "lr": 0.0009894402561283222, "epoch": 3.18010372465818, "percentage": 15.9, "elapsed_time": "0:21:54", "remaining_time": "1:55:52", "throughput": 5158.35, "total_tokens": 6780864} {"current_steps": 6750, "total_steps": 42420, "loss": 0.3114, "lr": 0.0009893981586812793, "epoch": 3.1824611032531824, "percentage": 15.91, "elapsed_time": "0:21:55", "remaining_time": "1:55:50", "throughput": 5158.34, "total_tokens": 6784992} {"current_steps": 6755, "total_steps": 42420, "loss": 0.2994, "lr": 0.0009893559783874376, "epoch": 3.184818481848185, "percentage": 15.92, "elapsed_time": "0:21:56", "remaining_time": "1:55:49", "throughput": 5158.42, "total_tokens": 6789632} {"current_steps": 6760, "total_steps": 42420, "loss": 0.3136, "lr": 0.0009893137152539374, "epoch": 3.1871758604431872, "percentage": 15.94, "elapsed_time": "0:21:57", "remaining_time": "1:55:48", "throughput": 5158.57, "total_tokens": 6794752} {"current_steps": 6765, "total_steps": 42420, "loss": 0.4289, "lr": 0.0009892713692879335, "epoch": 3.1895332390381896, "percentage": 15.95, "elapsed_time": "0:21:58", "remaining_time": "1:55:48", "throughput": 5159.38, "total_tokens": 6802400} {"current_steps": 6770, "total_steps": 42420, "loss": 0.3382, "lr": 0.000989228940496594, "epoch": 3.191890617633192, "percentage": 15.96, "elapsed_time": "0:21:59", "remaining_time": "1:55:47", "throughput": 5159.37, "total_tokens": 6806976} {"current_steps": 6775, "total_steps": 42420, "loss": 0.3267, "lr": 0.0009891864288871016, "epoch": 3.1942479962281944, "percentage": 15.97, "elapsed_time": "0:22:00", "remaining_time": "1:55:45", "throughput": 5159.32, "total_tokens": 6811296} {"current_steps": 6780, "total_steps": 42420, "loss": 0.2557, "lr": 0.0009891438344666525, "epoch": 3.1966053748231964, "percentage": 15.98, "elapsed_time": "0:22:01", "remaining_time": "1:55:44", "throughput": 5159.6, "total_tokens": 6816544} {"current_steps": 6785, "total_steps": 42420, "loss": 0.3854, "lr": 0.0009891011572424576, "epoch": 3.198962753418199, "percentage": 15.99, "elapsed_time": "0:22:02", "remaining_time": "1:55:43", "throughput": 5159.88, "total_tokens": 6821920} {"current_steps": 6790, "total_steps": 42420, "loss": 0.3376, "lr": 0.0009890583972217413, "epoch": 3.201320132013201, "percentage": 16.01, "elapsed_time": "0:22:03", "remaining_time": "1:55:42", "throughput": 5160.14, "total_tokens": 6827040} {"current_steps": 6795, "total_steps": 42420, "loss": 0.3634, "lr": 0.0009890155544117418, "epoch": 3.2036775106082036, "percentage": 16.02, "elapsed_time": "0:22:04", "remaining_time": "1:55:41", "throughput": 5160.72, "total_tokens": 6833280} {"current_steps": 6800, "total_steps": 42420, "loss": 0.3805, "lr": 0.0009889726288197123, "epoch": 3.206034889203206, "percentage": 16.03, "elapsed_time": "0:22:04", "remaining_time": "1:55:40", "throughput": 5160.72, "total_tokens": 6837600} {"current_steps": 6805, "total_steps": 42420, "loss": 0.3889, "lr": 0.0009889296204529188, "epoch": 3.2083922677982084, "percentage": 16.04, "elapsed_time": "0:22:05", "remaining_time": "1:55:39", "throughput": 5161.05, "total_tokens": 6842912} {"current_steps": 6810, "total_steps": 42420, "loss": 0.3579, "lr": 0.000988886529318642, "epoch": 3.210749646393211, "percentage": 16.05, "elapsed_time": "0:22:06", "remaining_time": "1:55:37", "throughput": 5161.22, "total_tokens": 6847872} {"current_steps": 6815, "total_steps": 42420, "loss": 0.4382, "lr": 0.000988843355424177, "epoch": 3.213107024988213, "percentage": 16.07, "elapsed_time": "0:22:07", "remaining_time": "1:55:36", "throughput": 5161.34, "total_tokens": 6852704} {"current_steps": 6820, "total_steps": 42420, "loss": 0.3613, "lr": 0.0009888000987768317, "epoch": 3.2154644035832156, "percentage": 16.08, "elapsed_time": "0:22:08", "remaining_time": "1:55:35", "throughput": 5161.86, "total_tokens": 6858720} {"current_steps": 6825, "total_steps": 42420, "loss": 0.362, "lr": 0.0009887567593839291, "epoch": 3.217821782178218, "percentage": 16.09, "elapsed_time": "0:22:09", "remaining_time": "1:55:34", "throughput": 5161.9, "total_tokens": 6863328} {"current_steps": 6830, "total_steps": 42420, "loss": 0.392, "lr": 0.000988713337252806, "epoch": 3.22017916077322, "percentage": 16.1, "elapsed_time": "0:22:10", "remaining_time": "1:55:33", "throughput": 5162.12, "total_tokens": 6868448} {"current_steps": 6835, "total_steps": 42420, "loss": 0.3399, "lr": 0.0009886698323908125, "epoch": 3.2225365393682224, "percentage": 16.11, "elapsed_time": "0:22:11", "remaining_time": "1:55:32", "throughput": 5162.5, "total_tokens": 6873920} {"current_steps": 6840, "total_steps": 42420, "loss": 0.3214, "lr": 0.0009886262448053134, "epoch": 3.2248939179632248, "percentage": 16.12, "elapsed_time": "0:22:12", "remaining_time": "1:55:31", "throughput": 5162.68, "total_tokens": 6879104} {"current_steps": 6845, "total_steps": 42420, "loss": 0.2482, "lr": 0.0009885825745036877, "epoch": 3.227251296558227, "percentage": 16.14, "elapsed_time": "0:22:13", "remaining_time": "1:55:30", "throughput": 5163.01, "total_tokens": 6884608} {"current_steps": 6850, "total_steps": 42420, "loss": 0.4636, "lr": 0.0009885388214933278, "epoch": 3.2296086751532296, "percentage": 16.15, "elapsed_time": "0:22:14", "remaining_time": "1:55:29", "throughput": 5163.38, "total_tokens": 6890400} {"current_steps": 6855, "total_steps": 42420, "loss": 0.3688, "lr": 0.0009884949857816403, "epoch": 3.231966053748232, "percentage": 16.16, "elapsed_time": "0:22:15", "remaining_time": "1:55:28", "throughput": 5163.69, "total_tokens": 6895936} {"current_steps": 6860, "total_steps": 42420, "loss": 0.3436, "lr": 0.000988451067376046, "epoch": 3.2343234323432344, "percentage": 16.17, "elapsed_time": "0:22:16", "remaining_time": "1:55:26", "throughput": 5163.69, "total_tokens": 6900160} {"current_steps": 6865, "total_steps": 42420, "loss": 0.3639, "lr": 0.0009884070662839791, "epoch": 3.236680810938237, "percentage": 16.18, "elapsed_time": "0:22:17", "remaining_time": "1:55:26", "throughput": 5164.11, "total_tokens": 6906080} {"current_steps": 6870, "total_steps": 42420, "loss": 0.3715, "lr": 0.0009883629825128888, "epoch": 3.239038189533239, "percentage": 16.2, "elapsed_time": "0:22:18", "remaining_time": "1:55:25", "throughput": 5164.45, "total_tokens": 6911680} {"current_steps": 6875, "total_steps": 42420, "loss": 0.3015, "lr": 0.0009883188160702372, "epoch": 3.2413955681282416, "percentage": 16.21, "elapsed_time": "0:22:19", "remaining_time": "1:55:24", "throughput": 5164.77, "total_tokens": 6917280} {"current_steps": 6880, "total_steps": 42420, "loss": 0.3768, "lr": 0.0009882745669635017, "epoch": 3.2437529467232435, "percentage": 16.22, "elapsed_time": "0:22:20", "remaining_time": "1:55:23", "throughput": 5164.88, "total_tokens": 6921984} {"current_steps": 6885, "total_steps": 42420, "loss": 0.3536, "lr": 0.000988230235200172, "epoch": 3.246110325318246, "percentage": 16.23, "elapsed_time": "0:22:21", "remaining_time": "1:55:22", "throughput": 5165.38, "total_tokens": 6928064} {"current_steps": 6890, "total_steps": 42420, "loss": 0.3486, "lr": 0.0009881858207877535, "epoch": 3.2484677039132484, "percentage": 16.24, "elapsed_time": "0:22:22", "remaining_time": "1:55:20", "throughput": 5165.35, "total_tokens": 6932128} {"current_steps": 6895, "total_steps": 42420, "loss": 0.3032, "lr": 0.0009881413237337643, "epoch": 3.2508250825082508, "percentage": 16.25, "elapsed_time": "0:22:22", "remaining_time": "1:55:19", "throughput": 5165.4, "total_tokens": 6936960} {"current_steps": 6900, "total_steps": 42420, "loss": 0.4555, "lr": 0.000988096744045737, "epoch": 3.253182461103253, "percentage": 16.27, "elapsed_time": "0:22:23", "remaining_time": "1:55:17", "throughput": 5165.5, "total_tokens": 6941664} {"current_steps": 6905, "total_steps": 42420, "loss": 0.3493, "lr": 0.0009880520817312185, "epoch": 3.2555398396982556, "percentage": 16.28, "elapsed_time": "0:22:24", "remaining_time": "1:55:15", "throughput": 5165.27, "total_tokens": 6945280} {"current_steps": 6910, "total_steps": 42420, "loss": 0.3416, "lr": 0.0009880073367977693, "epoch": 3.257897218293258, "percentage": 16.29, "elapsed_time": "0:22:25", "remaining_time": "1:55:14", "throughput": 5165.46, "total_tokens": 6950368} {"current_steps": 6915, "total_steps": 42420, "loss": 0.3464, "lr": 0.0009879625092529638, "epoch": 3.2602545968882604, "percentage": 16.3, "elapsed_time": "0:22:26", "remaining_time": "1:55:14", "throughput": 5166.15, "total_tokens": 6957248} {"current_steps": 6920, "total_steps": 42420, "loss": 0.3676, "lr": 0.0009879175991043907, "epoch": 3.2626119754832628, "percentage": 16.31, "elapsed_time": "0:22:27", "remaining_time": "1:55:12", "throughput": 5166.1, "total_tokens": 6961536} {"current_steps": 6925, "total_steps": 42420, "loss": 0.2802, "lr": 0.0009878726063596523, "epoch": 3.264969354078265, "percentage": 16.32, "elapsed_time": "0:22:28", "remaining_time": "1:55:11", "throughput": 5166.15, "total_tokens": 6965952} {"current_steps": 6930, "total_steps": 42420, "loss": 0.3356, "lr": 0.0009878275310263652, "epoch": 3.2673267326732676, "percentage": 16.34, "elapsed_time": "0:22:29", "remaining_time": "1:55:09", "throughput": 5166.26, "total_tokens": 6970720} {"current_steps": 6935, "total_steps": 42420, "loss": 0.1717, "lr": 0.00098778237311216, "epoch": 3.2696841112682695, "percentage": 16.35, "elapsed_time": "0:22:30", "remaining_time": "1:55:08", "throughput": 5166.45, "total_tokens": 6975904} {"current_steps": 6940, "total_steps": 42420, "loss": 0.3354, "lr": 0.0009877371326246812, "epoch": 3.272041489863272, "percentage": 16.36, "elapsed_time": "0:22:31", "remaining_time": "1:55:07", "throughput": 5166.41, "total_tokens": 6980192} {"current_steps": 6945, "total_steps": 42420, "loss": 0.5536, "lr": 0.000987691809571587, "epoch": 3.2743988684582743, "percentage": 16.37, "elapsed_time": "0:22:32", "remaining_time": "1:55:06", "throughput": 5166.61, "total_tokens": 6985440} {"current_steps": 6950, "total_steps": 42420, "loss": 0.3256, "lr": 0.0009876464039605501, "epoch": 3.2767562470532767, "percentage": 16.38, "elapsed_time": "0:22:32", "remaining_time": "1:55:05", "throughput": 5166.83, "total_tokens": 6990624} {"current_steps": 6955, "total_steps": 42420, "loss": 0.4728, "lr": 0.000987600915799257, "epoch": 3.279113625648279, "percentage": 16.4, "elapsed_time": "0:22:33", "remaining_time": "1:55:03", "throughput": 5167.1, "total_tokens": 6995840} {"current_steps": 6960, "total_steps": 42420, "loss": 0.3382, "lr": 0.0009875553450954075, "epoch": 3.2814710042432815, "percentage": 16.41, "elapsed_time": "0:22:34", "remaining_time": "1:55:02", "throughput": 5167.2, "total_tokens": 7000640} {"current_steps": 6965, "total_steps": 42420, "loss": 0.4097, "lr": 0.0009875096918567164, "epoch": 3.283828382838284, "percentage": 16.42, "elapsed_time": "0:22:35", "remaining_time": "1:55:00", "throughput": 5167.18, "total_tokens": 7004992} {"current_steps": 6970, "total_steps": 42420, "loss": 0.3731, "lr": 0.0009874639560909118, "epoch": 3.2861857614332863, "percentage": 16.43, "elapsed_time": "0:22:36", "remaining_time": "1:54:59", "throughput": 5167.28, "total_tokens": 7009536} {"current_steps": 6975, "total_steps": 42420, "loss": 0.3549, "lr": 0.0009874181378057363, "epoch": 3.2885431400282887, "percentage": 16.44, "elapsed_time": "0:22:37", "remaining_time": "1:54:58", "throughput": 5167.32, "total_tokens": 7014240} {"current_steps": 6980, "total_steps": 42420, "loss": 0.3825, "lr": 0.0009873722370089457, "epoch": 3.2909005186232907, "percentage": 16.45, "elapsed_time": "0:22:38", "remaining_time": "1:54:59", "throughput": 5168.49, "total_tokens": 7023200} {"current_steps": 6985, "total_steps": 42420, "loss": 0.4026, "lr": 0.0009873262537083108, "epoch": 3.293257897218293, "percentage": 16.47, "elapsed_time": "0:22:39", "remaining_time": "1:54:57", "throughput": 5168.48, "total_tokens": 7027552} {"current_steps": 6990, "total_steps": 42420, "loss": 0.3756, "lr": 0.0009872801879116152, "epoch": 3.2956152758132955, "percentage": 16.48, "elapsed_time": "0:22:40", "remaining_time": "1:54:56", "throughput": 5168.67, "total_tokens": 7032544} {"current_steps": 6995, "total_steps": 42420, "loss": 0.3618, "lr": 0.0009872340396266575, "epoch": 3.297972654408298, "percentage": 16.49, "elapsed_time": "0:22:41", "remaining_time": "1:54:55", "throughput": 5168.8, "total_tokens": 7037408} {"current_steps": 7000, "total_steps": 42420, "loss": 0.3609, "lr": 0.0009871878088612495, "epoch": 3.3003300330033003, "percentage": 16.5, "elapsed_time": "0:22:42", "remaining_time": "1:54:54", "throughput": 5169.27, "total_tokens": 7043712} {"current_steps": 7005, "total_steps": 42420, "loss": 0.3628, "lr": 0.0009871414956232176, "epoch": 3.3026874115983027, "percentage": 16.51, "elapsed_time": "0:22:43", "remaining_time": "1:54:53", "throughput": 5169.71, "total_tokens": 7049472} {"current_steps": 7010, "total_steps": 42420, "loss": 0.3317, "lr": 0.0009870950999204017, "epoch": 3.305044790193305, "percentage": 16.53, "elapsed_time": "0:22:44", "remaining_time": "1:54:52", "throughput": 5169.74, "total_tokens": 7054080} {"current_steps": 7015, "total_steps": 42420, "loss": 0.4158, "lr": 0.0009870486217606557, "epoch": 3.3074021687883075, "percentage": 16.54, "elapsed_time": "0:22:45", "remaining_time": "1:54:50", "throughput": 5169.82, "total_tokens": 7058560} {"current_steps": 7020, "total_steps": 42420, "loss": 0.3365, "lr": 0.0009870020611518478, "epoch": 3.30975954738331, "percentage": 16.55, "elapsed_time": "0:22:46", "remaining_time": "1:54:49", "throughput": 5169.82, "total_tokens": 7062752} {"current_steps": 7025, "total_steps": 42420, "loss": 0.3706, "lr": 0.0009869554181018594, "epoch": 3.3121169259783123, "percentage": 16.56, "elapsed_time": "0:22:47", "remaining_time": "1:54:47", "throughput": 5169.99, "total_tokens": 7067680} {"current_steps": 7030, "total_steps": 42420, "loss": 0.3538, "lr": 0.0009869086926185871, "epoch": 3.3144743045733147, "percentage": 16.57, "elapsed_time": "0:22:47", "remaining_time": "1:54:46", "throughput": 5170.12, "total_tokens": 7072352} {"current_steps": 7035, "total_steps": 42420, "loss": 0.3439, "lr": 0.0009868618847099402, "epoch": 3.3168316831683167, "percentage": 16.58, "elapsed_time": "0:22:48", "remaining_time": "1:54:45", "throughput": 5170.36, "total_tokens": 7077408} {"current_steps": 7040, "total_steps": 42420, "loss": 0.3749, "lr": 0.0009868149943838428, "epoch": 3.319189061763319, "percentage": 16.6, "elapsed_time": "0:22:49", "remaining_time": "1:54:44", "throughput": 5170.8, "total_tokens": 7083392} {"current_steps": 7045, "total_steps": 42420, "loss": 0.322, "lr": 0.0009867680216482325, "epoch": 3.3215464403583215, "percentage": 16.61, "elapsed_time": "0:22:50", "remaining_time": "1:54:42", "throughput": 5170.96, "total_tokens": 7088096} {"current_steps": 7050, "total_steps": 42420, "loss": 0.3565, "lr": 0.0009867209665110608, "epoch": 3.323903818953324, "percentage": 16.62, "elapsed_time": "0:22:51", "remaining_time": "1:54:41", "throughput": 5171.07, "total_tokens": 7092736} {"current_steps": 7055, "total_steps": 42420, "loss": 0.3287, "lr": 0.0009866738289802937, "epoch": 3.3262611975483263, "percentage": 16.63, "elapsed_time": "0:22:52", "remaining_time": "1:54:39", "throughput": 5171.11, "total_tokens": 7097280} {"current_steps": 7060, "total_steps": 42420, "loss": 0.3313, "lr": 0.0009866266090639104, "epoch": 3.3286185761433287, "percentage": 16.64, "elapsed_time": "0:22:53", "remaining_time": "1:54:38", "throughput": 5171.44, "total_tokens": 7102752} {"current_steps": 7065, "total_steps": 42420, "loss": 0.2765, "lr": 0.0009865793067699048, "epoch": 3.330975954738331, "percentage": 16.65, "elapsed_time": "0:22:54", "remaining_time": "1:54:37", "throughput": 5171.46, "total_tokens": 7107264} {"current_steps": 7070, "total_steps": 42420, "loss": 0.402, "lr": 0.0009865319221062843, "epoch": 3.3333333333333335, "percentage": 16.67, "elapsed_time": "0:22:55", "remaining_time": "1:54:35", "throughput": 5171.28, "total_tokens": 7111168} {"current_steps": 7075, "total_steps": 42420, "loss": 0.3427, "lr": 0.0009864844550810702, "epoch": 3.335690711928336, "percentage": 16.68, "elapsed_time": "0:22:56", "remaining_time": "1:54:34", "throughput": 5171.49, "total_tokens": 7116288} {"current_steps": 7080, "total_steps": 42420, "loss": 0.3539, "lr": 0.0009864369057022977, "epoch": 3.338048090523338, "percentage": 16.69, "elapsed_time": "0:22:56", "remaining_time": "1:54:32", "throughput": 5171.41, "total_tokens": 7120320} {"current_steps": 7085, "total_steps": 42420, "loss": 0.2904, "lr": 0.0009863892739780163, "epoch": 3.3404054691183402, "percentage": 16.7, "elapsed_time": "0:22:57", "remaining_time": "1:54:31", "throughput": 5171.52, "total_tokens": 7125088} {"current_steps": 7090, "total_steps": 42420, "loss": 0.4316, "lr": 0.0009863415599162893, "epoch": 3.3427628477133426, "percentage": 16.71, "elapsed_time": "0:22:58", "remaining_time": "1:54:30", "throughput": 5171.82, "total_tokens": 7130528} {"current_steps": 7095, "total_steps": 42420, "loss": 0.3388, "lr": 0.0009862937635251939, "epoch": 3.345120226308345, "percentage": 16.73, "elapsed_time": "0:22:59", "remaining_time": "1:54:28", "throughput": 5172.0, "total_tokens": 7135456} {"current_steps": 7100, "total_steps": 42420, "loss": 0.3394, "lr": 0.000986245884812821, "epoch": 3.3474776049033474, "percentage": 16.74, "elapsed_time": "0:23:00", "remaining_time": "1:54:27", "throughput": 5172.17, "total_tokens": 7140416} {"current_steps": 7105, "total_steps": 42420, "loss": 0.3095, "lr": 0.000986197923787276, "epoch": 3.34983498349835, "percentage": 16.75, "elapsed_time": "0:23:01", "remaining_time": "1:54:26", "throughput": 5172.07, "total_tokens": 7144576} {"current_steps": 7110, "total_steps": 42420, "loss": 0.4383, "lr": 0.0009861498804566776, "epoch": 3.3521923620933523, "percentage": 16.76, "elapsed_time": "0:23:02", "remaining_time": "1:54:24", "throughput": 5172.01, "total_tokens": 7148672} {"current_steps": 7115, "total_steps": 42420, "loss": 0.3247, "lr": 0.0009861017548291587, "epoch": 3.3545497406883547, "percentage": 16.77, "elapsed_time": "0:23:03", "remaining_time": "1:54:22", "throughput": 5172.1, "total_tokens": 7153504} {"current_steps": 7120, "total_steps": 42420, "loss": 0.3513, "lr": 0.000986053546912866, "epoch": 3.356907119283357, "percentage": 16.78, "elapsed_time": "0:23:03", "remaining_time": "1:54:21", "throughput": 5172.07, "total_tokens": 7157696} {"current_steps": 7125, "total_steps": 42420, "loss": 0.3449, "lr": 0.000986005256715961, "epoch": 3.3592644978783595, "percentage": 16.8, "elapsed_time": "0:23:04", "remaining_time": "1:54:19", "throughput": 5172.24, "total_tokens": 7162528} {"current_steps": 7130, "total_steps": 42420, "loss": 0.3512, "lr": 0.0009859568842466178, "epoch": 3.361621876473362, "percentage": 16.81, "elapsed_time": "0:23:05", "remaining_time": "1:54:18", "throughput": 5172.24, "total_tokens": 7166656} {"current_steps": 7135, "total_steps": 42420, "loss": 0.3266, "lr": 0.0009859084295130252, "epoch": 3.363979255068364, "percentage": 16.82, "elapsed_time": "0:23:06", "remaining_time": "1:54:16", "throughput": 5172.35, "total_tokens": 7171328} {"current_steps": 7140, "total_steps": 42420, "loss": 0.361, "lr": 0.0009858598925233854, "epoch": 3.366336633663366, "percentage": 16.83, "elapsed_time": "0:23:07", "remaining_time": "1:54:14", "throughput": 5172.46, "total_tokens": 7175712} {"current_steps": 7145, "total_steps": 42420, "loss": 0.4181, "lr": 0.0009858112732859157, "epoch": 3.3686940122583686, "percentage": 16.84, "elapsed_time": "0:23:08", "remaining_time": "1:54:13", "throughput": 5172.53, "total_tokens": 7180128} {"current_steps": 7150, "total_steps": 42420, "loss": 0.3347, "lr": 0.0009857625718088456, "epoch": 3.371051390853371, "percentage": 16.86, "elapsed_time": "0:23:08", "remaining_time": "1:54:11", "throughput": 5172.46, "total_tokens": 7184128} {"current_steps": 7155, "total_steps": 42420, "loss": 0.3857, "lr": 0.0009857137881004203, "epoch": 3.3734087694483734, "percentage": 16.87, "elapsed_time": "0:23:09", "remaining_time": "1:54:09", "throughput": 5172.62, "total_tokens": 7188864} {"current_steps": 7160, "total_steps": 42420, "loss": 0.354, "lr": 0.0009856649221688973, "epoch": 3.375766148043376, "percentage": 16.88, "elapsed_time": "0:23:10", "remaining_time": "1:54:08", "throughput": 5172.9, "total_tokens": 7194336} {"current_steps": 7165, "total_steps": 42420, "loss": 0.3887, "lr": 0.0009856159740225493, "epoch": 3.3781235266383782, "percentage": 16.89, "elapsed_time": "0:23:11", "remaining_time": "1:54:07", "throughput": 5173.08, "total_tokens": 7199296} {"current_steps": 7170, "total_steps": 42420, "loss": 0.3343, "lr": 0.0009855669436696622, "epoch": 3.3804809052333806, "percentage": 16.9, "elapsed_time": "0:23:12", "remaining_time": "1:54:07", "throughput": 5173.44, "total_tokens": 7205344} {"current_steps": 7175, "total_steps": 42420, "loss": 0.3435, "lr": 0.0009855178311185357, "epoch": 3.382838283828383, "percentage": 16.91, "elapsed_time": "0:23:13", "remaining_time": "1:54:06", "throughput": 5173.73, "total_tokens": 7211008} {"current_steps": 7180, "total_steps": 42420, "loss": 0.4003, "lr": 0.0009854686363774843, "epoch": 3.385195662423385, "percentage": 16.93, "elapsed_time": "0:23:14", "remaining_time": "1:54:04", "throughput": 5173.81, "total_tokens": 7215552} {"current_steps": 7185, "total_steps": 42420, "loss": 0.362, "lr": 0.0009854193594548354, "epoch": 3.3875530410183874, "percentage": 16.94, "elapsed_time": "0:23:15", "remaining_time": "1:54:03", "throughput": 5174.15, "total_tokens": 7220960} {"current_steps": 7190, "total_steps": 42420, "loss": 0.3995, "lr": 0.000985370000358931, "epoch": 3.38991041961339, "percentage": 16.95, "elapsed_time": "0:23:16", "remaining_time": "1:54:02", "throughput": 5174.14, "total_tokens": 7225312} {"current_steps": 7195, "total_steps": 42420, "loss": 0.3693, "lr": 0.0009853205590981268, "epoch": 3.392267798208392, "percentage": 16.96, "elapsed_time": "0:23:17", "remaining_time": "1:54:01", "throughput": 5174.4, "total_tokens": 7230624} {"current_steps": 7200, "total_steps": 42420, "loss": 0.3448, "lr": 0.0009852710356807919, "epoch": 3.3946251768033946, "percentage": 16.97, "elapsed_time": "0:23:18", "remaining_time": "1:53:59", "throughput": 5174.42, "total_tokens": 7235008} {"current_steps": 7205, "total_steps": 42420, "loss": 0.3078, "lr": 0.0009852214301153102, "epoch": 3.396982555398397, "percentage": 16.98, "elapsed_time": "0:23:19", "remaining_time": "1:53:58", "throughput": 5174.61, "total_tokens": 7239808} {"current_steps": 7210, "total_steps": 42420, "loss": 0.4463, "lr": 0.000985171742410079, "epoch": 3.3993399339933994, "percentage": 17.0, "elapsed_time": "0:23:19", "remaining_time": "1:53:56", "throughput": 5174.61, "total_tokens": 7244096} {"current_steps": 7215, "total_steps": 42420, "loss": 0.3322, "lr": 0.0009851219725735096, "epoch": 3.401697312588402, "percentage": 17.01, "elapsed_time": "0:23:20", "remaining_time": "1:53:56", "throughput": 5174.97, "total_tokens": 7250112} {"current_steps": 7220, "total_steps": 42420, "loss": 0.4069, "lr": 0.000985072120614027, "epoch": 3.404054691183404, "percentage": 17.02, "elapsed_time": "0:23:21", "remaining_time": "1:53:55", "throughput": 5175.34, "total_tokens": 7255776} {"current_steps": 7225, "total_steps": 42420, "loss": 0.3105, "lr": 0.0009850221865400706, "epoch": 3.4064120697784066, "percentage": 17.03, "elapsed_time": "0:23:22", "remaining_time": "1:53:53", "throughput": 5175.35, "total_tokens": 7260320} {"current_steps": 7230, "total_steps": 42420, "loss": 0.3318, "lr": 0.0009849721703600932, "epoch": 3.408769448373409, "percentage": 17.04, "elapsed_time": "0:23:23", "remaining_time": "1:53:53", "throughput": 5175.71, "total_tokens": 7266080} {"current_steps": 7235, "total_steps": 42420, "loss": 0.3279, "lr": 0.0009849220720825615, "epoch": 3.411126826968411, "percentage": 17.06, "elapsed_time": "0:23:24", "remaining_time": "1:53:51", "throughput": 5175.76, "total_tokens": 7270528} {"current_steps": 7240, "total_steps": 42420, "loss": 0.2656, "lr": 0.0009848718917159567, "epoch": 3.4134842055634134, "percentage": 17.07, "elapsed_time": "0:23:25", "remaining_time": "1:53:49", "throughput": 5175.68, "total_tokens": 7274624} {"current_steps": 7245, "total_steps": 42420, "loss": 0.4075, "lr": 0.0009848216292687732, "epoch": 3.4158415841584158, "percentage": 17.08, "elapsed_time": "0:23:26", "remaining_time": "1:53:48", "throughput": 5175.98, "total_tokens": 7280192} {"current_steps": 7250, "total_steps": 42420, "loss": 0.2885, "lr": 0.00098477128474952, "epoch": 3.418198962753418, "percentage": 17.09, "elapsed_time": "0:23:27", "remaining_time": "1:53:47", "throughput": 5176.08, "total_tokens": 7284896} {"current_steps": 7255, "total_steps": 42420, "loss": 0.3643, "lr": 0.0009847208581667188, "epoch": 3.4205563413484206, "percentage": 17.1, "elapsed_time": "0:23:28", "remaining_time": "1:53:45", "throughput": 5176.16, "total_tokens": 7289440} {"current_steps": 7260, "total_steps": 42420, "loss": 0.3039, "lr": 0.0009846703495289067, "epoch": 3.422913719943423, "percentage": 17.11, "elapsed_time": "0:23:29", "remaining_time": "1:53:45", "throughput": 5176.9, "total_tokens": 7296640} {"current_steps": 7265, "total_steps": 42420, "loss": 0.3563, "lr": 0.0009846197588446336, "epoch": 3.4252710985384254, "percentage": 17.13, "elapsed_time": "0:23:30", "remaining_time": "1:53:45", "throughput": 5177.25, "total_tokens": 7302272} {"current_steps": 7270, "total_steps": 42420, "loss": 0.3506, "lr": 0.0009845690861224636, "epoch": 3.4276284771334278, "percentage": 17.14, "elapsed_time": "0:23:31", "remaining_time": "1:53:43", "throughput": 5177.33, "total_tokens": 7306752} {"current_steps": 7275, "total_steps": 42420, "loss": 0.3511, "lr": 0.000984518331370975, "epoch": 3.42998585572843, "percentage": 17.15, "elapsed_time": "0:23:32", "remaining_time": "1:53:41", "throughput": 5177.39, "total_tokens": 7311232} {"current_steps": 7280, "total_steps": 42420, "loss": 0.3593, "lr": 0.0009844674945987592, "epoch": 3.432343234323432, "percentage": 17.16, "elapsed_time": "0:23:33", "remaining_time": "1:53:40", "throughput": 5177.65, "total_tokens": 7316480} {"current_steps": 7285, "total_steps": 42420, "loss": 0.3421, "lr": 0.0009844165758144225, "epoch": 3.4347006129184345, "percentage": 17.17, "elapsed_time": "0:23:33", "remaining_time": "1:53:39", "throughput": 5177.71, "total_tokens": 7320928} {"current_steps": 7290, "total_steps": 42420, "loss": 0.3909, "lr": 0.0009843655750265848, "epoch": 3.437057991513437, "percentage": 17.19, "elapsed_time": "0:23:34", "remaining_time": "1:53:38", "throughput": 5178.09, "total_tokens": 7326528} {"current_steps": 7295, "total_steps": 42420, "loss": 0.3751, "lr": 0.000984314492243879, "epoch": 3.4394153701084393, "percentage": 17.2, "elapsed_time": "0:23:35", "remaining_time": "1:53:37", "throughput": 5178.24, "total_tokens": 7331584} {"current_steps": 7300, "total_steps": 42420, "loss": 0.3602, "lr": 0.000984263327474953, "epoch": 3.4417727487034417, "percentage": 17.21, "elapsed_time": "0:23:36", "remaining_time": "1:53:35", "throughput": 5178.23, "total_tokens": 7335776} {"current_steps": 7305, "total_steps": 42420, "loss": 0.3962, "lr": 0.0009842120807284677, "epoch": 3.444130127298444, "percentage": 17.22, "elapsed_time": "0:23:37", "remaining_time": "1:53:33", "throughput": 5178.22, "total_tokens": 7339968} {"current_steps": 7310, "total_steps": 42420, "loss": 0.3788, "lr": 0.0009841607520130987, "epoch": 3.4464875058934465, "percentage": 17.23, "elapsed_time": "0:23:38", "remaining_time": "1:53:32", "throughput": 5178.37, "total_tokens": 7345088} {"current_steps": 7315, "total_steps": 42420, "loss": 0.3268, "lr": 0.000984109341337535, "epoch": 3.448844884488449, "percentage": 17.24, "elapsed_time": "0:23:39", "remaining_time": "1:53:30", "throughput": 5178.37, "total_tokens": 7349216} {"current_steps": 7320, "total_steps": 42420, "loss": 0.3033, "lr": 0.0009840578487104797, "epoch": 3.4512022630834513, "percentage": 17.26, "elapsed_time": "0:23:40", "remaining_time": "1:53:29", "throughput": 5178.61, "total_tokens": 7354528} {"current_steps": 7325, "total_steps": 42420, "loss": 0.4094, "lr": 0.0009840062741406491, "epoch": 3.4535596416784538, "percentage": 17.27, "elapsed_time": "0:23:41", "remaining_time": "1:53:29", "throughput": 5179.09, "total_tokens": 7360896} {"current_steps": 7330, "total_steps": 42420, "loss": 0.355, "lr": 0.0009839546176367746, "epoch": 3.455917020273456, "percentage": 17.28, "elapsed_time": "0:23:42", "remaining_time": "1:53:28", "throughput": 5179.12, "total_tokens": 7365408} {"current_steps": 7335, "total_steps": 42420, "loss": 0.3466, "lr": 0.0009839028792076002, "epoch": 3.458274398868458, "percentage": 17.29, "elapsed_time": "0:23:43", "remaining_time": "1:53:27", "throughput": 5179.43, "total_tokens": 7371040} {"current_steps": 7340, "total_steps": 42420, "loss": 0.3321, "lr": 0.0009838510588618845, "epoch": 3.4606317774634605, "percentage": 17.3, "elapsed_time": "0:23:44", "remaining_time": "1:53:26", "throughput": 5179.68, "total_tokens": 7376448} {"current_steps": 7345, "total_steps": 42420, "loss": 0.2879, "lr": 0.0009837991566084001, "epoch": 3.462989156058463, "percentage": 17.31, "elapsed_time": "0:23:45", "remaining_time": "1:53:25", "throughput": 5180.02, "total_tokens": 7381728} {"current_steps": 7350, "total_steps": 42420, "loss": 0.367, "lr": 0.0009837471724559327, "epoch": 3.4653465346534653, "percentage": 17.33, "elapsed_time": "0:23:45", "remaining_time": "1:53:23", "throughput": 5180.02, "total_tokens": 7386016} {"current_steps": 7355, "total_steps": 42420, "loss": 0.2589, "lr": 0.0009836951064132826, "epoch": 3.4677039132484677, "percentage": 17.34, "elapsed_time": "0:23:46", "remaining_time": "1:53:21", "throughput": 5180.07, "total_tokens": 7390400} {"current_steps": 7360, "total_steps": 42420, "loss": 0.2894, "lr": 0.0009836429584892637, "epoch": 3.47006129184347, "percentage": 17.35, "elapsed_time": "0:23:47", "remaining_time": "1:53:19", "throughput": 5179.98, "total_tokens": 7394368} {"current_steps": 7365, "total_steps": 42420, "loss": 0.3932, "lr": 0.0009835907286927037, "epoch": 3.4724186704384725, "percentage": 17.36, "elapsed_time": "0:23:48", "remaining_time": "1:53:18", "throughput": 5179.85, "total_tokens": 7398592} {"current_steps": 7370, "total_steps": 42420, "loss": 0.4347, "lr": 0.0009835384170324443, "epoch": 3.474776049033475, "percentage": 17.37, "elapsed_time": "0:23:49", "remaining_time": "1:53:16", "throughput": 5179.87, "total_tokens": 7402880} {"current_steps": 7375, "total_steps": 42420, "loss": 0.4006, "lr": 0.0009834860235173408, "epoch": 3.4771334276284773, "percentage": 17.39, "elapsed_time": "0:23:49", "remaining_time": "1:53:15", "throughput": 5179.75, "total_tokens": 7406912} {"current_steps": 7380, "total_steps": 42420, "loss": 0.3998, "lr": 0.000983433548156263, "epoch": 3.4794908062234793, "percentage": 17.4, "elapsed_time": "0:23:50", "remaining_time": "1:53:13", "throughput": 5179.88, "total_tokens": 7411584} {"current_steps": 7385, "total_steps": 42420, "loss": 0.3238, "lr": 0.0009833809909580932, "epoch": 3.4818481848184817, "percentage": 17.41, "elapsed_time": "0:23:51", "remaining_time": "1:53:12", "throughput": 5180.21, "total_tokens": 7417312} {"current_steps": 7390, "total_steps": 42420, "loss": 0.2598, "lr": 0.0009833283519317293, "epoch": 3.484205563413484, "percentage": 17.42, "elapsed_time": "0:23:52", "remaining_time": "1:53:11", "throughput": 5180.5, "total_tokens": 7422624} {"current_steps": 7395, "total_steps": 42420, "loss": 0.3548, "lr": 0.0009832756310860819, "epoch": 3.4865629420084865, "percentage": 17.43, "elapsed_time": "0:23:53", "remaining_time": "1:53:10", "throughput": 5180.75, "total_tokens": 7427776} {"current_steps": 7400, "total_steps": 42420, "loss": 0.2641, "lr": 0.0009832228284300756, "epoch": 3.488920320603489, "percentage": 17.44, "elapsed_time": "0:23:54", "remaining_time": "1:53:09", "throughput": 5180.92, "total_tokens": 7432832} {"current_steps": 7405, "total_steps": 42420, "loss": 0.4439, "lr": 0.000983169943972649, "epoch": 3.4912776991984913, "percentage": 17.46, "elapsed_time": "0:23:55", "remaining_time": "1:53:08", "throughput": 5181.46, "total_tokens": 7439168} {"current_steps": 7410, "total_steps": 42420, "loss": 0.3828, "lr": 0.0009831169777227548, "epoch": 3.4936350777934937, "percentage": 17.47, "elapsed_time": "0:23:56", "remaining_time": "1:53:08", "throughput": 5181.85, "total_tokens": 7444960} {"current_steps": 7415, "total_steps": 42420, "loss": 0.4122, "lr": 0.000983063929689359, "epoch": 3.495992456388496, "percentage": 17.48, "elapsed_time": "0:23:57", "remaining_time": "1:53:06", "throughput": 5181.95, "total_tokens": 7449696} {"current_steps": 7420, "total_steps": 42420, "loss": 0.3406, "lr": 0.000983010799881442, "epoch": 3.4983498349834985, "percentage": 17.49, "elapsed_time": "0:23:58", "remaining_time": "1:53:06", "throughput": 5182.18, "total_tokens": 7455296} {"current_steps": 7425, "total_steps": 42420, "loss": 0.3153, "lr": 0.0009829575883079976, "epoch": 3.500707213578501, "percentage": 17.5, "elapsed_time": "0:23:59", "remaining_time": "1:53:05", "throughput": 5182.54, "total_tokens": 7461344} {"current_steps": 7430, "total_steps": 42420, "loss": 0.2954, "lr": 0.0009829042949780335, "epoch": 3.5030645921735033, "percentage": 17.52, "elapsed_time": "0:24:00", "remaining_time": "1:53:04", "throughput": 5182.71, "total_tokens": 7466368} {"current_steps": 7435, "total_steps": 42420, "loss": 0.3635, "lr": 0.0009828509199005717, "epoch": 3.5054219707685053, "percentage": 17.53, "elapsed_time": "0:24:01", "remaining_time": "1:53:02", "throughput": 5182.73, "total_tokens": 7470624} {"current_steps": 7440, "total_steps": 42420, "loss": 0.3519, "lr": 0.0009827974630846475, "epoch": 3.5077793493635077, "percentage": 17.54, "elapsed_time": "0:24:02", "remaining_time": "1:53:01", "throughput": 5182.91, "total_tokens": 7475744} {"current_steps": 7445, "total_steps": 42420, "loss": 0.3253, "lr": 0.0009827439245393102, "epoch": 3.51013672795851, "percentage": 17.55, "elapsed_time": "0:24:03", "remaining_time": "1:53:00", "throughput": 5183.22, "total_tokens": 7481216} {"current_steps": 7450, "total_steps": 42420, "loss": 0.3446, "lr": 0.000982690304273623, "epoch": 3.5124941065535125, "percentage": 17.56, "elapsed_time": "0:24:04", "remaining_time": "1:52:59", "throughput": 5183.31, "total_tokens": 7485952} {"current_steps": 7455, "total_steps": 42420, "loss": 0.3444, "lr": 0.000982636602296663, "epoch": 3.514851485148515, "percentage": 17.57, "elapsed_time": "0:24:05", "remaining_time": "1:52:58", "throughput": 5183.54, "total_tokens": 7491456} {"current_steps": 7460, "total_steps": 42420, "loss": 0.3299, "lr": 0.0009825828186175208, "epoch": 3.5172088637435173, "percentage": 17.59, "elapsed_time": "0:24:06", "remaining_time": "1:52:56", "throughput": 5183.55, "total_tokens": 7495872} {"current_steps": 7465, "total_steps": 42420, "loss": 0.3988, "lr": 0.000982528953245301, "epoch": 3.5195662423385197, "percentage": 17.6, "elapsed_time": "0:24:06", "remaining_time": "1:52:55", "throughput": 5183.69, "total_tokens": 7500544} {"current_steps": 7470, "total_steps": 42420, "loss": 0.337, "lr": 0.0009824750061891226, "epoch": 3.521923620933522, "percentage": 17.61, "elapsed_time": "0:24:07", "remaining_time": "1:52:53", "throughput": 5183.91, "total_tokens": 7505376} {"current_steps": 7475, "total_steps": 42420, "loss": 0.3509, "lr": 0.0009824209774581174, "epoch": 3.524280999528524, "percentage": 17.62, "elapsed_time": "0:24:08", "remaining_time": "1:52:52", "throughput": 5184.12, "total_tokens": 7510720} {"current_steps": 7480, "total_steps": 42420, "loss": 0.3458, "lr": 0.000982366867061432, "epoch": 3.5266383781235264, "percentage": 17.63, "elapsed_time": "0:24:09", "remaining_time": "1:52:51", "throughput": 5184.3, "total_tokens": 7515680} {"current_steps": 7485, "total_steps": 42420, "loss": 0.3414, "lr": 0.000982312675008226, "epoch": 3.528995756718529, "percentage": 17.64, "elapsed_time": "0:24:10", "remaining_time": "1:52:50", "throughput": 5184.61, "total_tokens": 7521312} {"current_steps": 7490, "total_steps": 42420, "loss": 0.3185, "lr": 0.0009822584013076732, "epoch": 3.5313531353135312, "percentage": 17.66, "elapsed_time": "0:24:11", "remaining_time": "1:52:49", "throughput": 5184.86, "total_tokens": 7526560} {"current_steps": 7495, "total_steps": 42420, "loss": 0.3514, "lr": 0.0009822040459689617, "epoch": 3.5337105139085336, "percentage": 17.67, "elapsed_time": "0:24:12", "remaining_time": "1:52:48", "throughput": 5185.07, "total_tokens": 7531808} {"current_steps": 7500, "total_steps": 42420, "loss": 0.3423, "lr": 0.0009821496090012922, "epoch": 3.536067892503536, "percentage": 17.68, "elapsed_time": "0:24:13", "remaining_time": "1:52:47", "throughput": 5185.27, "total_tokens": 7536800} {"current_steps": 7505, "total_steps": 42420, "loss": 0.3733, "lr": 0.0009820950904138805, "epoch": 3.5384252710985384, "percentage": 17.69, "elapsed_time": "0:24:14", "remaining_time": "1:52:46", "throughput": 5185.6, "total_tokens": 7542496} {"current_steps": 7510, "total_steps": 42420, "loss": 0.3419, "lr": 0.0009820404902159555, "epoch": 3.540782649693541, "percentage": 17.7, "elapsed_time": "0:24:15", "remaining_time": "1:52:45", "throughput": 5185.82, "total_tokens": 7547968} {"current_steps": 7515, "total_steps": 42420, "loss": 0.3395, "lr": 0.00098198580841676, "epoch": 3.5431400282885432, "percentage": 17.72, "elapsed_time": "0:24:16", "remaining_time": "1:52:44", "throughput": 5185.98, "total_tokens": 7552960} {"current_steps": 7520, "total_steps": 42420, "loss": 0.35, "lr": 0.000981931045025551, "epoch": 3.5454974068835456, "percentage": 17.73, "elapsed_time": "0:24:17", "remaining_time": "1:52:43", "throughput": 5186.33, "total_tokens": 7558560} {"current_steps": 7525, "total_steps": 42420, "loss": 0.3437, "lr": 0.0009818762000515986, "epoch": 3.547854785478548, "percentage": 17.74, "elapsed_time": "0:24:18", "remaining_time": "1:52:42", "throughput": 5186.45, "total_tokens": 7563424} {"current_steps": 7530, "total_steps": 42420, "loss": 0.3264, "lr": 0.0009818212735041873, "epoch": 3.5502121640735504, "percentage": 17.75, "elapsed_time": "0:24:19", "remaining_time": "1:52:41", "throughput": 5186.66, "total_tokens": 7568704} {"current_steps": 7535, "total_steps": 42420, "loss": 0.3333, "lr": 0.0009817662653926155, "epoch": 3.5525695426685524, "percentage": 17.76, "elapsed_time": "0:24:20", "remaining_time": "1:52:40", "throughput": 5186.93, "total_tokens": 7574304} {"current_steps": 7540, "total_steps": 42420, "loss": 0.3685, "lr": 0.0009817111757261946, "epoch": 3.554926921263555, "percentage": 17.77, "elapsed_time": "0:24:21", "remaining_time": "1:52:39", "throughput": 5187.13, "total_tokens": 7579232} {"current_steps": 7545, "total_steps": 42420, "loss": 0.359, "lr": 0.0009816560045142508, "epoch": 3.557284299858557, "percentage": 17.79, "elapsed_time": "0:24:22", "remaining_time": "1:52:37", "throughput": 5187.15, "total_tokens": 7583744} {"current_steps": 7550, "total_steps": 42420, "loss": 0.3641, "lr": 0.0009816007517661233, "epoch": 3.5596416784535596, "percentage": 17.8, "elapsed_time": "0:24:22", "remaining_time": "1:52:36", "throughput": 5187.19, "total_tokens": 7588160} {"current_steps": 7555, "total_steps": 42420, "loss": 0.3539, "lr": 0.0009815454174911659, "epoch": 3.561999057048562, "percentage": 17.81, "elapsed_time": "0:24:23", "remaining_time": "1:52:34", "throughput": 5187.34, "total_tokens": 7592960} {"current_steps": 7560, "total_steps": 42420, "loss": 0.3569, "lr": 0.0009814900016987453, "epoch": 3.5643564356435644, "percentage": 17.82, "elapsed_time": "0:24:24", "remaining_time": "1:52:33", "throughput": 5187.33, "total_tokens": 7597248} {"current_steps": 7565, "total_steps": 42420, "loss": 0.3552, "lr": 0.0009814345043982428, "epoch": 3.566713814238567, "percentage": 17.83, "elapsed_time": "0:24:25", "remaining_time": "1:52:32", "throughput": 5187.53, "total_tokens": 7602368} {"current_steps": 7570, "total_steps": 42420, "loss": 0.3418, "lr": 0.000981378925599053, "epoch": 3.569071192833569, "percentage": 17.85, "elapsed_time": "0:24:26", "remaining_time": "1:52:30", "throughput": 5187.48, "total_tokens": 7606464} {"current_steps": 7575, "total_steps": 42420, "loss": 0.33, "lr": 0.0009813232653105844, "epoch": 3.571428571428571, "percentage": 17.86, "elapsed_time": "0:24:27", "remaining_time": "1:52:29", "throughput": 5187.62, "total_tokens": 7611456} {"current_steps": 7580, "total_steps": 42420, "loss": 0.4039, "lr": 0.0009812675235422593, "epoch": 3.5737859500235736, "percentage": 17.87, "elapsed_time": "0:24:28", "remaining_time": "1:52:28", "throughput": 5187.9, "total_tokens": 7617024} {"current_steps": 7585, "total_steps": 42420, "loss": 0.3465, "lr": 0.000981211700303514, "epoch": 3.576143328618576, "percentage": 17.88, "elapsed_time": "0:24:29", "remaining_time": "1:52:26", "throughput": 5187.91, "total_tokens": 7621440} {"current_steps": 7590, "total_steps": 42420, "loss": 0.3768, "lr": 0.0009811557956037984, "epoch": 3.5785007072135784, "percentage": 17.89, "elapsed_time": "0:24:30", "remaining_time": "1:52:26", "throughput": 5188.19, "total_tokens": 7626944} {"current_steps": 7595, "total_steps": 42420, "loss": 0.4097, "lr": 0.000981099809452576, "epoch": 3.580858085808581, "percentage": 17.9, "elapsed_time": "0:24:30", "remaining_time": "1:52:24", "throughput": 5188.35, "total_tokens": 7632000} {"current_steps": 7600, "total_steps": 42420, "loss": 0.3298, "lr": 0.0009810437418593246, "epoch": 3.583215464403583, "percentage": 17.92, "elapsed_time": "0:24:31", "remaining_time": "1:52:23", "throughput": 5188.43, "total_tokens": 7636576} {"current_steps": 7605, "total_steps": 42420, "loss": 0.4197, "lr": 0.0009809875928335353, "epoch": 3.5855728429985856, "percentage": 17.93, "elapsed_time": "0:24:32", "remaining_time": "1:52:23", "throughput": 5188.82, "total_tokens": 7643008} {"current_steps": 7610, "total_steps": 42420, "loss": 0.3889, "lr": 0.0009809313623847132, "epoch": 3.587930221593588, "percentage": 17.94, "elapsed_time": "0:24:34", "remaining_time": "1:52:22", "throughput": 5189.41, "total_tokens": 7649664} {"current_steps": 7615, "total_steps": 42420, "loss": 0.2961, "lr": 0.0009808750505223773, "epoch": 3.5902876001885904, "percentage": 17.95, "elapsed_time": "0:24:34", "remaining_time": "1:52:21", "throughput": 5189.28, "total_tokens": 7653600} {"current_steps": 7620, "total_steps": 42420, "loss": 0.3154, "lr": 0.0009808186572560602, "epoch": 3.592644978783593, "percentage": 17.96, "elapsed_time": "0:24:35", "remaining_time": "1:52:19", "throughput": 5189.16, "total_tokens": 7657632} {"current_steps": 7625, "total_steps": 42420, "loss": 0.3228, "lr": 0.000980762182595308, "epoch": 3.595002357378595, "percentage": 17.98, "elapsed_time": "0:24:36", "remaining_time": "1:52:18", "throughput": 5189.37, "total_tokens": 7662912} {"current_steps": 7630, "total_steps": 42420, "loss": 0.3998, "lr": 0.0009807056265496814, "epoch": 3.5973597359735976, "percentage": 17.99, "elapsed_time": "0:24:37", "remaining_time": "1:52:17", "throughput": 5189.55, "total_tokens": 7667936} {"current_steps": 7635, "total_steps": 42420, "loss": 0.3786, "lr": 0.0009806489891287541, "epoch": 3.5997171145685996, "percentage": 18.0, "elapsed_time": "0:24:38", "remaining_time": "1:52:16", "throughput": 5189.73, "total_tokens": 7673088} {"current_steps": 7640, "total_steps": 42420, "loss": 0.3812, "lr": 0.000980592270342114, "epoch": 3.602074493163602, "percentage": 18.01, "elapsed_time": "0:24:39", "remaining_time": "1:52:14", "throughput": 5189.89, "total_tokens": 7678144} {"current_steps": 7645, "total_steps": 42420, "loss": 0.3733, "lr": 0.0009805354701993623, "epoch": 3.6044318717586044, "percentage": 18.02, "elapsed_time": "0:24:40", "remaining_time": "1:52:14", "throughput": 5190.11, "total_tokens": 7683808} {"current_steps": 7650, "total_steps": 42420, "loss": 0.2889, "lr": 0.0009804785887101146, "epoch": 3.6067892503536068, "percentage": 18.03, "elapsed_time": "0:24:41", "remaining_time": "1:52:12", "throughput": 5190.26, "total_tokens": 7688704} {"current_steps": 7655, "total_steps": 42420, "loss": 0.3868, "lr": 0.000980421625884, "epoch": 3.609146628948609, "percentage": 18.05, "elapsed_time": "0:24:42", "remaining_time": "1:52:11", "throughput": 5190.34, "total_tokens": 7693632} {"current_steps": 7660, "total_steps": 42420, "loss": 0.2957, "lr": 0.000980364581730661, "epoch": 3.6115040075436116, "percentage": 18.06, "elapsed_time": "0:24:43", "remaining_time": "1:52:10", "throughput": 5190.22, "total_tokens": 7697696} {"current_steps": 7665, "total_steps": 42420, "loss": 0.3356, "lr": 0.0009803074562597547, "epoch": 3.613861386138614, "percentage": 18.07, "elapsed_time": "0:24:43", "remaining_time": "1:52:08", "throughput": 5190.16, "total_tokens": 7701824} {"current_steps": 7670, "total_steps": 42420, "loss": 0.3274, "lr": 0.000980250249480951, "epoch": 3.6162187647336164, "percentage": 18.08, "elapsed_time": "0:24:44", "remaining_time": "1:52:07", "throughput": 5190.51, "total_tokens": 7707456} {"current_steps": 7675, "total_steps": 42420, "loss": 0.3426, "lr": 0.0009801929614039345, "epoch": 3.6185761433286183, "percentage": 18.09, "elapsed_time": "0:24:45", "remaining_time": "1:52:06", "throughput": 5190.85, "total_tokens": 7712992} {"current_steps": 7680, "total_steps": 42420, "loss": 0.2793, "lr": 0.0009801355920384025, "epoch": 3.6209335219236207, "percentage": 18.1, "elapsed_time": "0:24:46", "remaining_time": "1:52:05", "throughput": 5190.9, "total_tokens": 7717664} {"current_steps": 7685, "total_steps": 42420, "loss": 0.366, "lr": 0.0009800781413940674, "epoch": 3.623290900518623, "percentage": 18.12, "elapsed_time": "0:24:47", "remaining_time": "1:52:03", "throughput": 5190.95, "total_tokens": 7722208} {"current_steps": 7690, "total_steps": 42420, "loss": 0.3822, "lr": 0.000980020609480654, "epoch": 3.6256482791136255, "percentage": 18.13, "elapsed_time": "0:24:48", "remaining_time": "1:52:02", "throughput": 5191.23, "total_tokens": 7727648} {"current_steps": 7695, "total_steps": 42420, "loss": 0.2586, "lr": 0.0009799629963079017, "epoch": 3.628005657708628, "percentage": 18.14, "elapsed_time": "0:24:49", "remaining_time": "1:52:01", "throughput": 5191.26, "total_tokens": 7732128} {"current_steps": 7700, "total_steps": 42420, "loss": 0.3137, "lr": 0.0009799053018855633, "epoch": 3.6303630363036303, "percentage": 18.15, "elapsed_time": "0:24:50", "remaining_time": "1:52:00", "throughput": 5191.47, "total_tokens": 7737344} {"current_steps": 7705, "total_steps": 42420, "loss": 0.4099, "lr": 0.0009798475262234058, "epoch": 3.6327204148986327, "percentage": 18.16, "elapsed_time": "0:24:51", "remaining_time": "1:51:58", "throughput": 5191.61, "total_tokens": 7741984} {"current_steps": 7710, "total_steps": 42420, "loss": 0.3436, "lr": 0.0009797896693312096, "epoch": 3.635077793493635, "percentage": 18.18, "elapsed_time": "0:24:52", "remaining_time": "1:51:58", "throughput": 5191.89, "total_tokens": 7747584} {"current_steps": 7715, "total_steps": 42420, "loss": 0.3464, "lr": 0.0009797317312187684, "epoch": 3.6374351720886375, "percentage": 18.19, "elapsed_time": "0:24:53", "remaining_time": "1:51:57", "throughput": 5192.31, "total_tokens": 7753504} {"current_steps": 7720, "total_steps": 42420, "loss": 0.3666, "lr": 0.0009796737118958907, "epoch": 3.63979255068364, "percentage": 18.2, "elapsed_time": "0:24:54", "remaining_time": "1:51:55", "throughput": 5192.24, "total_tokens": 7757600} {"current_steps": 7725, "total_steps": 42420, "loss": 0.3437, "lr": 0.0009796156113723979, "epoch": 3.6421499292786423, "percentage": 18.21, "elapsed_time": "0:24:54", "remaining_time": "1:51:54", "throughput": 5192.41, "total_tokens": 7762496} {"current_steps": 7730, "total_steps": 42420, "loss": 0.3635, "lr": 0.0009795574296581255, "epoch": 3.6445073078736447, "percentage": 18.22, "elapsed_time": "0:24:55", "remaining_time": "1:51:53", "throughput": 5192.42, "total_tokens": 7767168} {"current_steps": 7735, "total_steps": 42420, "loss": 0.257, "lr": 0.000979499166762923, "epoch": 3.6468646864686467, "percentage": 18.23, "elapsed_time": "0:24:56", "remaining_time": "1:51:51", "throughput": 5192.53, "total_tokens": 7771680} {"current_steps": 7740, "total_steps": 42420, "loss": 0.3964, "lr": 0.0009794408226966524, "epoch": 3.649222065063649, "percentage": 18.25, "elapsed_time": "0:24:57", "remaining_time": "1:51:50", "throughput": 5192.71, "total_tokens": 7776896} {"current_steps": 7745, "total_steps": 42420, "loss": 0.4223, "lr": 0.0009793823974691915, "epoch": 3.6515794436586515, "percentage": 18.26, "elapsed_time": "0:24:58", "remaining_time": "1:51:49", "throughput": 5193.05, "total_tokens": 7782496} {"current_steps": 7750, "total_steps": 42420, "loss": 0.4222, "lr": 0.0009793238910904298, "epoch": 3.653936822253654, "percentage": 18.27, "elapsed_time": "0:24:59", "remaining_time": "1:51:48", "throughput": 5193.41, "total_tokens": 7788192} {"current_steps": 7755, "total_steps": 42420, "loss": 0.3508, "lr": 0.000979265303570272, "epoch": 3.6562942008486563, "percentage": 18.28, "elapsed_time": "0:25:00", "remaining_time": "1:51:47", "throughput": 5193.51, "total_tokens": 7793088} {"current_steps": 7760, "total_steps": 42420, "loss": 0.2848, "lr": 0.0009792066349186357, "epoch": 3.6586515794436587, "percentage": 18.29, "elapsed_time": "0:25:01", "remaining_time": "1:51:45", "throughput": 5193.55, "total_tokens": 7797504} {"current_steps": 7765, "total_steps": 42420, "loss": 0.3653, "lr": 0.0009791478851454525, "epoch": 3.661008958038661, "percentage": 18.31, "elapsed_time": "0:25:02", "remaining_time": "1:51:44", "throughput": 5193.49, "total_tokens": 7801760} {"current_steps": 7770, "total_steps": 42420, "loss": 0.4535, "lr": 0.000979089054260668, "epoch": 3.6633663366336635, "percentage": 18.32, "elapsed_time": "0:25:03", "remaining_time": "1:51:43", "throughput": 5193.68, "total_tokens": 7806688} {"current_steps": 7775, "total_steps": 42420, "loss": 0.3489, "lr": 0.0009790301422742409, "epoch": 3.6657237152286655, "percentage": 18.33, "elapsed_time": "0:25:03", "remaining_time": "1:51:41", "throughput": 5193.52, "total_tokens": 7810656} {"current_steps": 7780, "total_steps": 42420, "loss": 0.3556, "lr": 0.0009789711491961443, "epoch": 3.668081093823668, "percentage": 18.34, "elapsed_time": "0:25:04", "remaining_time": "1:51:39", "throughput": 5193.4, "total_tokens": 7814848} {"current_steps": 7785, "total_steps": 42420, "loss": 0.3237, "lr": 0.0009789120750363645, "epoch": 3.6704384724186703, "percentage": 18.35, "elapsed_time": "0:25:05", "remaining_time": "1:51:38", "throughput": 5193.32, "total_tokens": 7818816} {"current_steps": 7790, "total_steps": 42420, "loss": 0.4129, "lr": 0.0009788529198049018, "epoch": 3.6727958510136727, "percentage": 18.36, "elapsed_time": "0:25:06", "remaining_time": "1:51:36", "throughput": 5193.37, "total_tokens": 7823392} {"current_steps": 7795, "total_steps": 42420, "loss": 0.3392, "lr": 0.0009787936835117706, "epoch": 3.675153229608675, "percentage": 18.38, "elapsed_time": "0:25:07", "remaining_time": "1:51:35", "throughput": 5193.7, "total_tokens": 7829120} {"current_steps": 7800, "total_steps": 42420, "loss": 0.3555, "lr": 0.0009787343661669977, "epoch": 3.6775106082036775, "percentage": 18.39, "elapsed_time": "0:25:08", "remaining_time": "1:51:34", "throughput": 5193.8, "total_tokens": 7833792} {"current_steps": 7805, "total_steps": 42420, "loss": 0.3541, "lr": 0.0009786749677806255, "epoch": 3.67986798679868, "percentage": 18.4, "elapsed_time": "0:25:09", "remaining_time": "1:51:33", "throughput": 5193.88, "total_tokens": 7838432} {"current_steps": 7810, "total_steps": 42420, "loss": 0.3434, "lr": 0.0009786154883627085, "epoch": 3.6822253653936823, "percentage": 18.41, "elapsed_time": "0:25:10", "remaining_time": "1:51:32", "throughput": 5194.24, "total_tokens": 7844192} {"current_steps": 7815, "total_steps": 42420, "loss": 0.3516, "lr": 0.0009785559279233158, "epoch": 3.6845827439886847, "percentage": 18.42, "elapsed_time": "0:25:11", "remaining_time": "1:51:31", "throughput": 5194.43, "total_tokens": 7849280} {"current_steps": 7820, "total_steps": 42420, "loss": 0.3563, "lr": 0.0009784962864725299, "epoch": 3.686940122583687, "percentage": 18.43, "elapsed_time": "0:25:11", "remaining_time": "1:51:29", "throughput": 5194.56, "total_tokens": 7853920} {"current_steps": 7825, "total_steps": 42420, "loss": 0.3468, "lr": 0.0009784365640204472, "epoch": 3.6892975011786895, "percentage": 18.45, "elapsed_time": "0:25:12", "remaining_time": "1:51:28", "throughput": 5194.54, "total_tokens": 7858144} {"current_steps": 7830, "total_steps": 42420, "loss": 0.32, "lr": 0.0009783767605771776, "epoch": 3.691654879773692, "percentage": 18.46, "elapsed_time": "0:25:13", "remaining_time": "1:51:26", "throughput": 5194.73, "total_tokens": 7863072} {"current_steps": 7835, "total_steps": 42420, "loss": 0.2919, "lr": 0.0009783168761528446, "epoch": 3.694012258368694, "percentage": 18.47, "elapsed_time": "0:25:14", "remaining_time": "1:51:26", "throughput": 5195.05, "total_tokens": 7868800} {"current_steps": 7840, "total_steps": 42420, "loss": 0.3478, "lr": 0.0009782569107575861, "epoch": 3.6963696369636962, "percentage": 18.48, "elapsed_time": "0:25:15", "remaining_time": "1:51:24", "throughput": 5195.19, "total_tokens": 7873760} {"current_steps": 7845, "total_steps": 42420, "loss": 0.4165, "lr": 0.0009781968644015527, "epoch": 3.6987270155586986, "percentage": 18.49, "elapsed_time": "0:25:16", "remaining_time": "1:51:23", "throughput": 5195.18, "total_tokens": 7877920} {"current_steps": 7850, "total_steps": 42420, "loss": 0.3365, "lr": 0.0009781367370949095, "epoch": 3.701084394153701, "percentage": 18.51, "elapsed_time": "0:25:17", "remaining_time": "1:51:22", "throughput": 5195.36, "total_tokens": 7883104} {"current_steps": 7855, "total_steps": 42420, "loss": 0.3476, "lr": 0.0009780765288478351, "epoch": 3.7034417727487035, "percentage": 18.52, "elapsed_time": "0:25:18", "remaining_time": "1:51:20", "throughput": 5195.6, "total_tokens": 7888192} {"current_steps": 7860, "total_steps": 42420, "loss": 0.4207, "lr": 0.0009780162396705217, "epoch": 3.705799151343706, "percentage": 18.53, "elapsed_time": "0:25:19", "remaining_time": "1:51:19", "throughput": 5195.77, "total_tokens": 7893248} {"current_steps": 7865, "total_steps": 42420, "loss": 0.3404, "lr": 0.0009779558695731752, "epoch": 3.7081565299387083, "percentage": 18.54, "elapsed_time": "0:25:19", "remaining_time": "1:51:18", "throughput": 5195.74, "total_tokens": 7897472} {"current_steps": 7870, "total_steps": 42420, "loss": 0.3438, "lr": 0.0009778954185660149, "epoch": 3.7105139085337107, "percentage": 18.55, "elapsed_time": "0:25:20", "remaining_time": "1:51:16", "throughput": 5195.94, "total_tokens": 7902464} {"current_steps": 7875, "total_steps": 42420, "loss": 0.4336, "lr": 0.0009778348866592748, "epoch": 3.7128712871287126, "percentage": 18.56, "elapsed_time": "0:25:21", "remaining_time": "1:51:15", "throughput": 5195.99, "total_tokens": 7907200} {"current_steps": 7880, "total_steps": 42420, "loss": 0.3462, "lr": 0.0009777742738632015, "epoch": 3.715228665723715, "percentage": 18.58, "elapsed_time": "0:25:22", "remaining_time": "1:51:14", "throughput": 5196.36, "total_tokens": 7912992} {"current_steps": 7885, "total_steps": 42420, "loss": 0.3448, "lr": 0.0009777135801880556, "epoch": 3.7175860443187174, "percentage": 18.59, "elapsed_time": "0:25:23", "remaining_time": "1:51:13", "throughput": 5196.45, "total_tokens": 7917760} {"current_steps": 7890, "total_steps": 42420, "loss": 0.4166, "lr": 0.0009776528056441118, "epoch": 3.71994342291372, "percentage": 18.6, "elapsed_time": "0:25:24", "remaining_time": "1:51:13", "throughput": 5196.87, "total_tokens": 7924096} {"current_steps": 7895, "total_steps": 42420, "loss": 0.3664, "lr": 0.000977591950241658, "epoch": 3.7223008015087222, "percentage": 18.61, "elapsed_time": "0:25:25", "remaining_time": "1:51:11", "throughput": 5197.12, "total_tokens": 7929344} {"current_steps": 7900, "total_steps": 42420, "loss": 0.3323, "lr": 0.000977531013990996, "epoch": 3.7246581801037246, "percentage": 18.62, "elapsed_time": "0:25:26", "remaining_time": "1:51:10", "throughput": 5197.13, "total_tokens": 7933664} {"current_steps": 7905, "total_steps": 42420, "loss": 0.3976, "lr": 0.0009774699969024416, "epoch": 3.727015558698727, "percentage": 18.64, "elapsed_time": "0:25:27", "remaining_time": "1:51:09", "throughput": 5197.39, "total_tokens": 7939040} {"current_steps": 7910, "total_steps": 42420, "loss": 0.3895, "lr": 0.0009774088989863237, "epoch": 3.7293729372937294, "percentage": 18.65, "elapsed_time": "0:25:28", "remaining_time": "1:51:08", "throughput": 5197.53, "total_tokens": 7943968} {"current_steps": 7915, "total_steps": 42420, "loss": 0.368, "lr": 0.000977347720252985, "epoch": 3.731730315888732, "percentage": 18.66, "elapsed_time": "0:25:29", "remaining_time": "1:51:07", "throughput": 5197.78, "total_tokens": 7949312} {"current_steps": 7920, "total_steps": 42420, "loss": 0.32, "lr": 0.0009772864607127823, "epoch": 3.7340876944837342, "percentage": 18.67, "elapsed_time": "0:25:30", "remaining_time": "1:51:06", "throughput": 5198.03, "total_tokens": 7954656} {"current_steps": 7925, "total_steps": 42420, "loss": 0.3419, "lr": 0.0009772251203760856, "epoch": 3.7364450730787366, "percentage": 18.68, "elapsed_time": "0:25:31", "remaining_time": "1:51:04", "throughput": 5198.18, "total_tokens": 7959520} {"current_steps": 7930, "total_steps": 42420, "loss": 0.2742, "lr": 0.0009771636992532788, "epoch": 3.738802451673739, "percentage": 18.69, "elapsed_time": "0:25:32", "remaining_time": "1:51:03", "throughput": 5198.27, "total_tokens": 7964064} {"current_steps": 7935, "total_steps": 42420, "loss": 0.3408, "lr": 0.0009771021973547594, "epoch": 3.741159830268741, "percentage": 18.71, "elapsed_time": "0:25:32", "remaining_time": "1:51:01", "throughput": 5198.31, "total_tokens": 7968608} {"current_steps": 7940, "total_steps": 42420, "loss": 0.3572, "lr": 0.0009770406146909388, "epoch": 3.7435172088637434, "percentage": 18.72, "elapsed_time": "0:25:33", "remaining_time": "1:51:00", "throughput": 5198.4, "total_tokens": 7973312} {"current_steps": 7945, "total_steps": 42420, "loss": 0.2982, "lr": 0.0009769789512722418, "epoch": 3.745874587458746, "percentage": 18.73, "elapsed_time": "0:25:34", "remaining_time": "1:51:00", "throughput": 5198.8, "total_tokens": 7979552} {"current_steps": 7950, "total_steps": 42420, "loss": 0.3641, "lr": 0.000976917207109107, "epoch": 3.748231966053748, "percentage": 18.74, "elapsed_time": "0:25:35", "remaining_time": "1:50:58", "throughput": 5198.89, "total_tokens": 7984448} {"current_steps": 7955, "total_steps": 42420, "loss": 0.3447, "lr": 0.0009768553822119869, "epoch": 3.7505893446487506, "percentage": 18.75, "elapsed_time": "0:25:36", "remaining_time": "1:50:58", "throughput": 5199.18, "total_tokens": 7989920} {"current_steps": 7960, "total_steps": 42420, "loss": 0.3526, "lr": 0.0009767934765913468, "epoch": 3.752946723243753, "percentage": 18.76, "elapsed_time": "0:25:37", "remaining_time": "1:50:56", "throughput": 5199.35, "total_tokens": 7995072} {"current_steps": 7965, "total_steps": 42420, "loss": 0.309, "lr": 0.0009767314902576667, "epoch": 3.7553041018387554, "percentage": 18.78, "elapsed_time": "0:25:38", "remaining_time": "1:50:55", "throughput": 5199.51, "total_tokens": 7999968} {"current_steps": 7970, "total_steps": 42420, "loss": 0.3758, "lr": 0.0009766694232214398, "epoch": 3.757661480433758, "percentage": 18.79, "elapsed_time": "0:25:39", "remaining_time": "1:50:54", "throughput": 5199.62, "total_tokens": 8004928} {"current_steps": 7975, "total_steps": 42420, "loss": 0.2918, "lr": 0.0009766072754931728, "epoch": 3.7600188590287598, "percentage": 18.8, "elapsed_time": "0:25:40", "remaining_time": "1:50:53", "throughput": 5199.99, "total_tokens": 8011040} {"current_steps": 7980, "total_steps": 42420, "loss": 0.3779, "lr": 0.0009765450470833866, "epoch": 3.762376237623762, "percentage": 18.81, "elapsed_time": "0:25:41", "remaining_time": "1:50:52", "throughput": 5200.26, "total_tokens": 8016320} {"current_steps": 7985, "total_steps": 42420, "loss": 0.3734, "lr": 0.0009764827380026152, "epoch": 3.7647336162187646, "percentage": 18.82, "elapsed_time": "0:25:42", "remaining_time": "1:50:52", "throughput": 5200.7, "total_tokens": 8022496} {"current_steps": 7990, "total_steps": 42420, "loss": 0.3723, "lr": 0.0009764203482614062, "epoch": 3.767090994813767, "percentage": 18.84, "elapsed_time": "0:25:43", "remaining_time": "1:50:51", "throughput": 5200.9, "total_tokens": 8027552} {"current_steps": 7995, "total_steps": 42420, "loss": 0.3589, "lr": 0.0009763578778703215, "epoch": 3.7694483734087694, "percentage": 18.85, "elapsed_time": "0:25:44", "remaining_time": "1:50:50", "throughput": 5201.25, "total_tokens": 8033152} {"current_steps": 8000, "total_steps": 42420, "loss": 0.3575, "lr": 0.0009762953268399363, "epoch": 3.7718057520037718, "percentage": 18.86, "elapsed_time": "0:25:45", "remaining_time": "1:50:48", "throughput": 5201.41, "total_tokens": 8038016} {"current_steps": 8005, "total_steps": 42420, "loss": 0.3258, "lr": 0.0009762326951808392, "epoch": 3.774163130598774, "percentage": 18.87, "elapsed_time": "0:25:46", "remaining_time": "1:50:47", "throughput": 5201.47, "total_tokens": 8042528} {"current_steps": 8010, "total_steps": 42420, "loss": 0.4456, "lr": 0.0009761699829036328, "epoch": 3.7765205091937766, "percentage": 18.88, "elapsed_time": "0:25:47", "remaining_time": "1:50:46", "throughput": 5201.64, "total_tokens": 8047456} {"current_steps": 8015, "total_steps": 42420, "loss": 0.4469, "lr": 0.0009761071900189331, "epoch": 3.778877887788779, "percentage": 18.89, "elapsed_time": "0:25:47", "remaining_time": "1:50:44", "throughput": 5201.67, "total_tokens": 8051872} {"current_steps": 8020, "total_steps": 42420, "loss": 0.3565, "lr": 0.0009760443165373701, "epoch": 3.7812352663837814, "percentage": 18.91, "elapsed_time": "0:25:48", "remaining_time": "1:50:43", "throughput": 5201.9, "total_tokens": 8057056} {"current_steps": 8025, "total_steps": 42420, "loss": 0.3489, "lr": 0.000975981362469587, "epoch": 3.783592644978784, "percentage": 18.92, "elapsed_time": "0:25:49", "remaining_time": "1:50:42", "throughput": 5202.09, "total_tokens": 8062176} {"current_steps": 8030, "total_steps": 42420, "loss": 0.381, "lr": 0.0009759183278262411, "epoch": 3.785950023573786, "percentage": 18.93, "elapsed_time": "0:25:50", "remaining_time": "1:50:42", "throughput": 5202.72, "total_tokens": 8069344} {"current_steps": 8035, "total_steps": 42420, "loss": 0.3602, "lr": 0.0009758552126180028, "epoch": 3.7883074021687886, "percentage": 18.94, "elapsed_time": "0:25:51", "remaining_time": "1:50:40", "throughput": 5202.69, "total_tokens": 8073600} {"current_steps": 8040, "total_steps": 42420, "loss": 0.3275, "lr": 0.0009757920168555565, "epoch": 3.7906647807637905, "percentage": 18.95, "elapsed_time": "0:25:52", "remaining_time": "1:50:39", "throughput": 5202.89, "total_tokens": 8078720} {"current_steps": 8045, "total_steps": 42420, "loss": 0.3174, "lr": 0.0009757287405496005, "epoch": 3.793022159358793, "percentage": 18.97, "elapsed_time": "0:25:53", "remaining_time": "1:50:38", "throughput": 5202.96, "total_tokens": 8083520} {"current_steps": 8050, "total_steps": 42420, "loss": 0.402, "lr": 0.0009756653837108458, "epoch": 3.7953795379537953, "percentage": 18.98, "elapsed_time": "0:25:54", "remaining_time": "1:50:37", "throughput": 5203.01, "total_tokens": 8088416} {"current_steps": 8055, "total_steps": 42420, "loss": 0.3082, "lr": 0.0009756019463500182, "epoch": 3.7977369165487977, "percentage": 18.99, "elapsed_time": "0:25:55", "remaining_time": "1:50:35", "throughput": 5202.98, "total_tokens": 8092704} {"current_steps": 8060, "total_steps": 42420, "loss": 0.3851, "lr": 0.0009755384284778564, "epoch": 3.8000942951438, "percentage": 19.0, "elapsed_time": "0:25:56", "remaining_time": "1:50:35", "throughput": 5203.31, "total_tokens": 8098528} {"current_steps": 8065, "total_steps": 42420, "loss": 0.3411, "lr": 0.0009754748301051126, "epoch": 3.8024516737388026, "percentage": 19.01, "elapsed_time": "0:25:57", "remaining_time": "1:50:34", "throughput": 5203.63, "total_tokens": 8104320} {"current_steps": 8070, "total_steps": 42420, "loss": 0.3429, "lr": 0.0009754111512425533, "epoch": 3.804809052333805, "percentage": 19.02, "elapsed_time": "0:25:58", "remaining_time": "1:50:32", "throughput": 5203.65, "total_tokens": 8108736} {"current_steps": 8075, "total_steps": 42420, "loss": 0.3569, "lr": 0.0009753473919009581, "epoch": 3.807166430928807, "percentage": 19.04, "elapsed_time": "0:25:59", "remaining_time": "1:50:31", "throughput": 5203.62, "total_tokens": 8113024} {"current_steps": 8080, "total_steps": 42420, "loss": 0.3192, "lr": 0.0009752835520911204, "epoch": 3.8095238095238093, "percentage": 19.05, "elapsed_time": "0:26:00", "remaining_time": "1:50:30", "throughput": 5203.96, "total_tokens": 8119040} {"current_steps": 8085, "total_steps": 42420, "loss": 0.3113, "lr": 0.0009752196318238472, "epoch": 3.8118811881188117, "percentage": 19.06, "elapsed_time": "0:26:00", "remaining_time": "1:50:29", "throughput": 5204.02, "total_tokens": 8123456} {"current_steps": 8090, "total_steps": 42420, "loss": 0.3141, "lr": 0.0009751556311099591, "epoch": 3.814238566713814, "percentage": 19.07, "elapsed_time": "0:26:01", "remaining_time": "1:50:28", "throughput": 5204.27, "total_tokens": 8128736} {"current_steps": 8095, "total_steps": 42420, "loss": 0.378, "lr": 0.0009750915499602903, "epoch": 3.8165959453088165, "percentage": 19.08, "elapsed_time": "0:26:02", "remaining_time": "1:50:27", "throughput": 5204.39, "total_tokens": 8133888} {"current_steps": 8100, "total_steps": 42420, "loss": 0.3484, "lr": 0.0009750273883856887, "epoch": 3.818953323903819, "percentage": 19.09, "elapsed_time": "0:26:03", "remaining_time": "1:50:26", "throughput": 5204.61, "total_tokens": 8139264} {"current_steps": 8105, "total_steps": 42420, "loss": 0.3383, "lr": 0.0009749631463970157, "epoch": 3.8213107024988213, "percentage": 19.11, "elapsed_time": "0:26:04", "remaining_time": "1:50:25", "throughput": 5204.91, "total_tokens": 8144864} {"current_steps": 8110, "total_steps": 42420, "loss": 0.3382, "lr": 0.0009748988240051465, "epoch": 3.8236680810938237, "percentage": 19.12, "elapsed_time": "0:26:05", "remaining_time": "1:50:23", "throughput": 5204.88, "total_tokens": 8149152} {"current_steps": 8115, "total_steps": 42420, "loss": 0.336, "lr": 0.0009748344212209697, "epoch": 3.826025459688826, "percentage": 19.13, "elapsed_time": "0:26:06", "remaining_time": "1:50:22", "throughput": 5204.9, "total_tokens": 8153760} {"current_steps": 8120, "total_steps": 42420, "loss": 0.3854, "lr": 0.0009747699380553876, "epoch": 3.8283828382838285, "percentage": 19.14, "elapsed_time": "0:26:07", "remaining_time": "1:50:23", "throughput": 5205.69, "total_tokens": 8162240} {"current_steps": 8125, "total_steps": 42420, "loss": 0.3524, "lr": 0.0009747053745193161, "epoch": 3.830740216878831, "percentage": 19.15, "elapsed_time": "0:26:08", "remaining_time": "1:50:22", "throughput": 5205.97, "total_tokens": 8168096} {"current_steps": 8130, "total_steps": 42420, "loss": 0.3487, "lr": 0.0009746407306236848, "epoch": 3.8330975954738333, "percentage": 19.17, "elapsed_time": "0:26:09", "remaining_time": "1:50:21", "throughput": 5206.05, "total_tokens": 8172800} {"current_steps": 8135, "total_steps": 42420, "loss": 0.3553, "lr": 0.0009745760063794369, "epoch": 3.8354549740688357, "percentage": 19.18, "elapsed_time": "0:26:10", "remaining_time": "1:50:19", "throughput": 5206.11, "total_tokens": 8177504} {"current_steps": 8140, "total_steps": 42420, "loss": 0.3242, "lr": 0.0009745112017975287, "epoch": 3.8378123526638377, "percentage": 19.19, "elapsed_time": "0:26:11", "remaining_time": "1:50:19", "throughput": 5206.37, "total_tokens": 8183200} {"current_steps": 8145, "total_steps": 42420, "loss": 0.2988, "lr": 0.0009744463168889308, "epoch": 3.84016973125884, "percentage": 19.2, "elapsed_time": "0:26:12", "remaining_time": "1:50:17", "throughput": 5206.39, "total_tokens": 8187456} {"current_steps": 8150, "total_steps": 42420, "loss": 0.4251, "lr": 0.0009743813516646273, "epoch": 3.8425271098538425, "percentage": 19.21, "elapsed_time": "0:26:13", "remaining_time": "1:50:16", "throughput": 5206.77, "total_tokens": 8193440} {"current_steps": 8155, "total_steps": 42420, "loss": 0.3386, "lr": 0.0009743163061356153, "epoch": 3.844884488448845, "percentage": 19.22, "elapsed_time": "0:26:14", "remaining_time": "1:50:15", "throughput": 5206.96, "total_tokens": 8198560} {"current_steps": 8160, "total_steps": 42420, "loss": 0.4157, "lr": 0.0009742511803129062, "epoch": 3.8472418670438473, "percentage": 19.24, "elapsed_time": "0:26:15", "remaining_time": "1:50:14", "throughput": 5207.12, "total_tokens": 8203424} {"current_steps": 8165, "total_steps": 42420, "loss": 0.3486, "lr": 0.0009741859742075246, "epoch": 3.8495992456388497, "percentage": 19.25, "elapsed_time": "0:26:16", "remaining_time": "1:50:13", "throughput": 5207.33, "total_tokens": 8208832} {"current_steps": 8170, "total_steps": 42420, "loss": 0.437, "lr": 0.0009741206878305087, "epoch": 3.851956624233852, "percentage": 19.26, "elapsed_time": "0:26:17", "remaining_time": "1:50:12", "throughput": 5207.47, "total_tokens": 8213504} {"current_steps": 8175, "total_steps": 42420, "loss": 0.3409, "lr": 0.0009740553211929104, "epoch": 3.854314002828854, "percentage": 19.27, "elapsed_time": "0:26:18", "remaining_time": "1:50:11", "throughput": 5207.69, "total_tokens": 8218816} {"current_steps": 8180, "total_steps": 42420, "loss": 0.3483, "lr": 0.0009739898743057954, "epoch": 3.8566713814238565, "percentage": 19.28, "elapsed_time": "0:26:19", "remaining_time": "1:50:10", "throughput": 5207.86, "total_tokens": 8223968} {"current_steps": 8185, "total_steps": 42420, "loss": 0.354, "lr": 0.0009739243471802426, "epoch": 3.859028760018859, "percentage": 19.3, "elapsed_time": "0:26:20", "remaining_time": "1:50:09", "throughput": 5208.02, "total_tokens": 8229408} {"current_steps": 8190, "total_steps": 42420, "loss": 0.3671, "lr": 0.0009738587398273444, "epoch": 3.8613861386138613, "percentage": 19.31, "elapsed_time": "0:26:20", "remaining_time": "1:50:07", "throughput": 5207.94, "total_tokens": 8233440} {"current_steps": 8195, "total_steps": 42420, "loss": 0.3408, "lr": 0.0009737930522582072, "epoch": 3.8637435172088637, "percentage": 19.32, "elapsed_time": "0:26:21", "remaining_time": "1:50:05", "throughput": 5207.95, "total_tokens": 8237760} {"current_steps": 8200, "total_steps": 42420, "loss": 0.3104, "lr": 0.0009737272844839511, "epoch": 3.866100895803866, "percentage": 19.33, "elapsed_time": "0:26:22", "remaining_time": "1:50:04", "throughput": 5208.01, "total_tokens": 8242400} {"current_steps": 8205, "total_steps": 42420, "loss": 0.2869, "lr": 0.0009736614365157089, "epoch": 3.8684582743988685, "percentage": 19.34, "elapsed_time": "0:26:23", "remaining_time": "1:50:03", "throughput": 5208.15, "total_tokens": 8247520} {"current_steps": 8210, "total_steps": 42420, "loss": 0.3271, "lr": 0.0009735955083646279, "epoch": 3.870815652993871, "percentage": 19.35, "elapsed_time": "0:26:24", "remaining_time": "1:50:02", "throughput": 5208.23, "total_tokens": 8252192} {"current_steps": 8215, "total_steps": 42420, "loss": 0.3526, "lr": 0.0009735295000418685, "epoch": 3.8731730315888733, "percentage": 19.37, "elapsed_time": "0:26:25", "remaining_time": "1:50:01", "throughput": 5208.48, "total_tokens": 8258016} {"current_steps": 8220, "total_steps": 42420, "loss": 0.3011, "lr": 0.0009734634115586048, "epoch": 3.8755304101838757, "percentage": 19.38, "elapsed_time": "0:26:26", "remaining_time": "1:50:00", "throughput": 5208.6, "total_tokens": 8263136} {"current_steps": 8225, "total_steps": 42420, "loss": 0.4114, "lr": 0.0009733972429260245, "epoch": 3.877887788778878, "percentage": 19.39, "elapsed_time": "0:26:27", "remaining_time": "1:49:58", "throughput": 5208.54, "total_tokens": 8267200} {"current_steps": 8230, "total_steps": 42420, "loss": 0.3619, "lr": 0.0009733309941553289, "epoch": 3.8802451673738805, "percentage": 19.4, "elapsed_time": "0:26:28", "remaining_time": "1:49:57", "throughput": 5208.57, "total_tokens": 8271744} {"current_steps": 8235, "total_steps": 42420, "loss": 0.3239, "lr": 0.0009732646652577326, "epoch": 3.882602545968883, "percentage": 19.41, "elapsed_time": "0:26:29", "remaining_time": "1:49:56", "throughput": 5208.71, "total_tokens": 8276896} {"current_steps": 8240, "total_steps": 42420, "loss": 0.3913, "lr": 0.000973198256244464, "epoch": 3.884959924563885, "percentage": 19.42, "elapsed_time": "0:26:29", "remaining_time": "1:49:54", "throughput": 5208.68, "total_tokens": 8281216} {"current_steps": 8245, "total_steps": 42420, "loss": 0.4141, "lr": 0.0009731317671267652, "epoch": 3.8873173031588872, "percentage": 19.44, "elapsed_time": "0:26:30", "remaining_time": "1:49:53", "throughput": 5208.81, "total_tokens": 8286208} {"current_steps": 8250, "total_steps": 42420, "loss": 0.3709, "lr": 0.0009730651979158913, "epoch": 3.8896746817538896, "percentage": 19.45, "elapsed_time": "0:26:31", "remaining_time": "1:49:52", "throughput": 5209.07, "total_tokens": 8291648} {"current_steps": 8255, "total_steps": 42420, "loss": 0.2967, "lr": 0.0009729985486231119, "epoch": 3.892032060348892, "percentage": 19.46, "elapsed_time": "0:26:32", "remaining_time": "1:49:51", "throughput": 5209.11, "total_tokens": 8296128} {"current_steps": 8260, "total_steps": 42420, "loss": 0.5307, "lr": 0.0009729318192597091, "epoch": 3.8943894389438944, "percentage": 19.47, "elapsed_time": "0:26:33", "remaining_time": "1:49:49", "throughput": 5209.11, "total_tokens": 8300576} {"current_steps": 8265, "total_steps": 42420, "loss": 0.3762, "lr": 0.0009728650098369793, "epoch": 3.896746817538897, "percentage": 19.48, "elapsed_time": "0:26:34", "remaining_time": "1:49:49", "throughput": 5209.58, "total_tokens": 8307200} {"current_steps": 8270, "total_steps": 42420, "loss": 0.4111, "lr": 0.0009727981203662319, "epoch": 3.8991041961338992, "percentage": 19.5, "elapsed_time": "0:26:35", "remaining_time": "1:49:48", "throughput": 5209.81, "total_tokens": 8312800} {"current_steps": 8275, "total_steps": 42420, "loss": 0.4491, "lr": 0.0009727311508587906, "epoch": 3.901461574728901, "percentage": 19.51, "elapsed_time": "0:26:36", "remaining_time": "1:49:47", "throughput": 5209.89, "total_tokens": 8317728} {"current_steps": 8280, "total_steps": 42420, "loss": 0.3682, "lr": 0.000972664101325992, "epoch": 3.9038189533239036, "percentage": 19.52, "elapsed_time": "0:26:37", "remaining_time": "1:49:46", "throughput": 5210.14, "total_tokens": 8323008} {"current_steps": 8285, "total_steps": 42420, "loss": 0.4218, "lr": 0.0009725969717791863, "epoch": 3.906176331918906, "percentage": 19.53, "elapsed_time": "0:26:38", "remaining_time": "1:49:44", "throughput": 5210.01, "total_tokens": 8326784} {"current_steps": 8290, "total_steps": 42420, "loss": 0.3534, "lr": 0.0009725297622297375, "epoch": 3.9085337105139084, "percentage": 19.54, "elapsed_time": "0:26:38", "remaining_time": "1:49:43", "throughput": 5209.83, "total_tokens": 8330496} {"current_steps": 8295, "total_steps": 42420, "loss": 0.2871, "lr": 0.0009724624726890232, "epoch": 3.910891089108911, "percentage": 19.55, "elapsed_time": "0:26:40", "remaining_time": "1:49:42", "throughput": 5210.16, "total_tokens": 8336576} {"current_steps": 8300, "total_steps": 42420, "loss": 0.352, "lr": 0.0009723951031684341, "epoch": 3.913248467703913, "percentage": 19.57, "elapsed_time": "0:26:40", "remaining_time": "1:49:40", "throughput": 5210.02, "total_tokens": 8340352} {"current_steps": 8305, "total_steps": 42420, "loss": 0.3562, "lr": 0.000972327653679375, "epoch": 3.9156058462989156, "percentage": 19.58, "elapsed_time": "0:26:41", "remaining_time": "1:49:39", "throughput": 5210.08, "total_tokens": 8345056} {"current_steps": 8310, "total_steps": 42420, "loss": 0.3356, "lr": 0.0009722601242332636, "epoch": 3.917963224893918, "percentage": 19.59, "elapsed_time": "0:26:42", "remaining_time": "1:49:39", "throughput": 5210.49, "total_tokens": 8351680} {"current_steps": 8315, "total_steps": 42420, "loss": 0.3625, "lr": 0.0009721925148415318, "epoch": 3.9203206034889204, "percentage": 19.6, "elapsed_time": "0:26:43", "remaining_time": "1:49:38", "throughput": 5210.62, "total_tokens": 8356768} {"current_steps": 8320, "total_steps": 42420, "loss": 0.3172, "lr": 0.0009721248255156246, "epoch": 3.922677982083923, "percentage": 19.61, "elapsed_time": "0:26:44", "remaining_time": "1:49:37", "throughput": 5211.02, "total_tokens": 8363040} {"current_steps": 8325, "total_steps": 42420, "loss": 0.3839, "lr": 0.0009720570562670007, "epoch": 3.9250353606789252, "percentage": 19.63, "elapsed_time": "0:26:45", "remaining_time": "1:49:36", "throughput": 5211.15, "total_tokens": 8368000} {"current_steps": 8330, "total_steps": 42420, "loss": 0.3496, "lr": 0.0009719892071071323, "epoch": 3.9273927392739276, "percentage": 19.64, "elapsed_time": "0:26:46", "remaining_time": "1:49:35", "throughput": 5211.5, "total_tokens": 8373952} {"current_steps": 8335, "total_steps": 42420, "loss": 0.4028, "lr": 0.0009719212780475049, "epoch": 3.92975011786893, "percentage": 19.65, "elapsed_time": "0:26:47", "remaining_time": "1:49:35", "throughput": 5211.84, "total_tokens": 8379776} {"current_steps": 8340, "total_steps": 42420, "loss": 0.341, "lr": 0.0009718532690996179, "epoch": 3.932107496463932, "percentage": 19.66, "elapsed_time": "0:26:48", "remaining_time": "1:49:33", "throughput": 5212.02, "total_tokens": 8384832} {"current_steps": 8345, "total_steps": 42420, "loss": 0.3548, "lr": 0.0009717851802749841, "epoch": 3.9344648750589344, "percentage": 19.67, "elapsed_time": "0:26:49", "remaining_time": "1:49:32", "throughput": 5212.17, "total_tokens": 8389664} {"current_steps": 8350, "total_steps": 42420, "loss": 0.3012, "lr": 0.0009717170115851298, "epoch": 3.936822253653937, "percentage": 19.68, "elapsed_time": "0:26:50", "remaining_time": "1:49:31", "throughput": 5212.27, "total_tokens": 8394400} {"current_steps": 8355, "total_steps": 42420, "loss": 0.3409, "lr": 0.0009716487630415946, "epoch": 3.939179632248939, "percentage": 19.7, "elapsed_time": "0:26:51", "remaining_time": "1:49:29", "throughput": 5212.19, "total_tokens": 8398400} {"current_steps": 8360, "total_steps": 42420, "loss": 0.3383, "lr": 0.000971580434655932, "epoch": 3.9415370108439416, "percentage": 19.71, "elapsed_time": "0:26:52", "remaining_time": "1:49:28", "throughput": 5212.38, "total_tokens": 8403424} {"current_steps": 8365, "total_steps": 42420, "loss": 0.2852, "lr": 0.0009715120264397089, "epoch": 3.943894389438944, "percentage": 19.72, "elapsed_time": "0:26:53", "remaining_time": "1:49:27", "throughput": 5212.49, "total_tokens": 8408160} {"current_steps": 8370, "total_steps": 42420, "loss": 0.3826, "lr": 0.0009714435384045056, "epoch": 3.9462517680339464, "percentage": 19.73, "elapsed_time": "0:26:53", "remaining_time": "1:49:25", "throughput": 5212.24, "total_tokens": 8411616} {"current_steps": 8375, "total_steps": 42420, "loss": 0.3798, "lr": 0.0009713749705619157, "epoch": 3.9486091466289484, "percentage": 19.74, "elapsed_time": "0:26:54", "remaining_time": "1:49:23", "throughput": 5212.36, "total_tokens": 8416448} {"current_steps": 8380, "total_steps": 42420, "loss": 0.3838, "lr": 0.0009713063229235469, "epoch": 3.9509665252239508, "percentage": 19.75, "elapsed_time": "0:26:55", "remaining_time": "1:49:22", "throughput": 5212.56, "total_tokens": 8421632} {"current_steps": 8385, "total_steps": 42420, "loss": 0.3538, "lr": 0.0009712375955010198, "epoch": 3.953323903818953, "percentage": 19.77, "elapsed_time": "0:26:56", "remaining_time": "1:49:21", "throughput": 5212.64, "total_tokens": 8426528} {"current_steps": 8390, "total_steps": 42420, "loss": 0.3366, "lr": 0.0009711687883059691, "epoch": 3.9556812824139556, "percentage": 19.78, "elapsed_time": "0:26:57", "remaining_time": "1:49:20", "throughput": 5212.77, "total_tokens": 8431360} {"current_steps": 8395, "total_steps": 42420, "loss": 0.3473, "lr": 0.0009710999013500426, "epoch": 3.958038661008958, "percentage": 19.79, "elapsed_time": "0:26:58", "remaining_time": "1:49:18", "throughput": 5212.79, "total_tokens": 8435872} {"current_steps": 8400, "total_steps": 42420, "loss": 0.3371, "lr": 0.0009710309346449013, "epoch": 3.9603960396039604, "percentage": 19.8, "elapsed_time": "0:26:59", "remaining_time": "1:49:18", "throughput": 5213.05, "total_tokens": 8441504} {"current_steps": 8405, "total_steps": 42420, "loss": 0.3432, "lr": 0.0009709618882022207, "epoch": 3.9627534181989628, "percentage": 19.81, "elapsed_time": "0:27:00", "remaining_time": "1:49:16", "throughput": 5213.07, "total_tokens": 8445888} {"current_steps": 8410, "total_steps": 42420, "loss": 0.3667, "lr": 0.0009708927620336889, "epoch": 3.965110796793965, "percentage": 19.83, "elapsed_time": "0:27:01", "remaining_time": "1:49:15", "throughput": 5213.3, "total_tokens": 8451360} {"current_steps": 8415, "total_steps": 42420, "loss": 0.2787, "lr": 0.0009708235561510077, "epoch": 3.9674681753889676, "percentage": 19.84, "elapsed_time": "0:27:01", "remaining_time": "1:49:14", "throughput": 5213.44, "total_tokens": 8456128} {"current_steps": 8420, "total_steps": 42420, "loss": 0.268, "lr": 0.0009707542705658925, "epoch": 3.96982555398397, "percentage": 19.85, "elapsed_time": "0:27:02", "remaining_time": "1:49:13", "throughput": 5213.61, "total_tokens": 8461312} {"current_steps": 8425, "total_steps": 42420, "loss": 0.4322, "lr": 0.0009706849052900724, "epoch": 3.9721829325789724, "percentage": 19.86, "elapsed_time": "0:27:03", "remaining_time": "1:49:12", "throughput": 5214.03, "total_tokens": 8467584} {"current_steps": 8430, "total_steps": 42420, "loss": 0.3465, "lr": 0.0009706154603352895, "epoch": 3.9745403111739748, "percentage": 19.87, "elapsed_time": "0:27:04", "remaining_time": "1:49:11", "throughput": 5214.31, "total_tokens": 8473056} {"current_steps": 8435, "total_steps": 42420, "loss": 0.3126, "lr": 0.0009705459357132998, "epoch": 3.976897689768977, "percentage": 19.88, "elapsed_time": "0:27:05", "remaining_time": "1:49:11", "throughput": 5214.55, "total_tokens": 8478688} {"current_steps": 8440, "total_steps": 42420, "loss": 0.3382, "lr": 0.0009704763314358726, "epoch": 3.979255068363979, "percentage": 19.9, "elapsed_time": "0:27:06", "remaining_time": "1:49:10", "throughput": 5214.76, "total_tokens": 8484032} {"current_steps": 8445, "total_steps": 42420, "loss": 0.3289, "lr": 0.0009704066475147907, "epoch": 3.9816124469589815, "percentage": 19.91, "elapsed_time": "0:27:07", "remaining_time": "1:49:09", "throughput": 5215.05, "total_tokens": 8489600} {"current_steps": 8450, "total_steps": 42420, "loss": 0.4173, "lr": 0.0009703368839618505, "epoch": 3.983969825553984, "percentage": 19.92, "elapsed_time": "0:27:08", "remaining_time": "1:49:08", "throughput": 5215.13, "total_tokens": 8494496} {"current_steps": 8455, "total_steps": 42420, "loss": 0.31, "lr": 0.0009702670407888617, "epoch": 3.9863272041489863, "percentage": 19.93, "elapsed_time": "0:27:09", "remaining_time": "1:49:06", "throughput": 5215.26, "total_tokens": 8499424} {"current_steps": 8460, "total_steps": 42420, "loss": 0.3703, "lr": 0.0009701971180076474, "epoch": 3.9886845827439887, "percentage": 19.94, "elapsed_time": "0:27:10", "remaining_time": "1:49:05", "throughput": 5215.47, "total_tokens": 8504608} {"current_steps": 8465, "total_steps": 42420, "loss": 0.3385, "lr": 0.0009701271156300447, "epoch": 3.991041961338991, "percentage": 19.96, "elapsed_time": "0:27:11", "remaining_time": "1:49:04", "throughput": 5215.52, "total_tokens": 8509056} {"current_steps": 8470, "total_steps": 42420, "loss": 0.3369, "lr": 0.0009700570336679035, "epoch": 3.9933993399339935, "percentage": 19.97, "elapsed_time": "0:27:12", "remaining_time": "1:49:02", "throughput": 5215.51, "total_tokens": 8513248} {"current_steps": 8475, "total_steps": 42420, "loss": 0.3491, "lr": 0.0009699868721330878, "epoch": 3.9957567185289955, "percentage": 19.98, "elapsed_time": "0:27:13", "remaining_time": "1:49:01", "throughput": 5215.58, "total_tokens": 8517760} {"current_steps": 8480, "total_steps": 42420, "loss": 0.3401, "lr": 0.0009699166310374745, "epoch": 3.998114097123998, "percentage": 19.99, "elapsed_time": "0:27:14", "remaining_time": "1:49:00", "throughput": 5215.7, "total_tokens": 8522816} {"current_steps": 8484, "total_steps": 42420, "eval_loss": 0.3406714200973511, "epoch": 4.0, "percentage": 20.0, "elapsed_time": "0:27:33", "remaining_time": "1:50:15", "throughput": 5155.75, "total_tokens": 8527168} {"current_steps": 8485, "total_steps": 42420, "loss": 0.3525, "lr": 0.0009698463103929542, "epoch": 4.000471475719, "percentage": 20.0, "elapsed_time": "0:27:36", "remaining_time": "1:50:25", "throughput": 5148.16, "total_tokens": 8528128} {"current_steps": 8490, "total_steps": 42420, "loss": 0.3264, "lr": 0.0009697759102114312, "epoch": 4.002828854314003, "percentage": 20.01, "elapsed_time": "0:27:37", "remaining_time": "1:50:23", "throughput": 5148.18, "total_tokens": 8532608} {"current_steps": 8495, "total_steps": 42420, "loss": 0.3798, "lr": 0.0009697054305048228, "epoch": 4.005186232909005, "percentage": 20.03, "elapsed_time": "0:27:38", "remaining_time": "1:50:22", "throughput": 5148.43, "total_tokens": 8537952} {"current_steps": 8500, "total_steps": 42420, "loss": 0.368, "lr": 0.0009696348712850602, "epoch": 4.0075436115040075, "percentage": 20.04, "elapsed_time": "0:27:39", "remaining_time": "1:50:21", "throughput": 5148.8, "total_tokens": 8543872} {"current_steps": 8505, "total_steps": 42420, "loss": 0.3226, "lr": 0.000969564232564088, "epoch": 4.00990099009901, "percentage": 20.05, "elapsed_time": "0:27:40", "remaining_time": "1:50:21", "throughput": 5149.03, "total_tokens": 8549376} {"current_steps": 8510, "total_steps": 42420, "loss": 0.3909, "lr": 0.000969493514353864, "epoch": 4.012258368694012, "percentage": 20.06, "elapsed_time": "0:27:41", "remaining_time": "1:50:20", "throughput": 5149.32, "total_tokens": 8555040} {"current_steps": 8515, "total_steps": 42420, "loss": 0.3998, "lr": 0.0009694227166663591, "epoch": 4.014615747289015, "percentage": 20.07, "elapsed_time": "0:27:42", "remaining_time": "1:50:18", "throughput": 5149.15, "total_tokens": 8558624} {"current_steps": 8520, "total_steps": 42420, "loss": 0.2956, "lr": 0.0009693518395135591, "epoch": 4.016973125884017, "percentage": 20.08, "elapsed_time": "0:27:43", "remaining_time": "1:50:17", "throughput": 5149.33, "total_tokens": 8563552} {"current_steps": 8525, "total_steps": 42420, "loss": 0.3036, "lr": 0.0009692808829074616, "epoch": 4.0193305044790195, "percentage": 20.1, "elapsed_time": "0:27:44", "remaining_time": "1:50:16", "throughput": 5149.55, "total_tokens": 8569056} {"current_steps": 8530, "total_steps": 42420, "loss": 0.3043, "lr": 0.0009692098468600787, "epoch": 4.021687883074022, "percentage": 20.11, "elapsed_time": "0:27:44", "remaining_time": "1:50:14", "throughput": 5149.61, "total_tokens": 8573600} {"current_steps": 8535, "total_steps": 42420, "loss": 0.2529, "lr": 0.0009691387313834352, "epoch": 4.024045261669024, "percentage": 20.12, "elapsed_time": "0:27:45", "remaining_time": "1:50:13", "throughput": 5149.59, "total_tokens": 8577856} {"current_steps": 8540, "total_steps": 42420, "loss": 0.463, "lr": 0.0009690675364895702, "epoch": 4.026402640264027, "percentage": 20.13, "elapsed_time": "0:27:46", "remaining_time": "1:50:12", "throughput": 5149.8, "total_tokens": 8583168} {"current_steps": 8545, "total_steps": 42420, "loss": 0.3071, "lr": 0.0009689962621905356, "epoch": 4.028760018859029, "percentage": 20.14, "elapsed_time": "0:27:47", "remaining_time": "1:50:10", "throughput": 5149.97, "total_tokens": 8588064} {"current_steps": 8550, "total_steps": 42420, "loss": 0.3385, "lr": 0.0009689249084983969, "epoch": 4.0311173974540315, "percentage": 20.16, "elapsed_time": "0:27:48", "remaining_time": "1:50:09", "throughput": 5150.23, "total_tokens": 8593472} {"current_steps": 8555, "total_steps": 42420, "loss": 0.3063, "lr": 0.0009688534754252329, "epoch": 4.033474776049033, "percentage": 20.17, "elapsed_time": "0:27:49", "remaining_time": "1:50:08", "throughput": 5150.4, "total_tokens": 8598560} {"current_steps": 8560, "total_steps": 42420, "loss": 0.3533, "lr": 0.0009687819629831365, "epoch": 4.035832154644035, "percentage": 20.18, "elapsed_time": "0:27:50", "remaining_time": "1:50:07", "throughput": 5150.47, "total_tokens": 8603104} {"current_steps": 8565, "total_steps": 42420, "loss": 0.3162, "lr": 0.000968710371184213, "epoch": 4.038189533239038, "percentage": 20.19, "elapsed_time": "0:27:51", "remaining_time": "1:50:06", "throughput": 5150.75, "total_tokens": 8608416} {"current_steps": 8570, "total_steps": 42420, "loss": 0.4914, "lr": 0.000968638700040582, "epoch": 4.04054691183404, "percentage": 20.2, "elapsed_time": "0:27:52", "remaining_time": "1:50:04", "throughput": 5150.77, "total_tokens": 8612896} {"current_steps": 8575, "total_steps": 42420, "loss": 0.3634, "lr": 0.000968566949564376, "epoch": 4.042904290429043, "percentage": 20.21, "elapsed_time": "0:27:52", "remaining_time": "1:50:03", "throughput": 5150.78, "total_tokens": 8617120} {"current_steps": 8580, "total_steps": 42420, "loss": 0.4528, "lr": 0.0009684951197677414, "epoch": 4.045261669024045, "percentage": 20.23, "elapsed_time": "0:27:53", "remaining_time": "1:50:01", "throughput": 5150.98, "total_tokens": 8622208} {"current_steps": 8585, "total_steps": 42420, "loss": 0.3501, "lr": 0.0009684232106628376, "epoch": 4.0476190476190474, "percentage": 20.24, "elapsed_time": "0:27:54", "remaining_time": "1:50:00", "throughput": 5151.22, "total_tokens": 8627552} {"current_steps": 8590, "total_steps": 42420, "loss": 0.4102, "lr": 0.0009683512222618376, "epoch": 4.04997642621405, "percentage": 20.25, "elapsed_time": "0:27:55", "remaining_time": "1:50:00", "throughput": 5151.61, "total_tokens": 8633824} {"current_steps": 8595, "total_steps": 42420, "loss": 0.374, "lr": 0.0009682791545769277, "epoch": 4.052333804809052, "percentage": 20.26, "elapsed_time": "0:27:57", "remaining_time": "1:49:59", "throughput": 5152.01, "total_tokens": 8640032} {"current_steps": 8600, "total_steps": 42420, "loss": 0.3598, "lr": 0.000968207007620308, "epoch": 4.054691183404055, "percentage": 20.27, "elapsed_time": "0:27:58", "remaining_time": "1:49:59", "throughput": 5152.36, "total_tokens": 8645952} {"current_steps": 8605, "total_steps": 42420, "loss": 0.3593, "lr": 0.0009681347814041915, "epoch": 4.057048561999057, "percentage": 20.29, "elapsed_time": "0:27:58", "remaining_time": "1:49:57", "throughput": 5152.36, "total_tokens": 8650112} {"current_steps": 8610, "total_steps": 42420, "loss": 0.3494, "lr": 0.0009680624759408051, "epoch": 4.0594059405940595, "percentage": 20.3, "elapsed_time": "0:28:00", "remaining_time": "1:49:57", "throughput": 5152.81, "total_tokens": 8656896} {"current_steps": 8615, "total_steps": 42420, "loss": 0.3492, "lr": 0.0009679900912423887, "epoch": 4.061763319189062, "percentage": 20.31, "elapsed_time": "0:28:00", "remaining_time": "1:49:55", "throughput": 5152.74, "total_tokens": 8660896} {"current_steps": 8620, "total_steps": 42420, "loss": 0.4394, "lr": 0.0009679176273211959, "epoch": 4.064120697784064, "percentage": 20.32, "elapsed_time": "0:28:01", "remaining_time": "1:49:54", "throughput": 5152.71, "total_tokens": 8665184} {"current_steps": 8625, "total_steps": 42420, "loss": 0.3256, "lr": 0.0009678450841894935, "epoch": 4.066478076379067, "percentage": 20.33, "elapsed_time": "0:28:02", "remaining_time": "1:49:52", "throughput": 5152.7, "total_tokens": 8669824} {"current_steps": 8630, "total_steps": 42420, "loss": 0.3479, "lr": 0.000967772461859562, "epoch": 4.068835454974069, "percentage": 20.34, "elapsed_time": "0:28:03", "remaining_time": "1:49:51", "throughput": 5152.9, "total_tokens": 8675008} {"current_steps": 8635, "total_steps": 42420, "loss": 0.3512, "lr": 0.000967699760343695, "epoch": 4.0711928335690715, "percentage": 20.36, "elapsed_time": "0:28:04", "remaining_time": "1:49:50", "throughput": 5152.91, "total_tokens": 8679264} {"current_steps": 8640, "total_steps": 42420, "loss": 0.3109, "lr": 0.0009676269796541996, "epoch": 4.073550212164074, "percentage": 20.37, "elapsed_time": "0:28:05", "remaining_time": "1:49:48", "throughput": 5153.0, "total_tokens": 8683808} {"current_steps": 8645, "total_steps": 42420, "loss": 0.432, "lr": 0.0009675541198033966, "epoch": 4.075907590759076, "percentage": 20.38, "elapsed_time": "0:28:06", "remaining_time": "1:49:47", "throughput": 5153.16, "total_tokens": 8688896} {"current_steps": 8650, "total_steps": 42420, "loss": 0.4107, "lr": 0.0009674811808036195, "epoch": 4.078264969354079, "percentage": 20.39, "elapsed_time": "0:28:07", "remaining_time": "1:49:46", "throughput": 5153.23, "total_tokens": 8693856} {"current_steps": 8655, "total_steps": 42420, "loss": 0.3739, "lr": 0.000967408162667216, "epoch": 4.08062234794908, "percentage": 20.4, "elapsed_time": "0:28:07", "remaining_time": "1:49:44", "throughput": 5153.18, "total_tokens": 8697984} {"current_steps": 8660, "total_steps": 42420, "loss": 0.3494, "lr": 0.0009673350654065467, "epoch": 4.082979726544083, "percentage": 20.41, "elapsed_time": "0:28:08", "remaining_time": "1:49:43", "throughput": 5153.19, "total_tokens": 8702464} {"current_steps": 8665, "total_steps": 42420, "loss": 0.4283, "lr": 0.0009672618890339857, "epoch": 4.085337105139085, "percentage": 20.43, "elapsed_time": "0:28:09", "remaining_time": "1:49:42", "throughput": 5153.29, "total_tokens": 8707200} {"current_steps": 8670, "total_steps": 42420, "loss": 0.3622, "lr": 0.0009671886335619206, "epoch": 4.087694483734087, "percentage": 20.44, "elapsed_time": "0:28:10", "remaining_time": "1:49:41", "throughput": 5153.54, "total_tokens": 8712768} {"current_steps": 8675, "total_steps": 42420, "loss": 0.3643, "lr": 0.0009671152990027521, "epoch": 4.09005186232909, "percentage": 20.45, "elapsed_time": "0:28:11", "remaining_time": "1:49:39", "throughput": 5153.59, "total_tokens": 8717472} {"current_steps": 8680, "total_steps": 42420, "loss": 0.3437, "lr": 0.0009670418853688947, "epoch": 4.092409240924092, "percentage": 20.46, "elapsed_time": "0:28:12", "remaining_time": "1:49:38", "throughput": 5153.74, "total_tokens": 8722400} {"current_steps": 8685, "total_steps": 42420, "loss": 0.2859, "lr": 0.0009669683926727761, "epoch": 4.094766619519095, "percentage": 20.47, "elapsed_time": "0:28:13", "remaining_time": "1:49:37", "throughput": 5153.92, "total_tokens": 8727648} {"current_steps": 8690, "total_steps": 42420, "loss": 0.1241, "lr": 0.0009668948209268373, "epoch": 4.097123998114097, "percentage": 20.49, "elapsed_time": "0:28:14", "remaining_time": "1:49:36", "throughput": 5154.03, "total_tokens": 8732320} {"current_steps": 8695, "total_steps": 42420, "loss": 0.471, "lr": 0.0009668211701435327, "epoch": 4.099481376709099, "percentage": 20.5, "elapsed_time": "0:28:15", "remaining_time": "1:49:35", "throughput": 5154.19, "total_tokens": 8737472} {"current_steps": 8700, "total_steps": 42420, "loss": 0.6585, "lr": 0.0009667474403353302, "epoch": 4.101838755304102, "percentage": 20.51, "elapsed_time": "0:28:16", "remaining_time": "1:49:34", "throughput": 5154.55, "total_tokens": 8743456} {"current_steps": 8705, "total_steps": 42420, "loss": 0.3028, "lr": 0.0009666736315147109, "epoch": 4.104196133899104, "percentage": 20.52, "elapsed_time": "0:28:17", "remaining_time": "1:49:33", "throughput": 5154.87, "total_tokens": 8749120} {"current_steps": 8710, "total_steps": 42420, "loss": 0.3625, "lr": 0.0009665997436941694, "epoch": 4.106553512494107, "percentage": 20.53, "elapsed_time": "0:28:18", "remaining_time": "1:49:31", "throughput": 5154.79, "total_tokens": 8753216} {"current_steps": 8715, "total_steps": 42420, "loss": 0.4163, "lr": 0.0009665257768862138, "epoch": 4.108910891089109, "percentage": 20.54, "elapsed_time": "0:28:19", "remaining_time": "1:49:31", "throughput": 5155.27, "total_tokens": 8760128} {"current_steps": 8720, "total_steps": 42420, "loss": 0.4207, "lr": 0.0009664517311033654, "epoch": 4.111268269684111, "percentage": 20.56, "elapsed_time": "0:28:20", "remaining_time": "1:49:30", "throughput": 5155.15, "total_tokens": 8763872} {"current_steps": 8725, "total_steps": 42420, "loss": 0.3344, "lr": 0.0009663776063581586, "epoch": 4.113625648279114, "percentage": 20.57, "elapsed_time": "0:28:20", "remaining_time": "1:49:28", "throughput": 5155.29, "total_tokens": 8768960} {"current_steps": 8730, "total_steps": 42420, "loss": 0.3562, "lr": 0.0009663034026631418, "epoch": 4.115983026874116, "percentage": 20.58, "elapsed_time": "0:28:21", "remaining_time": "1:49:27", "throughput": 5155.3, "total_tokens": 8773312} {"current_steps": 8735, "total_steps": 42420, "loss": 0.3657, "lr": 0.0009662291200308764, "epoch": 4.118340405469119, "percentage": 20.59, "elapsed_time": "0:28:22", "remaining_time": "1:49:26", "throughput": 5155.41, "total_tokens": 8777920} {"current_steps": 8740, "total_steps": 42420, "loss": 0.3916, "lr": 0.0009661547584739371, "epoch": 4.120697784064121, "percentage": 20.6, "elapsed_time": "0:28:23", "remaining_time": "1:49:25", "throughput": 5155.86, "total_tokens": 8784544} {"current_steps": 8745, "total_steps": 42420, "loss": 0.349, "lr": 0.0009660803180049119, "epoch": 4.123055162659123, "percentage": 20.62, "elapsed_time": "0:28:24", "remaining_time": "1:49:24", "throughput": 5155.9, "total_tokens": 8789024} {"current_steps": 8750, "total_steps": 42420, "loss": 0.4142, "lr": 0.0009660057986364026, "epoch": 4.125412541254126, "percentage": 20.63, "elapsed_time": "0:28:25", "remaining_time": "1:49:23", "throughput": 5156.12, "total_tokens": 8794272} {"current_steps": 8755, "total_steps": 42420, "loss": 0.3791, "lr": 0.0009659312003810239, "epoch": 4.127769919849127, "percentage": 20.64, "elapsed_time": "0:28:26", "remaining_time": "1:49:22", "throughput": 5156.32, "total_tokens": 8799552} {"current_steps": 8760, "total_steps": 42420, "loss": 0.3272, "lr": 0.000965856523251404, "epoch": 4.13012729844413, "percentage": 20.65, "elapsed_time": "0:28:27", "remaining_time": "1:49:20", "throughput": 5156.59, "total_tokens": 8804800} {"current_steps": 8765, "total_steps": 42420, "loss": 0.3489, "lr": 0.0009657817672601844, "epoch": 4.132484677039132, "percentage": 20.66, "elapsed_time": "0:28:28", "remaining_time": "1:49:19", "throughput": 5156.79, "total_tokens": 8810176} {"current_steps": 8770, "total_steps": 42420, "loss": 0.3512, "lr": 0.0009657069324200203, "epoch": 4.1348420556341345, "percentage": 20.67, "elapsed_time": "0:28:29", "remaining_time": "1:49:19", "throughput": 5157.03, "total_tokens": 8815680} {"current_steps": 8775, "total_steps": 42420, "loss": 0.3263, "lr": 0.0009656320187435796, "epoch": 4.137199434229137, "percentage": 20.69, "elapsed_time": "0:28:30", "remaining_time": "1:49:17", "throughput": 5156.99, "total_tokens": 8819872} {"current_steps": 8780, "total_steps": 42420, "loss": 0.2825, "lr": 0.0009655570262435443, "epoch": 4.139556812824139, "percentage": 20.7, "elapsed_time": "0:28:31", "remaining_time": "1:49:16", "throughput": 5157.02, "total_tokens": 8824288} {"current_steps": 8785, "total_steps": 42420, "loss": 0.3531, "lr": 0.000965481954932609, "epoch": 4.141914191419142, "percentage": 20.71, "elapsed_time": "0:28:32", "remaining_time": "1:49:14", "throughput": 5157.15, "total_tokens": 8829312} {"current_steps": 8790, "total_steps": 42420, "loss": 0.3065, "lr": 0.0009654068048234823, "epoch": 4.144271570014144, "percentage": 20.72, "elapsed_time": "0:28:33", "remaining_time": "1:49:14", "throughput": 5157.53, "total_tokens": 8835392} {"current_steps": 8795, "total_steps": 42420, "loss": 0.369, "lr": 0.0009653315759288857, "epoch": 4.1466289486091465, "percentage": 20.73, "elapsed_time": "0:28:34", "remaining_time": "1:49:13", "throughput": 5157.78, "total_tokens": 8840896} {"current_steps": 8800, "total_steps": 42420, "loss": 0.3412, "lr": 0.0009652562682615542, "epoch": 4.148986327204149, "percentage": 20.74, "elapsed_time": "0:28:34", "remaining_time": "1:49:12", "throughput": 5157.94, "total_tokens": 8845792} {"current_steps": 8805, "total_steps": 42420, "loss": 0.3594, "lr": 0.000965180881834236, "epoch": 4.151343705799151, "percentage": 20.76, "elapsed_time": "0:28:35", "remaining_time": "1:49:10", "throughput": 5157.86, "total_tokens": 8849888} {"current_steps": 8810, "total_steps": 42420, "loss": 0.3743, "lr": 0.0009651054166596929, "epoch": 4.153701084394154, "percentage": 20.77, "elapsed_time": "0:28:36", "remaining_time": "1:49:09", "throughput": 5158.06, "total_tokens": 8855296} {"current_steps": 8815, "total_steps": 42420, "loss": 0.3428, "lr": 0.0009650298727506997, "epoch": 4.156058462989156, "percentage": 20.78, "elapsed_time": "0:28:37", "remaining_time": "1:49:08", "throughput": 5158.11, "total_tokens": 8859712} {"current_steps": 8820, "total_steps": 42420, "loss": 0.4127, "lr": 0.0009649542501200448, "epoch": 4.158415841584159, "percentage": 20.79, "elapsed_time": "0:28:38", "remaining_time": "1:49:06", "throughput": 5158.25, "total_tokens": 8864736} {"current_steps": 8825, "total_steps": 42420, "loss": 0.2935, "lr": 0.0009648785487805299, "epoch": 4.160773220179161, "percentage": 20.8, "elapsed_time": "0:28:39", "remaining_time": "1:49:05", "throughput": 5158.43, "total_tokens": 8870048} {"current_steps": 8830, "total_steps": 42420, "loss": 0.3348, "lr": 0.0009648027687449699, "epoch": 4.163130598774163, "percentage": 20.82, "elapsed_time": "0:28:40", "remaining_time": "1:49:04", "throughput": 5158.46, "total_tokens": 8874528} {"current_steps": 8835, "total_steps": 42420, "loss": 0.3344, "lr": 0.0009647269100261929, "epoch": 4.165487977369166, "percentage": 20.83, "elapsed_time": "0:28:41", "remaining_time": "1:49:03", "throughput": 5158.59, "total_tokens": 8879392} {"current_steps": 8840, "total_steps": 42420, "loss": 0.3117, "lr": 0.0009646509726370408, "epoch": 4.167845355964168, "percentage": 20.84, "elapsed_time": "0:28:42", "remaining_time": "1:49:02", "throughput": 5158.86, "total_tokens": 8885152} {"current_steps": 8845, "total_steps": 42420, "loss": 0.3, "lr": 0.0009645749565903683, "epoch": 4.170202734559171, "percentage": 20.85, "elapsed_time": "0:28:43", "remaining_time": "1:49:01", "throughput": 5158.86, "total_tokens": 8889696} {"current_steps": 8850, "total_steps": 42420, "loss": 0.3188, "lr": 0.0009644988618990438, "epoch": 4.172560113154173, "percentage": 20.86, "elapsed_time": "0:28:44", "remaining_time": "1:49:00", "throughput": 5159.23, "total_tokens": 8896064} {"current_steps": 8855, "total_steps": 42420, "loss": 0.4005, "lr": 0.0009644226885759486, "epoch": 4.174917491749175, "percentage": 20.87, "elapsed_time": "0:28:45", "remaining_time": "1:48:59", "throughput": 5159.46, "total_tokens": 8901312} {"current_steps": 8860, "total_steps": 42420, "loss": 0.3557, "lr": 0.0009643464366339776, "epoch": 4.177274870344177, "percentage": 20.89, "elapsed_time": "0:28:46", "remaining_time": "1:48:58", "throughput": 5159.66, "total_tokens": 8906400} {"current_steps": 8865, "total_steps": 42420, "loss": 0.3728, "lr": 0.0009642701060860392, "epoch": 4.179632248939179, "percentage": 20.9, "elapsed_time": "0:28:47", "remaining_time": "1:48:56", "throughput": 5159.67, "total_tokens": 8910880} {"current_steps": 8870, "total_steps": 42420, "loss": 0.3978, "lr": 0.0009641936969450546, "epoch": 4.181989627534182, "percentage": 20.91, "elapsed_time": "0:28:47", "remaining_time": "1:48:55", "throughput": 5159.84, "total_tokens": 8916032} {"current_steps": 8875, "total_steps": 42420, "loss": 0.4073, "lr": 0.0009641172092239587, "epoch": 4.184347006129184, "percentage": 20.92, "elapsed_time": "0:28:48", "remaining_time": "1:48:54", "throughput": 5160.0, "total_tokens": 8921248} {"current_steps": 8880, "total_steps": 42420, "loss": 0.3518, "lr": 0.0009640406429356995, "epoch": 4.1867043847241865, "percentage": 20.93, "elapsed_time": "0:28:49", "remaining_time": "1:48:53", "throughput": 5160.17, "total_tokens": 8926272} {"current_steps": 8885, "total_steps": 42420, "loss": 0.319, "lr": 0.0009639639980932385, "epoch": 4.189061763319189, "percentage": 20.95, "elapsed_time": "0:28:50", "remaining_time": "1:48:52", "throughput": 5160.14, "total_tokens": 8930432} {"current_steps": 8890, "total_steps": 42420, "loss": 0.3217, "lr": 0.0009638872747095502, "epoch": 4.191419141914191, "percentage": 20.96, "elapsed_time": "0:28:51", "remaining_time": "1:48:51", "throughput": 5160.39, "total_tokens": 8935744} {"current_steps": 8895, "total_steps": 42420, "loss": 0.2858, "lr": 0.0009638104727976227, "epoch": 4.193776520509194, "percentage": 20.97, "elapsed_time": "0:28:52", "remaining_time": "1:48:49", "throughput": 5160.41, "total_tokens": 8940160} {"current_steps": 8900, "total_steps": 42420, "loss": 0.3904, "lr": 0.0009637335923704571, "epoch": 4.196133899104196, "percentage": 20.98, "elapsed_time": "0:28:53", "remaining_time": "1:48:47", "throughput": 5160.36, "total_tokens": 8944192} {"current_steps": 8905, "total_steps": 42420, "loss": 0.3368, "lr": 0.0009636566334410682, "epoch": 4.1984912776991985, "percentage": 20.99, "elapsed_time": "0:28:54", "remaining_time": "1:48:46", "throughput": 5160.57, "total_tokens": 8949632} {"current_steps": 8910, "total_steps": 42420, "loss": 0.3519, "lr": 0.0009635795960224835, "epoch": 4.200848656294201, "percentage": 21.0, "elapsed_time": "0:28:55", "remaining_time": "1:48:45", "throughput": 5160.71, "total_tokens": 8954656} {"current_steps": 8915, "total_steps": 42420, "loss": 0.3088, "lr": 0.0009635024801277445, "epoch": 4.203206034889203, "percentage": 21.02, "elapsed_time": "0:28:56", "remaining_time": "1:48:44", "throughput": 5160.88, "total_tokens": 8959488} {"current_steps": 8920, "total_steps": 42420, "loss": 0.2804, "lr": 0.0009634252857699055, "epoch": 4.205563413484206, "percentage": 21.03, "elapsed_time": "0:28:56", "remaining_time": "1:48:43", "throughput": 5161.02, "total_tokens": 8964256} {"current_steps": 8925, "total_steps": 42420, "loss": 0.3059, "lr": 0.000963348012962034, "epoch": 4.207920792079208, "percentage": 21.04, "elapsed_time": "0:28:58", "remaining_time": "1:48:42", "throughput": 5161.45, "total_tokens": 8970688} {"current_steps": 8930, "total_steps": 42420, "loss": 0.3051, "lr": 0.0009632706617172111, "epoch": 4.2102781706742105, "percentage": 21.05, "elapsed_time": "0:28:58", "remaining_time": "1:48:41", "throughput": 5161.48, "total_tokens": 8975456} {"current_steps": 8935, "total_steps": 42420, "loss": 0.3734, "lr": 0.0009631932320485312, "epoch": 4.212635549269213, "percentage": 21.06, "elapsed_time": "0:28:59", "remaining_time": "1:48:40", "throughput": 5161.57, "total_tokens": 8980064} {"current_steps": 8940, "total_steps": 42420, "loss": 0.4057, "lr": 0.0009631157239691015, "epoch": 4.214992927864215, "percentage": 21.07, "elapsed_time": "0:29:00", "remaining_time": "1:48:38", "throughput": 5161.58, "total_tokens": 8984480} {"current_steps": 8945, "total_steps": 42420, "loss": 0.3573, "lr": 0.0009630381374920432, "epoch": 4.217350306459218, "percentage": 21.09, "elapsed_time": "0:29:01", "remaining_time": "1:48:37", "throughput": 5161.65, "total_tokens": 8989280} {"current_steps": 8950, "total_steps": 42420, "loss": 0.3696, "lr": 0.0009629604726304899, "epoch": 4.21970768505422, "percentage": 21.1, "elapsed_time": "0:29:02", "remaining_time": "1:48:36", "throughput": 5161.83, "total_tokens": 8994400} {"current_steps": 8955, "total_steps": 42420, "loss": 0.2465, "lr": 0.0009628827293975895, "epoch": 4.222065063649222, "percentage": 21.11, "elapsed_time": "0:29:03", "remaining_time": "1:48:35", "throughput": 5161.91, "total_tokens": 8999264} {"current_steps": 8960, "total_steps": 42420, "loss": 0.3981, "lr": 0.0009628049078065021, "epoch": 4.224422442244224, "percentage": 21.12, "elapsed_time": "0:29:04", "remaining_time": "1:48:34", "throughput": 5162.07, "total_tokens": 9004544} {"current_steps": 8965, "total_steps": 42420, "loss": 0.5298, "lr": 0.0009627270078704018, "epoch": 4.226779820839226, "percentage": 21.13, "elapsed_time": "0:29:05", "remaining_time": "1:48:33", "throughput": 5162.43, "total_tokens": 9010688} {"current_steps": 8970, "total_steps": 42420, "loss": 0.4205, "lr": 0.0009626490296024758, "epoch": 4.229137199434229, "percentage": 21.15, "elapsed_time": "0:29:06", "remaining_time": "1:48:32", "throughput": 5162.63, "total_tokens": 9015904} {"current_steps": 8975, "total_steps": 42420, "loss": 0.4202, "lr": 0.0009625709730159244, "epoch": 4.231494578029231, "percentage": 21.16, "elapsed_time": "0:29:07", "remaining_time": "1:48:31", "throughput": 5162.89, "total_tokens": 9021408} {"current_steps": 8980, "total_steps": 42420, "loss": 0.4031, "lr": 0.0009624928381239613, "epoch": 4.233851956624234, "percentage": 21.17, "elapsed_time": "0:29:08", "remaining_time": "1:48:30", "throughput": 5162.97, "total_tokens": 9026272} {"current_steps": 8985, "total_steps": 42420, "loss": 0.4119, "lr": 0.0009624146249398133, "epoch": 4.236209335219236, "percentage": 21.18, "elapsed_time": "0:29:09", "remaining_time": "1:48:29", "throughput": 5163.21, "total_tokens": 9032128} {"current_steps": 8990, "total_steps": 42420, "loss": 0.3465, "lr": 0.0009623363334767208, "epoch": 4.238566713814238, "percentage": 21.19, "elapsed_time": "0:29:10", "remaining_time": "1:48:28", "throughput": 5163.36, "total_tokens": 9037024} {"current_steps": 8995, "total_steps": 42420, "loss": 0.34, "lr": 0.0009622579637479369, "epoch": 4.240924092409241, "percentage": 21.2, "elapsed_time": "0:29:11", "remaining_time": "1:48:27", "throughput": 5163.54, "total_tokens": 9042240} {"current_steps": 9000, "total_steps": 42420, "loss": 0.402, "lr": 0.0009621795157667286, "epoch": 4.243281471004243, "percentage": 21.22, "elapsed_time": "0:29:12", "remaining_time": "1:48:27", "throughput": 5164.11, "total_tokens": 9049408} {"current_steps": 9005, "total_steps": 42420, "loss": 0.3575, "lr": 0.0009621009895463755, "epoch": 4.245638849599246, "percentage": 21.23, "elapsed_time": "0:29:13", "remaining_time": "1:48:25", "throughput": 5164.23, "total_tokens": 9054304} {"current_steps": 9010, "total_steps": 42420, "loss": 0.3603, "lr": 0.000962022385100171, "epoch": 4.247996228194248, "percentage": 21.24, "elapsed_time": "0:29:14", "remaining_time": "1:48:24", "throughput": 5164.2, "total_tokens": 9058496} {"current_steps": 9015, "total_steps": 42420, "loss": 0.346, "lr": 0.0009619437024414213, "epoch": 4.2503536067892504, "percentage": 21.25, "elapsed_time": "0:29:14", "remaining_time": "1:48:22", "throughput": 5164.27, "total_tokens": 9062944} {"current_steps": 9020, "total_steps": 42420, "loss": 0.3388, "lr": 0.0009618649415834462, "epoch": 4.252710985384253, "percentage": 21.26, "elapsed_time": "0:29:15", "remaining_time": "1:48:21", "throughput": 5164.36, "total_tokens": 9067584} {"current_steps": 9025, "total_steps": 42420, "loss": 0.4456, "lr": 0.0009617861025395785, "epoch": 4.255068363979255, "percentage": 21.28, "elapsed_time": "0:29:16", "remaining_time": "1:48:20", "throughput": 5164.56, "total_tokens": 9072800} {"current_steps": 9030, "total_steps": 42420, "loss": 0.3092, "lr": 0.0009617071853231643, "epoch": 4.257425742574258, "percentage": 21.29, "elapsed_time": "0:29:17", "remaining_time": "1:48:19", "throughput": 5164.57, "total_tokens": 9077216} {"current_steps": 9035, "total_steps": 42420, "loss": 0.3604, "lr": 0.000961628189947563, "epoch": 4.25978312116926, "percentage": 21.3, "elapsed_time": "0:29:18", "remaining_time": "1:48:17", "throughput": 5164.77, "total_tokens": 9082400} {"current_steps": 9040, "total_steps": 42420, "loss": 0.3655, "lr": 0.0009615491164261472, "epoch": 4.2621404997642625, "percentage": 21.31, "elapsed_time": "0:29:19", "remaining_time": "1:48:16", "throughput": 5164.73, "total_tokens": 9086432} {"current_steps": 9045, "total_steps": 42420, "loss": 0.3399, "lr": 0.0009614699647723025, "epoch": 4.264497878359265, "percentage": 21.32, "elapsed_time": "0:29:20", "remaining_time": "1:48:15", "throughput": 5164.88, "total_tokens": 9091616} {"current_steps": 9050, "total_steps": 42420, "loss": 0.3599, "lr": 0.0009613907349994282, "epoch": 4.266855256954267, "percentage": 21.33, "elapsed_time": "0:29:21", "remaining_time": "1:48:13", "throughput": 5164.82, "total_tokens": 9095840} {"current_steps": 9055, "total_steps": 42420, "loss": 0.3665, "lr": 0.0009613114271209363, "epoch": 4.26921263554927, "percentage": 21.35, "elapsed_time": "0:29:21", "remaining_time": "1:48:12", "throughput": 5164.74, "total_tokens": 9099744} {"current_steps": 9060, "total_steps": 42420, "loss": 0.3654, "lr": 0.0009612320411502525, "epoch": 4.271570014144271, "percentage": 21.36, "elapsed_time": "0:29:22", "remaining_time": "1:48:10", "throughput": 5164.69, "total_tokens": 9103744} {"current_steps": 9065, "total_steps": 42420, "loss": 0.3118, "lr": 0.0009611525771008154, "epoch": 4.273927392739274, "percentage": 21.37, "elapsed_time": "0:29:23", "remaining_time": "1:48:08", "throughput": 5164.7, "total_tokens": 9108064} {"current_steps": 9070, "total_steps": 42420, "loss": 0.3513, "lr": 0.000961073034986077, "epoch": 4.276284771334276, "percentage": 21.38, "elapsed_time": "0:29:24", "remaining_time": "1:48:07", "throughput": 5164.62, "total_tokens": 9112000} {"current_steps": 9075, "total_steps": 42420, "loss": 0.3201, "lr": 0.0009609934148195021, "epoch": 4.278642149929278, "percentage": 21.39, "elapsed_time": "0:29:25", "remaining_time": "1:48:06", "throughput": 5164.91, "total_tokens": 9117632} {"current_steps": 9080, "total_steps": 42420, "loss": 0.3527, "lr": 0.0009609137166145696, "epoch": 4.280999528524281, "percentage": 21.4, "elapsed_time": "0:29:26", "remaining_time": "1:48:05", "throughput": 5165.14, "total_tokens": 9123008} {"current_steps": 9085, "total_steps": 42420, "loss": 0.3, "lr": 0.0009608339403847706, "epoch": 4.283356907119283, "percentage": 21.42, "elapsed_time": "0:29:27", "remaining_time": "1:48:03", "throughput": 5165.08, "total_tokens": 9126976} {"current_steps": 9090, "total_steps": 42420, "loss": 0.3851, "lr": 0.0009607540861436099, "epoch": 4.285714285714286, "percentage": 21.43, "elapsed_time": "0:29:27", "remaining_time": "1:48:02", "throughput": 5165.2, "total_tokens": 9131840} {"current_steps": 9095, "total_steps": 42420, "loss": 0.3434, "lr": 0.0009606741539046058, "epoch": 4.288071664309288, "percentage": 21.44, "elapsed_time": "0:29:28", "remaining_time": "1:48:00", "throughput": 5165.15, "total_tokens": 9135680} {"current_steps": 9100, "total_steps": 42420, "loss": 0.3797, "lr": 0.0009605941436812891, "epoch": 4.29042904290429, "percentage": 21.45, "elapsed_time": "0:29:29", "remaining_time": "1:47:59", "throughput": 5165.23, "total_tokens": 9140640} {"current_steps": 9105, "total_steps": 42420, "loss": 0.3819, "lr": 0.0009605140554872043, "epoch": 4.292786421499293, "percentage": 21.46, "elapsed_time": "0:29:30", "remaining_time": "1:47:58", "throughput": 5165.38, "total_tokens": 9145632} {"current_steps": 9110, "total_steps": 42420, "loss": 0.3575, "lr": 0.0009604338893359092, "epoch": 4.295143800094295, "percentage": 21.48, "elapsed_time": "0:29:31", "remaining_time": "1:47:57", "throughput": 5165.48, "total_tokens": 9150336} {"current_steps": 9115, "total_steps": 42420, "loss": 0.3559, "lr": 0.0009603536452409742, "epoch": 4.297501178689298, "percentage": 21.49, "elapsed_time": "0:29:32", "remaining_time": "1:47:56", "throughput": 5165.86, "total_tokens": 9156672} {"current_steps": 9120, "total_steps": 42420, "loss": 0.2869, "lr": 0.0009602733232159834, "epoch": 4.2998585572843, "percentage": 21.5, "elapsed_time": "0:29:33", "remaining_time": "1:47:55", "throughput": 5166.03, "total_tokens": 9161632} {"current_steps": 9125, "total_steps": 42420, "loss": 0.3451, "lr": 0.000960192923274534, "epoch": 4.302215935879302, "percentage": 21.51, "elapsed_time": "0:29:34", "remaining_time": "1:47:54", "throughput": 5166.16, "total_tokens": 9166688} {"current_steps": 9130, "total_steps": 42420, "loss": 0.3404, "lr": 0.0009601124454302364, "epoch": 4.304573314474305, "percentage": 21.52, "elapsed_time": "0:29:35", "remaining_time": "1:47:53", "throughput": 5166.34, "total_tokens": 9171616} {"current_steps": 9135, "total_steps": 42420, "loss": 0.3381, "lr": 0.000960031889696714, "epoch": 4.306930693069307, "percentage": 21.53, "elapsed_time": "0:29:36", "remaining_time": "1:47:51", "throughput": 5166.39, "total_tokens": 9176224} {"current_steps": 9140, "total_steps": 42420, "loss": 0.3473, "lr": 0.0009599512560876034, "epoch": 4.30928807166431, "percentage": 21.55, "elapsed_time": "0:29:37", "remaining_time": "1:47:50", "throughput": 5166.61, "total_tokens": 9181472} {"current_steps": 9145, "total_steps": 42420, "loss": 0.3511, "lr": 0.0009598705446165549, "epoch": 4.311645450259312, "percentage": 21.56, "elapsed_time": "0:29:37", "remaining_time": "1:47:49", "throughput": 5166.71, "total_tokens": 9186240} {"current_steps": 9150, "total_steps": 42420, "loss": 0.3614, "lr": 0.0009597897552972312, "epoch": 4.314002828854314, "percentage": 21.57, "elapsed_time": "0:29:38", "remaining_time": "1:47:48", "throughput": 5166.83, "total_tokens": 9191136} {"current_steps": 9155, "total_steps": 42420, "loss": 0.3386, "lr": 0.0009597088881433088, "epoch": 4.316360207449316, "percentage": 21.58, "elapsed_time": "0:29:39", "remaining_time": "1:47:47", "throughput": 5167.04, "total_tokens": 9196544} {"current_steps": 9160, "total_steps": 42420, "loss": 0.3051, "lr": 0.0009596279431684771, "epoch": 4.318717586044318, "percentage": 21.59, "elapsed_time": "0:29:40", "remaining_time": "1:47:46", "throughput": 5167.27, "total_tokens": 9202144} {"current_steps": 9165, "total_steps": 42420, "loss": 0.4067, "lr": 0.0009595469203864386, "epoch": 4.321074964639321, "percentage": 21.61, "elapsed_time": "0:29:41", "remaining_time": "1:47:44", "throughput": 5167.39, "total_tokens": 9206880} {"current_steps": 9170, "total_steps": 42420, "loss": 0.3776, "lr": 0.0009594658198109092, "epoch": 4.323432343234323, "percentage": 21.62, "elapsed_time": "0:29:42", "remaining_time": "1:47:43", "throughput": 5167.36, "total_tokens": 9210912} {"current_steps": 9175, "total_steps": 42420, "loss": 0.375, "lr": 0.0009593846414556177, "epoch": 4.3257897218293255, "percentage": 21.63, "elapsed_time": "0:29:43", "remaining_time": "1:47:42", "throughput": 5167.5, "total_tokens": 9215872} {"current_steps": 9180, "total_steps": 42420, "loss": 0.353, "lr": 0.0009593033853343062, "epoch": 4.328147100424328, "percentage": 21.64, "elapsed_time": "0:29:44", "remaining_time": "1:47:41", "throughput": 5167.85, "total_tokens": 9221824} {"current_steps": 9185, "total_steps": 42420, "loss": 0.3406, "lr": 0.0009592220514607303, "epoch": 4.33050447901933, "percentage": 21.65, "elapsed_time": "0:29:45", "remaining_time": "1:47:40", "throughput": 5168.08, "total_tokens": 9227168} {"current_steps": 9190, "total_steps": 42420, "loss": 0.3601, "lr": 0.0009591406398486581, "epoch": 4.332861857614333, "percentage": 21.66, "elapsed_time": "0:29:46", "remaining_time": "1:47:39", "throughput": 5168.2, "total_tokens": 9232000} {"current_steps": 9195, "total_steps": 42420, "loss": 0.3446, "lr": 0.0009590591505118716, "epoch": 4.335219236209335, "percentage": 21.68, "elapsed_time": "0:29:47", "remaining_time": "1:47:37", "throughput": 5168.32, "total_tokens": 9237056} {"current_steps": 9200, "total_steps": 42420, "loss": 0.3802, "lr": 0.000958977583464165, "epoch": 4.3375766148043375, "percentage": 21.69, "elapsed_time": "0:29:48", "remaining_time": "1:47:36", "throughput": 5168.39, "total_tokens": 9241760} {"current_steps": 9205, "total_steps": 42420, "loss": 0.4415, "lr": 0.0009588959387193466, "epoch": 4.33993399339934, "percentage": 21.7, "elapsed_time": "0:29:49", "remaining_time": "1:47:35", "throughput": 5168.61, "total_tokens": 9247264} {"current_steps": 9210, "total_steps": 42420, "loss": 0.364, "lr": 0.0009588142162912374, "epoch": 4.342291371994342, "percentage": 21.71, "elapsed_time": "0:29:50", "remaining_time": "1:47:34", "throughput": 5168.78, "total_tokens": 9252384} {"current_steps": 9215, "total_steps": 42420, "loss": 0.3624, "lr": 0.0009587324161936715, "epoch": 4.344648750589345, "percentage": 21.72, "elapsed_time": "0:29:51", "remaining_time": "1:47:34", "throughput": 5169.18, "total_tokens": 9258752} {"current_steps": 9220, "total_steps": 42420, "loss": 0.3677, "lr": 0.0009586505384404966, "epoch": 4.347006129184347, "percentage": 21.74, "elapsed_time": "0:29:52", "remaining_time": "1:47:33", "throughput": 5169.53, "total_tokens": 9264704} {"current_steps": 9225, "total_steps": 42420, "loss": 0.3579, "lr": 0.0009585685830455728, "epoch": 4.3493635077793495, "percentage": 21.75, "elapsed_time": "0:29:53", "remaining_time": "1:47:32", "throughput": 5169.72, "total_tokens": 9269696} {"current_steps": 9230, "total_steps": 42420, "loss": 0.336, "lr": 0.000958486550022774, "epoch": 4.351720886374352, "percentage": 21.76, "elapsed_time": "0:29:54", "remaining_time": "1:47:31", "throughput": 5169.98, "total_tokens": 9275520} {"current_steps": 9235, "total_steps": 42420, "loss": 0.3414, "lr": 0.000958404439385987, "epoch": 4.354078264969354, "percentage": 21.77, "elapsed_time": "0:29:54", "remaining_time": "1:47:30", "throughput": 5170.04, "total_tokens": 9280032} {"current_steps": 9240, "total_steps": 42420, "loss": 0.3638, "lr": 0.0009583222511491117, "epoch": 4.356435643564357, "percentage": 21.78, "elapsed_time": "0:29:55", "remaining_time": "1:47:28", "throughput": 5170.1, "total_tokens": 9284672} {"current_steps": 9245, "total_steps": 42420, "loss": 0.3611, "lr": 0.0009582399853260609, "epoch": 4.358793022159359, "percentage": 21.79, "elapsed_time": "0:29:56", "remaining_time": "1:47:28", "throughput": 5170.5, "total_tokens": 9291232} {"current_steps": 9250, "total_steps": 42420, "loss": 0.3449, "lr": 0.0009581576419307613, "epoch": 4.3611504007543616, "percentage": 21.81, "elapsed_time": "0:29:57", "remaining_time": "1:47:27", "throughput": 5170.72, "total_tokens": 9296768} {"current_steps": 9255, "total_steps": 42420, "loss": 0.347, "lr": 0.0009580752209771519, "epoch": 4.363507779349364, "percentage": 21.82, "elapsed_time": "0:29:58", "remaining_time": "1:47:26", "throughput": 5170.84, "total_tokens": 9301696} {"current_steps": 9260, "total_steps": 42420, "loss": 0.2937, "lr": 0.0009579927224791854, "epoch": 4.3658651579443655, "percentage": 21.83, "elapsed_time": "0:29:59", "remaining_time": "1:47:25", "throughput": 5171.1, "total_tokens": 9307232} {"current_steps": 9265, "total_steps": 42420, "loss": 0.2963, "lr": 0.0009579101464508271, "epoch": 4.368222536539368, "percentage": 21.84, "elapsed_time": "0:30:00", "remaining_time": "1:47:24", "throughput": 5171.15, "total_tokens": 9311936} {"current_steps": 9270, "total_steps": 42420, "loss": 0.4428, "lr": 0.0009578274929060559, "epoch": 4.37057991513437, "percentage": 21.85, "elapsed_time": "0:30:01", "remaining_time": "1:47:23", "throughput": 5171.37, "total_tokens": 9317568} {"current_steps": 9275, "total_steps": 42420, "loss": 0.3363, "lr": 0.0009577447618588636, "epoch": 4.372937293729373, "percentage": 21.86, "elapsed_time": "0:30:02", "remaining_time": "1:47:21", "throughput": 5171.37, "total_tokens": 9321792} {"current_steps": 9280, "total_steps": 42420, "loss": 0.3496, "lr": 0.0009576619533232552, "epoch": 4.375294672324375, "percentage": 21.88, "elapsed_time": "0:30:03", "remaining_time": "1:47:21", "throughput": 5171.69, "total_tokens": 9328000} {"current_steps": 9285, "total_steps": 42420, "loss": 0.3394, "lr": 0.0009575790673132487, "epoch": 4.3776520509193775, "percentage": 21.89, "elapsed_time": "0:30:04", "remaining_time": "1:47:19", "throughput": 5171.78, "total_tokens": 9332800} {"current_steps": 9290, "total_steps": 42420, "loss": 0.3679, "lr": 0.0009574961038428756, "epoch": 4.38000942951438, "percentage": 21.9, "elapsed_time": "0:30:05", "remaining_time": "1:47:18", "throughput": 5171.71, "total_tokens": 9336640} {"current_steps": 9295, "total_steps": 42420, "loss": 0.3309, "lr": 0.0009574130629261797, "epoch": 4.382366808109382, "percentage": 21.91, "elapsed_time": "0:30:06", "remaining_time": "1:47:16", "throughput": 5171.59, "total_tokens": 9340608} {"current_steps": 9300, "total_steps": 42420, "loss": 0.3534, "lr": 0.0009573299445772188, "epoch": 4.384724186704385, "percentage": 21.92, "elapsed_time": "0:30:07", "remaining_time": "1:47:15", "throughput": 5171.93, "total_tokens": 9346688} {"current_steps": 9305, "total_steps": 42420, "loss": 0.3089, "lr": 0.0009572467488100634, "epoch": 4.387081565299387, "percentage": 21.94, "elapsed_time": "0:30:07", "remaining_time": "1:47:14", "throughput": 5171.88, "total_tokens": 9350752} {"current_steps": 9310, "total_steps": 42420, "loss": 0.3571, "lr": 0.0009571634756387968, "epoch": 4.3894389438943895, "percentage": 21.95, "elapsed_time": "0:30:09", "remaining_time": "1:47:13", "throughput": 5172.1, "total_tokens": 9356544} {"current_steps": 9315, "total_steps": 42420, "loss": 0.3483, "lr": 0.0009570801250775161, "epoch": 4.391796322489392, "percentage": 21.96, "elapsed_time": "0:30:09", "remaining_time": "1:47:12", "throughput": 5172.11, "total_tokens": 9360800} {"current_steps": 9320, "total_steps": 42420, "loss": 0.3564, "lr": 0.000956996697140331, "epoch": 4.394153701084394, "percentage": 21.97, "elapsed_time": "0:30:10", "remaining_time": "1:47:10", "throughput": 5172.04, "total_tokens": 9365088} {"current_steps": 9325, "total_steps": 42420, "loss": 0.3668, "lr": 0.0009569131918413644, "epoch": 4.396511079679397, "percentage": 21.98, "elapsed_time": "0:30:11", "remaining_time": "1:47:09", "throughput": 5172.31, "total_tokens": 9370752} {"current_steps": 9330, "total_steps": 42420, "loss": 0.3463, "lr": 0.0009568296091947523, "epoch": 4.398868458274399, "percentage": 21.99, "elapsed_time": "0:30:12", "remaining_time": "1:47:08", "throughput": 5172.25, "total_tokens": 9374848} {"current_steps": 9335, "total_steps": 42420, "loss": 0.3626, "lr": 0.000956745949214644, "epoch": 4.4012258368694015, "percentage": 22.01, "elapsed_time": "0:30:13", "remaining_time": "1:47:07", "throughput": 5172.3, "total_tokens": 9379456} {"current_steps": 9340, "total_steps": 42420, "loss": 0.3798, "lr": 0.0009566622119152016, "epoch": 4.403583215464404, "percentage": 22.02, "elapsed_time": "0:30:14", "remaining_time": "1:47:05", "throughput": 5172.41, "total_tokens": 9384512} {"current_steps": 9345, "total_steps": 42420, "loss": 0.2939, "lr": 0.0009565783973106004, "epoch": 4.405940594059406, "percentage": 22.03, "elapsed_time": "0:30:15", "remaining_time": "1:47:04", "throughput": 5172.39, "total_tokens": 9388832} {"current_steps": 9350, "total_steps": 42420, "loss": 0.3614, "lr": 0.0009564945054150287, "epoch": 4.408297972654409, "percentage": 22.04, "elapsed_time": "0:30:16", "remaining_time": "1:47:03", "throughput": 5172.53, "total_tokens": 9393760} {"current_steps": 9355, "total_steps": 42420, "loss": 0.5882, "lr": 0.0009564105362426883, "epoch": 4.41065535124941, "percentage": 22.05, "elapsed_time": "0:30:16", "remaining_time": "1:47:01", "throughput": 5172.62, "total_tokens": 9398432} {"current_steps": 9360, "total_steps": 42420, "loss": 0.3387, "lr": 0.0009563264898077932, "epoch": 4.413012729844413, "percentage": 22.07, "elapsed_time": "0:30:17", "remaining_time": "1:47:00", "throughput": 5172.8, "total_tokens": 9403456} {"current_steps": 9365, "total_steps": 42420, "loss": 0.3665, "lr": 0.0009562423661245716, "epoch": 4.415370108439415, "percentage": 22.08, "elapsed_time": "0:30:18", "remaining_time": "1:46:59", "throughput": 5172.87, "total_tokens": 9408064} {"current_steps": 9370, "total_steps": 42420, "loss": 0.351, "lr": 0.0009561581652072639, "epoch": 4.417727487034417, "percentage": 22.09, "elapsed_time": "0:30:19", "remaining_time": "1:46:58", "throughput": 5172.89, "total_tokens": 9412672} {"current_steps": 9375, "total_steps": 42420, "loss": 0.3365, "lr": 0.0009560738870701239, "epoch": 4.42008486562942, "percentage": 22.1, "elapsed_time": "0:30:20", "remaining_time": "1:46:56", "throughput": 5172.96, "total_tokens": 9417408} {"current_steps": 9380, "total_steps": 42420, "loss": 0.3941, "lr": 0.0009559895317274186, "epoch": 4.422442244224422, "percentage": 22.11, "elapsed_time": "0:30:21", "remaining_time": "1:46:55", "throughput": 5172.98, "total_tokens": 9421888} {"current_steps": 9385, "total_steps": 42420, "loss": 0.3411, "lr": 0.0009559050991934278, "epoch": 4.424799622819425, "percentage": 22.12, "elapsed_time": "0:30:22", "remaining_time": "1:46:54", "throughput": 5173.08, "total_tokens": 9426944} {"current_steps": 9390, "total_steps": 42420, "loss": 0.3382, "lr": 0.0009558205894824445, "epoch": 4.427157001414427, "percentage": 22.14, "elapsed_time": "0:30:23", "remaining_time": "1:46:53", "throughput": 5173.25, "total_tokens": 9432256} {"current_steps": 9395, "total_steps": 42420, "loss": 0.3243, "lr": 0.000955736002608775, "epoch": 4.429514380009429, "percentage": 22.15, "elapsed_time": "0:30:24", "remaining_time": "1:46:52", "throughput": 5173.27, "total_tokens": 9436608} {"current_steps": 9400, "total_steps": 42420, "loss": 0.3105, "lr": 0.0009556513385867379, "epoch": 4.431871758604432, "percentage": 22.16, "elapsed_time": "0:30:25", "remaining_time": "1:46:51", "throughput": 5173.5, "total_tokens": 9442048} {"current_steps": 9405, "total_steps": 42420, "loss": 0.339, "lr": 0.0009555665974306658, "epoch": 4.434229137199434, "percentage": 22.17, "elapsed_time": "0:30:25", "remaining_time": "1:46:49", "throughput": 5173.6, "total_tokens": 9446656} {"current_steps": 9410, "total_steps": 42420, "loss": 0.3856, "lr": 0.000955481779154904, "epoch": 4.436586515794437, "percentage": 22.18, "elapsed_time": "0:30:26", "remaining_time": "1:46:48", "throughput": 5173.77, "total_tokens": 9451840} {"current_steps": 9415, "total_steps": 42420, "loss": 0.3732, "lr": 0.0009553968837738105, "epoch": 4.438943894389439, "percentage": 22.19, "elapsed_time": "0:30:27", "remaining_time": "1:46:47", "throughput": 5173.99, "total_tokens": 9457312} {"current_steps": 9420, "total_steps": 42420, "loss": 0.3639, "lr": 0.0009553119113017567, "epoch": 4.441301272984441, "percentage": 22.21, "elapsed_time": "0:30:28", "remaining_time": "1:46:46", "throughput": 5174.32, "total_tokens": 9463232} {"current_steps": 9425, "total_steps": 42420, "loss": 0.3298, "lr": 0.0009552268617531272, "epoch": 4.443658651579444, "percentage": 22.22, "elapsed_time": "0:30:29", "remaining_time": "1:46:45", "throughput": 5174.53, "total_tokens": 9468416} {"current_steps": 9430, "total_steps": 42420, "loss": 0.3033, "lr": 0.0009551417351423193, "epoch": 4.446016030174446, "percentage": 22.23, "elapsed_time": "0:30:30", "remaining_time": "1:46:44", "throughput": 5174.57, "total_tokens": 9472928} {"current_steps": 9435, "total_steps": 42420, "loss": 0.3658, "lr": 0.0009550565314837435, "epoch": 4.448373408769449, "percentage": 22.24, "elapsed_time": "0:30:31", "remaining_time": "1:46:43", "throughput": 5174.73, "total_tokens": 9477888} {"current_steps": 9440, "total_steps": 42420, "loss": 0.3707, "lr": 0.0009549712507918234, "epoch": 4.450730787364451, "percentage": 22.25, "elapsed_time": "0:30:32", "remaining_time": "1:46:41", "throughput": 5174.63, "total_tokens": 9481888} {"current_steps": 9445, "total_steps": 42420, "loss": 0.3174, "lr": 0.0009548858930809954, "epoch": 4.4530881659594534, "percentage": 22.27, "elapsed_time": "0:30:33", "remaining_time": "1:46:40", "throughput": 5174.83, "total_tokens": 9487072} {"current_steps": 9450, "total_steps": 42420, "loss": 0.3334, "lr": 0.0009548004583657092, "epoch": 4.455445544554456, "percentage": 22.28, "elapsed_time": "0:30:34", "remaining_time": "1:46:39", "throughput": 5174.89, "total_tokens": 9491808} {"current_steps": 9455, "total_steps": 42420, "loss": 0.2662, "lr": 0.0009547149466604274, "epoch": 4.457802923149458, "percentage": 22.29, "elapsed_time": "0:30:35", "remaining_time": "1:46:37", "throughput": 5174.78, "total_tokens": 9495904} {"current_steps": 9460, "total_steps": 42420, "loss": 0.3607, "lr": 0.000954629357979626, "epoch": 4.46016030174446, "percentage": 22.3, "elapsed_time": "0:30:35", "remaining_time": "1:46:36", "throughput": 5174.9, "total_tokens": 9500896} {"current_steps": 9465, "total_steps": 42420, "loss": 0.4012, "lr": 0.0009545436923377934, "epoch": 4.462517680339462, "percentage": 22.31, "elapsed_time": "0:30:37", "remaining_time": "1:46:36", "throughput": 5175.31, "total_tokens": 9507168} {"current_steps": 9470, "total_steps": 42420, "loss": 0.3374, "lr": 0.0009544579497494313, "epoch": 4.464875058934465, "percentage": 22.32, "elapsed_time": "0:30:37", "remaining_time": "1:46:34", "throughput": 5175.51, "total_tokens": 9512192} {"current_steps": 9475, "total_steps": 42420, "loss": 0.3506, "lr": 0.0009543721302290546, "epoch": 4.467232437529467, "percentage": 22.34, "elapsed_time": "0:30:38", "remaining_time": "1:46:33", "throughput": 5175.55, "total_tokens": 9516864} {"current_steps": 9480, "total_steps": 42420, "loss": 0.3723, "lr": 0.0009542862337911911, "epoch": 4.469589816124469, "percentage": 22.35, "elapsed_time": "0:30:39", "remaining_time": "1:46:32", "throughput": 5175.48, "total_tokens": 9520928} {"current_steps": 9485, "total_steps": 42420, "loss": 0.3584, "lr": 0.0009542002604503814, "epoch": 4.471947194719472, "percentage": 22.36, "elapsed_time": "0:30:40", "remaining_time": "1:46:31", "throughput": 5175.81, "total_tokens": 9526624} {"current_steps": 9490, "total_steps": 42420, "loss": 0.3944, "lr": 0.0009541142102211797, "epoch": 4.474304573314474, "percentage": 22.37, "elapsed_time": "0:30:41", "remaining_time": "1:46:30", "throughput": 5176.03, "total_tokens": 9532000} {"current_steps": 9495, "total_steps": 42420, "loss": 0.3795, "lr": 0.0009540280831181524, "epoch": 4.476661951909477, "percentage": 22.38, "elapsed_time": "0:30:42", "remaining_time": "1:46:29", "throughput": 5176.14, "total_tokens": 9537280} {"current_steps": 9500, "total_steps": 42420, "loss": 0.3579, "lr": 0.0009539418791558799, "epoch": 4.479019330504479, "percentage": 22.4, "elapsed_time": "0:30:43", "remaining_time": "1:46:28", "throughput": 5176.31, "total_tokens": 9542304} {"current_steps": 9505, "total_steps": 42420, "loss": 0.351, "lr": 0.0009538555983489546, "epoch": 4.481376709099481, "percentage": 22.41, "elapsed_time": "0:30:44", "remaining_time": "1:46:27", "throughput": 5176.52, "total_tokens": 9547904} {"current_steps": 9510, "total_steps": 42420, "loss": 0.3608, "lr": 0.0009537692407119825, "epoch": 4.483734087694484, "percentage": 22.42, "elapsed_time": "0:30:45", "remaining_time": "1:46:26", "throughput": 5176.62, "total_tokens": 9552960} {"current_steps": 9515, "total_steps": 42420, "loss": 0.3518, "lr": 0.0009536828062595824, "epoch": 4.486091466289486, "percentage": 22.43, "elapsed_time": "0:30:46", "remaining_time": "1:46:25", "throughput": 5176.77, "total_tokens": 9558144} {"current_steps": 9520, "total_steps": 42420, "loss": 0.3642, "lr": 0.0009535962950063864, "epoch": 4.488448844884489, "percentage": 22.44, "elapsed_time": "0:30:47", "remaining_time": "1:46:24", "throughput": 5177.06, "total_tokens": 9563840} {"current_steps": 9525, "total_steps": 42420, "loss": 0.3811, "lr": 0.0009535097069670393, "epoch": 4.490806223479491, "percentage": 22.45, "elapsed_time": "0:30:48", "remaining_time": "1:46:22", "throughput": 5177.08, "total_tokens": 9568352} {"current_steps": 9530, "total_steps": 42420, "loss": 0.3409, "lr": 0.0009534230421561989, "epoch": 4.493163602074493, "percentage": 22.47, "elapsed_time": "0:30:49", "remaining_time": "1:46:21", "throughput": 5177.16, "total_tokens": 9573280} {"current_steps": 9535, "total_steps": 42420, "loss": 0.358, "lr": 0.0009533363005885363, "epoch": 4.495520980669496, "percentage": 22.48, "elapsed_time": "0:30:49", "remaining_time": "1:46:20", "throughput": 5177.19, "total_tokens": 9577632} {"current_steps": 9540, "total_steps": 42420, "loss": 0.351, "lr": 0.0009532494822787352, "epoch": 4.497878359264498, "percentage": 22.49, "elapsed_time": "0:30:50", "remaining_time": "1:46:19", "throughput": 5177.29, "total_tokens": 9582432} {"current_steps": 9545, "total_steps": 42420, "loss": 0.3409, "lr": 0.0009531625872414924, "epoch": 4.500235737859501, "percentage": 22.5, "elapsed_time": "0:30:51", "remaining_time": "1:46:17", "throughput": 5177.3, "total_tokens": 9586848} {"current_steps": 9550, "total_steps": 42420, "loss": 0.3161, "lr": 0.0009530756154915178, "epoch": 4.502593116454502, "percentage": 22.51, "elapsed_time": "0:30:52", "remaining_time": "1:46:16", "throughput": 5177.34, "total_tokens": 9591360} {"current_steps": 9555, "total_steps": 42420, "loss": 0.3671, "lr": 0.0009529885670435343, "epoch": 4.5049504950495045, "percentage": 22.52, "elapsed_time": "0:30:53", "remaining_time": "1:46:15", "throughput": 5177.52, "total_tokens": 9596576} {"current_steps": 9560, "total_steps": 42420, "loss": 0.3258, "lr": 0.0009529014419122779, "epoch": 4.507307873644507, "percentage": 22.54, "elapsed_time": "0:30:54", "remaining_time": "1:46:13", "throughput": 5177.54, "total_tokens": 9601056} {"current_steps": 9565, "total_steps": 42420, "loss": 0.4086, "lr": 0.000952814240112497, "epoch": 4.509665252239509, "percentage": 22.55, "elapsed_time": "0:30:55", "remaining_time": "1:46:12", "throughput": 5177.51, "total_tokens": 9605248} {"current_steps": 9570, "total_steps": 42420, "loss": 0.323, "lr": 0.0009527269616589535, "epoch": 4.512022630834512, "percentage": 22.56, "elapsed_time": "0:30:56", "remaining_time": "1:46:11", "throughput": 5177.73, "total_tokens": 9610784} {"current_steps": 9575, "total_steps": 42420, "loss": 0.3646, "lr": 0.0009526396065664225, "epoch": 4.514380009429514, "percentage": 22.57, "elapsed_time": "0:30:57", "remaining_time": "1:46:10", "throughput": 5178.01, "total_tokens": 9616352} {"current_steps": 9580, "total_steps": 42420, "loss": 0.3678, "lr": 0.0009525521748496914, "epoch": 4.5167373880245165, "percentage": 22.58, "elapsed_time": "0:30:58", "remaining_time": "1:46:09", "throughput": 5178.09, "total_tokens": 9621088} {"current_steps": 9585, "total_steps": 42420, "loss": 0.3166, "lr": 0.000952464666523561, "epoch": 4.519094766619519, "percentage": 22.6, "elapsed_time": "0:30:58", "remaining_time": "1:46:08", "throughput": 5178.18, "total_tokens": 9625952} {"current_steps": 9590, "total_steps": 42420, "loss": 0.3679, "lr": 0.0009523770816028449, "epoch": 4.521452145214521, "percentage": 22.61, "elapsed_time": "0:30:59", "remaining_time": "1:46:06", "throughput": 5178.28, "total_tokens": 9630848} {"current_steps": 9595, "total_steps": 42420, "loss": 0.288, "lr": 0.00095228942010237, "epoch": 4.523809523809524, "percentage": 22.62, "elapsed_time": "0:31:00", "remaining_time": "1:46:05", "throughput": 5178.33, "total_tokens": 9635360} {"current_steps": 9600, "total_steps": 42420, "loss": 0.2725, "lr": 0.0009522016820369757, "epoch": 4.526166902404526, "percentage": 22.63, "elapsed_time": "0:31:01", "remaining_time": "1:46:04", "throughput": 5178.56, "total_tokens": 9641120} {"current_steps": 9605, "total_steps": 42420, "loss": 0.4071, "lr": 0.0009521138674215144, "epoch": 4.5285242809995285, "percentage": 22.64, "elapsed_time": "0:31:02", "remaining_time": "1:46:04", "throughput": 5178.96, "total_tokens": 9647808} {"current_steps": 9610, "total_steps": 42420, "loss": 0.2917, "lr": 0.0009520259762708518, "epoch": 4.530881659594531, "percentage": 22.65, "elapsed_time": "0:31:03", "remaining_time": "1:46:03", "throughput": 5179.1, "total_tokens": 9653344} {"current_steps": 9615, "total_steps": 42420, "loss": 0.2726, "lr": 0.0009519380085998665, "epoch": 4.533239038189533, "percentage": 22.67, "elapsed_time": "0:31:04", "remaining_time": "1:46:02", "throughput": 5179.23, "total_tokens": 9658208} {"current_steps": 9620, "total_steps": 42420, "loss": 0.3163, "lr": 0.0009518499644234498, "epoch": 4.535596416784536, "percentage": 22.68, "elapsed_time": "0:31:05", "remaining_time": "1:46:01", "throughput": 5179.51, "total_tokens": 9664000} {"current_steps": 9625, "total_steps": 42420, "loss": 0.2924, "lr": 0.0009517618437565061, "epoch": 4.537953795379538, "percentage": 22.69, "elapsed_time": "0:31:06", "remaining_time": "1:46:00", "throughput": 5179.55, "total_tokens": 9668832} {"current_steps": 9630, "total_steps": 42420, "loss": 0.3149, "lr": 0.0009516736466139527, "epoch": 4.5403111739745405, "percentage": 22.7, "elapsed_time": "0:31:07", "remaining_time": "1:45:59", "throughput": 5179.59, "total_tokens": 9673280} {"current_steps": 9635, "total_steps": 42420, "loss": 0.4124, "lr": 0.00095158537301072, "epoch": 4.542668552569543, "percentage": 22.71, "elapsed_time": "0:31:08", "remaining_time": "1:45:58", "throughput": 5179.78, "total_tokens": 9678624} {"current_steps": 9640, "total_steps": 42420, "loss": 0.3457, "lr": 0.000951497022961751, "epoch": 4.545025931164545, "percentage": 22.73, "elapsed_time": "0:31:09", "remaining_time": "1:45:56", "throughput": 5179.83, "total_tokens": 9683168} {"current_steps": 9645, "total_steps": 42420, "loss": 0.4029, "lr": 0.0009514085964820021, "epoch": 4.547383309759548, "percentage": 22.74, "elapsed_time": "0:31:10", "remaining_time": "1:45:55", "throughput": 5179.88, "total_tokens": 9687904} {"current_steps": 9650, "total_steps": 42420, "loss": 0.3718, "lr": 0.0009513200935864423, "epoch": 4.54974068835455, "percentage": 22.75, "elapsed_time": "0:31:11", "remaining_time": "1:45:54", "throughput": 5180.07, "total_tokens": 9693344} {"current_steps": 9655, "total_steps": 42420, "loss": 0.3359, "lr": 0.0009512315142900536, "epoch": 4.5520980669495525, "percentage": 22.76, "elapsed_time": "0:31:12", "remaining_time": "1:45:53", "throughput": 5180.2, "total_tokens": 9698208} {"current_steps": 9660, "total_steps": 42420, "loss": 0.2573, "lr": 0.0009511428586078309, "epoch": 4.554455445544555, "percentage": 22.77, "elapsed_time": "0:31:13", "remaining_time": "1:45:52", "throughput": 5180.26, "total_tokens": 9702848} {"current_steps": 9665, "total_steps": 42420, "loss": 0.2827, "lr": 0.0009510541265547823, "epoch": 4.5568128241395565, "percentage": 22.78, "elapsed_time": "0:31:13", "remaining_time": "1:45:50", "throughput": 5180.23, "total_tokens": 9707168} {"current_steps": 9670, "total_steps": 42420, "loss": 0.4336, "lr": 0.0009509653181459286, "epoch": 4.559170202734559, "percentage": 22.8, "elapsed_time": "0:31:14", "remaining_time": "1:45:49", "throughput": 5180.31, "total_tokens": 9712096} {"current_steps": 9675, "total_steps": 42420, "loss": 0.3329, "lr": 0.0009508764333963034, "epoch": 4.561527581329561, "percentage": 22.81, "elapsed_time": "0:31:15", "remaining_time": "1:45:48", "throughput": 5180.35, "total_tokens": 9716448} {"current_steps": 9680, "total_steps": 42420, "loss": 0.35, "lr": 0.0009507874723209534, "epoch": 4.563884959924564, "percentage": 22.82, "elapsed_time": "0:31:16", "remaining_time": "1:45:47", "throughput": 5180.67, "total_tokens": 9722560} {"current_steps": 9685, "total_steps": 42420, "loss": 0.3631, "lr": 0.0009506984349349384, "epoch": 4.566242338519566, "percentage": 22.83, "elapsed_time": "0:31:17", "remaining_time": "1:45:46", "throughput": 5180.75, "total_tokens": 9727232} {"current_steps": 9690, "total_steps": 42420, "loss": 0.347, "lr": 0.0009506093212533307, "epoch": 4.5685997171145685, "percentage": 22.84, "elapsed_time": "0:31:18", "remaining_time": "1:45:44", "throughput": 5180.79, "total_tokens": 9731712} {"current_steps": 9695, "total_steps": 42420, "loss": 0.3006, "lr": 0.0009505201312912159, "epoch": 4.570957095709571, "percentage": 22.85, "elapsed_time": "0:31:19", "remaining_time": "1:45:43", "throughput": 5180.97, "total_tokens": 9736896} {"current_steps": 9700, "total_steps": 42420, "loss": 0.3652, "lr": 0.0009504308650636923, "epoch": 4.573314474304573, "percentage": 22.87, "elapsed_time": "0:31:20", "remaining_time": "1:45:43", "throughput": 5181.24, "total_tokens": 9742880} {"current_steps": 9705, "total_steps": 42420, "loss": 0.3832, "lr": 0.000950341522585871, "epoch": 4.575671852899576, "percentage": 22.88, "elapsed_time": "0:31:21", "remaining_time": "1:45:42", "throughput": 5181.43, "total_tokens": 9748224} {"current_steps": 9710, "total_steps": 42420, "loss": 0.322, "lr": 0.0009502521038728764, "epoch": 4.578029231494578, "percentage": 22.89, "elapsed_time": "0:31:22", "remaining_time": "1:45:40", "throughput": 5181.62, "total_tokens": 9753408} {"current_steps": 9715, "total_steps": 42420, "loss": 0.367, "lr": 0.0009501626089398454, "epoch": 4.5803866100895805, "percentage": 22.9, "elapsed_time": "0:31:23", "remaining_time": "1:45:39", "throughput": 5181.6, "total_tokens": 9757728} {"current_steps": 9720, "total_steps": 42420, "loss": 0.3522, "lr": 0.0009500730378019281, "epoch": 4.582743988684583, "percentage": 22.91, "elapsed_time": "0:31:24", "remaining_time": "1:45:38", "throughput": 5181.62, "total_tokens": 9762272} {"current_steps": 9725, "total_steps": 42420, "loss": 0.3334, "lr": 0.0009499833904742873, "epoch": 4.585101367279585, "percentage": 22.93, "elapsed_time": "0:31:24", "remaining_time": "1:45:37", "throughput": 5181.78, "total_tokens": 9767456} {"current_steps": 9730, "total_steps": 42420, "loss": 0.3882, "lr": 0.0009498936669720987, "epoch": 4.587458745874588, "percentage": 22.94, "elapsed_time": "0:31:25", "remaining_time": "1:45:36", "throughput": 5182.01, "total_tokens": 9772928} {"current_steps": 9735, "total_steps": 42420, "loss": 0.3244, "lr": 0.0009498038673105511, "epoch": 4.58981612446959, "percentage": 22.95, "elapsed_time": "0:31:26", "remaining_time": "1:45:35", "throughput": 5182.16, "total_tokens": 9777984} {"current_steps": 9740, "total_steps": 42420, "loss": 0.3346, "lr": 0.0009497139915048459, "epoch": 4.5921735030645925, "percentage": 22.96, "elapsed_time": "0:31:27", "remaining_time": "1:45:33", "throughput": 5182.24, "total_tokens": 9782688} {"current_steps": 9745, "total_steps": 42420, "loss": 0.3396, "lr": 0.0009496240395701979, "epoch": 4.594530881659595, "percentage": 22.97, "elapsed_time": "0:31:28", "remaining_time": "1:45:32", "throughput": 5182.26, "total_tokens": 9787328} {"current_steps": 9750, "total_steps": 42420, "loss": 0.3361, "lr": 0.0009495340115218339, "epoch": 4.596888260254596, "percentage": 22.98, "elapsed_time": "0:31:29", "remaining_time": "1:45:31", "throughput": 5182.45, "total_tokens": 9792480} {"current_steps": 9755, "total_steps": 42420, "loss": 0.3405, "lr": 0.0009494439073749945, "epoch": 4.599245638849599, "percentage": 23.0, "elapsed_time": "0:31:30", "remaining_time": "1:45:30", "throughput": 5182.48, "total_tokens": 9797024} {"current_steps": 9760, "total_steps": 42420, "loss": 0.3334, "lr": 0.0009493537271449326, "epoch": 4.601603017444601, "percentage": 23.01, "elapsed_time": "0:31:31", "remaining_time": "1:45:29", "throughput": 5182.86, "total_tokens": 9803136} {"current_steps": 9765, "total_steps": 42420, "loss": 0.327, "lr": 0.0009492634708469145, "epoch": 4.603960396039604, "percentage": 23.02, "elapsed_time": "0:31:32", "remaining_time": "1:45:28", "throughput": 5183.05, "total_tokens": 9808288} {"current_steps": 9770, "total_steps": 42420, "loss": 0.3451, "lr": 0.0009491731384962188, "epoch": 4.606317774634606, "percentage": 23.03, "elapsed_time": "0:31:33", "remaining_time": "1:45:26", "throughput": 5183.06, "total_tokens": 9812768} {"current_steps": 9775, "total_steps": 42420, "loss": 0.3611, "lr": 0.0009490827301081372, "epoch": 4.608675153229608, "percentage": 23.04, "elapsed_time": "0:31:34", "remaining_time": "1:45:26", "throughput": 5183.48, "total_tokens": 9819424} {"current_steps": 9780, "total_steps": 42420, "loss": 0.3599, "lr": 0.0009489922456979744, "epoch": 4.611032531824611, "percentage": 23.06, "elapsed_time": "0:31:35", "remaining_time": "1:45:25", "throughput": 5183.64, "total_tokens": 9824672} {"current_steps": 9785, "total_steps": 42420, "loss": 0.3304, "lr": 0.0009489016852810479, "epoch": 4.613389910419613, "percentage": 23.07, "elapsed_time": "0:31:36", "remaining_time": "1:45:24", "throughput": 5183.85, "total_tokens": 9830304} {"current_steps": 9790, "total_steps": 42420, "loss": 0.3667, "lr": 0.0009488110488726879, "epoch": 4.615747289014616, "percentage": 23.08, "elapsed_time": "0:31:37", "remaining_time": "1:45:24", "throughput": 5184.09, "total_tokens": 9836288} {"current_steps": 9795, "total_steps": 42420, "loss": 0.3711, "lr": 0.0009487203364882378, "epoch": 4.618104667609618, "percentage": 23.09, "elapsed_time": "0:31:38", "remaining_time": "1:45:22", "throughput": 5184.28, "total_tokens": 9841568} {"current_steps": 9800, "total_steps": 42420, "loss": 0.3424, "lr": 0.0009486295481430536, "epoch": 4.62046204620462, "percentage": 23.1, "elapsed_time": "0:31:39", "remaining_time": "1:45:21", "throughput": 5184.47, "total_tokens": 9846592} {"current_steps": 9805, "total_steps": 42420, "loss": 0.3375, "lr": 0.0009485386838525041, "epoch": 4.622819424799623, "percentage": 23.11, "elapsed_time": "0:31:40", "remaining_time": "1:45:20", "throughput": 5184.54, "total_tokens": 9851488} {"current_steps": 9810, "total_steps": 42420, "loss": 0.3503, "lr": 0.0009484477436319712, "epoch": 4.625176803394625, "percentage": 23.13, "elapsed_time": "0:31:41", "remaining_time": "1:45:19", "throughput": 5184.72, "total_tokens": 9856672} {"current_steps": 9815, "total_steps": 42420, "loss": 0.3837, "lr": 0.0009483567274968495, "epoch": 4.627534181989628, "percentage": 23.14, "elapsed_time": "0:31:42", "remaining_time": "1:45:18", "throughput": 5184.86, "total_tokens": 9861664} {"current_steps": 9820, "total_steps": 42420, "loss": 0.3623, "lr": 0.0009482656354625465, "epoch": 4.62989156058463, "percentage": 23.15, "elapsed_time": "0:31:43", "remaining_time": "1:45:17", "throughput": 5185.34, "total_tokens": 9868480} {"current_steps": 9825, "total_steps": 42420, "loss": 0.3473, "lr": 0.0009481744675444826, "epoch": 4.632248939179632, "percentage": 23.16, "elapsed_time": "0:31:44", "remaining_time": "1:45:16", "throughput": 5185.44, "total_tokens": 9873216} {"current_steps": 9830, "total_steps": 42420, "loss": 0.3311, "lr": 0.0009480832237580909, "epoch": 4.634606317774635, "percentage": 23.17, "elapsed_time": "0:31:44", "remaining_time": "1:45:15", "throughput": 5185.64, "total_tokens": 9878528} {"current_steps": 9835, "total_steps": 42420, "loss": 0.3798, "lr": 0.0009479919041188174, "epoch": 4.636963696369637, "percentage": 23.18, "elapsed_time": "0:31:45", "remaining_time": "1:45:14", "throughput": 5185.72, "total_tokens": 9883264} {"current_steps": 9840, "total_steps": 42420, "loss": 0.3524, "lr": 0.0009479005086421209, "epoch": 4.63932107496464, "percentage": 23.2, "elapsed_time": "0:31:46", "remaining_time": "1:45:13", "throughput": 5185.85, "total_tokens": 9888352} {"current_steps": 9845, "total_steps": 42420, "loss": 0.2627, "lr": 0.0009478090373434733, "epoch": 4.641678453559642, "percentage": 23.21, "elapsed_time": "0:31:47", "remaining_time": "1:45:12", "throughput": 5186.2, "total_tokens": 9894464} {"current_steps": 9850, "total_steps": 42420, "loss": 0.3423, "lr": 0.0009477174902383592, "epoch": 4.644035832154644, "percentage": 23.22, "elapsed_time": "0:31:48", "remaining_time": "1:45:11", "throughput": 5186.41, "total_tokens": 9900064} {"current_steps": 9855, "total_steps": 42420, "loss": 0.4012, "lr": 0.0009476258673422755, "epoch": 4.646393210749647, "percentage": 23.23, "elapsed_time": "0:31:49", "remaining_time": "1:45:10", "throughput": 5186.4, "total_tokens": 9904576} {"current_steps": 9860, "total_steps": 42420, "loss": 0.3854, "lr": 0.000947534168670733, "epoch": 4.648750589344649, "percentage": 23.24, "elapsed_time": "0:31:50", "remaining_time": "1:45:09", "throughput": 5186.49, "total_tokens": 9909440} {"current_steps": 9865, "total_steps": 42420, "loss": 0.3519, "lr": 0.0009474423942392543, "epoch": 4.651107967939651, "percentage": 23.26, "elapsed_time": "0:31:51", "remaining_time": "1:45:07", "throughput": 5186.48, "total_tokens": 9913600} {"current_steps": 9870, "total_steps": 42420, "loss": 0.3415, "lr": 0.0009473505440633754, "epoch": 4.653465346534653, "percentage": 23.27, "elapsed_time": "0:31:52", "remaining_time": "1:45:06", "throughput": 5186.57, "total_tokens": 9918464} {"current_steps": 9875, "total_steps": 42420, "loss": 0.4089, "lr": 0.000947258618158645, "epoch": 4.655822725129656, "percentage": 23.28, "elapsed_time": "0:31:53", "remaining_time": "1:45:06", "throughput": 5187.04, "total_tokens": 9925152} {"current_steps": 9880, "total_steps": 42420, "loss": 0.3368, "lr": 0.0009471666165406245, "epoch": 4.658180103724658, "percentage": 23.29, "elapsed_time": "0:31:54", "remaining_time": "1:45:04", "throughput": 5187.1, "total_tokens": 9929664} {"current_steps": 9885, "total_steps": 42420, "loss": 0.3672, "lr": 0.0009470745392248884, "epoch": 4.66053748231966, "percentage": 23.3, "elapsed_time": "0:31:55", "remaining_time": "1:45:03", "throughput": 5187.12, "total_tokens": 9934144} {"current_steps": 9890, "total_steps": 42420, "loss": 0.3954, "lr": 0.0009469823862270236, "epoch": 4.662894860914663, "percentage": 23.31, "elapsed_time": "0:31:56", "remaining_time": "1:45:02", "throughput": 5187.39, "total_tokens": 9939712} {"current_steps": 9895, "total_steps": 42420, "loss": 0.3851, "lr": 0.0009468901575626302, "epoch": 4.665252239509665, "percentage": 23.33, "elapsed_time": "0:31:56", "remaining_time": "1:45:01", "throughput": 5187.36, "total_tokens": 9943968} {"current_steps": 9900, "total_steps": 42420, "loss": 0.3407, "lr": 0.000946797853247321, "epoch": 4.667609618104668, "percentage": 23.34, "elapsed_time": "0:31:57", "remaining_time": "1:44:59", "throughput": 5187.33, "total_tokens": 9948288} {"current_steps": 9905, "total_steps": 42420, "loss": 0.375, "lr": 0.0009467054732967211, "epoch": 4.66996699669967, "percentage": 23.35, "elapsed_time": "0:31:58", "remaining_time": "1:44:58", "throughput": 5187.5, "total_tokens": 9953376} {"current_steps": 9910, "total_steps": 42420, "loss": 0.313, "lr": 0.0009466130177264695, "epoch": 4.672324375294672, "percentage": 23.36, "elapsed_time": "0:31:59", "remaining_time": "1:44:57", "throughput": 5187.55, "total_tokens": 9958048} {"current_steps": 9915, "total_steps": 42420, "loss": 0.339, "lr": 0.0009465204865522169, "epoch": 4.674681753889675, "percentage": 23.37, "elapsed_time": "0:32:00", "remaining_time": "1:44:56", "throughput": 5187.57, "total_tokens": 9962720} {"current_steps": 9920, "total_steps": 42420, "loss": 0.3261, "lr": 0.0009464278797896276, "epoch": 4.677039132484677, "percentage": 23.39, "elapsed_time": "0:32:01", "remaining_time": "1:44:54", "throughput": 5187.61, "total_tokens": 9967104} {"current_steps": 9925, "total_steps": 42420, "loss": 0.2784, "lr": 0.0009463351974543779, "epoch": 4.67939651107968, "percentage": 23.4, "elapsed_time": "0:32:02", "remaining_time": "1:44:53", "throughput": 5187.79, "total_tokens": 9972352} {"current_steps": 9930, "total_steps": 42420, "loss": 0.4247, "lr": 0.0009462424395621579, "epoch": 4.681753889674682, "percentage": 23.41, "elapsed_time": "0:32:03", "remaining_time": "1:44:52", "throughput": 5188.07, "total_tokens": 9978144} {"current_steps": 9935, "total_steps": 42420, "loss": 0.2642, "lr": 0.0009461496061286694, "epoch": 4.684111268269684, "percentage": 23.42, "elapsed_time": "0:32:04", "remaining_time": "1:44:51", "throughput": 5188.12, "total_tokens": 9982784} {"current_steps": 9940, "total_steps": 42420, "loss": 0.3092, "lr": 0.0009460566971696278, "epoch": 4.686468646864687, "percentage": 23.43, "elapsed_time": "0:32:04", "remaining_time": "1:44:49", "throughput": 5188.1, "total_tokens": 9986752} {"current_steps": 9945, "total_steps": 42420, "loss": 0.3673, "lr": 0.0009459637127007611, "epoch": 4.688826025459689, "percentage": 23.44, "elapsed_time": "0:32:05", "remaining_time": "1:44:48", "throughput": 5188.25, "total_tokens": 9991936} {"current_steps": 9950, "total_steps": 42420, "loss": 0.2469, "lr": 0.0009458706527378098, "epoch": 4.691183404054691, "percentage": 23.46, "elapsed_time": "0:32:06", "remaining_time": "1:44:47", "throughput": 5188.45, "total_tokens": 9997408} {"current_steps": 9955, "total_steps": 42420, "loss": 0.343, "lr": 0.0009457775172965275, "epoch": 4.693540782649693, "percentage": 23.47, "elapsed_time": "0:32:07", "remaining_time": "1:44:47", "throughput": 5188.79, "total_tokens": 10003872} {"current_steps": 9960, "total_steps": 42420, "loss": 0.2724, "lr": 0.0009456843063926803, "epoch": 4.6958981612446955, "percentage": 23.48, "elapsed_time": "0:32:08", "remaining_time": "1:44:46", "throughput": 5188.7, "total_tokens": 10008032} {"current_steps": 9965, "total_steps": 42420, "loss": 0.423, "lr": 0.0009455910200420474, "epoch": 4.698255539839698, "percentage": 23.49, "elapsed_time": "0:32:09", "remaining_time": "1:44:44", "throughput": 5188.85, "total_tokens": 10013056} {"current_steps": 9970, "total_steps": 42420, "loss": 0.3238, "lr": 0.0009454976582604205, "epoch": 4.7006129184347, "percentage": 23.5, "elapsed_time": "0:32:10", "remaining_time": "1:44:44", "throughput": 5189.08, "total_tokens": 10018624} {"current_steps": 9975, "total_steps": 42420, "loss": 0.3579, "lr": 0.0009454042210636041, "epoch": 4.702970297029703, "percentage": 23.51, "elapsed_time": "0:32:11", "remaining_time": "1:44:43", "throughput": 5189.25, "total_tokens": 10024128} {"current_steps": 9980, "total_steps": 42420, "loss": 0.3564, "lr": 0.0009453107084674157, "epoch": 4.705327675624705, "percentage": 23.53, "elapsed_time": "0:32:12", "remaining_time": "1:44:41", "throughput": 5189.24, "total_tokens": 10028320} {"current_steps": 9985, "total_steps": 42420, "loss": 0.3416, "lr": 0.0009452171204876853, "epoch": 4.7076850542197075, "percentage": 23.54, "elapsed_time": "0:32:13", "remaining_time": "1:44:40", "throughput": 5189.33, "total_tokens": 10033312} {"current_steps": 9990, "total_steps": 42420, "loss": 0.3504, "lr": 0.0009451234571402557, "epoch": 4.71004243281471, "percentage": 23.55, "elapsed_time": "0:32:14", "remaining_time": "1:44:39", "throughput": 5189.38, "total_tokens": 10037792} {"current_steps": 9995, "total_steps": 42420, "loss": 0.3648, "lr": 0.0009450297184409826, "epoch": 4.712399811409712, "percentage": 23.56, "elapsed_time": "0:32:15", "remaining_time": "1:44:38", "throughput": 5189.58, "total_tokens": 10043040} {"current_steps": 10000, "total_steps": 42420, "loss": 0.4003, "lr": 0.0009449359044057344, "epoch": 4.714757190004715, "percentage": 23.57, "elapsed_time": "0:32:16", "remaining_time": "1:44:37", "throughput": 5189.72, "total_tokens": 10048192} {"current_steps": 10005, "total_steps": 42420, "loss": 0.3509, "lr": 0.000944842015050392, "epoch": 4.717114568599717, "percentage": 23.59, "elapsed_time": "0:32:17", "remaining_time": "1:44:35", "throughput": 5189.8, "total_tokens": 10052928} {"current_steps": 10010, "total_steps": 42420, "loss": 0.3492, "lr": 0.0009447480503908496, "epoch": 4.7194719471947195, "percentage": 23.6, "elapsed_time": "0:32:18", "remaining_time": "1:44:34", "throughput": 5190.01, "total_tokens": 10058368} {"current_steps": 10015, "total_steps": 42420, "loss": 0.3202, "lr": 0.0009446540104430136, "epoch": 4.721829325789722, "percentage": 23.61, "elapsed_time": "0:32:18", "remaining_time": "1:44:33", "throughput": 5190.07, "total_tokens": 10062848} {"current_steps": 10020, "total_steps": 42420, "loss": 0.4073, "lr": 0.0009445598952228034, "epoch": 4.724186704384724, "percentage": 23.62, "elapsed_time": "0:32:19", "remaining_time": "1:44:32", "throughput": 5190.17, "total_tokens": 10067584} {"current_steps": 10025, "total_steps": 42420, "loss": 0.3037, "lr": 0.0009444657047461511, "epoch": 4.726544082979727, "percentage": 23.63, "elapsed_time": "0:32:20", "remaining_time": "1:44:30", "throughput": 5190.21, "total_tokens": 10072160} {"current_steps": 10030, "total_steps": 42420, "loss": 0.3393, "lr": 0.0009443714390290017, "epoch": 4.728901461574729, "percentage": 23.64, "elapsed_time": "0:32:21", "remaining_time": "1:44:29", "throughput": 5190.37, "total_tokens": 10077248} {"current_steps": 10035, "total_steps": 42420, "loss": 0.3384, "lr": 0.0009442770980873127, "epoch": 4.7312588401697315, "percentage": 23.66, "elapsed_time": "0:32:22", "remaining_time": "1:44:28", "throughput": 5190.42, "total_tokens": 10081696} {"current_steps": 10040, "total_steps": 42420, "loss": 0.2468, "lr": 0.0009441826819370543, "epoch": 4.733616218764734, "percentage": 23.67, "elapsed_time": "0:32:23", "remaining_time": "1:44:27", "throughput": 5190.66, "total_tokens": 10087520} {"current_steps": 10045, "total_steps": 42420, "loss": 0.3676, "lr": 0.0009440881905942096, "epoch": 4.735973597359736, "percentage": 23.68, "elapsed_time": "0:32:24", "remaining_time": "1:44:26", "throughput": 5190.96, "total_tokens": 10093568} {"current_steps": 10050, "total_steps": 42420, "loss": 0.3331, "lr": 0.0009439936240747746, "epoch": 4.738330975954739, "percentage": 23.69, "elapsed_time": "0:32:25", "remaining_time": "1:44:25", "throughput": 5191.06, "total_tokens": 10098336} {"current_steps": 10055, "total_steps": 42420, "loss": 0.2949, "lr": 0.0009438989823947574, "epoch": 4.740688354549741, "percentage": 23.7, "elapsed_time": "0:32:26", "remaining_time": "1:44:24", "throughput": 5191.1, "total_tokens": 10103072} {"current_steps": 10060, "total_steps": 42420, "loss": 0.3341, "lr": 0.0009438042655701796, "epoch": 4.7430457331447435, "percentage": 23.72, "elapsed_time": "0:32:27", "remaining_time": "1:44:23", "throughput": 5191.29, "total_tokens": 10108704} {"current_steps": 10065, "total_steps": 42420, "loss": 0.3744, "lr": 0.0009437094736170751, "epoch": 4.745403111739745, "percentage": 23.73, "elapsed_time": "0:32:28", "remaining_time": "1:44:22", "throughput": 5191.42, "total_tokens": 10113952} {"current_steps": 10070, "total_steps": 42420, "loss": 0.3075, "lr": 0.0009436146065514902, "epoch": 4.7477604903347475, "percentage": 23.74, "elapsed_time": "0:32:29", "remaining_time": "1:44:21", "throughput": 5191.47, "total_tokens": 10118656} {"current_steps": 10075, "total_steps": 42420, "loss": 0.3191, "lr": 0.0009435196643894848, "epoch": 4.75011786892975, "percentage": 23.75, "elapsed_time": "0:32:29", "remaining_time": "1:44:19", "throughput": 5191.31, "total_tokens": 10122208} {"current_steps": 10080, "total_steps": 42420, "loss": 0.4145, "lr": 0.0009434246471471306, "epoch": 4.752475247524752, "percentage": 23.76, "elapsed_time": "0:32:30", "remaining_time": "1:44:18", "throughput": 5191.39, "total_tokens": 10126912} {"current_steps": 10085, "total_steps": 42420, "loss": 0.3191, "lr": 0.0009433295548405125, "epoch": 4.754832626119755, "percentage": 23.77, "elapsed_time": "0:32:31", "remaining_time": "1:44:17", "throughput": 5191.32, "total_tokens": 10130912} {"current_steps": 10090, "total_steps": 42420, "loss": 0.3459, "lr": 0.0009432343874857283, "epoch": 4.757190004714757, "percentage": 23.79, "elapsed_time": "0:32:32", "remaining_time": "1:44:15", "throughput": 5191.33, "total_tokens": 10135264} {"current_steps": 10095, "total_steps": 42420, "loss": 0.349, "lr": 0.0009431391450988877, "epoch": 4.7595473833097595, "percentage": 23.8, "elapsed_time": "0:32:33", "remaining_time": "1:44:14", "throughput": 5191.46, "total_tokens": 10140160} {"current_steps": 10100, "total_steps": 42420, "loss": 0.3464, "lr": 0.0009430438276961142, "epoch": 4.761904761904762, "percentage": 23.81, "elapsed_time": "0:32:34", "remaining_time": "1:44:13", "throughput": 5191.56, "total_tokens": 10145024} {"current_steps": 10105, "total_steps": 42420, "loss": 0.3084, "lr": 0.0009429484352935431, "epoch": 4.764262140499764, "percentage": 23.82, "elapsed_time": "0:32:35", "remaining_time": "1:44:11", "throughput": 5191.65, "total_tokens": 10149696} {"current_steps": 10110, "total_steps": 42420, "loss": 0.2889, "lr": 0.0009428529679073225, "epoch": 4.766619519094767, "percentage": 23.83, "elapsed_time": "0:32:36", "remaining_time": "1:44:11", "throughput": 5191.92, "total_tokens": 10155648} {"current_steps": 10115, "total_steps": 42420, "loss": 0.2944, "lr": 0.0009427574255536138, "epoch": 4.768976897689769, "percentage": 23.84, "elapsed_time": "0:32:37", "remaining_time": "1:44:10", "throughput": 5192.19, "total_tokens": 10161184} {"current_steps": 10120, "total_steps": 42420, "loss": 0.3619, "lr": 0.0009426618082485907, "epoch": 4.7713342762847715, "percentage": 23.86, "elapsed_time": "0:32:37", "remaining_time": "1:44:08", "throughput": 5192.18, "total_tokens": 10165728} {"current_steps": 10125, "total_steps": 42420, "loss": 0.4014, "lr": 0.0009425661160084392, "epoch": 4.773691654879774, "percentage": 23.87, "elapsed_time": "0:32:38", "remaining_time": "1:44:07", "throughput": 5192.37, "total_tokens": 10171072} {"current_steps": 10130, "total_steps": 42420, "loss": 0.3351, "lr": 0.0009424703488493586, "epoch": 4.776049033474776, "percentage": 23.88, "elapsed_time": "0:32:39", "remaining_time": "1:44:07", "throughput": 5192.44, "total_tokens": 10176192} {"current_steps": 10135, "total_steps": 42420, "loss": 0.3585, "lr": 0.0009423745067875609, "epoch": 4.778406412069779, "percentage": 23.89, "elapsed_time": "0:32:40", "remaining_time": "1:44:06", "throughput": 5192.75, "total_tokens": 10182400} {"current_steps": 10140, "total_steps": 42420, "loss": 0.3676, "lr": 0.00094227858983927, "epoch": 4.780763790664781, "percentage": 23.9, "elapsed_time": "0:32:42", "remaining_time": "1:44:05", "throughput": 5193.14, "total_tokens": 10188960} {"current_steps": 10145, "total_steps": 42420, "loss": 0.3497, "lr": 0.0009421825980207235, "epoch": 4.7831211692597835, "percentage": 23.92, "elapsed_time": "0:32:43", "remaining_time": "1:44:05", "throughput": 5193.42, "total_tokens": 10194784} {"current_steps": 10150, "total_steps": 42420, "loss": 0.3206, "lr": 0.000942086531348171, "epoch": 4.785478547854785, "percentage": 23.93, "elapsed_time": "0:32:43", "remaining_time": "1:44:03", "throughput": 5193.33, "total_tokens": 10198816} {"current_steps": 10155, "total_steps": 42420, "loss": 0.3598, "lr": 0.0009419903898378747, "epoch": 4.787835926449787, "percentage": 23.94, "elapsed_time": "0:32:44", "remaining_time": "1:44:02", "throughput": 5193.51, "total_tokens": 10204352} {"current_steps": 10160, "total_steps": 42420, "loss": 0.3113, "lr": 0.0009418941735061103, "epoch": 4.79019330504479, "percentage": 23.95, "elapsed_time": "0:32:45", "remaining_time": "1:44:01", "throughput": 5193.61, "total_tokens": 10209088} {"current_steps": 10165, "total_steps": 42420, "loss": 0.3805, "lr": 0.0009417978823691651, "epoch": 4.792550683639792, "percentage": 23.96, "elapsed_time": "0:32:46", "remaining_time": "1:44:00", "throughput": 5193.67, "total_tokens": 10213696} {"current_steps": 10170, "total_steps": 42420, "loss": 0.33, "lr": 0.0009417015164433398, "epoch": 4.794908062234795, "percentage": 23.97, "elapsed_time": "0:32:47", "remaining_time": "1:43:58", "throughput": 5193.79, "total_tokens": 10218560} {"current_steps": 10175, "total_steps": 42420, "loss": 0.3192, "lr": 0.0009416050757449473, "epoch": 4.797265440829797, "percentage": 23.99, "elapsed_time": "0:32:48", "remaining_time": "1:43:57", "throughput": 5193.82, "total_tokens": 10222912} {"current_steps": 10180, "total_steps": 42420, "loss": 0.4004, "lr": 0.0009415085602903135, "epoch": 4.799622819424799, "percentage": 24.0, "elapsed_time": "0:32:49", "remaining_time": "1:43:56", "throughput": 5194.06, "total_tokens": 10228768} {"current_steps": 10185, "total_steps": 42420, "loss": 0.3491, "lr": 0.0009414119700957769, "epoch": 4.801980198019802, "percentage": 24.01, "elapsed_time": "0:32:50", "remaining_time": "1:43:55", "throughput": 5194.09, "total_tokens": 10233472} {"current_steps": 10190, "total_steps": 42420, "loss": 0.3487, "lr": 0.0009413153051776885, "epoch": 4.804337576614804, "percentage": 24.02, "elapsed_time": "0:32:51", "remaining_time": "1:43:54", "throughput": 5194.33, "total_tokens": 10239200} {"current_steps": 10195, "total_steps": 42420, "loss": 0.2883, "lr": 0.000941218565552412, "epoch": 4.806694955209807, "percentage": 24.03, "elapsed_time": "0:32:52", "remaining_time": "1:43:53", "throughput": 5194.52, "total_tokens": 10244640} {"current_steps": 10200, "total_steps": 42420, "loss": 0.2862, "lr": 0.000941121751236324, "epoch": 4.809052333804809, "percentage": 24.05, "elapsed_time": "0:32:53", "remaining_time": "1:43:53", "throughput": 5194.95, "total_tokens": 10251232} {"current_steps": 10205, "total_steps": 42420, "loss": 0.5152, "lr": 0.0009410248622458131, "epoch": 4.811409712399811, "percentage": 24.06, "elapsed_time": "0:32:54", "remaining_time": "1:43:52", "throughput": 5195.15, "total_tokens": 10256416} {"current_steps": 10210, "total_steps": 42420, "loss": 0.2978, "lr": 0.0009409278985972814, "epoch": 4.813767090994814, "percentage": 24.07, "elapsed_time": "0:32:55", "remaining_time": "1:43:50", "throughput": 5195.1, "total_tokens": 10260512} {"current_steps": 10215, "total_steps": 42420, "loss": 0.3339, "lr": 0.0009408308603071427, "epoch": 4.816124469589816, "percentage": 24.08, "elapsed_time": "0:32:56", "remaining_time": "1:43:49", "throughput": 5195.13, "total_tokens": 10265632} {"current_steps": 10220, "total_steps": 42420, "loss": 0.3286, "lr": 0.0009407337473918244, "epoch": 4.818481848184819, "percentage": 24.09, "elapsed_time": "0:32:56", "remaining_time": "1:43:48", "throughput": 5195.24, "total_tokens": 10270528} {"current_steps": 10225, "total_steps": 42420, "loss": 0.3467, "lr": 0.000940636559867766, "epoch": 4.820839226779821, "percentage": 24.1, "elapsed_time": "0:32:57", "remaining_time": "1:43:47", "throughput": 5195.47, "total_tokens": 10276128} {"current_steps": 10230, "total_steps": 42420, "loss": 0.3492, "lr": 0.0009405392977514196, "epoch": 4.823196605374823, "percentage": 24.12, "elapsed_time": "0:32:58", "remaining_time": "1:43:46", "throughput": 5195.46, "total_tokens": 10280544} {"current_steps": 10235, "total_steps": 42420, "loss": 0.3517, "lr": 0.0009404419610592498, "epoch": 4.825553983969826, "percentage": 24.13, "elapsed_time": "0:32:59", "remaining_time": "1:43:45", "throughput": 5195.47, "total_tokens": 10284992} {"current_steps": 10240, "total_steps": 42420, "loss": 0.3142, "lr": 0.0009403445498077346, "epoch": 4.827911362564828, "percentage": 24.14, "elapsed_time": "0:33:00", "remaining_time": "1:43:43", "throughput": 5195.49, "total_tokens": 10289664} {"current_steps": 10245, "total_steps": 42420, "loss": 0.4091, "lr": 0.0009402470640133634, "epoch": 4.830268741159831, "percentage": 24.15, "elapsed_time": "0:33:01", "remaining_time": "1:43:43", "throughput": 5195.87, "total_tokens": 10296224} {"current_steps": 10250, "total_steps": 42420, "loss": 0.2928, "lr": 0.0009401495036926394, "epoch": 4.832626119754833, "percentage": 24.16, "elapsed_time": "0:33:02", "remaining_time": "1:43:42", "throughput": 5196.08, "total_tokens": 10301600} {"current_steps": 10255, "total_steps": 42420, "loss": 0.3479, "lr": 0.0009400518688620777, "epoch": 4.834983498349835, "percentage": 24.17, "elapsed_time": "0:33:03", "remaining_time": "1:43:41", "throughput": 5196.19, "total_tokens": 10306624} {"current_steps": 10260, "total_steps": 42420, "loss": 0.3654, "lr": 0.0009399541595382063, "epoch": 4.837340876944838, "percentage": 24.19, "elapsed_time": "0:33:04", "remaining_time": "1:43:40", "throughput": 5196.28, "total_tokens": 10311392} {"current_steps": 10265, "total_steps": 42420, "loss": 0.3276, "lr": 0.0009398563757375656, "epoch": 4.839698255539839, "percentage": 24.2, "elapsed_time": "0:33:05", "remaining_time": "1:43:38", "throughput": 5196.38, "total_tokens": 10316320} {"current_steps": 10270, "total_steps": 42420, "loss": 0.3851, "lr": 0.0009397585174767089, "epoch": 4.842055634134842, "percentage": 24.21, "elapsed_time": "0:33:06", "remaining_time": "1:43:38", "throughput": 5196.68, "total_tokens": 10322368} {"current_steps": 10275, "total_steps": 42420, "loss": 0.3797, "lr": 0.000939660584772202, "epoch": 4.844413012729844, "percentage": 24.22, "elapsed_time": "0:33:07", "remaining_time": "1:43:38", "throughput": 5197.17, "total_tokens": 10329888} {"current_steps": 10280, "total_steps": 42420, "loss": 0.3589, "lr": 0.000939562577640623, "epoch": 4.8467703913248465, "percentage": 24.23, "elapsed_time": "0:33:08", "remaining_time": "1:43:36", "throughput": 5197.11, "total_tokens": 10333888} {"current_steps": 10285, "total_steps": 42420, "loss": 0.3636, "lr": 0.0009394644960985627, "epoch": 4.849127769919849, "percentage": 24.25, "elapsed_time": "0:33:09", "remaining_time": "1:43:35", "throughput": 5197.15, "total_tokens": 10338304} {"current_steps": 10290, "total_steps": 42420, "loss": 0.3538, "lr": 0.0009393663401626252, "epoch": 4.851485148514851, "percentage": 24.26, "elapsed_time": "0:33:10", "remaining_time": "1:43:33", "throughput": 5197.04, "total_tokens": 10342336} {"current_steps": 10295, "total_steps": 42420, "loss": 0.3357, "lr": 0.0009392681098494263, "epoch": 4.853842527109854, "percentage": 24.27, "elapsed_time": "0:33:10", "remaining_time": "1:43:32", "throughput": 5197.06, "total_tokens": 10346752} {"current_steps": 10300, "total_steps": 42420, "loss": 0.2952, "lr": 0.0009391698051755947, "epoch": 4.856199905704856, "percentage": 24.28, "elapsed_time": "0:33:11", "remaining_time": "1:43:31", "throughput": 5197.17, "total_tokens": 10351552} {"current_steps": 10305, "total_steps": 42420, "loss": 0.3973, "lr": 0.0009390714261577717, "epoch": 4.858557284299859, "percentage": 24.29, "elapsed_time": "0:33:12", "remaining_time": "1:43:30", "throughput": 5197.45, "total_tokens": 10357248} {"current_steps": 10310, "total_steps": 42420, "loss": 0.4021, "lr": 0.0009389729728126113, "epoch": 4.860914662894861, "percentage": 24.3, "elapsed_time": "0:33:13", "remaining_time": "1:43:29", "throughput": 5197.51, "total_tokens": 10361824} {"current_steps": 10315, "total_steps": 42420, "loss": 0.3584, "lr": 0.00093887444515678, "epoch": 4.863272041489863, "percentage": 24.32, "elapsed_time": "0:33:14", "remaining_time": "1:43:27", "throughput": 5197.5, "total_tokens": 10366112} {"current_steps": 10320, "total_steps": 42420, "loss": 0.3632, "lr": 0.0009387758432069567, "epoch": 4.865629420084866, "percentage": 24.33, "elapsed_time": "0:33:15", "remaining_time": "1:43:26", "throughput": 5197.71, "total_tokens": 10371552} {"current_steps": 10325, "total_steps": 42420, "loss": 0.3339, "lr": 0.0009386771669798333, "epoch": 4.867986798679868, "percentage": 24.34, "elapsed_time": "0:33:16", "remaining_time": "1:43:25", "throughput": 5197.83, "total_tokens": 10376992} {"current_steps": 10330, "total_steps": 42420, "loss": 0.3628, "lr": 0.0009385784164921136, "epoch": 4.870344177274871, "percentage": 24.35, "elapsed_time": "0:33:17", "remaining_time": "1:43:24", "throughput": 5198.06, "total_tokens": 10382624} {"current_steps": 10335, "total_steps": 42420, "loss": 0.3714, "lr": 0.0009384795917605148, "epoch": 4.872701555869873, "percentage": 24.36, "elapsed_time": "0:33:18", "remaining_time": "1:43:23", "throughput": 5198.25, "total_tokens": 10388032} {"current_steps": 10340, "total_steps": 42420, "loss": 0.3443, "lr": 0.000938380692801766, "epoch": 4.875058934464875, "percentage": 24.38, "elapsed_time": "0:33:19", "remaining_time": "1:43:23", "throughput": 5198.49, "total_tokens": 10393696} {"current_steps": 10345, "total_steps": 42420, "loss": 0.3572, "lr": 0.0009382817196326093, "epoch": 4.877416313059878, "percentage": 24.39, "elapsed_time": "0:33:20", "remaining_time": "1:43:21", "throughput": 5198.48, "total_tokens": 10398112} {"current_steps": 10350, "total_steps": 42420, "loss": 0.3604, "lr": 0.000938182672269799, "epoch": 4.879773691654879, "percentage": 24.4, "elapsed_time": "0:33:21", "remaining_time": "1:43:20", "throughput": 5198.65, "total_tokens": 10403296} {"current_steps": 10355, "total_steps": 42420, "loss": 0.3121, "lr": 0.0009380835507301021, "epoch": 4.882131070249882, "percentage": 24.41, "elapsed_time": "0:33:22", "remaining_time": "1:43:19", "throughput": 5198.8, "total_tokens": 10408512} {"current_steps": 10360, "total_steps": 42420, "loss": 0.3083, "lr": 0.0009379843550302985, "epoch": 4.884488448844884, "percentage": 24.42, "elapsed_time": "0:33:22", "remaining_time": "1:43:18", "throughput": 5198.82, "total_tokens": 10413024} {"current_steps": 10365, "total_steps": 42420, "loss": 0.2526, "lr": 0.0009378850851871799, "epoch": 4.8868458274398865, "percentage": 24.43, "elapsed_time": "0:33:23", "remaining_time": "1:43:17", "throughput": 5199.02, "total_tokens": 10418304} {"current_steps": 10370, "total_steps": 42420, "loss": 0.3868, "lr": 0.0009377857412175514, "epoch": 4.889203206034889, "percentage": 24.45, "elapsed_time": "0:33:24", "remaining_time": "1:43:16", "throughput": 5199.24, "total_tokens": 10423744} {"current_steps": 10375, "total_steps": 42420, "loss": 0.3592, "lr": 0.00093768632313823, "epoch": 4.891560584629891, "percentage": 24.46, "elapsed_time": "0:33:25", "remaining_time": "1:43:15", "throughput": 5199.33, "total_tokens": 10428512} {"current_steps": 10380, "total_steps": 42420, "loss": 0.3483, "lr": 0.0009375868309660458, "epoch": 4.893917963224894, "percentage": 24.47, "elapsed_time": "0:33:26", "remaining_time": "1:43:13", "throughput": 5199.34, "total_tokens": 10433120} {"current_steps": 10385, "total_steps": 42420, "loss": 0.3487, "lr": 0.0009374872647178407, "epoch": 4.896275341819896, "percentage": 24.48, "elapsed_time": "0:33:27", "remaining_time": "1:43:12", "throughput": 5199.37, "total_tokens": 10437664} {"current_steps": 10390, "total_steps": 42420, "loss": 0.3375, "lr": 0.0009373876244104699, "epoch": 4.8986327204148985, "percentage": 24.49, "elapsed_time": "0:33:28", "remaining_time": "1:43:11", "throughput": 5199.4, "total_tokens": 10442528} {"current_steps": 10395, "total_steps": 42420, "loss": 0.356, "lr": 0.0009372879100608006, "epoch": 4.900990099009901, "percentage": 24.5, "elapsed_time": "0:33:29", "remaining_time": "1:43:10", "throughput": 5199.57, "total_tokens": 10447776} {"current_steps": 10400, "total_steps": 42420, "loss": 0.3494, "lr": 0.000937188121685713, "epoch": 4.903347477604903, "percentage": 24.52, "elapsed_time": "0:33:30", "remaining_time": "1:43:09", "throughput": 5199.73, "total_tokens": 10452736} {"current_steps": 10405, "total_steps": 42420, "loss": 0.3333, "lr": 0.0009370882593020993, "epoch": 4.905704856199906, "percentage": 24.53, "elapsed_time": "0:33:31", "remaining_time": "1:43:08", "throughput": 5199.84, "total_tokens": 10457728} {"current_steps": 10410, "total_steps": 42420, "loss": 0.4441, "lr": 0.0009369883229268648, "epoch": 4.908062234794908, "percentage": 24.54, "elapsed_time": "0:33:32", "remaining_time": "1:43:07", "throughput": 5200.2, "total_tokens": 10463968} {"current_steps": 10415, "total_steps": 42420, "loss": 0.357, "lr": 0.0009368883125769267, "epoch": 4.9104196133899105, "percentage": 24.55, "elapsed_time": "0:33:33", "remaining_time": "1:43:06", "throughput": 5200.22, "total_tokens": 10468320} {"current_steps": 10420, "total_steps": 42420, "loss": 0.4173, "lr": 0.0009367882282692153, "epoch": 4.912776991984913, "percentage": 24.56, "elapsed_time": "0:33:33", "remaining_time": "1:43:04", "throughput": 5200.15, "total_tokens": 10472576} {"current_steps": 10425, "total_steps": 42420, "loss": 0.3466, "lr": 0.000936688070020673, "epoch": 4.915134370579915, "percentage": 24.58, "elapsed_time": "0:33:34", "remaining_time": "1:43:03", "throughput": 5200.36, "total_tokens": 10478080} {"current_steps": 10430, "total_steps": 42420, "loss": 0.395, "lr": 0.0009365878378482551, "epoch": 4.917491749174918, "percentage": 24.59, "elapsed_time": "0:33:35", "remaining_time": "1:43:03", "throughput": 5200.59, "total_tokens": 10483936} {"current_steps": 10435, "total_steps": 42420, "loss": 0.2836, "lr": 0.000936487531768929, "epoch": 4.91984912776992, "percentage": 24.6, "elapsed_time": "0:33:36", "remaining_time": "1:43:02", "throughput": 5200.75, "total_tokens": 10489248} {"current_steps": 10440, "total_steps": 42420, "loss": 0.3524, "lr": 0.0009363871517996749, "epoch": 4.9222065063649225, "percentage": 24.61, "elapsed_time": "0:33:37", "remaining_time": "1:43:00", "throughput": 5200.8, "total_tokens": 10493696} {"current_steps": 10445, "total_steps": 42420, "loss": 0.2917, "lr": 0.0009362866979574856, "epoch": 4.924563884959925, "percentage": 24.62, "elapsed_time": "0:33:38", "remaining_time": "1:42:59", "throughput": 5201.01, "total_tokens": 10499040} {"current_steps": 10450, "total_steps": 42420, "loss": 0.2905, "lr": 0.0009361861702593659, "epoch": 4.926921263554927, "percentage": 24.63, "elapsed_time": "0:33:39", "remaining_time": "1:42:58", "throughput": 5201.04, "total_tokens": 10503424} {"current_steps": 10455, "total_steps": 42420, "loss": 0.3758, "lr": 0.0009360855687223337, "epoch": 4.92927864214993, "percentage": 24.65, "elapsed_time": "0:33:40", "remaining_time": "1:42:56", "throughput": 5201.08, "total_tokens": 10507968} {"current_steps": 10460, "total_steps": 42420, "loss": 0.3288, "lr": 0.000935984893363419, "epoch": 4.931636020744932, "percentage": 24.66, "elapsed_time": "0:33:41", "remaining_time": "1:42:56", "throughput": 5201.43, "total_tokens": 10514496} {"current_steps": 10465, "total_steps": 42420, "loss": 0.2837, "lr": 0.0009358841441996644, "epoch": 4.933993399339934, "percentage": 24.67, "elapsed_time": "0:33:42", "remaining_time": "1:42:55", "throughput": 5201.5, "total_tokens": 10519328} {"current_steps": 10470, "total_steps": 42420, "loss": 0.3211, "lr": 0.000935783321248125, "epoch": 4.936350777934936, "percentage": 24.68, "elapsed_time": "0:33:43", "remaining_time": "1:42:53", "throughput": 5201.51, "total_tokens": 10523616} {"current_steps": 10475, "total_steps": 42420, "loss": 0.3199, "lr": 0.0009356824245258686, "epoch": 4.938708156529938, "percentage": 24.69, "elapsed_time": "0:33:44", "remaining_time": "1:42:53", "throughput": 5201.73, "total_tokens": 10529248} {"current_steps": 10480, "total_steps": 42420, "loss": 0.411, "lr": 0.0009355814540499751, "epoch": 4.941065535124941, "percentage": 24.71, "elapsed_time": "0:33:45", "remaining_time": "1:42:51", "throughput": 5201.86, "total_tokens": 10534336} {"current_steps": 10485, "total_steps": 42420, "loss": 0.3624, "lr": 0.0009354804098375373, "epoch": 4.943422913719943, "percentage": 24.72, "elapsed_time": "0:33:46", "remaining_time": "1:42:51", "throughput": 5202.44, "total_tokens": 10542176} {"current_steps": 10490, "total_steps": 42420, "loss": 0.3584, "lr": 0.0009353792919056603, "epoch": 4.945780292314946, "percentage": 24.73, "elapsed_time": "0:33:47", "remaining_time": "1:42:50", "throughput": 5202.32, "total_tokens": 10546112} {"current_steps": 10495, "total_steps": 42420, "loss": 0.3192, "lr": 0.0009352781002714612, "epoch": 4.948137670909948, "percentage": 24.74, "elapsed_time": "0:33:48", "remaining_time": "1:42:49", "throughput": 5202.39, "total_tokens": 10550656} {"current_steps": 10500, "total_steps": 42420, "loss": 0.3131, "lr": 0.0009351768349520706, "epoch": 4.9504950495049505, "percentage": 24.75, "elapsed_time": "0:33:49", "remaining_time": "1:42:48", "throughput": 5202.78, "total_tokens": 10557248} {"current_steps": 10505, "total_steps": 42420, "loss": 0.2873, "lr": 0.0009350754959646306, "epoch": 4.952852428099953, "percentage": 24.76, "elapsed_time": "0:33:50", "remaining_time": "1:42:47", "throughput": 5203.04, "total_tokens": 10563104} {"current_steps": 10510, "total_steps": 42420, "loss": 0.4413, "lr": 0.0009349740833262961, "epoch": 4.955209806694955, "percentage": 24.78, "elapsed_time": "0:33:51", "remaining_time": "1:42:46", "throughput": 5203.16, "total_tokens": 10568128} {"current_steps": 10515, "total_steps": 42420, "loss": 0.3547, "lr": 0.000934872597054235, "epoch": 4.957567185289958, "percentage": 24.79, "elapsed_time": "0:33:51", "remaining_time": "1:42:45", "throughput": 5203.2, "total_tokens": 10572800} {"current_steps": 10520, "total_steps": 42420, "loss": 0.3866, "lr": 0.0009347710371656268, "epoch": 4.95992456388496, "percentage": 24.8, "elapsed_time": "0:33:52", "remaining_time": "1:42:44", "throughput": 5203.24, "total_tokens": 10577664} {"current_steps": 10525, "total_steps": 42420, "loss": 0.3832, "lr": 0.0009346694036776641, "epoch": 4.9622819424799625, "percentage": 24.81, "elapsed_time": "0:33:53", "remaining_time": "1:42:43", "throughput": 5203.37, "total_tokens": 10582560} {"current_steps": 10530, "total_steps": 42420, "loss": 0.365, "lr": 0.0009345676966075516, "epoch": 4.964639321074965, "percentage": 24.82, "elapsed_time": "0:33:54", "remaining_time": "1:42:42", "throughput": 5203.71, "total_tokens": 10589056} {"current_steps": 10535, "total_steps": 42420, "loss": 0.3384, "lr": 0.0009344659159725065, "epoch": 4.966996699669967, "percentage": 24.83, "elapsed_time": "0:33:55", "remaining_time": "1:42:41", "throughput": 5203.9, "total_tokens": 10594496} {"current_steps": 10540, "total_steps": 42420, "loss": 0.3481, "lr": 0.0009343640617897587, "epoch": 4.96935407826497, "percentage": 24.85, "elapsed_time": "0:33:56", "remaining_time": "1:42:40", "throughput": 5203.92, "total_tokens": 10598912} {"current_steps": 10545, "total_steps": 42420, "loss": 0.3025, "lr": 0.0009342621340765503, "epoch": 4.971711456859972, "percentage": 24.86, "elapsed_time": "0:33:57", "remaining_time": "1:42:39", "throughput": 5204.11, "total_tokens": 10604320} {"current_steps": 10550, "total_steps": 42420, "loss": 0.3831, "lr": 0.000934160132850136, "epoch": 4.974068835454974, "percentage": 24.87, "elapsed_time": "0:33:58", "remaining_time": "1:42:38", "throughput": 5204.14, "total_tokens": 10608992} {"current_steps": 10555, "total_steps": 42420, "loss": 0.3112, "lr": 0.0009340580581277828, "epoch": 4.976426214049976, "percentage": 24.88, "elapsed_time": "0:33:59", "remaining_time": "1:42:37", "throughput": 5204.46, "total_tokens": 10615264} {"current_steps": 10560, "total_steps": 42420, "loss": 0.331, "lr": 0.0009339559099267703, "epoch": 4.978783592644978, "percentage": 24.89, "elapsed_time": "0:34:00", "remaining_time": "1:42:36", "throughput": 5204.49, "total_tokens": 10620128} {"current_steps": 10565, "total_steps": 42420, "loss": 0.3581, "lr": 0.0009338536882643904, "epoch": 4.981140971239981, "percentage": 24.91, "elapsed_time": "0:34:01", "remaining_time": "1:42:35", "throughput": 5204.71, "total_tokens": 10625568} {"current_steps": 10570, "total_steps": 42420, "loss": 0.4004, "lr": 0.0009337513931579473, "epoch": 4.983498349834983, "percentage": 24.92, "elapsed_time": "0:34:02", "remaining_time": "1:42:34", "throughput": 5204.88, "total_tokens": 10630976} {"current_steps": 10575, "total_steps": 42420, "loss": 0.3707, "lr": 0.0009336490246247582, "epoch": 4.985855728429986, "percentage": 24.93, "elapsed_time": "0:34:03", "remaining_time": "1:42:33", "throughput": 5204.89, "total_tokens": 10635424} {"current_steps": 10580, "total_steps": 42420, "loss": 0.4087, "lr": 0.000933546582682152, "epoch": 4.988213107024988, "percentage": 24.94, "elapsed_time": "0:34:04", "remaining_time": "1:42:31", "throughput": 5204.88, "total_tokens": 10639776} {"current_steps": 10585, "total_steps": 42420, "loss": 0.3695, "lr": 0.0009334440673474709, "epoch": 4.99057048561999, "percentage": 24.95, "elapsed_time": "0:34:05", "remaining_time": "1:42:30", "throughput": 5205.03, "total_tokens": 10644800} {"current_steps": 10590, "total_steps": 42420, "loss": 0.3206, "lr": 0.0009333414786380684, "epoch": 4.992927864214993, "percentage": 24.96, "elapsed_time": "0:34:05", "remaining_time": "1:42:29", "throughput": 5205.1, "total_tokens": 10649440} {"current_steps": 10595, "total_steps": 42420, "loss": 0.3334, "lr": 0.0009332388165713113, "epoch": 4.995285242809995, "percentage": 24.98, "elapsed_time": "0:34:06", "remaining_time": "1:42:28", "throughput": 5205.34, "total_tokens": 10654912} {"current_steps": 10600, "total_steps": 42420, "loss": 0.3721, "lr": 0.0009331360811645786, "epoch": 4.997642621404998, "percentage": 24.99, "elapsed_time": "0:34:07", "remaining_time": "1:42:27", "throughput": 5205.4, "total_tokens": 10659648} {"current_steps": 10605, "total_steps": 42420, "loss": 0.3356, "lr": 0.0009330332724352615, "epoch": 5.0, "percentage": 25.0, "elapsed_time": "0:34:08", "remaining_time": "1:42:26", "throughput": 5205.07, "total_tokens": 10663616} {"current_steps": 10605, "total_steps": 42420, "eval_loss": 0.3410816192626953, "epoch": 5.0, "percentage": 25.0, "elapsed_time": "0:34:27", "remaining_time": "1:43:22", "throughput": 5157.46, "total_tokens": 10663616} {"current_steps": 10610, "total_steps": 42420, "loss": 0.3362, "lr": 0.000932930390400764, "epoch": 5.002357378595002, "percentage": 25.01, "elapsed_time": "0:34:30", "remaining_time": "1:43:27", "throughput": 5152.07, "total_tokens": 10667904} {"current_steps": 10615, "total_steps": 42420, "loss": 0.3553, "lr": 0.000932827435078502, "epoch": 5.004714757190005, "percentage": 25.02, "elapsed_time": "0:34:31", "remaining_time": "1:43:26", "throughput": 5152.16, "total_tokens": 10672576} {"current_steps": 10620, "total_steps": 42420, "loss": 0.2935, "lr": 0.0009327244064859044, "epoch": 5.007072135785007, "percentage": 25.04, "elapsed_time": "0:34:32", "remaining_time": "1:43:25", "throughput": 5152.34, "total_tokens": 10677888} {"current_steps": 10625, "total_steps": 42420, "loss": 0.4139, "lr": 0.0009326213046404119, "epoch": 5.00942951438001, "percentage": 25.05, "elapsed_time": "0:34:33", "remaining_time": "1:43:24", "throughput": 5152.32, "total_tokens": 10682112} {"current_steps": 10630, "total_steps": 42420, "loss": 0.3525, "lr": 0.000932518129559478, "epoch": 5.011786892975012, "percentage": 25.06, "elapsed_time": "0:34:34", "remaining_time": "1:43:22", "throughput": 5152.39, "total_tokens": 10686656} {"current_steps": 10635, "total_steps": 42420, "loss": 0.3626, "lr": 0.0009324148812605687, "epoch": 5.014144271570014, "percentage": 25.07, "elapsed_time": "0:34:34", "remaining_time": "1:43:21", "throughput": 5152.44, "total_tokens": 10691104} {"current_steps": 10640, "total_steps": 42420, "loss": 0.3586, "lr": 0.0009323115597611617, "epoch": 5.016501650165017, "percentage": 25.08, "elapsed_time": "0:34:35", "remaining_time": "1:43:20", "throughput": 5152.59, "total_tokens": 10696160} {"current_steps": 10645, "total_steps": 42420, "loss": 0.283, "lr": 0.000932208165078748, "epoch": 5.018859028760019, "percentage": 25.09, "elapsed_time": "0:34:36", "remaining_time": "1:43:18", "throughput": 5152.61, "total_tokens": 10700448} {"current_steps": 10650, "total_steps": 42420, "loss": 0.4402, "lr": 0.0009321046972308305, "epoch": 5.021216407355022, "percentage": 25.11, "elapsed_time": "0:34:37", "remaining_time": "1:43:17", "throughput": 5152.58, "total_tokens": 10704640} {"current_steps": 10655, "total_steps": 42420, "loss": 0.2627, "lr": 0.0009320011562349243, "epoch": 5.023573785950024, "percentage": 25.12, "elapsed_time": "0:34:38", "remaining_time": "1:43:16", "throughput": 5152.77, "total_tokens": 10709792} {"current_steps": 10660, "total_steps": 42420, "loss": 0.3682, "lr": 0.0009318975421085574, "epoch": 5.0259311645450255, "percentage": 25.13, "elapsed_time": "0:34:39", "remaining_time": "1:43:15", "throughput": 5152.8, "total_tokens": 10714240} {"current_steps": 10665, "total_steps": 42420, "loss": 0.3388, "lr": 0.00093179385486927, "epoch": 5.028288543140028, "percentage": 25.14, "elapsed_time": "0:34:40", "remaining_time": "1:43:13", "throughput": 5153.0, "total_tokens": 10719520} {"current_steps": 10670, "total_steps": 42420, "loss": 0.3403, "lr": 0.0009316900945346141, "epoch": 5.03064592173503, "percentage": 25.15, "elapsed_time": "0:34:41", "remaining_time": "1:43:13", "throughput": 5153.39, "total_tokens": 10726176} {"current_steps": 10675, "total_steps": 42420, "loss": 0.3127, "lr": 0.000931586261122155, "epoch": 5.033003300330033, "percentage": 25.17, "elapsed_time": "0:34:42", "remaining_time": "1:43:12", "throughput": 5153.57, "total_tokens": 10731616} {"current_steps": 10680, "total_steps": 42420, "loss": 0.3357, "lr": 0.0009314823546494699, "epoch": 5.035360678925035, "percentage": 25.18, "elapsed_time": "0:34:43", "remaining_time": "1:43:11", "throughput": 5153.74, "total_tokens": 10736960} {"current_steps": 10685, "total_steps": 42420, "loss": 0.3856, "lr": 0.0009313783751341482, "epoch": 5.0377180575200375, "percentage": 25.19, "elapsed_time": "0:34:44", "remaining_time": "1:43:10", "throughput": 5153.98, "total_tokens": 10742816} {"current_steps": 10690, "total_steps": 42420, "loss": 0.3535, "lr": 0.0009312743225937921, "epoch": 5.04007543611504, "percentage": 25.2, "elapsed_time": "0:34:45", "remaining_time": "1:43:09", "throughput": 5154.22, "total_tokens": 10748480} {"current_steps": 10695, "total_steps": 42420, "loss": 0.3215, "lr": 0.0009311701970460157, "epoch": 5.042432814710042, "percentage": 25.21, "elapsed_time": "0:34:46", "remaining_time": "1:43:08", "throughput": 5154.17, "total_tokens": 10752512} {"current_steps": 10700, "total_steps": 42420, "loss": 0.4064, "lr": 0.0009310659985084458, "epoch": 5.044790193305045, "percentage": 25.22, "elapsed_time": "0:34:47", "remaining_time": "1:43:07", "throughput": 5154.35, "total_tokens": 10757888} {"current_steps": 10705, "total_steps": 42420, "loss": 0.3567, "lr": 0.0009309617269987215, "epoch": 5.047147571900047, "percentage": 25.24, "elapsed_time": "0:34:47", "remaining_time": "1:43:05", "throughput": 5154.28, "total_tokens": 10761888} {"current_steps": 10710, "total_steps": 42420, "loss": 0.3455, "lr": 0.0009308573825344941, "epoch": 5.0495049504950495, "percentage": 25.25, "elapsed_time": "0:34:48", "remaining_time": "1:43:04", "throughput": 5154.35, "total_tokens": 10766592} {"current_steps": 10715, "total_steps": 42420, "loss": 0.3512, "lr": 0.0009307529651334274, "epoch": 5.051862329090052, "percentage": 25.26, "elapsed_time": "0:34:49", "remaining_time": "1:43:03", "throughput": 5154.46, "total_tokens": 10771488} {"current_steps": 10720, "total_steps": 42420, "loss": 0.3283, "lr": 0.0009306484748131974, "epoch": 5.054219707685054, "percentage": 25.27, "elapsed_time": "0:34:50", "remaining_time": "1:43:01", "throughput": 5154.38, "total_tokens": 10775392} {"current_steps": 10725, "total_steps": 42420, "loss": 0.3264, "lr": 0.0009305439115914927, "epoch": 5.056577086280057, "percentage": 25.28, "elapsed_time": "0:34:51", "remaining_time": "1:43:00", "throughput": 5154.44, "total_tokens": 10779872} {"current_steps": 10730, "total_steps": 42420, "loss": 0.3455, "lr": 0.0009304392754860139, "epoch": 5.058934464875059, "percentage": 25.29, "elapsed_time": "0:34:52", "remaining_time": "1:42:59", "throughput": 5154.64, "total_tokens": 10785440} {"current_steps": 10735, "total_steps": 42420, "loss": 0.3112, "lr": 0.0009303345665144742, "epoch": 5.061291843470062, "percentage": 25.31, "elapsed_time": "0:34:53", "remaining_time": "1:42:58", "throughput": 5154.6, "total_tokens": 10789664} {"current_steps": 10740, "total_steps": 42420, "loss": 0.2661, "lr": 0.0009302297846945991, "epoch": 5.063649222065064, "percentage": 25.32, "elapsed_time": "0:34:54", "remaining_time": "1:42:57", "throughput": 5154.64, "total_tokens": 10794368} {"current_steps": 10745, "total_steps": 42420, "loss": 0.427, "lr": 0.0009301249300441264, "epoch": 5.066006600660066, "percentage": 25.33, "elapsed_time": "0:34:55", "remaining_time": "1:42:56", "throughput": 5155.06, "total_tokens": 10801440} {"current_steps": 10750, "total_steps": 42420, "loss": 0.3213, "lr": 0.0009300200025808061, "epoch": 5.068363979255069, "percentage": 25.34, "elapsed_time": "0:34:56", "remaining_time": "1:42:55", "throughput": 5155.25, "total_tokens": 10806752} {"current_steps": 10755, "total_steps": 42420, "loss": 0.3391, "lr": 0.0009299150023224006, "epoch": 5.07072135785007, "percentage": 25.35, "elapsed_time": "0:34:57", "remaining_time": "1:42:54", "throughput": 5155.17, "total_tokens": 10810496} {"current_steps": 10760, "total_steps": 42420, "loss": 0.3272, "lr": 0.0009298099292866848, "epoch": 5.073078736445073, "percentage": 25.37, "elapsed_time": "0:34:57", "remaining_time": "1:42:52", "throughput": 5155.3, "total_tokens": 10815360} {"current_steps": 10765, "total_steps": 42420, "loss": 0.3648, "lr": 0.0009297047834914458, "epoch": 5.075436115040075, "percentage": 25.38, "elapsed_time": "0:34:58", "remaining_time": "1:42:52", "throughput": 5155.55, "total_tokens": 10821152} {"current_steps": 10770, "total_steps": 42420, "loss": 0.3522, "lr": 0.0009295995649544829, "epoch": 5.0777934936350775, "percentage": 25.39, "elapsed_time": "0:34:59", "remaining_time": "1:42:50", "throughput": 5155.59, "total_tokens": 10825536} {"current_steps": 10775, "total_steps": 42420, "loss": 0.3331, "lr": 0.0009294942736936078, "epoch": 5.08015087223008, "percentage": 25.4, "elapsed_time": "0:35:00", "remaining_time": "1:42:50", "throughput": 5156.01, "total_tokens": 10832416} {"current_steps": 10780, "total_steps": 42420, "loss": 0.3043, "lr": 0.0009293889097266446, "epoch": 5.082508250825082, "percentage": 25.41, "elapsed_time": "0:35:01", "remaining_time": "1:42:49", "throughput": 5156.08, "total_tokens": 10837216} {"current_steps": 10785, "total_steps": 42420, "loss": 0.362, "lr": 0.0009292834730714296, "epoch": 5.084865629420085, "percentage": 25.42, "elapsed_time": "0:35:02", "remaining_time": "1:42:48", "throughput": 5156.38, "total_tokens": 10843040} {"current_steps": 10790, "total_steps": 42420, "loss": 0.3359, "lr": 0.0009291779637458113, "epoch": 5.087223008015087, "percentage": 25.44, "elapsed_time": "0:35:03", "remaining_time": "1:42:47", "throughput": 5156.47, "total_tokens": 10848064} {"current_steps": 10795, "total_steps": 42420, "loss": 0.3626, "lr": 0.000929072381767651, "epoch": 5.0895803866100895, "percentage": 25.45, "elapsed_time": "0:35:04", "remaining_time": "1:42:46", "throughput": 5156.76, "total_tokens": 10853984} {"current_steps": 10800, "total_steps": 42420, "loss": 0.3293, "lr": 0.0009289667271548214, "epoch": 5.091937765205092, "percentage": 25.46, "elapsed_time": "0:35:05", "remaining_time": "1:42:45", "throughput": 5157.06, "total_tokens": 10860192} {"current_steps": 10805, "total_steps": 42420, "loss": 0.3163, "lr": 0.0009288609999252087, "epoch": 5.094295143800094, "percentage": 25.47, "elapsed_time": "0:35:06", "remaining_time": "1:42:44", "throughput": 5157.05, "total_tokens": 10864512} {"current_steps": 10810, "total_steps": 42420, "loss": 0.3319, "lr": 0.0009287552000967101, "epoch": 5.096652522395097, "percentage": 25.48, "elapsed_time": "0:35:07", "remaining_time": "1:42:42", "throughput": 5157.14, "total_tokens": 10869120} {"current_steps": 10815, "total_steps": 42420, "loss": 0.3616, "lr": 0.000928649327687236, "epoch": 5.099009900990099, "percentage": 25.5, "elapsed_time": "0:35:08", "remaining_time": "1:42:41", "throughput": 5157.11, "total_tokens": 10873568} {"current_steps": 10820, "total_steps": 42420, "loss": 0.3697, "lr": 0.0009285433827147088, "epoch": 5.1013672795851015, "percentage": 25.51, "elapsed_time": "0:35:09", "remaining_time": "1:42:40", "throughput": 5157.12, "total_tokens": 10877888} {"current_steps": 10825, "total_steps": 42420, "loss": 0.3562, "lr": 0.0009284373651970631, "epoch": 5.103724658180104, "percentage": 25.52, "elapsed_time": "0:35:10", "remaining_time": "1:42:39", "throughput": 5157.38, "total_tokens": 10883680} {"current_steps": 10830, "total_steps": 42420, "loss": 0.3609, "lr": 0.000928331275152246, "epoch": 5.106082036775106, "percentage": 25.53, "elapsed_time": "0:35:11", "remaining_time": "1:42:38", "throughput": 5157.71, "total_tokens": 10889792} {"current_steps": 10835, "total_steps": 42420, "loss": 0.3272, "lr": 0.0009282251125982166, "epoch": 5.108439415370109, "percentage": 25.54, "elapsed_time": "0:35:12", "remaining_time": "1:42:37", "throughput": 5157.91, "total_tokens": 10895136} {"current_steps": 10840, "total_steps": 42420, "loss": 0.3753, "lr": 0.0009281188775529466, "epoch": 5.110796793965111, "percentage": 25.55, "elapsed_time": "0:35:13", "remaining_time": "1:42:36", "throughput": 5157.98, "total_tokens": 10899872} {"current_steps": 10845, "total_steps": 42420, "loss": 0.349, "lr": 0.0009280125700344196, "epoch": 5.1131541725601135, "percentage": 25.57, "elapsed_time": "0:35:14", "remaining_time": "1:42:35", "throughput": 5158.06, "total_tokens": 10904864} {"current_steps": 10850, "total_steps": 42420, "loss": 0.2967, "lr": 0.0009279061900606318, "epoch": 5.115511551155116, "percentage": 25.58, "elapsed_time": "0:35:14", "remaining_time": "1:42:33", "throughput": 5157.99, "total_tokens": 10908992} {"current_steps": 10855, "total_steps": 42420, "loss": 0.3476, "lr": 0.0009277997376495914, "epoch": 5.117868929750118, "percentage": 25.59, "elapsed_time": "0:35:15", "remaining_time": "1:42:32", "throughput": 5158.14, "total_tokens": 10914176} {"current_steps": 10860, "total_steps": 42420, "loss": 0.3148, "lr": 0.000927693212819319, "epoch": 5.12022630834512, "percentage": 25.6, "elapsed_time": "0:35:16", "remaining_time": "1:42:31", "throughput": 5158.39, "total_tokens": 10919808} {"current_steps": 10865, "total_steps": 42420, "loss": 0.3246, "lr": 0.0009275866155878475, "epoch": 5.122583686940122, "percentage": 25.61, "elapsed_time": "0:35:17", "remaining_time": "1:42:30", "throughput": 5158.54, "total_tokens": 10924960} {"current_steps": 10870, "total_steps": 42420, "loss": 0.416, "lr": 0.000927479945973222, "epoch": 5.124941065535125, "percentage": 25.62, "elapsed_time": "0:35:18", "remaining_time": "1:42:29", "throughput": 5158.52, "total_tokens": 10929312} {"current_steps": 10875, "total_steps": 42420, "loss": 0.3608, "lr": 0.0009273732039934999, "epoch": 5.127298444130127, "percentage": 25.64, "elapsed_time": "0:35:19", "remaining_time": "1:42:28", "throughput": 5158.66, "total_tokens": 10934176} {"current_steps": 10880, "total_steps": 42420, "loss": 0.3575, "lr": 0.0009272663896667507, "epoch": 5.129655822725129, "percentage": 25.65, "elapsed_time": "0:35:20", "remaining_time": "1:42:27", "throughput": 5158.82, "total_tokens": 10939296} {"current_steps": 10885, "total_steps": 42420, "loss": 0.3557, "lr": 0.0009271595030110563, "epoch": 5.132013201320132, "percentage": 25.66, "elapsed_time": "0:35:21", "remaining_time": "1:42:26", "throughput": 5158.94, "total_tokens": 10944320} {"current_steps": 10890, "total_steps": 42420, "loss": 0.3068, "lr": 0.0009270525440445108, "epoch": 5.134370579915134, "percentage": 25.67, "elapsed_time": "0:35:22", "remaining_time": "1:42:24", "throughput": 5159.01, "total_tokens": 10948896} {"current_steps": 10895, "total_steps": 42420, "loss": 0.3093, "lr": 0.0009269455127852206, "epoch": 5.136727958510137, "percentage": 25.68, "elapsed_time": "0:35:23", "remaining_time": "1:42:24", "throughput": 5159.65, "total_tokens": 10957504} {"current_steps": 10900, "total_steps": 42420, "loss": 0.3329, "lr": 0.0009268384092513041, "epoch": 5.139085337105139, "percentage": 25.7, "elapsed_time": "0:35:24", "remaining_time": "1:42:24", "throughput": 5159.87, "total_tokens": 10963104} {"current_steps": 10905, "total_steps": 42420, "loss": 0.3717, "lr": 0.0009267312334608923, "epoch": 5.141442715700141, "percentage": 25.71, "elapsed_time": "0:35:25", "remaining_time": "1:42:22", "throughput": 5159.97, "total_tokens": 10967872} {"current_steps": 10910, "total_steps": 42420, "loss": 0.3006, "lr": 0.0009266239854321282, "epoch": 5.143800094295144, "percentage": 25.72, "elapsed_time": "0:35:26", "remaining_time": "1:42:21", "throughput": 5159.93, "total_tokens": 10971808} {"current_steps": 10915, "total_steps": 42420, "loss": 0.2925, "lr": 0.0009265166651831669, "epoch": 5.146157472890146, "percentage": 25.73, "elapsed_time": "0:35:27", "remaining_time": "1:42:20", "throughput": 5160.06, "total_tokens": 10976928} {"current_steps": 10920, "total_steps": 42420, "loss": 0.3439, "lr": 0.000926409272732176, "epoch": 5.148514851485149, "percentage": 25.74, "elapsed_time": "0:35:28", "remaining_time": "1:42:19", "throughput": 5160.23, "total_tokens": 10982080} {"current_steps": 10925, "total_steps": 42420, "loss": 0.338, "lr": 0.0009263018080973353, "epoch": 5.150872230080151, "percentage": 25.75, "elapsed_time": "0:35:29", "remaining_time": "1:42:17", "throughput": 5160.22, "total_tokens": 10986336} {"current_steps": 10930, "total_steps": 42420, "loss": 0.3546, "lr": 0.0009261942712968368, "epoch": 5.1532296086751535, "percentage": 25.77, "elapsed_time": "0:35:30", "remaining_time": "1:42:16", "throughput": 5160.56, "total_tokens": 10992576} {"current_steps": 10935, "total_steps": 42420, "loss": 0.3394, "lr": 0.0009260866623488844, "epoch": 5.155586987270156, "percentage": 25.78, "elapsed_time": "0:35:31", "remaining_time": "1:42:15", "throughput": 5160.7, "total_tokens": 10997728} {"current_steps": 10940, "total_steps": 42420, "loss": 0.3747, "lr": 0.0009259789812716947, "epoch": 5.157944365865158, "percentage": 25.79, "elapsed_time": "0:35:31", "remaining_time": "1:42:14", "throughput": 5160.72, "total_tokens": 11002240} {"current_steps": 10945, "total_steps": 42420, "loss": 0.3341, "lr": 0.000925871228083496, "epoch": 5.160301744460161, "percentage": 25.8, "elapsed_time": "0:35:32", "remaining_time": "1:42:13", "throughput": 5160.68, "total_tokens": 11006528} {"current_steps": 10950, "total_steps": 42420, "loss": 0.3059, "lr": 0.0009257634028025294, "epoch": 5.162659123055163, "percentage": 25.81, "elapsed_time": "0:35:33", "remaining_time": "1:42:11", "throughput": 5160.66, "total_tokens": 11010592} {"current_steps": 10955, "total_steps": 42420, "loss": 0.4033, "lr": 0.0009256555054470478, "epoch": 5.165016501650165, "percentage": 25.83, "elapsed_time": "0:35:34", "remaining_time": "1:42:10", "throughput": 5160.79, "total_tokens": 11015584} {"current_steps": 10960, "total_steps": 42420, "loss": 0.3774, "lr": 0.0009255475360353163, "epoch": 5.167373880245167, "percentage": 25.84, "elapsed_time": "0:35:35", "remaining_time": "1:42:09", "throughput": 5160.96, "total_tokens": 11021088} {"current_steps": 10965, "total_steps": 42420, "loss": 0.3305, "lr": 0.0009254394945856124, "epoch": 5.169731258840169, "percentage": 25.85, "elapsed_time": "0:35:36", "remaining_time": "1:42:08", "throughput": 5161.14, "total_tokens": 11026368} {"current_steps": 10970, "total_steps": 42420, "loss": 0.3572, "lr": 0.0009253313811162255, "epoch": 5.172088637435172, "percentage": 25.86, "elapsed_time": "0:35:37", "remaining_time": "1:42:07", "throughput": 5161.26, "total_tokens": 11031456} {"current_steps": 10975, "total_steps": 42420, "loss": 0.3275, "lr": 0.0009252231956454575, "epoch": 5.174446016030174, "percentage": 25.87, "elapsed_time": "0:35:38", "remaining_time": "1:42:06", "throughput": 5161.3, "total_tokens": 11035904} {"current_steps": 10980, "total_steps": 42420, "loss": 0.287, "lr": 0.0009251149381916224, "epoch": 5.176803394625177, "percentage": 25.88, "elapsed_time": "0:35:39", "remaining_time": "1:42:05", "throughput": 5161.61, "total_tokens": 11041856} {"current_steps": 10985, "total_steps": 42420, "loss": 0.3577, "lr": 0.0009250066087730463, "epoch": 5.179160773220179, "percentage": 25.9, "elapsed_time": "0:35:40", "remaining_time": "1:42:04", "throughput": 5161.71, "total_tokens": 11046528} {"current_steps": 10990, "total_steps": 42420, "loss": 0.4277, "lr": 0.0009248982074080677, "epoch": 5.181518151815181, "percentage": 25.91, "elapsed_time": "0:35:40", "remaining_time": "1:42:02", "throughput": 5161.72, "total_tokens": 11050816} {"current_steps": 10995, "total_steps": 42420, "loss": 0.3313, "lr": 0.0009247897341150366, "epoch": 5.183875530410184, "percentage": 25.92, "elapsed_time": "0:35:41", "remaining_time": "1:42:01", "throughput": 5161.85, "total_tokens": 11055776} {"current_steps": 11000, "total_steps": 42420, "loss": 0.3407, "lr": 0.0009246811889123162, "epoch": 5.186232909005186, "percentage": 25.93, "elapsed_time": "0:35:42", "remaining_time": "1:42:01", "throughput": 5162.25, "total_tokens": 11062496} {"current_steps": 11005, "total_steps": 42420, "loss": 0.3202, "lr": 0.0009245725718182813, "epoch": 5.188590287600189, "percentage": 25.94, "elapsed_time": "0:35:43", "remaining_time": "1:42:00", "throughput": 5162.46, "total_tokens": 11068128} {"current_steps": 11010, "total_steps": 42420, "loss": 0.3163, "lr": 0.0009244638828513188, "epoch": 5.190947666195191, "percentage": 25.95, "elapsed_time": "0:35:44", "remaining_time": "1:41:59", "throughput": 5162.61, "total_tokens": 11073440} {"current_steps": 11015, "total_steps": 42420, "loss": 0.3496, "lr": 0.0009243551220298278, "epoch": 5.193305044790193, "percentage": 25.97, "elapsed_time": "0:35:45", "remaining_time": "1:41:57", "throughput": 5162.63, "total_tokens": 11077920} {"current_steps": 11020, "total_steps": 42420, "loss": 0.3286, "lr": 0.0009242462893722199, "epoch": 5.195662423385196, "percentage": 25.98, "elapsed_time": "0:35:46", "remaining_time": "1:41:56", "throughput": 5162.62, "total_tokens": 11082048} {"current_steps": 11025, "total_steps": 42420, "loss": 0.4055, "lr": 0.0009241373848969184, "epoch": 5.198019801980198, "percentage": 25.99, "elapsed_time": "0:35:47", "remaining_time": "1:41:55", "throughput": 5162.77, "total_tokens": 11087104} {"current_steps": 11030, "total_steps": 42420, "loss": 0.4208, "lr": 0.0009240284086223591, "epoch": 5.200377180575201, "percentage": 26.0, "elapsed_time": "0:35:48", "remaining_time": "1:41:53", "throughput": 5162.85, "total_tokens": 11091680} {"current_steps": 11035, "total_steps": 42420, "loss": 0.3389, "lr": 0.0009239193605669898, "epoch": 5.202734559170203, "percentage": 26.01, "elapsed_time": "0:35:49", "remaining_time": "1:41:52", "throughput": 5162.99, "total_tokens": 11096640} {"current_steps": 11040, "total_steps": 42420, "loss": 0.3551, "lr": 0.0009238102407492705, "epoch": 5.205091937765205, "percentage": 26.03, "elapsed_time": "0:35:50", "remaining_time": "1:41:51", "throughput": 5163.03, "total_tokens": 11101088} {"current_steps": 11045, "total_steps": 42420, "loss": 0.2879, "lr": 0.0009237010491876734, "epoch": 5.207449316360208, "percentage": 26.04, "elapsed_time": "0:35:50", "remaining_time": "1:41:50", "throughput": 5163.1, "total_tokens": 11105696} {"current_steps": 11050, "total_steps": 42420, "loss": 0.318, "lr": 0.0009235917859006826, "epoch": 5.20980669495521, "percentage": 26.05, "elapsed_time": "0:35:51", "remaining_time": "1:41:48", "throughput": 5163.16, "total_tokens": 11110336} {"current_steps": 11055, "total_steps": 42420, "loss": 0.3823, "lr": 0.0009234824509067948, "epoch": 5.212164073550213, "percentage": 26.06, "elapsed_time": "0:35:52", "remaining_time": "1:41:47", "throughput": 5163.22, "total_tokens": 11115040} {"current_steps": 11060, "total_steps": 42420, "loss": 0.3422, "lr": 0.0009233730442245183, "epoch": 5.214521452145214, "percentage": 26.07, "elapsed_time": "0:35:53", "remaining_time": "1:41:46", "throughput": 5163.22, "total_tokens": 11119424} {"current_steps": 11065, "total_steps": 42420, "loss": 0.411, "lr": 0.0009232635658723738, "epoch": 5.2168788307402165, "percentage": 26.08, "elapsed_time": "0:35:54", "remaining_time": "1:41:46", "throughput": 5163.79, "total_tokens": 11127968} {"current_steps": 11070, "total_steps": 42420, "loss": 0.3556, "lr": 0.0009231540158688942, "epoch": 5.219236209335219, "percentage": 26.1, "elapsed_time": "0:35:55", "remaining_time": "1:41:45", "throughput": 5163.85, "total_tokens": 11132992} {"current_steps": 11075, "total_steps": 42420, "loss": 0.3429, "lr": 0.0009230443942326246, "epoch": 5.221593587930221, "percentage": 26.11, "elapsed_time": "0:35:56", "remaining_time": "1:41:44", "throughput": 5163.94, "total_tokens": 11137728} {"current_steps": 11080, "total_steps": 42420, "loss": 0.3812, "lr": 0.0009229347009821219, "epoch": 5.223950966525224, "percentage": 26.12, "elapsed_time": "0:35:57", "remaining_time": "1:41:43", "throughput": 5164.11, "total_tokens": 11143040} {"current_steps": 11085, "total_steps": 42420, "loss": 0.3515, "lr": 0.000922824936135955, "epoch": 5.226308345120226, "percentage": 26.13, "elapsed_time": "0:35:58", "remaining_time": "1:41:42", "throughput": 5164.3, "total_tokens": 11148960} {"current_steps": 11090, "total_steps": 42420, "loss": 0.3003, "lr": 0.000922715099712706, "epoch": 5.2286657237152285, "percentage": 26.14, "elapsed_time": "0:35:59", "remaining_time": "1:41:41", "throughput": 5164.27, "total_tokens": 11153280} {"current_steps": 11095, "total_steps": 42420, "loss": 0.2883, "lr": 0.0009226051917309678, "epoch": 5.231023102310231, "percentage": 26.16, "elapsed_time": "0:36:00", "remaining_time": "1:41:40", "throughput": 5164.53, "total_tokens": 11159520} {"current_steps": 11100, "total_steps": 42420, "loss": 0.389, "lr": 0.0009224952122093461, "epoch": 5.233380480905233, "percentage": 26.17, "elapsed_time": "0:36:01", "remaining_time": "1:41:39", "throughput": 5164.51, "total_tokens": 11163680} {"current_steps": 11105, "total_steps": 42420, "loss": 0.3793, "lr": 0.0009223851611664584, "epoch": 5.235737859500236, "percentage": 26.18, "elapsed_time": "0:36:02", "remaining_time": "1:41:38", "throughput": 5164.51, "total_tokens": 11168384} {"current_steps": 11110, "total_steps": 42420, "loss": 0.3458, "lr": 0.0009222750386209348, "epoch": 5.238095238095238, "percentage": 26.19, "elapsed_time": "0:36:03", "remaining_time": "1:41:36", "throughput": 5164.62, "total_tokens": 11173248} {"current_steps": 11115, "total_steps": 42420, "loss": 0.3485, "lr": 0.0009221648445914168, "epoch": 5.2404526166902405, "percentage": 26.2, "elapsed_time": "0:36:04", "remaining_time": "1:41:35", "throughput": 5164.69, "total_tokens": 11177984} {"current_steps": 11120, "total_steps": 42420, "loss": 0.3742, "lr": 0.0009220545790965586, "epoch": 5.242809995285243, "percentage": 26.21, "elapsed_time": "0:36:05", "remaining_time": "1:41:34", "throughput": 5164.81, "total_tokens": 11183008} {"current_steps": 11125, "total_steps": 42420, "loss": 0.4164, "lr": 0.0009219442421550263, "epoch": 5.245167373880245, "percentage": 26.23, "elapsed_time": "0:36:06", "remaining_time": "1:41:33", "throughput": 5164.92, "total_tokens": 11187936} {"current_steps": 11130, "total_steps": 42420, "loss": 0.3882, "lr": 0.0009218338337854982, "epoch": 5.247524752475248, "percentage": 26.24, "elapsed_time": "0:36:07", "remaining_time": "1:41:32", "throughput": 5165.07, "total_tokens": 11193248} {"current_steps": 11135, "total_steps": 42420, "loss": 0.3615, "lr": 0.000921723354006664, "epoch": 5.24988213107025, "percentage": 26.25, "elapsed_time": "0:36:08", "remaining_time": "1:41:31", "throughput": 5165.2, "total_tokens": 11198368} {"current_steps": 11140, "total_steps": 42420, "loss": 0.3713, "lr": 0.0009216128028372266, "epoch": 5.2522395096652525, "percentage": 26.26, "elapsed_time": "0:36:08", "remaining_time": "1:41:30", "throughput": 5165.29, "total_tokens": 11203424} {"current_steps": 11145, "total_steps": 42420, "loss": 0.3284, "lr": 0.0009215021802959003, "epoch": 5.254596888260255, "percentage": 26.27, "elapsed_time": "0:36:09", "remaining_time": "1:41:29", "throughput": 5165.43, "total_tokens": 11208512} {"current_steps": 11150, "total_steps": 42420, "loss": 0.3441, "lr": 0.0009213914864014115, "epoch": 5.256954266855257, "percentage": 26.28, "elapsed_time": "0:36:10", "remaining_time": "1:41:27", "throughput": 5165.41, "total_tokens": 11212768} {"current_steps": 11155, "total_steps": 42420, "loss": 0.3076, "lr": 0.0009212807211724987, "epoch": 5.259311645450259, "percentage": 26.3, "elapsed_time": "0:36:11", "remaining_time": "1:41:26", "throughput": 5165.52, "total_tokens": 11217792} {"current_steps": 11160, "total_steps": 42420, "loss": 0.4084, "lr": 0.000921169884627913, "epoch": 5.261669024045261, "percentage": 26.31, "elapsed_time": "0:36:12", "remaining_time": "1:41:25", "throughput": 5165.52, "total_tokens": 11221920} {"current_steps": 11165, "total_steps": 42420, "loss": 0.3327, "lr": 0.0009210589767864168, "epoch": 5.264026402640264, "percentage": 26.32, "elapsed_time": "0:36:13", "remaining_time": "1:41:24", "throughput": 5165.62, "total_tokens": 11226720} {"current_steps": 11170, "total_steps": 42420, "loss": 0.3461, "lr": 0.0009209479976667851, "epoch": 5.266383781235266, "percentage": 26.33, "elapsed_time": "0:36:14", "remaining_time": "1:41:23", "throughput": 5165.84, "total_tokens": 11232192} {"current_steps": 11175, "total_steps": 42420, "loss": 0.3519, "lr": 0.0009208369472878047, "epoch": 5.2687411598302685, "percentage": 26.34, "elapsed_time": "0:36:15", "remaining_time": "1:41:22", "throughput": 5166.08, "total_tokens": 11237856} {"current_steps": 11180, "total_steps": 42420, "loss": 0.3596, "lr": 0.0009207258256682743, "epoch": 5.271098538425271, "percentage": 26.36, "elapsed_time": "0:36:16", "remaining_time": "1:41:21", "throughput": 5166.24, "total_tokens": 11242976} {"current_steps": 11185, "total_steps": 42420, "loss": 0.3453, "lr": 0.0009206146328270053, "epoch": 5.273455917020273, "percentage": 26.37, "elapsed_time": "0:36:17", "remaining_time": "1:41:19", "throughput": 5166.27, "total_tokens": 11247456} {"current_steps": 11190, "total_steps": 42420, "loss": 0.3253, "lr": 0.0009205033687828206, "epoch": 5.275813295615276, "percentage": 26.38, "elapsed_time": "0:36:18", "remaining_time": "1:41:18", "throughput": 5166.49, "total_tokens": 11253152} {"current_steps": 11195, "total_steps": 42420, "loss": 0.3838, "lr": 0.0009203920335545552, "epoch": 5.278170674210278, "percentage": 26.39, "elapsed_time": "0:36:19", "remaining_time": "1:41:17", "throughput": 5166.61, "total_tokens": 11258304} {"current_steps": 11200, "total_steps": 42420, "loss": 0.3662, "lr": 0.0009202806271610565, "epoch": 5.2805280528052805, "percentage": 26.4, "elapsed_time": "0:36:19", "remaining_time": "1:41:16", "throughput": 5166.65, "total_tokens": 11262848} {"current_steps": 11205, "total_steps": 42420, "loss": 0.3303, "lr": 0.0009201691496211835, "epoch": 5.282885431400283, "percentage": 26.41, "elapsed_time": "0:36:20", "remaining_time": "1:41:15", "throughput": 5166.62, "total_tokens": 11266976} {"current_steps": 11210, "total_steps": 42420, "loss": 0.3666, "lr": 0.0009200576009538075, "epoch": 5.285242809995285, "percentage": 26.43, "elapsed_time": "0:36:21", "remaining_time": "1:41:14", "throughput": 5166.91, "total_tokens": 11273376} {"current_steps": 11215, "total_steps": 42420, "loss": 0.3645, "lr": 0.0009199459811778119, "epoch": 5.287600188590288, "percentage": 26.44, "elapsed_time": "0:36:22", "remaining_time": "1:41:13", "throughput": 5167.0, "total_tokens": 11278016} {"current_steps": 11220, "total_steps": 42420, "loss": 0.3573, "lr": 0.000919834290312092, "epoch": 5.28995756718529, "percentage": 26.45, "elapsed_time": "0:36:23", "remaining_time": "1:41:12", "throughput": 5167.24, "total_tokens": 11284000} {"current_steps": 11225, "total_steps": 42420, "loss": 0.3725, "lr": 0.000919722528375555, "epoch": 5.2923149457802925, "percentage": 26.46, "elapsed_time": "0:36:24", "remaining_time": "1:41:11", "throughput": 5167.36, "total_tokens": 11288928} {"current_steps": 11230, "total_steps": 42420, "loss": 0.3182, "lr": 0.0009196106953871205, "epoch": 5.294672324375295, "percentage": 26.47, "elapsed_time": "0:36:25", "remaining_time": "1:41:09", "throughput": 5167.39, "total_tokens": 11293248} {"current_steps": 11235, "total_steps": 42420, "loss": 0.3134, "lr": 0.0009194987913657198, "epoch": 5.297029702970297, "percentage": 26.49, "elapsed_time": "0:36:26", "remaining_time": "1:41:08", "throughput": 5167.43, "total_tokens": 11297696} {"current_steps": 11240, "total_steps": 42420, "loss": 0.2955, "lr": 0.0009193868163302966, "epoch": 5.2993870815653, "percentage": 26.5, "elapsed_time": "0:36:27", "remaining_time": "1:41:07", "throughput": 5167.62, "total_tokens": 11303168} {"current_steps": 11245, "total_steps": 42420, "loss": 0.3681, "lr": 0.000919274770299806, "epoch": 5.301744460160302, "percentage": 26.51, "elapsed_time": "0:36:28", "remaining_time": "1:41:06", "throughput": 5167.77, "total_tokens": 11308416} {"current_steps": 11250, "total_steps": 42420, "loss": 0.3663, "lr": 0.0009191626532932157, "epoch": 5.3041018387553045, "percentage": 26.52, "elapsed_time": "0:36:29", "remaining_time": "1:41:05", "throughput": 5167.96, "total_tokens": 11313984} {"current_steps": 11255, "total_steps": 42420, "loss": 0.3234, "lr": 0.0009190504653295052, "epoch": 5.306459217350307, "percentage": 26.53, "elapsed_time": "0:36:30", "remaining_time": "1:41:04", "throughput": 5168.21, "total_tokens": 11319776} {"current_steps": 11260, "total_steps": 42420, "loss": 0.3689, "lr": 0.0009189382064276659, "epoch": 5.308816595945308, "percentage": 26.54, "elapsed_time": "0:36:31", "remaining_time": "1:41:03", "throughput": 5168.47, "total_tokens": 11325568} {"current_steps": 11265, "total_steps": 42420, "loss": 0.3406, "lr": 0.0009188258766067016, "epoch": 5.311173974540311, "percentage": 26.56, "elapsed_time": "0:36:32", "remaining_time": "1:41:02", "throughput": 5168.49, "total_tokens": 11329920} {"current_steps": 11270, "total_steps": 42420, "loss": 0.3565, "lr": 0.0009187134758856275, "epoch": 5.313531353135313, "percentage": 26.57, "elapsed_time": "0:36:33", "remaining_time": "1:41:02", "throughput": 5168.94, "total_tokens": 11336928} {"current_steps": 11275, "total_steps": 42420, "loss": 0.3465, "lr": 0.0009186010042834714, "epoch": 5.315888731730316, "percentage": 26.58, "elapsed_time": "0:36:34", "remaining_time": "1:41:01", "throughput": 5169.17, "total_tokens": 11342624} {"current_steps": 11280, "total_steps": 42420, "loss": 0.3645, "lr": 0.0009184884618192726, "epoch": 5.318246110325318, "percentage": 26.59, "elapsed_time": "0:36:35", "remaining_time": "1:41:00", "throughput": 5169.29, "total_tokens": 11348064} {"current_steps": 11285, "total_steps": 42420, "loss": 0.396, "lr": 0.0009183758485120828, "epoch": 5.32060348892032, "percentage": 26.6, "elapsed_time": "0:36:36", "remaining_time": "1:40:59", "throughput": 5169.55, "total_tokens": 11354016} {"current_steps": 11290, "total_steps": 42420, "loss": 0.3372, "lr": 0.0009182631643809653, "epoch": 5.322960867515323, "percentage": 26.61, "elapsed_time": "0:36:37", "remaining_time": "1:40:58", "throughput": 5169.46, "total_tokens": 11357696} {"current_steps": 11295, "total_steps": 42420, "loss": 0.3087, "lr": 0.0009181504094449961, "epoch": 5.325318246110325, "percentage": 26.63, "elapsed_time": "0:36:37", "remaining_time": "1:40:56", "throughput": 5169.46, "total_tokens": 11361824} {"current_steps": 11300, "total_steps": 42420, "loss": 0.4248, "lr": 0.0009180375837232621, "epoch": 5.327675624705328, "percentage": 26.64, "elapsed_time": "0:36:38", "remaining_time": "1:40:55", "throughput": 5169.63, "total_tokens": 11366976} {"current_steps": 11305, "total_steps": 42420, "loss": 0.2624, "lr": 0.0009179246872348632, "epoch": 5.33003300330033, "percentage": 26.65, "elapsed_time": "0:36:39", "remaining_time": "1:40:54", "throughput": 5169.78, "total_tokens": 11372000} {"current_steps": 11310, "total_steps": 42420, "loss": 0.2861, "lr": 0.0009178117199989107, "epoch": 5.332390381895332, "percentage": 26.66, "elapsed_time": "0:36:40", "remaining_time": "1:40:53", "throughput": 5169.81, "total_tokens": 11376608} {"current_steps": 11315, "total_steps": 42420, "loss": 0.3939, "lr": 0.0009176986820345279, "epoch": 5.334747760490335, "percentage": 26.67, "elapsed_time": "0:36:41", "remaining_time": "1:40:51", "throughput": 5169.77, "total_tokens": 11380832} {"current_steps": 11320, "total_steps": 42420, "loss": 0.3063, "lr": 0.0009175855733608505, "epoch": 5.337105139085337, "percentage": 26.69, "elapsed_time": "0:36:42", "remaining_time": "1:40:50", "throughput": 5169.84, "total_tokens": 11385696} {"current_steps": 11325, "total_steps": 42420, "loss": 0.3859, "lr": 0.0009174723939970257, "epoch": 5.33946251768034, "percentage": 26.7, "elapsed_time": "0:36:43", "remaining_time": "1:40:49", "throughput": 5170.0, "total_tokens": 11391072} {"current_steps": 11330, "total_steps": 42420, "loss": 0.3088, "lr": 0.0009173591439622129, "epoch": 5.341819896275342, "percentage": 26.71, "elapsed_time": "0:36:44", "remaining_time": "1:40:48", "throughput": 5170.11, "total_tokens": 11396000} {"current_steps": 11335, "total_steps": 42420, "loss": 0.3203, "lr": 0.0009172458232755832, "epoch": 5.344177274870344, "percentage": 26.72, "elapsed_time": "0:36:45", "remaining_time": "1:40:47", "throughput": 5170.11, "total_tokens": 11400256} {"current_steps": 11340, "total_steps": 42420, "loss": 0.3031, "lr": 0.0009171324319563202, "epoch": 5.346534653465347, "percentage": 26.73, "elapsed_time": "0:36:46", "remaining_time": "1:40:46", "throughput": 5170.39, "total_tokens": 11406144} {"current_steps": 11345, "total_steps": 42420, "loss": 0.3416, "lr": 0.000917018970023619, "epoch": 5.348892032060349, "percentage": 26.74, "elapsed_time": "0:36:47", "remaining_time": "1:40:45", "throughput": 5170.56, "total_tokens": 11411616} {"current_steps": 11350, "total_steps": 42420, "loss": 0.3454, "lr": 0.0009169054374966868, "epoch": 5.351249410655352, "percentage": 26.76, "elapsed_time": "0:36:47", "remaining_time": "1:40:43", "throughput": 5170.49, "total_tokens": 11415552} {"current_steps": 11355, "total_steps": 42420, "loss": 0.4033, "lr": 0.0009167918343947426, "epoch": 5.353606789250353, "percentage": 26.77, "elapsed_time": "0:36:48", "remaining_time": "1:40:42", "throughput": 5170.71, "total_tokens": 11421248} {"current_steps": 11360, "total_steps": 42420, "loss": 0.3459, "lr": 0.0009166781607370175, "epoch": 5.355964167845356, "percentage": 26.78, "elapsed_time": "0:36:49", "remaining_time": "1:40:42", "throughput": 5170.97, "total_tokens": 11427264} {"current_steps": 11365, "total_steps": 42420, "loss": 0.3891, "lr": 0.0009165644165427545, "epoch": 5.358321546440358, "percentage": 26.79, "elapsed_time": "0:36:50", "remaining_time": "1:40:40", "throughput": 5171.1, "total_tokens": 11432032} {"current_steps": 11370, "total_steps": 42420, "loss": 0.3552, "lr": 0.0009164506018312089, "epoch": 5.36067892503536, "percentage": 26.8, "elapsed_time": "0:36:51", "remaining_time": "1:40:40", "throughput": 5171.41, "total_tokens": 11438080} {"current_steps": 11375, "total_steps": 42420, "loss": 0.2845, "lr": 0.0009163367166216472, "epoch": 5.363036303630363, "percentage": 26.82, "elapsed_time": "0:36:52", "remaining_time": "1:40:38", "throughput": 5171.56, "total_tokens": 11443168} {"current_steps": 11380, "total_steps": 42420, "loss": 0.29, "lr": 0.0009162227609333484, "epoch": 5.365393682225365, "percentage": 26.83, "elapsed_time": "0:36:53", "remaining_time": "1:40:37", "throughput": 5171.7, "total_tokens": 11448288} {"current_steps": 11385, "total_steps": 42420, "loss": 0.4432, "lr": 0.0009161087347856032, "epoch": 5.367751060820368, "percentage": 26.84, "elapsed_time": "0:36:54", "remaining_time": "1:40:36", "throughput": 5171.73, "total_tokens": 11452928} {"current_steps": 11390, "total_steps": 42420, "loss": 0.289, "lr": 0.0009159946381977143, "epoch": 5.37010843941537, "percentage": 26.85, "elapsed_time": "0:36:55", "remaining_time": "1:40:35", "throughput": 5171.71, "total_tokens": 11457408} {"current_steps": 11395, "total_steps": 42420, "loss": 0.3744, "lr": 0.0009158804711889965, "epoch": 5.372465818010372, "percentage": 26.86, "elapsed_time": "0:36:56", "remaining_time": "1:40:34", "throughput": 5171.88, "total_tokens": 11462912} {"current_steps": 11400, "total_steps": 42420, "loss": 0.3583, "lr": 0.000915766233778776, "epoch": 5.374823196605375, "percentage": 26.87, "elapsed_time": "0:36:57", "remaining_time": "1:40:33", "throughput": 5172.1, "total_tokens": 11468384} {"current_steps": 11405, "total_steps": 42420, "loss": 0.3679, "lr": 0.0009156519259863916, "epoch": 5.377180575200377, "percentage": 26.89, "elapsed_time": "0:36:58", "remaining_time": "1:40:32", "throughput": 5172.26, "total_tokens": 11473824} {"current_steps": 11410, "total_steps": 42420, "loss": 0.3401, "lr": 0.0009155375478311934, "epoch": 5.37953795379538, "percentage": 26.9, "elapsed_time": "0:36:59", "remaining_time": "1:40:31", "throughput": 5172.57, "total_tokens": 11479904} {"current_steps": 11415, "total_steps": 42420, "loss": 0.3672, "lr": 0.0009154230993325437, "epoch": 5.381895332390382, "percentage": 26.91, "elapsed_time": "0:37:00", "remaining_time": "1:40:30", "throughput": 5172.82, "total_tokens": 11485760} {"current_steps": 11420, "total_steps": 42420, "loss": 0.3273, "lr": 0.0009153085805098169, "epoch": 5.384252710985384, "percentage": 26.92, "elapsed_time": "0:37:01", "remaining_time": "1:40:30", "throughput": 5173.08, "total_tokens": 11491584} {"current_steps": 11425, "total_steps": 42420, "loss": 0.3455, "lr": 0.0009151939913823988, "epoch": 5.386610089580387, "percentage": 26.93, "elapsed_time": "0:37:02", "remaining_time": "1:40:29", "throughput": 5173.22, "total_tokens": 11496832} {"current_steps": 11430, "total_steps": 42420, "loss": 0.3431, "lr": 0.0009150793319696876, "epoch": 5.388967468175389, "percentage": 26.94, "elapsed_time": "0:37:03", "remaining_time": "1:40:28", "throughput": 5173.39, "total_tokens": 11502016} {"current_steps": 11435, "total_steps": 42420, "loss": 0.3283, "lr": 0.0009149646022910933, "epoch": 5.391324846770392, "percentage": 26.96, "elapsed_time": "0:37:04", "remaining_time": "1:40:27", "throughput": 5173.61, "total_tokens": 11507648} {"current_steps": 11440, "total_steps": 42420, "loss": 0.294, "lr": 0.0009148498023660373, "epoch": 5.393682225365394, "percentage": 26.97, "elapsed_time": "0:37:05", "remaining_time": "1:40:25", "throughput": 5173.7, "total_tokens": 11512480} {"current_steps": 11445, "total_steps": 42420, "loss": 0.3547, "lr": 0.0009147349322139537, "epoch": 5.396039603960396, "percentage": 26.98, "elapsed_time": "0:37:06", "remaining_time": "1:40:24", "throughput": 5173.69, "total_tokens": 11516864} {"current_steps": 11450, "total_steps": 42420, "loss": 0.35, "lr": 0.0009146199918542878, "epoch": 5.398396982555399, "percentage": 26.99, "elapsed_time": "0:37:07", "remaining_time": "1:40:23", "throughput": 5173.94, "total_tokens": 11522432} {"current_steps": 11455, "total_steps": 42420, "loss": 0.3646, "lr": 0.0009145049813064972, "epoch": 5.400754361150401, "percentage": 27.0, "elapsed_time": "0:37:07", "remaining_time": "1:40:22", "throughput": 5173.98, "total_tokens": 11526944} {"current_steps": 11460, "total_steps": 42420, "loss": 0.3193, "lr": 0.000914389900590051, "epoch": 5.403111739745403, "percentage": 27.02, "elapsed_time": "0:37:08", "remaining_time": "1:40:21", "throughput": 5174.14, "total_tokens": 11532576} {"current_steps": 11465, "total_steps": 42420, "loss": 0.354, "lr": 0.0009142747497244306, "epoch": 5.405469118340405, "percentage": 27.03, "elapsed_time": "0:37:09", "remaining_time": "1:40:20", "throughput": 5174.4, "total_tokens": 11538336} {"current_steps": 11470, "total_steps": 42420, "loss": 0.3613, "lr": 0.0009141595287291291, "epoch": 5.4078264969354075, "percentage": 27.04, "elapsed_time": "0:37:10", "remaining_time": "1:40:19", "throughput": 5174.34, "total_tokens": 11542176} {"current_steps": 11475, "total_steps": 42420, "loss": 0.3559, "lr": 0.0009140442376236515, "epoch": 5.41018387553041, "percentage": 27.05, "elapsed_time": "0:37:11", "remaining_time": "1:40:18", "throughput": 5174.67, "total_tokens": 11548960} {"current_steps": 11480, "total_steps": 42420, "loss": 0.3599, "lr": 0.0009139288764275144, "epoch": 5.412541254125412, "percentage": 27.06, "elapsed_time": "0:37:12", "remaining_time": "1:40:17", "throughput": 5174.83, "total_tokens": 11554144} {"current_steps": 11485, "total_steps": 42420, "loss": 0.3644, "lr": 0.0009138134451602466, "epoch": 5.414898632720415, "percentage": 27.07, "elapsed_time": "0:37:13", "remaining_time": "1:40:16", "throughput": 5175.05, "total_tokens": 11559936} {"current_steps": 11490, "total_steps": 42420, "loss": 0.3439, "lr": 0.0009136979438413889, "epoch": 5.417256011315417, "percentage": 27.09, "elapsed_time": "0:37:14", "remaining_time": "1:40:15", "throughput": 5175.11, "total_tokens": 11564640} {"current_steps": 11495, "total_steps": 42420, "loss": 0.3565, "lr": 0.0009135823724904934, "epoch": 5.4196133899104195, "percentage": 27.1, "elapsed_time": "0:37:15", "remaining_time": "1:40:14", "throughput": 5175.24, "total_tokens": 11569696} {"current_steps": 11500, "total_steps": 42420, "loss": 0.3937, "lr": 0.0009134667311271245, "epoch": 5.421970768505422, "percentage": 27.11, "elapsed_time": "0:37:16", "remaining_time": "1:40:13", "throughput": 5175.29, "total_tokens": 11574400} {"current_steps": 11505, "total_steps": 42420, "loss": 0.3582, "lr": 0.0009133510197708582, "epoch": 5.424328147100424, "percentage": 27.12, "elapsed_time": "0:37:17", "remaining_time": "1:40:11", "throughput": 5175.31, "total_tokens": 11578816} {"current_steps": 11510, "total_steps": 42420, "loss": 0.3543, "lr": 0.0009132352384412825, "epoch": 5.426685525695427, "percentage": 27.13, "elapsed_time": "0:37:18", "remaining_time": "1:40:10", "throughput": 5175.28, "total_tokens": 11582880} {"current_steps": 11515, "total_steps": 42420, "loss": 0.365, "lr": 0.0009131193871579975, "epoch": 5.429042904290429, "percentage": 27.15, "elapsed_time": "0:37:18", "remaining_time": "1:40:08", "throughput": 5175.2, "total_tokens": 11586816} {"current_steps": 11520, "total_steps": 42420, "loss": 0.3642, "lr": 0.0009130034659406144, "epoch": 5.4314002828854315, "percentage": 27.16, "elapsed_time": "0:37:19", "remaining_time": "1:40:07", "throughput": 5175.32, "total_tokens": 11591968} {"current_steps": 11525, "total_steps": 42420, "loss": 0.3192, "lr": 0.000912887474808757, "epoch": 5.433757661480434, "percentage": 27.17, "elapsed_time": "0:37:20", "remaining_time": "1:40:06", "throughput": 5175.31, "total_tokens": 11596384} {"current_steps": 11530, "total_steps": 42420, "loss": 0.326, "lr": 0.0009127714137820606, "epoch": 5.436115040075436, "percentage": 27.18, "elapsed_time": "0:37:21", "remaining_time": "1:40:05", "throughput": 5175.51, "total_tokens": 11601824} {"current_steps": 11535, "total_steps": 42420, "loss": 0.3713, "lr": 0.0009126552828801721, "epoch": 5.438472418670439, "percentage": 27.19, "elapsed_time": "0:37:22", "remaining_time": "1:40:04", "throughput": 5175.62, "total_tokens": 11606592} {"current_steps": 11540, "total_steps": 42420, "loss": 0.294, "lr": 0.0009125390821227511, "epoch": 5.440829797265441, "percentage": 27.2, "elapsed_time": "0:37:23", "remaining_time": "1:40:03", "throughput": 5175.73, "total_tokens": 11611904} {"current_steps": 11545, "total_steps": 42420, "loss": 0.3691, "lr": 0.0009124228115294677, "epoch": 5.4431871758604435, "percentage": 27.22, "elapsed_time": "0:37:24", "remaining_time": "1:40:02", "throughput": 5175.73, "total_tokens": 11616352} {"current_steps": 11550, "total_steps": 42420, "loss": 0.3653, "lr": 0.0009123064711200049, "epoch": 5.445544554455446, "percentage": 27.23, "elapsed_time": "0:37:25", "remaining_time": "1:40:00", "throughput": 5175.64, "total_tokens": 11620256} {"current_steps": 11555, "total_steps": 42420, "loss": 0.3912, "lr": 0.000912190060914057, "epoch": 5.4479019330504475, "percentage": 27.24, "elapsed_time": "0:37:26", "remaining_time": "1:39:59", "throughput": 5175.77, "total_tokens": 11625216} {"current_steps": 11560, "total_steps": 42420, "loss": 0.3528, "lr": 0.0009120735809313305, "epoch": 5.45025931164545, "percentage": 27.25, "elapsed_time": "0:37:26", "remaining_time": "1:39:58", "throughput": 5175.73, "total_tokens": 11629184} {"current_steps": 11565, "total_steps": 42420, "loss": 0.3244, "lr": 0.0009119570311915434, "epoch": 5.452616690240452, "percentage": 27.26, "elapsed_time": "0:37:27", "remaining_time": "1:39:57", "throughput": 5175.97, "total_tokens": 11634720} {"current_steps": 11570, "total_steps": 42420, "loss": 0.4297, "lr": 0.0009118404117144255, "epoch": 5.454974068835455, "percentage": 27.27, "elapsed_time": "0:37:28", "remaining_time": "1:39:55", "throughput": 5175.88, "total_tokens": 11638496} {"current_steps": 11575, "total_steps": 42420, "loss": 0.3642, "lr": 0.0009117237225197186, "epoch": 5.457331447430457, "percentage": 27.29, "elapsed_time": "0:37:29", "remaining_time": "1:39:54", "throughput": 5175.9, "total_tokens": 11642848} {"current_steps": 11580, "total_steps": 42420, "loss": 0.3414, "lr": 0.000911606963627176, "epoch": 5.4596888260254595, "percentage": 27.3, "elapsed_time": "0:37:30", "remaining_time": "1:39:53", "throughput": 5175.95, "total_tokens": 11647552} {"current_steps": 11585, "total_steps": 42420, "loss": 0.3498, "lr": 0.0009114901350565632, "epoch": 5.462046204620462, "percentage": 27.31, "elapsed_time": "0:37:31", "remaining_time": "1:39:51", "throughput": 5176.06, "total_tokens": 11652544} {"current_steps": 11590, "total_steps": 42420, "loss": 0.3378, "lr": 0.0009113732368276572, "epoch": 5.464403583215464, "percentage": 27.32, "elapsed_time": "0:37:32", "remaining_time": "1:39:50", "throughput": 5176.12, "total_tokens": 11657216} {"current_steps": 11595, "total_steps": 42420, "loss": 0.3766, "lr": 0.000911256268960247, "epoch": 5.466760961810467, "percentage": 27.33, "elapsed_time": "0:37:33", "remaining_time": "1:39:49", "throughput": 5176.39, "total_tokens": 11663072} {"current_steps": 11600, "total_steps": 42420, "loss": 0.3507, "lr": 0.0009111392314741333, "epoch": 5.469118340405469, "percentage": 27.35, "elapsed_time": "0:37:33", "remaining_time": "1:39:48", "throughput": 5176.45, "total_tokens": 11667552} {"current_steps": 11605, "total_steps": 42420, "loss": 0.3507, "lr": 0.0009110221243891284, "epoch": 5.4714757190004715, "percentage": 27.36, "elapsed_time": "0:37:34", "remaining_time": "1:39:47", "throughput": 5176.73, "total_tokens": 11673376} {"current_steps": 11610, "total_steps": 42420, "loss": 0.305, "lr": 0.0009109049477250566, "epoch": 5.473833097595474, "percentage": 27.37, "elapsed_time": "0:37:35", "remaining_time": "1:39:46", "throughput": 5176.7, "total_tokens": 11677696} {"current_steps": 11615, "total_steps": 42420, "loss": 0.3336, "lr": 0.0009107877015017539, "epoch": 5.476190476190476, "percentage": 27.38, "elapsed_time": "0:37:36", "remaining_time": "1:39:44", "throughput": 5176.63, "total_tokens": 11681696} {"current_steps": 11620, "total_steps": 42420, "loss": 0.3167, "lr": 0.000910670385739068, "epoch": 5.478547854785479, "percentage": 27.39, "elapsed_time": "0:37:37", "remaining_time": "1:39:43", "throughput": 5176.74, "total_tokens": 11686560} {"current_steps": 11625, "total_steps": 42420, "loss": 0.2007, "lr": 0.0009105530004568588, "epoch": 5.480905233380481, "percentage": 27.4, "elapsed_time": "0:37:38", "remaining_time": "1:39:42", "throughput": 5176.77, "total_tokens": 11691040} {"current_steps": 11630, "total_steps": 42420, "loss": 0.3242, "lr": 0.0009104355456749973, "epoch": 5.4832626119754835, "percentage": 27.42, "elapsed_time": "0:37:39", "remaining_time": "1:39:41", "throughput": 5176.83, "total_tokens": 11695904} {"current_steps": 11635, "total_steps": 42420, "loss": 0.3247, "lr": 0.0009103180214133667, "epoch": 5.485619990570486, "percentage": 27.43, "elapsed_time": "0:37:40", "remaining_time": "1:39:40", "throughput": 5177.11, "total_tokens": 11701696} {"current_steps": 11640, "total_steps": 42420, "loss": 0.3149, "lr": 0.0009102004276918619, "epoch": 5.487977369165488, "percentage": 27.44, "elapsed_time": "0:37:41", "remaining_time": "1:39:39", "throughput": 5177.3, "total_tokens": 11707264} {"current_steps": 11645, "total_steps": 42420, "loss": 0.3414, "lr": 0.0009100827645303894, "epoch": 5.490334747760491, "percentage": 27.45, "elapsed_time": "0:37:42", "remaining_time": "1:39:38", "throughput": 5177.38, "total_tokens": 11712320} {"current_steps": 11650, "total_steps": 42420, "loss": 0.3494, "lr": 0.0009099650319488677, "epoch": 5.492692126355493, "percentage": 27.46, "elapsed_time": "0:37:43", "remaining_time": "1:39:37", "throughput": 5177.49, "total_tokens": 11717216} {"current_steps": 11655, "total_steps": 42420, "loss": 0.3429, "lr": 0.000909847229967227, "epoch": 5.4950495049504955, "percentage": 27.48, "elapsed_time": "0:37:44", "remaining_time": "1:39:36", "throughput": 5177.66, "total_tokens": 11722752} {"current_steps": 11660, "total_steps": 42420, "loss": 0.3381, "lr": 0.0009097293586054091, "epoch": 5.497406883545497, "percentage": 27.49, "elapsed_time": "0:37:44", "remaining_time": "1:39:35", "throughput": 5177.64, "total_tokens": 11727040} {"current_steps": 11665, "total_steps": 42420, "loss": 0.2879, "lr": 0.0009096114178833674, "epoch": 5.499764262140499, "percentage": 27.5, "elapsed_time": "0:37:45", "remaining_time": "1:39:34", "throughput": 5177.85, "total_tokens": 11732448} {"current_steps": 11670, "total_steps": 42420, "loss": 0.4709, "lr": 0.0009094934078210678, "epoch": 5.502121640735502, "percentage": 27.51, "elapsed_time": "0:37:46", "remaining_time": "1:39:33", "throughput": 5178.17, "total_tokens": 11738752} {"current_steps": 11675, "total_steps": 42420, "loss": 0.3468, "lr": 0.0009093753284384867, "epoch": 5.504479019330504, "percentage": 27.52, "elapsed_time": "0:37:47", "remaining_time": "1:39:32", "throughput": 5178.18, "total_tokens": 11743328} {"current_steps": 11680, "total_steps": 42420, "loss": 0.3167, "lr": 0.0009092571797556135, "epoch": 5.506836397925507, "percentage": 27.53, "elapsed_time": "0:37:48", "remaining_time": "1:39:31", "throughput": 5178.46, "total_tokens": 11749568} {"current_steps": 11685, "total_steps": 42420, "loss": 0.3719, "lr": 0.0009091389617924485, "epoch": 5.509193776520509, "percentage": 27.55, "elapsed_time": "0:37:49", "remaining_time": "1:39:30", "throughput": 5178.53, "total_tokens": 11754432} {"current_steps": 11690, "total_steps": 42420, "loss": 0.3669, "lr": 0.0009090206745690039, "epoch": 5.511551155115511, "percentage": 27.56, "elapsed_time": "0:37:50", "remaining_time": "1:39:29", "throughput": 5178.75, "total_tokens": 11759936} {"current_steps": 11695, "total_steps": 42420, "loss": 0.346, "lr": 0.0009089023181053042, "epoch": 5.513908533710514, "percentage": 27.57, "elapsed_time": "0:37:51", "remaining_time": "1:39:28", "throughput": 5178.9, "total_tokens": 11765568} {"current_steps": 11700, "total_steps": 42420, "loss": 0.2842, "lr": 0.0009087838924213844, "epoch": 5.516265912305516, "percentage": 27.58, "elapsed_time": "0:37:52", "remaining_time": "1:39:27", "throughput": 5178.98, "total_tokens": 11770144} {"current_steps": 11705, "total_steps": 42420, "loss": 0.4059, "lr": 0.0009086653975372927, "epoch": 5.518623290900519, "percentage": 27.59, "elapsed_time": "0:37:53", "remaining_time": "1:39:26", "throughput": 5179.1, "total_tokens": 11774976} {"current_steps": 11710, "total_steps": 42420, "loss": 0.2603, "lr": 0.000908546833473088, "epoch": 5.520980669495521, "percentage": 27.6, "elapsed_time": "0:37:54", "remaining_time": "1:39:24", "throughput": 5179.05, "total_tokens": 11779136} {"current_steps": 11715, "total_steps": 42420, "loss": 0.4114, "lr": 0.0009084282002488409, "epoch": 5.523338048090523, "percentage": 27.62, "elapsed_time": "0:37:55", "remaining_time": "1:39:23", "throughput": 5178.97, "total_tokens": 11783104} {"current_steps": 11720, "total_steps": 42420, "loss": 0.4084, "lr": 0.0009083094978846343, "epoch": 5.525695426685526, "percentage": 27.63, "elapsed_time": "0:37:56", "remaining_time": "1:39:21", "throughput": 5178.99, "total_tokens": 11787456} {"current_steps": 11725, "total_steps": 42420, "loss": 0.3841, "lr": 0.0009081907264005623, "epoch": 5.528052805280528, "percentage": 27.64, "elapsed_time": "0:37:56", "remaining_time": "1:39:20", "throughput": 5179.03, "total_tokens": 11791968} {"current_steps": 11730, "total_steps": 42420, "loss": 0.3672, "lr": 0.0009080718858167312, "epoch": 5.530410183875531, "percentage": 27.65, "elapsed_time": "0:37:57", "remaining_time": "1:39:19", "throughput": 5179.24, "total_tokens": 11797472} {"current_steps": 11735, "total_steps": 42420, "loss": 0.3198, "lr": 0.0009079529761532583, "epoch": 5.532767562470533, "percentage": 27.66, "elapsed_time": "0:37:58", "remaining_time": "1:39:18", "throughput": 5179.27, "total_tokens": 11802144} {"current_steps": 11740, "total_steps": 42420, "loss": 0.4044, "lr": 0.0009078339974302734, "epoch": 5.535124941065535, "percentage": 27.68, "elapsed_time": "0:37:59", "remaining_time": "1:39:17", "throughput": 5179.34, "total_tokens": 11806944} {"current_steps": 11745, "total_steps": 42420, "loss": 0.4081, "lr": 0.0009077149496679171, "epoch": 5.537482319660538, "percentage": 27.69, "elapsed_time": "0:38:00", "remaining_time": "1:39:16", "throughput": 5179.39, "total_tokens": 11811712} {"current_steps": 11750, "total_steps": 42420, "loss": 0.3507, "lr": 0.0009075958328863427, "epoch": 5.539839698255539, "percentage": 27.7, "elapsed_time": "0:38:01", "remaining_time": "1:39:14", "throughput": 5179.4, "total_tokens": 11816064} {"current_steps": 11755, "total_steps": 42420, "loss": 0.3702, "lr": 0.0009074766471057142, "epoch": 5.542197076850542, "percentage": 27.71, "elapsed_time": "0:38:02", "remaining_time": "1:39:13", "throughput": 5179.46, "total_tokens": 11820736} {"current_steps": 11760, "total_steps": 42420, "loss": 0.3333, "lr": 0.000907357392346208, "epoch": 5.544554455445544, "percentage": 27.72, "elapsed_time": "0:38:03", "remaining_time": "1:39:12", "throughput": 5179.62, "total_tokens": 11826112} {"current_steps": 11765, "total_steps": 42420, "loss": 0.3834, "lr": 0.0009072380686280117, "epoch": 5.5469118340405466, "percentage": 27.73, "elapsed_time": "0:38:04", "remaining_time": "1:39:11", "throughput": 5179.83, "total_tokens": 11831776} {"current_steps": 11770, "total_steps": 42420, "loss": 0.3582, "lr": 0.0009071186759713249, "epoch": 5.549269212635549, "percentage": 27.75, "elapsed_time": "0:38:05", "remaining_time": "1:39:10", "throughput": 5179.99, "total_tokens": 11837152} {"current_steps": 11775, "total_steps": 42420, "loss": 0.2559, "lr": 0.0009069992143963588, "epoch": 5.551626591230551, "percentage": 27.76, "elapsed_time": "0:38:06", "remaining_time": "1:39:09", "throughput": 5180.19, "total_tokens": 11842688} {"current_steps": 11780, "total_steps": 42420, "loss": 0.4414, "lr": 0.0009068796839233362, "epoch": 5.553983969825554, "percentage": 27.77, "elapsed_time": "0:38:07", "remaining_time": "1:39:08", "throughput": 5180.25, "total_tokens": 11847488} {"current_steps": 11785, "total_steps": 42420, "loss": 0.3329, "lr": 0.0009067600845724915, "epoch": 5.556341348420556, "percentage": 27.78, "elapsed_time": "0:38:07", "remaining_time": "1:39:07", "throughput": 5180.31, "total_tokens": 11852096} {"current_steps": 11790, "total_steps": 42420, "loss": 0.3331, "lr": 0.0009066404163640708, "epoch": 5.558698727015559, "percentage": 27.79, "elapsed_time": "0:38:08", "remaining_time": "1:39:06", "throughput": 5180.47, "total_tokens": 11857152} {"current_steps": 11795, "total_steps": 42420, "loss": 0.3425, "lr": 0.000906520679318332, "epoch": 5.561056105610561, "percentage": 27.81, "elapsed_time": "0:38:09", "remaining_time": "1:39:05", "throughput": 5180.66, "total_tokens": 11862464} {"current_steps": 11800, "total_steps": 42420, "loss": 0.3021, "lr": 0.0009064008734555446, "epoch": 5.563413484205563, "percentage": 27.82, "elapsed_time": "0:38:10", "remaining_time": "1:39:04", "throughput": 5181.02, "total_tokens": 11868800} {"current_steps": 11805, "total_steps": 42420, "loss": 0.2498, "lr": 0.0009062809987959895, "epoch": 5.565770862800566, "percentage": 27.83, "elapsed_time": "0:38:11", "remaining_time": "1:39:03", "throughput": 5181.33, "total_tokens": 11875232} {"current_steps": 11810, "total_steps": 42420, "loss": 0.5113, "lr": 0.0009061610553599596, "epoch": 5.568128241395568, "percentage": 27.84, "elapsed_time": "0:38:12", "remaining_time": "1:39:02", "throughput": 5181.31, "total_tokens": 11879392} {"current_steps": 11815, "total_steps": 42420, "loss": 0.3902, "lr": 0.0009060410431677593, "epoch": 5.570485619990571, "percentage": 27.85, "elapsed_time": "0:38:13", "remaining_time": "1:39:01", "throughput": 5181.26, "total_tokens": 11883488} {"current_steps": 11820, "total_steps": 42420, "loss": 0.4204, "lr": 0.0009059209622397045, "epoch": 5.572842998585573, "percentage": 27.86, "elapsed_time": "0:38:14", "remaining_time": "1:38:59", "throughput": 5181.33, "total_tokens": 11888352} {"current_steps": 11825, "total_steps": 42420, "loss": 0.4082, "lr": 0.000905800812596123, "epoch": 5.575200377180575, "percentage": 27.88, "elapsed_time": "0:38:15", "remaining_time": "1:38:58", "throughput": 5181.46, "total_tokens": 11893600} {"current_steps": 11830, "total_steps": 42420, "loss": 0.3834, "lr": 0.000905680594257354, "epoch": 5.577557755775578, "percentage": 27.89, "elapsed_time": "0:38:16", "remaining_time": "1:38:58", "throughput": 5181.79, "total_tokens": 11900608} {"current_steps": 11835, "total_steps": 42420, "loss": 0.3333, "lr": 0.0009055603072437486, "epoch": 5.57991513437058, "percentage": 27.9, "elapsed_time": "0:38:17", "remaining_time": "1:38:57", "throughput": 5181.79, "total_tokens": 11904960} {"current_steps": 11840, "total_steps": 42420, "loss": 0.3292, "lr": 0.000905439951575669, "epoch": 5.582272512965583, "percentage": 27.91, "elapsed_time": "0:38:18", "remaining_time": "1:38:56", "throughput": 5181.82, "total_tokens": 11909472} {"current_steps": 11845, "total_steps": 42420, "loss": 0.3089, "lr": 0.0009053195272734898, "epoch": 5.584629891560585, "percentage": 27.92, "elapsed_time": "0:38:19", "remaining_time": "1:38:54", "throughput": 5181.85, "total_tokens": 11914048} {"current_steps": 11850, "total_steps": 42420, "loss": 0.3976, "lr": 0.0009051990343575963, "epoch": 5.586987270155587, "percentage": 27.93, "elapsed_time": "0:38:20", "remaining_time": "1:38:53", "throughput": 5182.02, "total_tokens": 11919552} {"current_steps": 11855, "total_steps": 42420, "loss": 0.3575, "lr": 0.0009050784728483863, "epoch": 5.58934464875059, "percentage": 27.95, "elapsed_time": "0:38:21", "remaining_time": "1:38:52", "throughput": 5182.2, "total_tokens": 11925024} {"current_steps": 11860, "total_steps": 42420, "loss": 0.3381, "lr": 0.0009049578427662687, "epoch": 5.591702027345592, "percentage": 27.96, "elapsed_time": "0:38:22", "remaining_time": "1:38:51", "throughput": 5182.36, "total_tokens": 11930048} {"current_steps": 11865, "total_steps": 42420, "loss": 0.3506, "lr": 0.0009048371441316641, "epoch": 5.594059405940594, "percentage": 27.97, "elapsed_time": "0:38:23", "remaining_time": "1:38:50", "throughput": 5182.54, "total_tokens": 11935584} {"current_steps": 11870, "total_steps": 42420, "loss": 0.3399, "lr": 0.0009047163769650047, "epoch": 5.596416784535596, "percentage": 27.98, "elapsed_time": "0:38:23", "remaining_time": "1:38:49", "throughput": 5182.66, "total_tokens": 11940544} {"current_steps": 11875, "total_steps": 42420, "loss": 0.3405, "lr": 0.0009045955412867344, "epoch": 5.5987741631305985, "percentage": 27.99, "elapsed_time": "0:38:24", "remaining_time": "1:38:48", "throughput": 5182.71, "total_tokens": 11944992} {"current_steps": 11880, "total_steps": 42420, "loss": 0.3059, "lr": 0.0009044746371173087, "epoch": 5.601131541725601, "percentage": 28.01, "elapsed_time": "0:38:25", "remaining_time": "1:38:47", "throughput": 5182.64, "total_tokens": 11949184} {"current_steps": 11885, "total_steps": 42420, "loss": 0.3112, "lr": 0.0009043536644771944, "epoch": 5.603488920320603, "percentage": 28.02, "elapsed_time": "0:38:26", "remaining_time": "1:38:45", "throughput": 5182.73, "total_tokens": 11954144} {"current_steps": 11890, "total_steps": 42420, "loss": 0.3168, "lr": 0.0009042326233868703, "epoch": 5.605846298915606, "percentage": 28.03, "elapsed_time": "0:38:27", "remaining_time": "1:38:44", "throughput": 5182.99, "total_tokens": 11959776} {"current_steps": 11895, "total_steps": 42420, "loss": 0.458, "lr": 0.0009041115138668266, "epoch": 5.608203677510608, "percentage": 28.04, "elapsed_time": "0:38:28", "remaining_time": "1:38:44", "throughput": 5183.15, "total_tokens": 11965152} {"current_steps": 11900, "total_steps": 42420, "loss": 0.3672, "lr": 0.0009039903359375649, "epoch": 5.6105610561056105, "percentage": 28.05, "elapsed_time": "0:38:29", "remaining_time": "1:38:42", "throughput": 5183.18, "total_tokens": 11969632} {"current_steps": 11905, "total_steps": 42420, "loss": 0.3778, "lr": 0.0009038690896195987, "epoch": 5.612918434700613, "percentage": 28.06, "elapsed_time": "0:38:30", "remaining_time": "1:38:41", "throughput": 5183.24, "total_tokens": 11974368} {"current_steps": 11910, "total_steps": 42420, "loss": 0.3516, "lr": 0.0009037477749334529, "epoch": 5.615275813295615, "percentage": 28.08, "elapsed_time": "0:38:31", "remaining_time": "1:38:40", "throughput": 5183.27, "total_tokens": 11978880} {"current_steps": 11915, "total_steps": 42420, "loss": 0.372, "lr": 0.0009036263918996641, "epoch": 5.617633191890618, "percentage": 28.09, "elapsed_time": "0:38:31", "remaining_time": "1:38:38", "throughput": 5183.18, "total_tokens": 11982720} {"current_steps": 11920, "total_steps": 42420, "loss": 0.2919, "lr": 0.0009035049405387804, "epoch": 5.61999057048562, "percentage": 28.1, "elapsed_time": "0:38:32", "remaining_time": "1:38:37", "throughput": 5183.26, "total_tokens": 11987776} {"current_steps": 11925, "total_steps": 42420, "loss": 0.4119, "lr": 0.0009033834208713614, "epoch": 5.6223479490806225, "percentage": 28.11, "elapsed_time": "0:38:33", "remaining_time": "1:38:37", "throughput": 5183.56, "total_tokens": 11994304} {"current_steps": 11930, "total_steps": 42420, "loss": 0.3536, "lr": 0.0009032618329179782, "epoch": 5.624705327675625, "percentage": 28.12, "elapsed_time": "0:38:34", "remaining_time": "1:38:36", "throughput": 5183.71, "total_tokens": 11999808} {"current_steps": 11935, "total_steps": 42420, "loss": 0.3536, "lr": 0.0009031401766992138, "epoch": 5.627062706270627, "percentage": 28.14, "elapsed_time": "0:38:35", "remaining_time": "1:38:35", "throughput": 5183.75, "total_tokens": 12004320} {"current_steps": 11940, "total_steps": 42420, "loss": 0.3805, "lr": 0.0009030184522356623, "epoch": 5.62942008486563, "percentage": 28.15, "elapsed_time": "0:38:36", "remaining_time": "1:38:34", "throughput": 5183.93, "total_tokens": 12009728} {"current_steps": 11945, "total_steps": 42420, "loss": 0.3599, "lr": 0.0009028966595479298, "epoch": 5.631777463460632, "percentage": 28.16, "elapsed_time": "0:38:37", "remaining_time": "1:38:32", "throughput": 5183.97, "total_tokens": 12014368} {"current_steps": 11950, "total_steps": 42420, "loss": 0.3479, "lr": 0.0009027747986566335, "epoch": 5.634134842055634, "percentage": 28.17, "elapsed_time": "0:38:38", "remaining_time": "1:38:31", "throughput": 5184.13, "total_tokens": 12019584} {"current_steps": 11955, "total_steps": 42420, "loss": 0.3601, "lr": 0.0009026528695824026, "epoch": 5.636492220650636, "percentage": 28.18, "elapsed_time": "0:38:39", "remaining_time": "1:38:30", "throughput": 5184.33, "total_tokens": 12025408} {"current_steps": 11960, "total_steps": 42420, "loss": 0.3371, "lr": 0.0009025308723458776, "epoch": 5.6388495992456384, "percentage": 28.19, "elapsed_time": "0:38:40", "remaining_time": "1:38:29", "throughput": 5184.47, "total_tokens": 12030464} {"current_steps": 11965, "total_steps": 42420, "loss": 0.3247, "lr": 0.0009024088069677105, "epoch": 5.641206977840641, "percentage": 28.21, "elapsed_time": "0:38:41", "remaining_time": "1:38:28", "throughput": 5184.5, "total_tokens": 12034944} {"current_steps": 11970, "total_steps": 42420, "loss": 0.2763, "lr": 0.0009022866734685647, "epoch": 5.643564356435643, "percentage": 28.22, "elapsed_time": "0:38:42", "remaining_time": "1:38:27", "throughput": 5184.59, "total_tokens": 12040160} {"current_steps": 11975, "total_steps": 42420, "loss": 0.3596, "lr": 0.0009021644718691157, "epoch": 5.645921735030646, "percentage": 28.23, "elapsed_time": "0:38:43", "remaining_time": "1:38:26", "throughput": 5184.76, "total_tokens": 12045504} {"current_steps": 11980, "total_steps": 42420, "loss": 0.2986, "lr": 0.0009020422021900499, "epoch": 5.648279113625648, "percentage": 28.24, "elapsed_time": "0:38:44", "remaining_time": "1:38:25", "throughput": 5184.77, "total_tokens": 12049952} {"current_steps": 11985, "total_steps": 42420, "loss": 0.3392, "lr": 0.0009019198644520656, "epoch": 5.6506364922206505, "percentage": 28.25, "elapsed_time": "0:38:44", "remaining_time": "1:38:24", "throughput": 5184.76, "total_tokens": 12054496} {"current_steps": 11990, "total_steps": 42420, "loss": 0.234, "lr": 0.0009017974586758723, "epoch": 5.652993870815653, "percentage": 28.26, "elapsed_time": "0:38:45", "remaining_time": "1:38:22", "throughput": 5184.67, "total_tokens": 12058208} {"current_steps": 11995, "total_steps": 42420, "loss": 0.3827, "lr": 0.0009016749848821917, "epoch": 5.655351249410655, "percentage": 28.28, "elapsed_time": "0:38:46", "remaining_time": "1:38:21", "throughput": 5184.7, "total_tokens": 12062592} {"current_steps": 12000, "total_steps": 42420, "loss": 0.3346, "lr": 0.000901552443091756, "epoch": 5.657708628005658, "percentage": 28.29, "elapsed_time": "0:38:47", "remaining_time": "1:38:20", "throughput": 5184.64, "total_tokens": 12066816} {"current_steps": 12005, "total_steps": 42420, "loss": 0.3343, "lr": 0.0009014298333253097, "epoch": 5.66006600660066, "percentage": 28.3, "elapsed_time": "0:38:48", "remaining_time": "1:38:18", "throughput": 5184.73, "total_tokens": 12071968} {"current_steps": 12010, "total_steps": 42420, "loss": 0.3802, "lr": 0.0009013071556036084, "epoch": 5.6624233851956625, "percentage": 28.31, "elapsed_time": "0:38:49", "remaining_time": "1:38:18", "throughput": 5184.93, "total_tokens": 12077568} {"current_steps": 12015, "total_steps": 42420, "loss": 0.3429, "lr": 0.0009011844099474197, "epoch": 5.664780763790665, "percentage": 28.32, "elapsed_time": "0:38:50", "remaining_time": "1:38:16", "throughput": 5185.0, "total_tokens": 12082400} {"current_steps": 12020, "total_steps": 42420, "loss": 0.3595, "lr": 0.0009010615963775219, "epoch": 5.667138142385667, "percentage": 28.34, "elapsed_time": "0:38:51", "remaining_time": "1:38:15", "throughput": 5184.99, "total_tokens": 12086528} {"current_steps": 12025, "total_steps": 42420, "loss": 0.3127, "lr": 0.0009009387149147056, "epoch": 5.66949552098067, "percentage": 28.35, "elapsed_time": "0:38:51", "remaining_time": "1:38:14", "throughput": 5185.02, "total_tokens": 12091040} {"current_steps": 12030, "total_steps": 42420, "loss": 0.322, "lr": 0.0009008157655797724, "epoch": 5.671852899575672, "percentage": 28.36, "elapsed_time": "0:38:52", "remaining_time": "1:38:13", "throughput": 5185.18, "total_tokens": 12096672} {"current_steps": 12035, "total_steps": 42420, "loss": 0.3516, "lr": 0.0009006927483935354, "epoch": 5.6742102781706745, "percentage": 28.37, "elapsed_time": "0:38:53", "remaining_time": "1:38:12", "throughput": 5185.37, "total_tokens": 12102272} {"current_steps": 12040, "total_steps": 42420, "loss": 0.4217, "lr": 0.0009005696633768197, "epoch": 5.676567656765677, "percentage": 28.38, "elapsed_time": "0:38:54", "remaining_time": "1:38:11", "throughput": 5185.53, "total_tokens": 12107488} {"current_steps": 12045, "total_steps": 42420, "loss": 0.3573, "lr": 0.0009004465105504611, "epoch": 5.678925035360679, "percentage": 28.39, "elapsed_time": "0:38:55", "remaining_time": "1:38:10", "throughput": 5185.73, "total_tokens": 12113152} {"current_steps": 12050, "total_steps": 42420, "loss": 0.3379, "lr": 0.0009003232899353075, "epoch": 5.681282413955682, "percentage": 28.41, "elapsed_time": "0:38:56", "remaining_time": "1:38:09", "throughput": 5185.81, "total_tokens": 12117920} {"current_steps": 12055, "total_steps": 42420, "loss": 0.3196, "lr": 0.0009002000015522181, "epoch": 5.683639792550684, "percentage": 28.42, "elapsed_time": "0:38:57", "remaining_time": "1:38:08", "throughput": 5185.79, "total_tokens": 12122144} {"current_steps": 12060, "total_steps": 42420, "loss": 0.3155, "lr": 0.0009000766454220633, "epoch": 5.6859971711456865, "percentage": 28.43, "elapsed_time": "0:38:58", "remaining_time": "1:38:06", "throughput": 5185.94, "total_tokens": 12127264} {"current_steps": 12065, "total_steps": 42420, "loss": 0.419, "lr": 0.0008999532215657256, "epoch": 5.688354549740688, "percentage": 28.44, "elapsed_time": "0:38:59", "remaining_time": "1:38:05", "throughput": 5186.04, "total_tokens": 12132416} {"current_steps": 12070, "total_steps": 42420, "loss": 0.3493, "lr": 0.0008998297300040981, "epoch": 5.69071192833569, "percentage": 28.45, "elapsed_time": "0:39:00", "remaining_time": "1:38:04", "throughput": 5186.05, "total_tokens": 12136768} {"current_steps": 12075, "total_steps": 42420, "loss": 0.3453, "lr": 0.0008997061707580861, "epoch": 5.693069306930693, "percentage": 28.47, "elapsed_time": "0:39:01", "remaining_time": "1:38:03", "throughput": 5186.21, "total_tokens": 12141952} {"current_steps": 12080, "total_steps": 42420, "loss": 0.358, "lr": 0.0008995825438486059, "epoch": 5.695426685525695, "percentage": 28.48, "elapsed_time": "0:39:02", "remaining_time": "1:38:02", "throughput": 5186.28, "total_tokens": 12146816} {"current_steps": 12085, "total_steps": 42420, "loss": 0.3677, "lr": 0.0008994588492965856, "epoch": 5.697784064120698, "percentage": 28.49, "elapsed_time": "0:39:03", "remaining_time": "1:38:01", "throughput": 5186.34, "total_tokens": 12151968} {"current_steps": 12090, "total_steps": 42420, "loss": 0.3619, "lr": 0.0008993350871229647, "epoch": 5.7001414427157, "percentage": 28.5, "elapsed_time": "0:39:03", "remaining_time": "1:38:00", "throughput": 5186.42, "total_tokens": 12156928} {"current_steps": 12095, "total_steps": 42420, "loss": 0.3722, "lr": 0.0008992112573486937, "epoch": 5.702498821310702, "percentage": 28.51, "elapsed_time": "0:39:04", "remaining_time": "1:37:59", "throughput": 5186.57, "total_tokens": 12162048} {"current_steps": 12100, "total_steps": 42420, "loss": 0.3313, "lr": 0.0008990873599947351, "epoch": 5.704856199905705, "percentage": 28.52, "elapsed_time": "0:39:05", "remaining_time": "1:37:58", "throughput": 5186.7, "total_tokens": 12167328} {"current_steps": 12105, "total_steps": 42420, "loss": 0.3475, "lr": 0.0008989633950820626, "epoch": 5.707213578500707, "percentage": 28.54, "elapsed_time": "0:39:07", "remaining_time": "1:37:58", "throughput": 5187.27, "total_tokens": 12175744} {"current_steps": 12110, "total_steps": 42420, "loss": 0.312, "lr": 0.0008988393626316614, "epoch": 5.70957095709571, "percentage": 28.55, "elapsed_time": "0:39:08", "remaining_time": "1:37:57", "throughput": 5187.38, "total_tokens": 12180992} {"current_steps": 12115, "total_steps": 42420, "loss": 0.2469, "lr": 0.0008987152626645278, "epoch": 5.711928335690712, "percentage": 28.56, "elapsed_time": "0:39:09", "remaining_time": "1:37:56", "throughput": 5187.47, "total_tokens": 12185888} {"current_steps": 12120, "total_steps": 42420, "loss": 0.389, "lr": 0.0008985910952016702, "epoch": 5.714285714285714, "percentage": 28.57, "elapsed_time": "0:39:09", "remaining_time": "1:37:54", "throughput": 5187.44, "total_tokens": 12190176} {"current_steps": 12125, "total_steps": 42420, "loss": 0.3421, "lr": 0.0008984668602641078, "epoch": 5.716643092880717, "percentage": 28.58, "elapsed_time": "0:39:10", "remaining_time": "1:37:53", "throughput": 5187.48, "total_tokens": 12194848} {"current_steps": 12130, "total_steps": 42420, "loss": 0.354, "lr": 0.0008983425578728717, "epoch": 5.719000471475719, "percentage": 28.6, "elapsed_time": "0:39:11", "remaining_time": "1:37:52", "throughput": 5187.54, "total_tokens": 12199680} {"current_steps": 12135, "total_steps": 42420, "loss": 0.341, "lr": 0.0008982181880490037, "epoch": 5.721357850070722, "percentage": 28.61, "elapsed_time": "0:39:12", "remaining_time": "1:37:51", "throughput": 5187.73, "total_tokens": 12205280} {"current_steps": 12140, "total_steps": 42420, "loss": 0.3581, "lr": 0.000898093750813558, "epoch": 5.723715228665724, "percentage": 28.62, "elapsed_time": "0:39:13", "remaining_time": "1:37:51", "throughput": 5187.99, "total_tokens": 12211680} {"current_steps": 12145, "total_steps": 42420, "loss": 0.3584, "lr": 0.0008979692461875997, "epoch": 5.726072607260726, "percentage": 28.63, "elapsed_time": "0:39:14", "remaining_time": "1:37:50", "throughput": 5188.14, "total_tokens": 12216960} {"current_steps": 12150, "total_steps": 42420, "loss": 0.3367, "lr": 0.000897844674192205, "epoch": 5.728429985855728, "percentage": 28.64, "elapsed_time": "0:39:15", "remaining_time": "1:37:49", "throughput": 5188.35, "total_tokens": 12222560} {"current_steps": 12155, "total_steps": 42420, "loss": 0.3233, "lr": 0.0008977200348484619, "epoch": 5.73078736445073, "percentage": 28.65, "elapsed_time": "0:39:16", "remaining_time": "1:37:48", "throughput": 5188.57, "total_tokens": 12228384} {"current_steps": 12160, "total_steps": 42420, "loss": 0.3739, "lr": 0.0008975953281774698, "epoch": 5.733144743045733, "percentage": 28.67, "elapsed_time": "0:39:17", "remaining_time": "1:37:46", "throughput": 5188.41, "total_tokens": 12232000} {"current_steps": 12165, "total_steps": 42420, "loss": 0.3949, "lr": 0.0008974705542003396, "epoch": 5.735502121640735, "percentage": 28.68, "elapsed_time": "0:39:18", "remaining_time": "1:37:46", "throughput": 5188.77, "total_tokens": 12238560} {"current_steps": 12170, "total_steps": 42420, "loss": 0.361, "lr": 0.0008973457129381931, "epoch": 5.7378595002357375, "percentage": 28.69, "elapsed_time": "0:39:19", "remaining_time": "1:37:44", "throughput": 5188.86, "total_tokens": 12243328} {"current_steps": 12175, "total_steps": 42420, "loss": 0.3951, "lr": 0.000897220804412164, "epoch": 5.74021687883074, "percentage": 28.7, "elapsed_time": "0:39:20", "remaining_time": "1:37:43", "throughput": 5188.86, "total_tokens": 12247808} {"current_steps": 12180, "total_steps": 42420, "loss": 0.367, "lr": 0.0008970958286433972, "epoch": 5.742574257425742, "percentage": 28.71, "elapsed_time": "0:39:21", "remaining_time": "1:37:42", "throughput": 5189.0, "total_tokens": 12253120} {"current_steps": 12185, "total_steps": 42420, "loss": 0.3178, "lr": 0.0008969707856530488, "epoch": 5.744931636020745, "percentage": 28.72, "elapsed_time": "0:39:22", "remaining_time": "1:37:41", "throughput": 5189.14, "total_tokens": 12258432} {"current_steps": 12190, "total_steps": 42420, "loss": 0.4346, "lr": 0.0008968456754622865, "epoch": 5.747289014615747, "percentage": 28.74, "elapsed_time": "0:39:23", "remaining_time": "1:37:40", "throughput": 5189.22, "total_tokens": 12263168} {"current_steps": 12195, "total_steps": 42420, "loss": 0.4099, "lr": 0.0008967204980922896, "epoch": 5.7496463932107496, "percentage": 28.75, "elapsed_time": "0:39:24", "remaining_time": "1:37:39", "throughput": 5189.4, "total_tokens": 12268928} {"current_steps": 12200, "total_steps": 42420, "loss": 0.3743, "lr": 0.0008965952535642481, "epoch": 5.752003771805752, "percentage": 28.76, "elapsed_time": "0:39:25", "remaining_time": "1:37:38", "throughput": 5189.42, "total_tokens": 12273504} {"current_steps": 12205, "total_steps": 42420, "loss": 0.3483, "lr": 0.0008964699418993642, "epoch": 5.754361150400754, "percentage": 28.77, "elapsed_time": "0:39:26", "remaining_time": "1:37:37", "throughput": 5189.53, "total_tokens": 12278688} {"current_steps": 12210, "total_steps": 42420, "loss": 0.3439, "lr": 0.0008963445631188507, "epoch": 5.756718528995757, "percentage": 28.78, "elapsed_time": "0:39:26", "remaining_time": "1:37:36", "throughput": 5189.54, "total_tokens": 12283168} {"current_steps": 12215, "total_steps": 42420, "loss": 0.3244, "lr": 0.0008962191172439322, "epoch": 5.759075907590759, "percentage": 28.8, "elapsed_time": "0:39:27", "remaining_time": "1:37:34", "throughput": 5189.56, "total_tokens": 12287616} {"current_steps": 12220, "total_steps": 42420, "loss": 0.3783, "lr": 0.0008960936042958445, "epoch": 5.761433286185762, "percentage": 28.81, "elapsed_time": "0:39:28", "remaining_time": "1:37:33", "throughput": 5189.66, "total_tokens": 12292544} {"current_steps": 12225, "total_steps": 42420, "loss": 0.3247, "lr": 0.000895968024295835, "epoch": 5.763790664780764, "percentage": 28.82, "elapsed_time": "0:39:29", "remaining_time": "1:37:32", "throughput": 5189.8, "total_tokens": 12297856} {"current_steps": 12230, "total_steps": 42420, "loss": 0.3132, "lr": 0.000895842377265162, "epoch": 5.766148043375766, "percentage": 28.83, "elapsed_time": "0:39:30", "remaining_time": "1:37:31", "throughput": 5189.91, "total_tokens": 12302688} {"current_steps": 12235, "total_steps": 42420, "loss": 0.3109, "lr": 0.0008957166632250957, "epoch": 5.768505421970769, "percentage": 28.84, "elapsed_time": "0:39:31", "remaining_time": "1:37:30", "throughput": 5190.04, "total_tokens": 12308064} {"current_steps": 12240, "total_steps": 42420, "loss": 0.3885, "lr": 0.0008955908821969172, "epoch": 5.770862800565771, "percentage": 28.85, "elapsed_time": "0:39:32", "remaining_time": "1:37:29", "throughput": 5190.12, "total_tokens": 12312800} {"current_steps": 12245, "total_steps": 42420, "loss": 0.3612, "lr": 0.0008954650342019191, "epoch": 5.773220179160774, "percentage": 28.87, "elapsed_time": "0:39:33", "remaining_time": "1:37:28", "throughput": 5190.04, "total_tokens": 12316800} {"current_steps": 12250, "total_steps": 42420, "loss": 0.3503, "lr": 0.0008953391192614053, "epoch": 5.775577557755776, "percentage": 28.88, "elapsed_time": "0:39:34", "remaining_time": "1:37:27", "throughput": 5190.31, "total_tokens": 12323264} {"current_steps": 12255, "total_steps": 42420, "loss": 0.3734, "lr": 0.000895213137396691, "epoch": 5.777934936350778, "percentage": 28.89, "elapsed_time": "0:39:35", "remaining_time": "1:37:26", "throughput": 5190.4, "total_tokens": 12328288} {"current_steps": 12260, "total_steps": 42420, "loss": 0.3418, "lr": 0.0008950870886291029, "epoch": 5.780292314945781, "percentage": 28.9, "elapsed_time": "0:39:36", "remaining_time": "1:37:25", "throughput": 5190.62, "total_tokens": 12333856} {"current_steps": 12265, "total_steps": 42420, "loss": 0.3808, "lr": 0.000894960972979979, "epoch": 5.782649693540782, "percentage": 28.91, "elapsed_time": "0:39:37", "remaining_time": "1:37:24", "throughput": 5190.85, "total_tokens": 12339488} {"current_steps": 12270, "total_steps": 42420, "loss": 0.2937, "lr": 0.0008948347904706683, "epoch": 5.785007072135785, "percentage": 28.93, "elapsed_time": "0:39:38", "remaining_time": "1:37:23", "throughput": 5190.89, "total_tokens": 12344160} {"current_steps": 12275, "total_steps": 42420, "loss": 0.3836, "lr": 0.0008947085411225316, "epoch": 5.787364450730787, "percentage": 28.94, "elapsed_time": "0:39:39", "remaining_time": "1:37:22", "throughput": 5191.05, "total_tokens": 12349760} {"current_steps": 12280, "total_steps": 42420, "loss": 0.3424, "lr": 0.0008945822249569405, "epoch": 5.7897218293257895, "percentage": 28.95, "elapsed_time": "0:39:39", "remaining_time": "1:37:21", "throughput": 5191.07, "total_tokens": 12354240} {"current_steps": 12285, "total_steps": 42420, "loss": 0.3456, "lr": 0.0008944558419952786, "epoch": 5.792079207920792, "percentage": 28.96, "elapsed_time": "0:39:40", "remaining_time": "1:37:20", "throughput": 5191.18, "total_tokens": 12359360} {"current_steps": 12290, "total_steps": 42420, "loss": 0.3501, "lr": 0.0008943293922589399, "epoch": 5.794436586515794, "percentage": 28.97, "elapsed_time": "0:39:41", "remaining_time": "1:37:19", "throughput": 5191.35, "total_tokens": 12364864} {"current_steps": 12295, "total_steps": 42420, "loss": 0.3113, "lr": 0.0008942028757693305, "epoch": 5.796793965110797, "percentage": 28.98, "elapsed_time": "0:39:42", "remaining_time": "1:37:18", "throughput": 5191.44, "total_tokens": 12369792} {"current_steps": 12300, "total_steps": 42420, "loss": 0.2894, "lr": 0.0008940762925478674, "epoch": 5.799151343705799, "percentage": 29.0, "elapsed_time": "0:39:43", "remaining_time": "1:37:16", "throughput": 5191.52, "total_tokens": 12374592} {"current_steps": 12305, "total_steps": 42420, "loss": 0.3503, "lr": 0.000893949642615979, "epoch": 5.8015087223008015, "percentage": 29.01, "elapsed_time": "0:39:44", "remaining_time": "1:37:16", "throughput": 5191.69, "total_tokens": 12380160} {"current_steps": 12310, "total_steps": 42420, "loss": 0.2897, "lr": 0.000893822925995105, "epoch": 5.803866100895804, "percentage": 29.02, "elapsed_time": "0:39:45", "remaining_time": "1:37:14", "throughput": 5191.77, "total_tokens": 12384928} {"current_steps": 12315, "total_steps": 42420, "loss": 0.3267, "lr": 0.0008936961427066963, "epoch": 5.806223479490806, "percentage": 29.03, "elapsed_time": "0:39:46", "remaining_time": "1:37:13", "throughput": 5191.92, "total_tokens": 12390016} {"current_steps": 12320, "total_steps": 42420, "loss": 0.2717, "lr": 0.0008935692927722153, "epoch": 5.808580858085809, "percentage": 29.04, "elapsed_time": "0:39:47", "remaining_time": "1:37:12", "throughput": 5192.16, "total_tokens": 12395840} {"current_steps": 12325, "total_steps": 42420, "loss": 0.3567, "lr": 0.0008934423762131354, "epoch": 5.810938236680811, "percentage": 29.05, "elapsed_time": "0:39:48", "remaining_time": "1:37:11", "throughput": 5192.34, "total_tokens": 12401344} {"current_steps": 12330, "total_steps": 42420, "loss": 0.3154, "lr": 0.0008933153930509415, "epoch": 5.8132956152758135, "percentage": 29.07, "elapsed_time": "0:39:49", "remaining_time": "1:37:10", "throughput": 5192.35, "total_tokens": 12406144} {"current_steps": 12335, "total_steps": 42420, "loss": 0.2996, "lr": 0.0008931883433071297, "epoch": 5.815652993870816, "percentage": 29.08, "elapsed_time": "0:39:50", "remaining_time": "1:37:09", "throughput": 5192.44, "total_tokens": 12410944} {"current_steps": 12340, "total_steps": 42420, "loss": 0.346, "lr": 0.0008930612270032075, "epoch": 5.818010372465818, "percentage": 29.09, "elapsed_time": "0:39:51", "remaining_time": "1:37:08", "throughput": 5192.62, "total_tokens": 12416320} {"current_steps": 12345, "total_steps": 42420, "loss": 0.4164, "lr": 0.0008929340441606931, "epoch": 5.820367751060821, "percentage": 29.1, "elapsed_time": "0:39:52", "remaining_time": "1:37:07", "throughput": 5192.82, "total_tokens": 12422016} {"current_steps": 12350, "total_steps": 42420, "loss": 0.3634, "lr": 0.0008928067948011167, "epoch": 5.822725129655822, "percentage": 29.11, "elapsed_time": "0:39:53", "remaining_time": "1:37:06", "throughput": 5193.06, "total_tokens": 12427968} {"current_steps": 12355, "total_steps": 42420, "loss": 0.3903, "lr": 0.0008926794789460195, "epoch": 5.825082508250825, "percentage": 29.13, "elapsed_time": "0:39:53", "remaining_time": "1:37:05", "throughput": 5192.98, "total_tokens": 12431936} {"current_steps": 12360, "total_steps": 42420, "loss": 0.3459, "lr": 0.0008925520966169538, "epoch": 5.827439886845827, "percentage": 29.14, "elapsed_time": "0:39:54", "remaining_time": "1:37:04", "throughput": 5193.07, "total_tokens": 12436768} {"current_steps": 12365, "total_steps": 42420, "loss": 0.2575, "lr": 0.0008924246478354833, "epoch": 5.829797265440829, "percentage": 29.15, "elapsed_time": "0:39:55", "remaining_time": "1:37:03", "throughput": 5193.09, "total_tokens": 12441344} {"current_steps": 12370, "total_steps": 42420, "loss": 0.4017, "lr": 0.0008922971326231831, "epoch": 5.832154644035832, "percentage": 29.16, "elapsed_time": "0:39:56", "remaining_time": "1:37:02", "throughput": 5193.18, "total_tokens": 12446336} {"current_steps": 12375, "total_steps": 42420, "loss": 0.3201, "lr": 0.0008921695510016389, "epoch": 5.834512022630834, "percentage": 29.17, "elapsed_time": "0:39:57", "remaining_time": "1:37:00", "throughput": 5193.24, "total_tokens": 12451008} {"current_steps": 12380, "total_steps": 42420, "loss": 0.3686, "lr": 0.0008920419029924485, "epoch": 5.836869401225837, "percentage": 29.18, "elapsed_time": "0:39:58", "remaining_time": "1:36:59", "throughput": 5193.28, "total_tokens": 12456032} {"current_steps": 12385, "total_steps": 42420, "loss": 0.3501, "lr": 0.0008919141886172204, "epoch": 5.839226779820839, "percentage": 29.2, "elapsed_time": "0:39:59", "remaining_time": "1:36:59", "throughput": 5193.58, "total_tokens": 12462496} {"current_steps": 12390, "total_steps": 42420, "loss": 0.3518, "lr": 0.0008917864078975743, "epoch": 5.841584158415841, "percentage": 29.21, "elapsed_time": "0:40:00", "remaining_time": "1:36:58", "throughput": 5193.66, "total_tokens": 12467520} {"current_steps": 12395, "total_steps": 42420, "loss": 0.3618, "lr": 0.0008916585608551418, "epoch": 5.843941537010844, "percentage": 29.22, "elapsed_time": "0:40:01", "remaining_time": "1:36:57", "throughput": 5193.79, "total_tokens": 12472832} {"current_steps": 12400, "total_steps": 42420, "loss": 0.3146, "lr": 0.0008915306475115647, "epoch": 5.846298915605846, "percentage": 29.23, "elapsed_time": "0:40:02", "remaining_time": "1:36:55", "throughput": 5193.75, "total_tokens": 12476960} {"current_steps": 12405, "total_steps": 42420, "loss": 0.4059, "lr": 0.000891402667888497, "epoch": 5.848656294200849, "percentage": 29.24, "elapsed_time": "0:40:03", "remaining_time": "1:36:54", "throughput": 5193.75, "total_tokens": 12481568} {"current_steps": 12410, "total_steps": 42420, "loss": 0.3882, "lr": 0.0008912746220076033, "epoch": 5.851013672795851, "percentage": 29.26, "elapsed_time": "0:40:04", "remaining_time": "1:36:53", "throughput": 5193.78, "total_tokens": 12486240} {"current_steps": 12415, "total_steps": 42420, "loss": 0.3467, "lr": 0.0008911465098905593, "epoch": 5.8533710513908535, "percentage": 29.27, "elapsed_time": "0:40:05", "remaining_time": "1:36:52", "throughput": 5193.95, "total_tokens": 12491744} {"current_steps": 12420, "total_steps": 42420, "loss": 0.3484, "lr": 0.0008910183315590525, "epoch": 5.855728429985856, "percentage": 29.28, "elapsed_time": "0:40:06", "remaining_time": "1:36:51", "throughput": 5194.04, "total_tokens": 12496992} {"current_steps": 12425, "total_steps": 42420, "loss": 0.2954, "lr": 0.0008908900870347815, "epoch": 5.858085808580858, "percentage": 29.29, "elapsed_time": "0:40:06", "remaining_time": "1:36:50", "throughput": 5194.2, "total_tokens": 12502400} {"current_steps": 12430, "total_steps": 42420, "loss": 0.2633, "lr": 0.0008907617763394555, "epoch": 5.860443187175861, "percentage": 29.3, "elapsed_time": "0:40:07", "remaining_time": "1:36:49", "throughput": 5194.32, "total_tokens": 12507584} {"current_steps": 12435, "total_steps": 42420, "loss": 0.2893, "lr": 0.0008906333994947956, "epoch": 5.862800565770863, "percentage": 29.31, "elapsed_time": "0:40:08", "remaining_time": "1:36:48", "throughput": 5194.4, "total_tokens": 12512384} {"current_steps": 12440, "total_steps": 42420, "loss": 0.4015, "lr": 0.0008905049565225338, "epoch": 5.8651579443658655, "percentage": 29.33, "elapsed_time": "0:40:09", "remaining_time": "1:36:47", "throughput": 5194.38, "total_tokens": 12516896} {"current_steps": 12445, "total_steps": 42420, "loss": 0.3647, "lr": 0.0008903764474444131, "epoch": 5.867515322960868, "percentage": 29.34, "elapsed_time": "0:40:10", "remaining_time": "1:36:46", "throughput": 5194.47, "total_tokens": 12521920} {"current_steps": 12450, "total_steps": 42420, "loss": 0.325, "lr": 0.0008902478722821881, "epoch": 5.86987270155587, "percentage": 29.35, "elapsed_time": "0:40:11", "remaining_time": "1:36:44", "throughput": 5194.37, "total_tokens": 12525792} {"current_steps": 12455, "total_steps": 42420, "loss": 0.356, "lr": 0.0008901192310576245, "epoch": 5.872230080150873, "percentage": 29.36, "elapsed_time": "0:40:12", "remaining_time": "1:36:43", "throughput": 5194.55, "total_tokens": 12531200} {"current_steps": 12460, "total_steps": 42420, "loss": 0.3614, "lr": 0.0008899905237924987, "epoch": 5.874587458745875, "percentage": 29.37, "elapsed_time": "0:40:13", "remaining_time": "1:36:42", "throughput": 5194.6, "total_tokens": 12535904} {"current_steps": 12465, "total_steps": 42420, "loss": 0.3409, "lr": 0.0008898617505085991, "epoch": 5.876944837340877, "percentage": 29.38, "elapsed_time": "0:40:14", "remaining_time": "1:36:41", "throughput": 5194.79, "total_tokens": 12541632} {"current_steps": 12470, "total_steps": 42420, "loss": 0.2718, "lr": 0.0008897329112277246, "epoch": 5.879302215935879, "percentage": 29.4, "elapsed_time": "0:40:15", "remaining_time": "1:36:40", "throughput": 5194.76, "total_tokens": 12545760} {"current_steps": 12475, "total_steps": 42420, "loss": 0.3924, "lr": 0.0008896040059716855, "epoch": 5.881659594530881, "percentage": 29.41, "elapsed_time": "0:40:16", "remaining_time": "1:36:39", "throughput": 5194.98, "total_tokens": 12551712} {"current_steps": 12480, "total_steps": 42420, "loss": 0.2829, "lr": 0.0008894750347623032, "epoch": 5.884016973125884, "percentage": 29.42, "elapsed_time": "0:40:16", "remaining_time": "1:36:38", "throughput": 5195.0, "total_tokens": 12556032} {"current_steps": 12485, "total_steps": 42420, "loss": 0.4973, "lr": 0.0008893459976214104, "epoch": 5.886374351720886, "percentage": 29.43, "elapsed_time": "0:40:17", "remaining_time": "1:36:37", "throughput": 5195.13, "total_tokens": 12561184} {"current_steps": 12490, "total_steps": 42420, "loss": 0.357, "lr": 0.000889216894570851, "epoch": 5.888731730315889, "percentage": 29.44, "elapsed_time": "0:40:18", "remaining_time": "1:36:36", "throughput": 5195.35, "total_tokens": 12566944} {"current_steps": 12495, "total_steps": 42420, "loss": 0.4183, "lr": 0.0008890877256324799, "epoch": 5.891089108910891, "percentage": 29.46, "elapsed_time": "0:40:20", "remaining_time": "1:36:35", "throughput": 5195.74, "total_tokens": 12573888} {"current_steps": 12500, "total_steps": 42420, "loss": 0.3741, "lr": 0.0008889584908281632, "epoch": 5.893446487505893, "percentage": 29.47, "elapsed_time": "0:40:20", "remaining_time": "1:36:34", "throughput": 5195.66, "total_tokens": 12577856} {"current_steps": 12505, "total_steps": 42420, "loss": 0.3036, "lr": 0.000888829190179778, "epoch": 5.895803866100896, "percentage": 29.48, "elapsed_time": "0:40:21", "remaining_time": "1:36:33", "throughput": 5195.88, "total_tokens": 12583648} {"current_steps": 12510, "total_steps": 42420, "loss": 0.327, "lr": 0.000888699823709213, "epoch": 5.898161244695898, "percentage": 29.49, "elapsed_time": "0:40:22", "remaining_time": "1:36:32", "throughput": 5195.9, "total_tokens": 12588032} {"current_steps": 12515, "total_steps": 42420, "loss": 0.4512, "lr": 0.0008885703914383675, "epoch": 5.900518623290901, "percentage": 29.5, "elapsed_time": "0:40:23", "remaining_time": "1:36:31", "throughput": 5196.07, "total_tokens": 12593728} {"current_steps": 12520, "total_steps": 42420, "loss": 0.4319, "lr": 0.0008884408933891524, "epoch": 5.902876001885903, "percentage": 29.51, "elapsed_time": "0:40:24", "remaining_time": "1:36:30", "throughput": 5196.2, "total_tokens": 12598816} {"current_steps": 12525, "total_steps": 42420, "loss": 0.3352, "lr": 0.0008883113295834892, "epoch": 5.905233380480905, "percentage": 29.53, "elapsed_time": "0:40:25", "remaining_time": "1:36:29", "throughput": 5196.31, "total_tokens": 12603936} {"current_steps": 12530, "total_steps": 42420, "loss": 0.3988, "lr": 0.0008881817000433112, "epoch": 5.907590759075908, "percentage": 29.54, "elapsed_time": "0:40:26", "remaining_time": "1:36:28", "throughput": 5196.36, "total_tokens": 12608480} {"current_steps": 12535, "total_steps": 42420, "loss": 0.3475, "lr": 0.0008880520047905622, "epoch": 5.90994813767091, "percentage": 29.55, "elapsed_time": "0:40:27", "remaining_time": "1:36:26", "throughput": 5196.35, "total_tokens": 12612640} {"current_steps": 12540, "total_steps": 42420, "loss": 0.3134, "lr": 0.0008879222438471976, "epoch": 5.912305516265913, "percentage": 29.56, "elapsed_time": "0:40:28", "remaining_time": "1:36:26", "throughput": 5196.82, "total_tokens": 12620320} {"current_steps": 12545, "total_steps": 42420, "loss": 0.3936, "lr": 0.0008877924172351839, "epoch": 5.914662894860915, "percentage": 29.57, "elapsed_time": "0:40:29", "remaining_time": "1:36:25", "throughput": 5196.89, "total_tokens": 12625120} {"current_steps": 12550, "total_steps": 42420, "loss": 0.3293, "lr": 0.000887662524976498, "epoch": 5.9170202734559165, "percentage": 29.59, "elapsed_time": "0:40:30", "remaining_time": "1:36:24", "throughput": 5197.02, "total_tokens": 12630400} {"current_steps": 12555, "total_steps": 42420, "loss": 0.3262, "lr": 0.0008875325670931289, "epoch": 5.919377652050919, "percentage": 29.6, "elapsed_time": "0:40:31", "remaining_time": "1:36:23", "throughput": 5197.05, "total_tokens": 12635104} {"current_steps": 12560, "total_steps": 42420, "loss": 0.3147, "lr": 0.000887402543607076, "epoch": 5.921735030645921, "percentage": 29.61, "elapsed_time": "0:40:32", "remaining_time": "1:36:22", "throughput": 5197.29, "total_tokens": 12641408} {"current_steps": 12565, "total_steps": 42420, "loss": 0.3663, "lr": 0.0008872724545403503, "epoch": 5.924092409240924, "percentage": 29.62, "elapsed_time": "0:40:33", "remaining_time": "1:36:21", "throughput": 5197.38, "total_tokens": 12646432} {"current_steps": 12570, "total_steps": 42420, "loss": 0.3757, "lr": 0.0008871422999149734, "epoch": 5.926449787835926, "percentage": 29.63, "elapsed_time": "0:40:34", "remaining_time": "1:36:20", "throughput": 5197.44, "total_tokens": 12651040} {"current_steps": 12575, "total_steps": 42420, "loss": 0.3282, "lr": 0.0008870120797529786, "epoch": 5.9288071664309285, "percentage": 29.64, "elapsed_time": "0:40:35", "remaining_time": "1:36:19", "throughput": 5197.55, "total_tokens": 12656160} {"current_steps": 12580, "total_steps": 42420, "loss": 0.3547, "lr": 0.0008868817940764098, "epoch": 5.931164545025931, "percentage": 29.66, "elapsed_time": "0:40:35", "remaining_time": "1:36:18", "throughput": 5197.66, "total_tokens": 12660992} {"current_steps": 12585, "total_steps": 42420, "loss": 0.2839, "lr": 0.000886751442907322, "epoch": 5.933521923620933, "percentage": 29.67, "elapsed_time": "0:40:36", "remaining_time": "1:36:16", "throughput": 5197.73, "total_tokens": 12665952} {"current_steps": 12590, "total_steps": 42420, "loss": 0.3622, "lr": 0.0008866210262677818, "epoch": 5.935879302215936, "percentage": 29.68, "elapsed_time": "0:40:37", "remaining_time": "1:36:16", "throughput": 5197.98, "total_tokens": 12672416} {"current_steps": 12595, "total_steps": 42420, "loss": 0.3823, "lr": 0.0008864905441798661, "epoch": 5.938236680810938, "percentage": 29.69, "elapsed_time": "0:40:38", "remaining_time": "1:36:15", "throughput": 5198.0, "total_tokens": 12676704} {"current_steps": 12600, "total_steps": 42420, "loss": 0.3456, "lr": 0.0008863599966656635, "epoch": 5.9405940594059405, "percentage": 29.7, "elapsed_time": "0:40:39", "remaining_time": "1:36:14", "throughput": 5198.14, "total_tokens": 12682144} {"current_steps": 12605, "total_steps": 42420, "loss": 0.3483, "lr": 0.0008862293837472735, "epoch": 5.942951438000943, "percentage": 29.71, "elapsed_time": "0:40:40", "remaining_time": "1:36:12", "throughput": 5198.18, "total_tokens": 12686784} {"current_steps": 12610, "total_steps": 42420, "loss": 0.3463, "lr": 0.0008860987054468069, "epoch": 5.945308816595945, "percentage": 29.73, "elapsed_time": "0:40:41", "remaining_time": "1:36:11", "throughput": 5198.35, "total_tokens": 12692192} {"current_steps": 12615, "total_steps": 42420, "loss": 0.3479, "lr": 0.0008859679617863847, "epoch": 5.947666195190948, "percentage": 29.74, "elapsed_time": "0:40:42", "remaining_time": "1:36:10", "throughput": 5198.36, "total_tokens": 12696704} {"current_steps": 12620, "total_steps": 42420, "loss": 0.3513, "lr": 0.0008858371527881401, "epoch": 5.95002357378595, "percentage": 29.75, "elapsed_time": "0:40:43", "remaining_time": "1:36:09", "throughput": 5198.41, "total_tokens": 12701600} {"current_steps": 12625, "total_steps": 42420, "loss": 0.3418, "lr": 0.0008857062784742166, "epoch": 5.9523809523809526, "percentage": 29.76, "elapsed_time": "0:40:44", "remaining_time": "1:36:08", "throughput": 5198.45, "total_tokens": 12706048} {"current_steps": 12630, "total_steps": 42420, "loss": 0.3433, "lr": 0.0008855753388667692, "epoch": 5.954738330975955, "percentage": 29.77, "elapsed_time": "0:40:45", "remaining_time": "1:36:07", "throughput": 5198.51, "total_tokens": 12710688} {"current_steps": 12635, "total_steps": 42420, "loss": 0.3472, "lr": 0.0008854443339879635, "epoch": 5.957095709570957, "percentage": 29.79, "elapsed_time": "0:40:46", "remaining_time": "1:36:06", "throughput": 5198.71, "total_tokens": 12716256} {"current_steps": 12640, "total_steps": 42420, "loss": 0.3548, "lr": 0.0008853132638599766, "epoch": 5.95945308816596, "percentage": 29.8, "elapsed_time": "0:40:46", "remaining_time": "1:36:05", "throughput": 5198.78, "total_tokens": 12721376} {"current_steps": 12645, "total_steps": 42420, "loss": 0.356, "lr": 0.0008851821285049963, "epoch": 5.961810466760962, "percentage": 29.81, "elapsed_time": "0:40:47", "remaining_time": "1:36:04", "throughput": 5198.88, "total_tokens": 12726400} {"current_steps": 12650, "total_steps": 42420, "loss": 0.3508, "lr": 0.0008850509279452217, "epoch": 5.964167845355965, "percentage": 29.82, "elapsed_time": "0:40:48", "remaining_time": "1:36:02", "throughput": 5198.87, "total_tokens": 12730880} {"current_steps": 12655, "total_steps": 42420, "loss": 0.3602, "lr": 0.0008849196622028628, "epoch": 5.966525223950967, "percentage": 29.83, "elapsed_time": "0:40:49", "remaining_time": "1:36:01", "throughput": 5198.98, "total_tokens": 12736032} {"current_steps": 12660, "total_steps": 42420, "loss": 0.3455, "lr": 0.0008847883313001407, "epoch": 5.968882602545969, "percentage": 29.84, "elapsed_time": "0:40:50", "remaining_time": "1:36:00", "throughput": 5199.19, "total_tokens": 12741888} {"current_steps": 12665, "total_steps": 42420, "loss": 0.3504, "lr": 0.0008846569352592874, "epoch": 5.971239981140971, "percentage": 29.86, "elapsed_time": "0:40:51", "remaining_time": "1:35:59", "throughput": 5199.19, "total_tokens": 12746560} {"current_steps": 12670, "total_steps": 42420, "loss": 0.3122, "lr": 0.000884525474102546, "epoch": 5.973597359735973, "percentage": 29.87, "elapsed_time": "0:40:52", "remaining_time": "1:35:58", "throughput": 5199.31, "total_tokens": 12751680} {"current_steps": 12675, "total_steps": 42420, "loss": 0.3699, "lr": 0.0008843939478521708, "epoch": 5.975954738330976, "percentage": 29.88, "elapsed_time": "0:40:53", "remaining_time": "1:35:57", "throughput": 5199.34, "total_tokens": 12756160} {"current_steps": 12680, "total_steps": 42420, "loss": 0.292, "lr": 0.0008842623565304268, "epoch": 5.978312116925978, "percentage": 29.89, "elapsed_time": "0:40:54", "remaining_time": "1:35:56", "throughput": 5199.43, "total_tokens": 12761312} {"current_steps": 12685, "total_steps": 42420, "loss": 0.3417, "lr": 0.0008841307001595903, "epoch": 5.9806694955209805, "percentage": 29.9, "elapsed_time": "0:40:55", "remaining_time": "1:35:55", "throughput": 5199.57, "total_tokens": 12766528} {"current_steps": 12690, "total_steps": 42420, "loss": 0.3707, "lr": 0.0008839989787619484, "epoch": 5.983026874115983, "percentage": 29.92, "elapsed_time": "0:40:56", "remaining_time": "1:35:54", "throughput": 5199.63, "total_tokens": 12771328} {"current_steps": 12695, "total_steps": 42420, "loss": 0.3149, "lr": 0.0008838671923597994, "epoch": 5.985384252710985, "percentage": 29.93, "elapsed_time": "0:40:57", "remaining_time": "1:35:53", "throughput": 5199.61, "total_tokens": 12775712} {"current_steps": 12700, "total_steps": 42420, "loss": 0.3143, "lr": 0.0008837353409754523, "epoch": 5.987741631305988, "percentage": 29.94, "elapsed_time": "0:40:57", "remaining_time": "1:35:51", "throughput": 5199.61, "total_tokens": 12780128} {"current_steps": 12705, "total_steps": 42420, "loss": 0.3719, "lr": 0.0008836034246312275, "epoch": 5.99009900990099, "percentage": 29.95, "elapsed_time": "0:40:58", "remaining_time": "1:35:50", "throughput": 5199.73, "total_tokens": 12785472} {"current_steps": 12710, "total_steps": 42420, "loss": 0.2943, "lr": 0.0008834714433494561, "epoch": 5.9924563884959925, "percentage": 29.96, "elapsed_time": "0:40:59", "remaining_time": "1:35:49", "throughput": 5199.66, "total_tokens": 12789568} {"current_steps": 12715, "total_steps": 42420, "loss": 0.3632, "lr": 0.0008833393971524803, "epoch": 5.994813767090995, "percentage": 29.97, "elapsed_time": "0:41:00", "remaining_time": "1:35:48", "throughput": 5199.66, "total_tokens": 12793888} {"current_steps": 12720, "total_steps": 42420, "loss": 0.2843, "lr": 0.0008832072860626534, "epoch": 5.997171145685997, "percentage": 29.99, "elapsed_time": "0:41:01", "remaining_time": "1:35:47", "throughput": 5199.67, "total_tokens": 12798464} {"current_steps": 12725, "total_steps": 42420, "loss": 0.3954, "lr": 0.0008830751101023393, "epoch": 5.999528524281, "percentage": 30.0, "elapsed_time": "0:41:02", "remaining_time": "1:35:45", "throughput": 5199.68, "total_tokens": 12802816} {"current_steps": 12726, "total_steps": 42420, "eval_loss": 0.34014013409614563, "epoch": 6.0, "percentage": 30.0, "elapsed_time": "0:41:21", "remaining_time": "1:36:30", "throughput": 5159.83, "total_tokens": 12804064} {"current_steps": 12730, "total_steps": 42420, "loss": 0.3087, "lr": 0.0008829428692939134, "epoch": 6.001885902876002, "percentage": 30.01, "elapsed_time": "0:41:24", "remaining_time": "1:36:34", "throughput": 5155.53, "total_tokens": 12808736} {"current_steps": 12735, "total_steps": 42420, "loss": 0.3671, "lr": 0.0008828105636597618, "epoch": 6.0042432814710045, "percentage": 30.02, "elapsed_time": "0:41:25", "remaining_time": "1:36:33", "throughput": 5155.64, "total_tokens": 12813696} {"current_steps": 12740, "total_steps": 42420, "loss": 0.3277, "lr": 0.0008826781932222814, "epoch": 6.006600660066007, "percentage": 30.03, "elapsed_time": "0:41:26", "remaining_time": "1:36:32", "throughput": 5155.67, "total_tokens": 12818400} {"current_steps": 12745, "total_steps": 42420, "loss": 0.3401, "lr": 0.0008825457580038805, "epoch": 6.008958038661009, "percentage": 30.04, "elapsed_time": "0:41:27", "remaining_time": "1:36:30", "throughput": 5155.66, "total_tokens": 12822816} {"current_steps": 12750, "total_steps": 42420, "loss": 0.3546, "lr": 0.0008824132580269779, "epoch": 6.011315417256012, "percentage": 30.06, "elapsed_time": "0:41:27", "remaining_time": "1:36:29", "throughput": 5155.67, "total_tokens": 12827296} {"current_steps": 12755, "total_steps": 42420, "loss": 0.3178, "lr": 0.0008822806933140038, "epoch": 6.013672795851014, "percentage": 30.07, "elapsed_time": "0:41:29", "remaining_time": "1:36:28", "throughput": 5155.96, "total_tokens": 12833408} {"current_steps": 12760, "total_steps": 42420, "loss": 0.3519, "lr": 0.0008821480638873991, "epoch": 6.016030174446016, "percentage": 30.08, "elapsed_time": "0:41:29", "remaining_time": "1:36:27", "throughput": 5155.99, "total_tokens": 12837760} {"current_steps": 12765, "total_steps": 42420, "loss": 0.3023, "lr": 0.0008820153697696156, "epoch": 6.018387553041018, "percentage": 30.09, "elapsed_time": "0:41:30", "remaining_time": "1:36:26", "throughput": 5155.96, "total_tokens": 12841728} {"current_steps": 12770, "total_steps": 42420, "loss": 0.2906, "lr": 0.0008818826109831164, "epoch": 6.02074493163602, "percentage": 30.1, "elapsed_time": "0:41:31", "remaining_time": "1:36:25", "throughput": 5156.06, "total_tokens": 12847136} {"current_steps": 12775, "total_steps": 42420, "loss": 0.4018, "lr": 0.0008817497875503751, "epoch": 6.023102310231023, "percentage": 30.12, "elapsed_time": "0:41:32", "remaining_time": "1:36:24", "throughput": 5156.3, "total_tokens": 12853280} {"current_steps": 12780, "total_steps": 42420, "loss": 0.3498, "lr": 0.0008816168994938764, "epoch": 6.025459688826025, "percentage": 30.13, "elapsed_time": "0:41:33", "remaining_time": "1:36:23", "throughput": 5156.39, "total_tokens": 12858048} {"current_steps": 12785, "total_steps": 42420, "loss": 0.3594, "lr": 0.0008814839468361162, "epoch": 6.027817067421028, "percentage": 30.14, "elapsed_time": "0:41:34", "remaining_time": "1:36:22", "throughput": 5156.44, "total_tokens": 12862720} {"current_steps": 12790, "total_steps": 42420, "loss": 0.3434, "lr": 0.000881350929599601, "epoch": 6.03017444601603, "percentage": 30.15, "elapsed_time": "0:41:35", "remaining_time": "1:36:21", "throughput": 5156.51, "total_tokens": 12867776} {"current_steps": 12795, "total_steps": 42420, "loss": 0.4173, "lr": 0.0008812178478068483, "epoch": 6.032531824611032, "percentage": 30.16, "elapsed_time": "0:41:36", "remaining_time": "1:36:19", "throughput": 5156.61, "total_tokens": 12872736} {"current_steps": 12800, "total_steps": 42420, "loss": 0.3736, "lr": 0.0008810847014803868, "epoch": 6.034889203206035, "percentage": 30.17, "elapsed_time": "0:41:37", "remaining_time": "1:36:18", "throughput": 5156.62, "total_tokens": 12877184} {"current_steps": 12805, "total_steps": 42420, "loss": 0.3226, "lr": 0.0008809514906427559, "epoch": 6.037246581801037, "percentage": 30.19, "elapsed_time": "0:41:38", "remaining_time": "1:36:17", "throughput": 5156.77, "total_tokens": 12882432} {"current_steps": 12810, "total_steps": 42420, "loss": 0.3725, "lr": 0.0008808182153165056, "epoch": 6.03960396039604, "percentage": 30.2, "elapsed_time": "0:41:39", "remaining_time": "1:36:16", "throughput": 5156.87, "total_tokens": 12887264} {"current_steps": 12815, "total_steps": 42420, "loss": 0.4664, "lr": 0.0008806848755241974, "epoch": 6.041961338991042, "percentage": 30.21, "elapsed_time": "0:41:39", "remaining_time": "1:36:15", "throughput": 5156.98, "total_tokens": 12892352} {"current_steps": 12820, "total_steps": 42420, "loss": 0.3281, "lr": 0.0008805514712884035, "epoch": 6.044318717586044, "percentage": 30.22, "elapsed_time": "0:41:40", "remaining_time": "1:36:14", "throughput": 5157.08, "total_tokens": 12897600} {"current_steps": 12825, "total_steps": 42420, "loss": 0.3608, "lr": 0.0008804180026317069, "epoch": 6.046676096181047, "percentage": 30.23, "elapsed_time": "0:41:41", "remaining_time": "1:36:13", "throughput": 5157.11, "total_tokens": 12902176} {"current_steps": 12830, "total_steps": 42420, "loss": 0.3529, "lr": 0.0008802844695767016, "epoch": 6.049033474776049, "percentage": 30.25, "elapsed_time": "0:41:42", "remaining_time": "1:36:12", "throughput": 5157.29, "total_tokens": 12907808} {"current_steps": 12835, "total_steps": 42420, "loss": 0.3764, "lr": 0.0008801508721459925, "epoch": 6.051390853371052, "percentage": 30.26, "elapsed_time": "0:41:43", "remaining_time": "1:36:11", "throughput": 5157.34, "total_tokens": 12912480} {"current_steps": 12840, "total_steps": 42420, "loss": 0.3421, "lr": 0.0008800172103621953, "epoch": 6.053748231966054, "percentage": 30.27, "elapsed_time": "0:41:44", "remaining_time": "1:36:10", "throughput": 5157.59, "total_tokens": 12918720} {"current_steps": 12845, "total_steps": 42420, "loss": 0.3514, "lr": 0.0008798834842479366, "epoch": 6.0561056105610565, "percentage": 30.28, "elapsed_time": "0:41:45", "remaining_time": "1:36:09", "throughput": 5157.87, "total_tokens": 12924736} {"current_steps": 12850, "total_steps": 42420, "loss": 0.3445, "lr": 0.0008797496938258543, "epoch": 6.058462989156059, "percentage": 30.29, "elapsed_time": "0:41:46", "remaining_time": "1:36:08", "throughput": 5158.01, "total_tokens": 12930048} {"current_steps": 12855, "total_steps": 42420, "loss": 0.3302, "lr": 0.0008796158391185965, "epoch": 6.060820367751061, "percentage": 30.3, "elapsed_time": "0:41:47", "remaining_time": "1:36:07", "throughput": 5158.03, "total_tokens": 12934400} {"current_steps": 12860, "total_steps": 42420, "loss": 0.3004, "lr": 0.0008794819201488225, "epoch": 6.063177746346063, "percentage": 30.32, "elapsed_time": "0:41:48", "remaining_time": "1:36:05", "throughput": 5158.06, "total_tokens": 12938816} {"current_steps": 12865, "total_steps": 42420, "loss": 0.2375, "lr": 0.0008793479369392028, "epoch": 6.065535124941065, "percentage": 30.33, "elapsed_time": "0:41:49", "remaining_time": "1:36:04", "throughput": 5158.13, "total_tokens": 12943520} {"current_steps": 12870, "total_steps": 42420, "loss": 0.3582, "lr": 0.0008792138895124184, "epoch": 6.067892503536068, "percentage": 30.34, "elapsed_time": "0:41:50", "remaining_time": "1:36:03", "throughput": 5158.28, "total_tokens": 12948736} {"current_steps": 12875, "total_steps": 42420, "loss": 0.4096, "lr": 0.000879079777891161, "epoch": 6.07024988213107, "percentage": 30.35, "elapsed_time": "0:41:51", "remaining_time": "1:36:02", "throughput": 5158.26, "total_tokens": 12953024} {"current_steps": 12880, "total_steps": 42420, "loss": 0.2939, "lr": 0.0008789456020981337, "epoch": 6.072607260726072, "percentage": 30.36, "elapsed_time": "0:41:51", "remaining_time": "1:36:01", "throughput": 5158.3, "total_tokens": 12957504} {"current_steps": 12885, "total_steps": 42420, "loss": 0.3636, "lr": 0.0008788113621560502, "epoch": 6.074964639321075, "percentage": 30.37, "elapsed_time": "0:41:52", "remaining_time": "1:36:00", "throughput": 5158.36, "total_tokens": 12962400} {"current_steps": 12890, "total_steps": 42420, "loss": 0.341, "lr": 0.0008786770580876349, "epoch": 6.077322017916077, "percentage": 30.39, "elapsed_time": "0:41:53", "remaining_time": "1:35:59", "throughput": 5158.65, "total_tokens": 12968576} {"current_steps": 12895, "total_steps": 42420, "loss": 0.3549, "lr": 0.0008785426899156231, "epoch": 6.07967939651108, "percentage": 30.4, "elapsed_time": "0:41:54", "remaining_time": "1:35:58", "throughput": 5158.73, "total_tokens": 12973632} {"current_steps": 12900, "total_steps": 42420, "loss": 0.3463, "lr": 0.0008784082576627613, "epoch": 6.082036775106082, "percentage": 30.41, "elapsed_time": "0:41:55", "remaining_time": "1:35:57", "throughput": 5158.84, "total_tokens": 12978592} {"current_steps": 12905, "total_steps": 42420, "loss": 0.3433, "lr": 0.0008782737613518064, "epoch": 6.084394153701084, "percentage": 30.42, "elapsed_time": "0:41:56", "remaining_time": "1:35:55", "throughput": 5158.99, "total_tokens": 12983648} {"current_steps": 12910, "total_steps": 42420, "loss": 0.3391, "lr": 0.0008781392010055266, "epoch": 6.086751532296087, "percentage": 30.43, "elapsed_time": "0:41:57", "remaining_time": "1:35:55", "throughput": 5159.18, "total_tokens": 12989472} {"current_steps": 12915, "total_steps": 42420, "loss": 0.3448, "lr": 0.0008780045766467004, "epoch": 6.089108910891089, "percentage": 30.45, "elapsed_time": "0:41:58", "remaining_time": "1:35:54", "throughput": 5159.47, "total_tokens": 12995808} {"current_steps": 12920, "total_steps": 42420, "loss": 0.3979, "lr": 0.0008778698882981176, "epoch": 6.091466289486092, "percentage": 30.46, "elapsed_time": "0:41:59", "remaining_time": "1:35:53", "throughput": 5159.68, "total_tokens": 13001664} {"current_steps": 12925, "total_steps": 42420, "loss": 0.3451, "lr": 0.0008777351359825783, "epoch": 6.093823668081094, "percentage": 30.47, "elapsed_time": "0:42:00", "remaining_time": "1:35:52", "throughput": 5159.84, "total_tokens": 13006912} {"current_steps": 12930, "total_steps": 42420, "loss": 0.3488, "lr": 0.0008776003197228943, "epoch": 6.096181046676096, "percentage": 30.48, "elapsed_time": "0:42:01", "remaining_time": "1:35:51", "throughput": 5159.81, "total_tokens": 13011136} {"current_steps": 12935, "total_steps": 42420, "loss": 0.3319, "lr": 0.0008774654395418875, "epoch": 6.098538425271099, "percentage": 30.49, "elapsed_time": "0:42:02", "remaining_time": "1:35:50", "throughput": 5159.79, "total_tokens": 13015648} {"current_steps": 12940, "total_steps": 42420, "loss": 0.3504, "lr": 0.0008773304954623907, "epoch": 6.100895803866101, "percentage": 30.5, "elapsed_time": "0:42:03", "remaining_time": "1:35:48", "throughput": 5159.82, "total_tokens": 13020448} {"current_steps": 12945, "total_steps": 42420, "loss": 0.3514, "lr": 0.0008771954875072477, "epoch": 6.103253182461104, "percentage": 30.52, "elapsed_time": "0:42:04", "remaining_time": "1:35:47", "throughput": 5159.91, "total_tokens": 13025312} {"current_steps": 12950, "total_steps": 42420, "loss": 0.3303, "lr": 0.0008770604156993131, "epoch": 6.105610561056106, "percentage": 30.53, "elapsed_time": "0:42:05", "remaining_time": "1:35:46", "throughput": 5160.11, "total_tokens": 13030912} {"current_steps": 12955, "total_steps": 42420, "loss": 0.2771, "lr": 0.0008769252800614522, "epoch": 6.107967939651108, "percentage": 30.54, "elapsed_time": "0:42:06", "remaining_time": "1:35:45", "throughput": 5160.33, "total_tokens": 13036832} {"current_steps": 12960, "total_steps": 42420, "loss": 0.4227, "lr": 0.0008767900806165413, "epoch": 6.11032531824611, "percentage": 30.55, "elapsed_time": "0:42:07", "remaining_time": "1:35:44", "throughput": 5160.38, "total_tokens": 13041600} {"current_steps": 12965, "total_steps": 42420, "loss": 0.3499, "lr": 0.0008766548173874672, "epoch": 6.112682696841112, "percentage": 30.56, "elapsed_time": "0:42:08", "remaining_time": "1:35:43", "throughput": 5160.49, "total_tokens": 13046720} {"current_steps": 12970, "total_steps": 42420, "loss": 0.385, "lr": 0.0008765194903971278, "epoch": 6.115040075436115, "percentage": 30.58, "elapsed_time": "0:42:09", "remaining_time": "1:35:42", "throughput": 5160.6, "total_tokens": 13051680} {"current_steps": 12975, "total_steps": 42420, "loss": 0.3444, "lr": 0.0008763840996684317, "epoch": 6.117397454031117, "percentage": 30.59, "elapsed_time": "0:42:10", "remaining_time": "1:35:42", "throughput": 5160.99, "total_tokens": 13058816} {"current_steps": 12980, "total_steps": 42420, "loss": 0.2607, "lr": 0.0008762486452242981, "epoch": 6.1197548326261195, "percentage": 30.6, "elapsed_time": "0:42:11", "remaining_time": "1:35:41", "throughput": 5161.19, "total_tokens": 13064256} {"current_steps": 12985, "total_steps": 42420, "loss": 0.2563, "lr": 0.0008761131270876571, "epoch": 6.122112211221122, "percentage": 30.61, "elapsed_time": "0:42:12", "remaining_time": "1:35:39", "throughput": 5161.2, "total_tokens": 13068736} {"current_steps": 12990, "total_steps": 42420, "loss": 0.4711, "lr": 0.0008759775452814498, "epoch": 6.124469589816124, "percentage": 30.62, "elapsed_time": "0:42:13", "remaining_time": "1:35:38", "throughput": 5161.25, "total_tokens": 13073504} {"current_steps": 12995, "total_steps": 42420, "loss": 0.3414, "lr": 0.0008758418998286279, "epoch": 6.126826968411127, "percentage": 30.63, "elapsed_time": "0:42:13", "remaining_time": "1:35:37", "throughput": 5161.21, "total_tokens": 13077568} {"current_steps": 13000, "total_steps": 42420, "loss": 0.306, "lr": 0.0008757061907521536, "epoch": 6.129184347006129, "percentage": 30.65, "elapsed_time": "0:42:14", "remaining_time": "1:35:36", "throughput": 5161.22, "total_tokens": 13081824} {"current_steps": 13005, "total_steps": 42420, "loss": 0.306, "lr": 0.0008755704180750005, "epoch": 6.1315417256011315, "percentage": 30.66, "elapsed_time": "0:42:15", "remaining_time": "1:35:35", "throughput": 5161.45, "total_tokens": 13087904} {"current_steps": 13010, "total_steps": 42420, "loss": 0.3257, "lr": 0.0008754345818201523, "epoch": 6.133899104196134, "percentage": 30.67, "elapsed_time": "0:42:16", "remaining_time": "1:35:34", "throughput": 5161.46, "total_tokens": 13092224} {"current_steps": 13015, "total_steps": 42420, "loss": 0.3537, "lr": 0.0008752986820106041, "epoch": 6.136256482791136, "percentage": 30.68, "elapsed_time": "0:42:17", "remaining_time": "1:35:32", "throughput": 5161.55, "total_tokens": 13097216} {"current_steps": 13020, "total_steps": 42420, "loss": 0.3676, "lr": 0.0008751627186693612, "epoch": 6.138613861386139, "percentage": 30.69, "elapsed_time": "0:42:18", "remaining_time": "1:35:31", "throughput": 5161.68, "total_tokens": 13102432} {"current_steps": 13025, "total_steps": 42420, "loss": 0.3212, "lr": 0.0008750266918194398, "epoch": 6.140971239981141, "percentage": 30.7, "elapsed_time": "0:42:19", "remaining_time": "1:35:30", "throughput": 5161.88, "total_tokens": 13108096} {"current_steps": 13030, "total_steps": 42420, "loss": 0.3462, "lr": 0.0008748906014838671, "epoch": 6.1433286185761435, "percentage": 30.72, "elapsed_time": "0:42:20", "remaining_time": "1:35:30", "throughput": 5162.07, "total_tokens": 13113824} {"current_steps": 13035, "total_steps": 42420, "loss": 0.3132, "lr": 0.0008747544476856809, "epoch": 6.145685997171146, "percentage": 30.73, "elapsed_time": "0:42:21", "remaining_time": "1:35:28", "throughput": 5162.1, "total_tokens": 13118656} {"current_steps": 13040, "total_steps": 42420, "loss": 0.3233, "lr": 0.0008746182304479297, "epoch": 6.148043375766148, "percentage": 30.74, "elapsed_time": "0:42:22", "remaining_time": "1:35:27", "throughput": 5162.06, "total_tokens": 13122688} {"current_steps": 13045, "total_steps": 42420, "loss": 0.3465, "lr": 0.0008744819497936727, "epoch": 6.150400754361151, "percentage": 30.75, "elapsed_time": "0:42:23", "remaining_time": "1:35:26", "throughput": 5162.23, "total_tokens": 13127872} {"current_steps": 13050, "total_steps": 42420, "loss": 0.4093, "lr": 0.00087434560574598, "epoch": 6.152758132956153, "percentage": 30.76, "elapsed_time": "0:42:24", "remaining_time": "1:35:25", "throughput": 5162.44, "total_tokens": 13133824} {"current_steps": 13055, "total_steps": 42420, "loss": 0.3556, "lr": 0.0008742091983279321, "epoch": 6.1551155115511555, "percentage": 30.78, "elapsed_time": "0:42:25", "remaining_time": "1:35:24", "throughput": 5162.57, "total_tokens": 13139008} {"current_steps": 13060, "total_steps": 42420, "loss": 0.3255, "lr": 0.0008740727275626208, "epoch": 6.157472890146157, "percentage": 30.79, "elapsed_time": "0:42:26", "remaining_time": "1:35:23", "throughput": 5162.72, "total_tokens": 13144416} {"current_steps": 13065, "total_steps": 42420, "loss": 0.3838, "lr": 0.0008739361934731481, "epoch": 6.1598302687411595, "percentage": 30.8, "elapsed_time": "0:42:26", "remaining_time": "1:35:22", "throughput": 5162.83, "total_tokens": 13149536} {"current_steps": 13070, "total_steps": 42420, "loss": 0.348, "lr": 0.0008737995960826268, "epoch": 6.162187647336162, "percentage": 30.81, "elapsed_time": "0:42:27", "remaining_time": "1:35:21", "throughput": 5162.98, "total_tokens": 13154752} {"current_steps": 13075, "total_steps": 42420, "loss": 0.3064, "lr": 0.0008736629354141808, "epoch": 6.164545025931164, "percentage": 30.82, "elapsed_time": "0:42:28", "remaining_time": "1:35:20", "throughput": 5162.9, "total_tokens": 13158560} {"current_steps": 13080, "total_steps": 42420, "loss": 0.3329, "lr": 0.0008735262114909441, "epoch": 6.166902404526167, "percentage": 30.83, "elapsed_time": "0:42:29", "remaining_time": "1:35:18", "throughput": 5162.88, "total_tokens": 13162592} {"current_steps": 13085, "total_steps": 42420, "loss": 0.318, "lr": 0.0008733894243360621, "epoch": 6.169259783121169, "percentage": 30.85, "elapsed_time": "0:42:30", "remaining_time": "1:35:17", "throughput": 5162.83, "total_tokens": 13166560} {"current_steps": 13090, "total_steps": 42420, "loss": 0.3723, "lr": 0.0008732525739726902, "epoch": 6.1716171617161715, "percentage": 30.86, "elapsed_time": "0:42:31", "remaining_time": "1:35:16", "throughput": 5162.86, "total_tokens": 13171136} {"current_steps": 13095, "total_steps": 42420, "loss": 0.3431, "lr": 0.0008731156604239952, "epoch": 6.173974540311174, "percentage": 30.87, "elapsed_time": "0:42:32", "remaining_time": "1:35:14", "throughput": 5162.88, "total_tokens": 13175680} {"current_steps": 13100, "total_steps": 42420, "loss": 0.3538, "lr": 0.0008729786837131537, "epoch": 6.176331918906176, "percentage": 30.88, "elapsed_time": "0:42:32", "remaining_time": "1:35:13", "throughput": 5162.93, "total_tokens": 13180224} {"current_steps": 13105, "total_steps": 42420, "loss": 0.3391, "lr": 0.0008728416438633542, "epoch": 6.178689297501179, "percentage": 30.89, "elapsed_time": "0:42:33", "remaining_time": "1:35:12", "throughput": 5163.16, "total_tokens": 13186304} {"current_steps": 13110, "total_steps": 42420, "loss": 0.3837, "lr": 0.0008727045408977947, "epoch": 6.181046676096181, "percentage": 30.91, "elapsed_time": "0:42:34", "remaining_time": "1:35:11", "throughput": 5163.14, "total_tokens": 13190592} {"current_steps": 13115, "total_steps": 42420, "loss": 0.3525, "lr": 0.0008725673748396846, "epoch": 6.1834040546911835, "percentage": 30.92, "elapsed_time": "0:42:35", "remaining_time": "1:35:10", "throughput": 5163.27, "total_tokens": 13195712} {"current_steps": 13120, "total_steps": 42420, "loss": 0.3569, "lr": 0.0008724301457122438, "epoch": 6.185761433286186, "percentage": 30.93, "elapsed_time": "0:42:36", "remaining_time": "1:35:09", "throughput": 5163.39, "total_tokens": 13200800} {"current_steps": 13125, "total_steps": 42420, "loss": 0.3606, "lr": 0.0008722928535387028, "epoch": 6.188118811881188, "percentage": 30.94, "elapsed_time": "0:42:37", "remaining_time": "1:35:08", "throughput": 5163.52, "total_tokens": 13205792} {"current_steps": 13130, "total_steps": 42420, "loss": 0.3328, "lr": 0.0008721554983423029, "epoch": 6.190476190476191, "percentage": 30.95, "elapsed_time": "0:42:38", "remaining_time": "1:35:07", "throughput": 5163.65, "total_tokens": 13210976} {"current_steps": 13135, "total_steps": 42420, "loss": 0.3774, "lr": 0.000872018080146296, "epoch": 6.192833569071193, "percentage": 30.96, "elapsed_time": "0:42:39", "remaining_time": "1:35:05", "throughput": 5163.6, "total_tokens": 13215040} {"current_steps": 13140, "total_steps": 42420, "loss": 0.2947, "lr": 0.0008718805989739446, "epoch": 6.1951909476661955, "percentage": 30.98, "elapsed_time": "0:42:40", "remaining_time": "1:35:05", "throughput": 5163.85, "total_tokens": 13220864} {"current_steps": 13145, "total_steps": 42420, "loss": 0.3404, "lr": 0.0008717430548485219, "epoch": 6.197548326261198, "percentage": 30.99, "elapsed_time": "0:42:41", "remaining_time": "1:35:04", "throughput": 5164.07, "total_tokens": 13226720} {"current_steps": 13150, "total_steps": 42420, "loss": 0.3975, "lr": 0.000871605447793312, "epoch": 6.1999057048562, "percentage": 31.0, "elapsed_time": "0:42:42", "remaining_time": "1:35:03", "throughput": 5164.21, "total_tokens": 13231840} {"current_steps": 13155, "total_steps": 42420, "loss": 0.3766, "lr": 0.0008714677778316092, "epoch": 6.202263083451202, "percentage": 31.01, "elapsed_time": "0:42:43", "remaining_time": "1:35:01", "throughput": 5164.13, "total_tokens": 13235712} {"current_steps": 13160, "total_steps": 42420, "loss": 0.3651, "lr": 0.0008713300449867189, "epoch": 6.204620462046204, "percentage": 31.02, "elapsed_time": "0:42:44", "remaining_time": "1:35:01", "throughput": 5164.56, "total_tokens": 13243584} {"current_steps": 13165, "total_steps": 42420, "loss": 0.4329, "lr": 0.0008711922492819566, "epoch": 6.206977840641207, "percentage": 31.03, "elapsed_time": "0:42:45", "remaining_time": "1:35:00", "throughput": 5164.61, "total_tokens": 13248416} {"current_steps": 13170, "total_steps": 42420, "loss": 0.395, "lr": 0.0008710543907406492, "epoch": 6.209335219236209, "percentage": 31.05, "elapsed_time": "0:42:46", "remaining_time": "1:34:59", "throughput": 5164.76, "total_tokens": 13253504} {"current_steps": 13175, "total_steps": 42420, "loss": 0.3606, "lr": 0.0008709164693861334, "epoch": 6.211692597831211, "percentage": 31.06, "elapsed_time": "0:42:46", "remaining_time": "1:34:57", "throughput": 5164.79, "total_tokens": 13257888} {"current_steps": 13180, "total_steps": 42420, "loss": 0.3668, "lr": 0.0008707784852417573, "epoch": 6.214049976426214, "percentage": 31.07, "elapsed_time": "0:42:47", "remaining_time": "1:34:56", "throughput": 5164.78, "total_tokens": 13261984} {"current_steps": 13185, "total_steps": 42420, "loss": 0.3199, "lr": 0.0008706404383308789, "epoch": 6.216407355021216, "percentage": 31.08, "elapsed_time": "0:42:48", "remaining_time": "1:34:55", "throughput": 5165.03, "total_tokens": 13268288} {"current_steps": 13190, "total_steps": 42420, "loss": 0.3572, "lr": 0.0008705023286768676, "epoch": 6.218764733616219, "percentage": 31.09, "elapsed_time": "0:42:49", "remaining_time": "1:34:54", "throughput": 5165.11, "total_tokens": 13273248} {"current_steps": 13195, "total_steps": 42420, "loss": 0.3798, "lr": 0.0008703641563031027, "epoch": 6.221122112211221, "percentage": 31.11, "elapsed_time": "0:42:50", "remaining_time": "1:34:53", "throughput": 5165.29, "total_tokens": 13278816} {"current_steps": 13200, "total_steps": 42420, "loss": 0.3622, "lr": 0.0008702259212329746, "epoch": 6.223479490806223, "percentage": 31.12, "elapsed_time": "0:42:51", "remaining_time": "1:34:53", "throughput": 5165.51, "total_tokens": 13284896} {"current_steps": 13205, "total_steps": 42420, "loss": 0.3756, "lr": 0.0008700876234898843, "epoch": 6.225836869401226, "percentage": 31.13, "elapsed_time": "0:42:52", "remaining_time": "1:34:52", "throughput": 5165.7, "total_tokens": 13290720} {"current_steps": 13210, "total_steps": 42420, "loss": 0.3577, "lr": 0.0008699492630972427, "epoch": 6.228194247996228, "percentage": 31.14, "elapsed_time": "0:42:53", "remaining_time": "1:34:51", "throughput": 5165.76, "total_tokens": 13295488} {"current_steps": 13215, "total_steps": 42420, "loss": 0.3426, "lr": 0.0008698108400784725, "epoch": 6.230551626591231, "percentage": 31.15, "elapsed_time": "0:42:54", "remaining_time": "1:34:50", "throughput": 5166.01, "total_tokens": 13301344} {"current_steps": 13220, "total_steps": 42420, "loss": 0.3556, "lr": 0.0008696723544570061, "epoch": 6.232909005186233, "percentage": 31.16, "elapsed_time": "0:42:55", "remaining_time": "1:34:48", "throughput": 5165.98, "total_tokens": 13305568} {"current_steps": 13225, "total_steps": 42420, "loss": 0.3525, "lr": 0.0008695338062562868, "epoch": 6.235266383781235, "percentage": 31.18, "elapsed_time": "0:42:56", "remaining_time": "1:34:47", "throughput": 5166.0, "total_tokens": 13310048} {"current_steps": 13230, "total_steps": 42420, "loss": 0.3989, "lr": 0.0008693951954997683, "epoch": 6.237623762376238, "percentage": 31.19, "elapsed_time": "0:42:57", "remaining_time": "1:34:46", "throughput": 5166.04, "total_tokens": 13314752} {"current_steps": 13235, "total_steps": 42420, "loss": 0.3368, "lr": 0.0008692565222109152, "epoch": 6.23998114097124, "percentage": 31.2, "elapsed_time": "0:42:58", "remaining_time": "1:34:45", "throughput": 5166.09, "total_tokens": 13319488} {"current_steps": 13240, "total_steps": 42420, "loss": 0.3506, "lr": 0.0008691177864132024, "epoch": 6.242338519566243, "percentage": 31.21, "elapsed_time": "0:42:59", "remaining_time": "1:34:44", "throughput": 5166.2, "total_tokens": 13324640} {"current_steps": 13245, "total_steps": 42420, "loss": 0.2536, "lr": 0.0008689789881301159, "epoch": 6.244695898161245, "percentage": 31.22, "elapsed_time": "0:43:00", "remaining_time": "1:34:43", "throughput": 5166.23, "total_tokens": 13329376} {"current_steps": 13250, "total_steps": 42420, "loss": 0.3101, "lr": 0.0008688401273851514, "epoch": 6.247053276756247, "percentage": 31.24, "elapsed_time": "0:43:01", "remaining_time": "1:34:42", "throughput": 5166.28, "total_tokens": 13334176} {"current_steps": 13255, "total_steps": 42420, "loss": 0.4268, "lr": 0.0008687012042018161, "epoch": 6.24941065535125, "percentage": 31.25, "elapsed_time": "0:43:01", "remaining_time": "1:34:40", "throughput": 5166.3, "total_tokens": 13338496} {"current_steps": 13260, "total_steps": 42420, "loss": 0.3945, "lr": 0.0008685622186036269, "epoch": 6.251768033946251, "percentage": 31.26, "elapsed_time": "0:43:02", "remaining_time": "1:34:39", "throughput": 5166.37, "total_tokens": 13343200} {"current_steps": 13265, "total_steps": 42420, "loss": 0.3542, "lr": 0.000868423170614112, "epoch": 6.254125412541254, "percentage": 31.27, "elapsed_time": "0:43:03", "remaining_time": "1:34:38", "throughput": 5166.44, "total_tokens": 13347776} {"current_steps": 13270, "total_steps": 42420, "loss": 0.3431, "lr": 0.0008682840602568098, "epoch": 6.256482791136256, "percentage": 31.28, "elapsed_time": "0:43:04", "remaining_time": "1:34:37", "throughput": 5166.51, "total_tokens": 13352576} {"current_steps": 13275, "total_steps": 42420, "loss": 0.3658, "lr": 0.0008681448875552694, "epoch": 6.258840169731259, "percentage": 31.29, "elapsed_time": "0:43:05", "remaining_time": "1:34:35", "throughput": 5166.44, "total_tokens": 13356448} {"current_steps": 13280, "total_steps": 42420, "loss": 0.3493, "lr": 0.00086800565253305, "epoch": 6.261197548326261, "percentage": 31.31, "elapsed_time": "0:43:06", "remaining_time": "1:34:34", "throughput": 5166.52, "total_tokens": 13361216} {"current_steps": 13285, "total_steps": 42420, "loss": 0.3643, "lr": 0.0008678663552137223, "epoch": 6.263554926921263, "percentage": 31.32, "elapsed_time": "0:43:07", "remaining_time": "1:34:33", "throughput": 5166.59, "total_tokens": 13366208} {"current_steps": 13290, "total_steps": 42420, "loss": 0.3654, "lr": 0.0008677269956208664, "epoch": 6.265912305516266, "percentage": 31.33, "elapsed_time": "0:43:07", "remaining_time": "1:34:32", "throughput": 5166.54, "total_tokens": 13370272} {"current_steps": 13295, "total_steps": 42420, "loss": 0.3104, "lr": 0.0008675875737780741, "epoch": 6.268269684111268, "percentage": 31.34, "elapsed_time": "0:43:08", "remaining_time": "1:34:31", "throughput": 5166.62, "total_tokens": 13375040} {"current_steps": 13300, "total_steps": 42420, "loss": 0.3346, "lr": 0.0008674480897089465, "epoch": 6.270627062706271, "percentage": 31.35, "elapsed_time": "0:43:09", "remaining_time": "1:34:30", "throughput": 5166.69, "total_tokens": 13380000} {"current_steps": 13305, "total_steps": 42420, "loss": 0.3523, "lr": 0.0008673085434370964, "epoch": 6.272984441301273, "percentage": 31.36, "elapsed_time": "0:43:10", "remaining_time": "1:34:29", "throughput": 5166.9, "total_tokens": 13385824} {"current_steps": 13310, "total_steps": 42420, "loss": 0.3353, "lr": 0.0008671689349861465, "epoch": 6.275341819896275, "percentage": 31.38, "elapsed_time": "0:43:11", "remaining_time": "1:34:28", "throughput": 5166.98, "total_tokens": 13390816} {"current_steps": 13315, "total_steps": 42420, "loss": 0.2957, "lr": 0.0008670292643797302, "epoch": 6.277699198491278, "percentage": 31.39, "elapsed_time": "0:43:12", "remaining_time": "1:34:26", "throughput": 5167.1, "total_tokens": 13395776} {"current_steps": 13320, "total_steps": 42420, "loss": 0.333, "lr": 0.000866889531641491, "epoch": 6.28005657708628, "percentage": 31.4, "elapsed_time": "0:43:13", "remaining_time": "1:34:25", "throughput": 5167.17, "total_tokens": 13400512} {"current_steps": 13325, "total_steps": 42420, "loss": 0.3033, "lr": 0.0008667497367950836, "epoch": 6.282413955681283, "percentage": 31.41, "elapsed_time": "0:43:14", "remaining_time": "1:34:24", "throughput": 5167.27, "total_tokens": 13405472} {"current_steps": 13330, "total_steps": 42420, "loss": 0.3098, "lr": 0.0008666098798641728, "epoch": 6.284771334276285, "percentage": 31.42, "elapsed_time": "0:43:15", "remaining_time": "1:34:23", "throughput": 5167.4, "total_tokens": 13410720} {"current_steps": 13335, "total_steps": 42420, "loss": 0.385, "lr": 0.0008664699608724341, "epoch": 6.287128712871287, "percentage": 31.44, "elapsed_time": "0:43:16", "remaining_time": "1:34:22", "throughput": 5167.61, "total_tokens": 13416576} {"current_steps": 13340, "total_steps": 42420, "loss": 0.2839, "lr": 0.0008663299798435531, "epoch": 6.28948609146629, "percentage": 31.45, "elapsed_time": "0:43:17", "remaining_time": "1:34:21", "throughput": 5167.65, "total_tokens": 13420928} {"current_steps": 13345, "total_steps": 42420, "loss": 0.2929, "lr": 0.0008661899368012267, "epoch": 6.291843470061292, "percentage": 31.46, "elapsed_time": "0:43:18", "remaining_time": "1:34:20", "throughput": 5167.71, "total_tokens": 13425792} {"current_steps": 13350, "total_steps": 42420, "loss": 0.3165, "lr": 0.0008660498317691614, "epoch": 6.294200848656295, "percentage": 31.47, "elapsed_time": "0:43:19", "remaining_time": "1:34:19", "throughput": 5167.94, "total_tokens": 13431616} {"current_steps": 13355, "total_steps": 42420, "loss": 0.3117, "lr": 0.0008659096647710749, "epoch": 6.296558227251296, "percentage": 31.48, "elapsed_time": "0:43:19", "remaining_time": "1:34:18", "throughput": 5168.01, "total_tokens": 13436384} {"current_steps": 13360, "total_steps": 42420, "loss": 0.3409, "lr": 0.0008657694358306947, "epoch": 6.2989156058462985, "percentage": 31.49, "elapsed_time": "0:43:20", "remaining_time": "1:34:17", "throughput": 5168.08, "total_tokens": 13441152} {"current_steps": 13365, "total_steps": 42420, "loss": 0.2379, "lr": 0.0008656291449717597, "epoch": 6.301272984441301, "percentage": 31.51, "elapsed_time": "0:43:21", "remaining_time": "1:34:16", "throughput": 5168.25, "total_tokens": 13446592} {"current_steps": 13370, "total_steps": 42420, "loss": 0.3158, "lr": 0.0008654887922180184, "epoch": 6.303630363036303, "percentage": 31.52, "elapsed_time": "0:43:22", "remaining_time": "1:34:15", "throughput": 5168.45, "total_tokens": 13452448} {"current_steps": 13375, "total_steps": 42420, "loss": 0.3192, "lr": 0.0008653483775932301, "epoch": 6.305987741631306, "percentage": 31.53, "elapsed_time": "0:43:23", "remaining_time": "1:34:14", "throughput": 5168.64, "total_tokens": 13458208} {"current_steps": 13380, "total_steps": 42420, "loss": 0.3412, "lr": 0.0008652079011211649, "epoch": 6.308345120226308, "percentage": 31.54, "elapsed_time": "0:43:24", "remaining_time": "1:34:13", "throughput": 5168.7, "total_tokens": 13463040} {"current_steps": 13385, "total_steps": 42420, "loss": 0.364, "lr": 0.0008650673628256027, "epoch": 6.3107024988213105, "percentage": 31.55, "elapsed_time": "0:43:25", "remaining_time": "1:34:12", "throughput": 5168.77, "total_tokens": 13467872} {"current_steps": 13390, "total_steps": 42420, "loss": 0.3569, "lr": 0.0008649267627303347, "epoch": 6.313059877416313, "percentage": 31.57, "elapsed_time": "0:43:26", "remaining_time": "1:34:10", "throughput": 5168.81, "total_tokens": 13472384} {"current_steps": 13395, "total_steps": 42420, "loss": 0.3547, "lr": 0.0008647861008591617, "epoch": 6.315417256011315, "percentage": 31.58, "elapsed_time": "0:43:27", "remaining_time": "1:34:09", "throughput": 5168.92, "total_tokens": 13477376} {"current_steps": 13400, "total_steps": 42420, "loss": 0.3416, "lr": 0.0008646453772358958, "epoch": 6.317774634606318, "percentage": 31.59, "elapsed_time": "0:43:28", "remaining_time": "1:34:08", "throughput": 5168.9, "total_tokens": 13481376} {"current_steps": 13405, "total_steps": 42420, "loss": 0.304, "lr": 0.0008645045918843586, "epoch": 6.32013201320132, "percentage": 31.6, "elapsed_time": "0:43:29", "remaining_time": "1:34:07", "throughput": 5169.04, "total_tokens": 13486592} {"current_steps": 13410, "total_steps": 42420, "loss": 0.3421, "lr": 0.0008643637448283829, "epoch": 6.3224893917963225, "percentage": 31.61, "elapsed_time": "0:43:30", "remaining_time": "1:34:06", "throughput": 5169.28, "total_tokens": 13492576} {"current_steps": 13415, "total_steps": 42420, "loss": 0.3912, "lr": 0.0008642228360918119, "epoch": 6.324846770391325, "percentage": 31.62, "elapsed_time": "0:43:31", "remaining_time": "1:34:05", "throughput": 5169.3, "total_tokens": 13497216} {"current_steps": 13420, "total_steps": 42420, "loss": 0.3028, "lr": 0.0008640818656984989, "epoch": 6.327204148986327, "percentage": 31.64, "elapsed_time": "0:43:31", "remaining_time": "1:34:04", "throughput": 5169.31, "total_tokens": 13501600} {"current_steps": 13425, "total_steps": 42420, "loss": 0.2791, "lr": 0.0008639408336723078, "epoch": 6.32956152758133, "percentage": 31.65, "elapsed_time": "0:43:32", "remaining_time": "1:34:02", "throughput": 5169.38, "total_tokens": 13506400} {"current_steps": 13430, "total_steps": 42420, "loss": 0.3436, "lr": 0.0008637997400371129, "epoch": 6.331918906176332, "percentage": 31.66, "elapsed_time": "0:43:33", "remaining_time": "1:34:02", "throughput": 5169.47, "total_tokens": 13511680} {"current_steps": 13435, "total_steps": 42420, "loss": 0.3935, "lr": 0.0008636585848167988, "epoch": 6.3342762847713345, "percentage": 31.67, "elapsed_time": "0:43:34", "remaining_time": "1:34:00", "throughput": 5169.49, "total_tokens": 13515904} {"current_steps": 13440, "total_steps": 42420, "loss": 0.3681, "lr": 0.0008635173680352611, "epoch": 6.336633663366337, "percentage": 31.68, "elapsed_time": "0:43:35", "remaining_time": "1:33:59", "throughput": 5169.53, "total_tokens": 13520576} {"current_steps": 13445, "total_steps": 42420, "loss": 0.3339, "lr": 0.0008633760897164051, "epoch": 6.338991041961339, "percentage": 31.69, "elapsed_time": "0:43:36", "remaining_time": "1:33:58", "throughput": 5169.69, "total_tokens": 13525888} {"current_steps": 13450, "total_steps": 42420, "loss": 0.3571, "lr": 0.0008632347498841469, "epoch": 6.341348420556342, "percentage": 31.71, "elapsed_time": "0:43:37", "remaining_time": "1:33:57", "throughput": 5169.69, "total_tokens": 13530528} {"current_steps": 13455, "total_steps": 42420, "loss": 0.3402, "lr": 0.0008630933485624129, "epoch": 6.343705799151344, "percentage": 31.72, "elapsed_time": "0:43:38", "remaining_time": "1:33:56", "throughput": 5169.83, "total_tokens": 13535808} {"current_steps": 13460, "total_steps": 42420, "loss": 0.3257, "lr": 0.0008629518857751399, "epoch": 6.346063177746346, "percentage": 31.73, "elapsed_time": "0:43:39", "remaining_time": "1:33:55", "throughput": 5170.09, "total_tokens": 13541920} {"current_steps": 13465, "total_steps": 42420, "loss": 0.3374, "lr": 0.0008628103615462752, "epoch": 6.348420556341348, "percentage": 31.74, "elapsed_time": "0:43:40", "remaining_time": "1:33:54", "throughput": 5170.12, "total_tokens": 13546592} {"current_steps": 13470, "total_steps": 42420, "loss": 0.4147, "lr": 0.0008626687758997766, "epoch": 6.3507779349363505, "percentage": 31.75, "elapsed_time": "0:43:40", "remaining_time": "1:33:53", "throughput": 5170.08, "total_tokens": 13550720} {"current_steps": 13475, "total_steps": 42420, "loss": 0.3251, "lr": 0.000862527128859612, "epoch": 6.353135313531353, "percentage": 31.77, "elapsed_time": "0:43:41", "remaining_time": "1:33:51", "throughput": 5170.08, "total_tokens": 13555168} {"current_steps": 13480, "total_steps": 42420, "loss": 0.3419, "lr": 0.0008623854204497597, "epoch": 6.355492692126355, "percentage": 31.78, "elapsed_time": "0:43:42", "remaining_time": "1:33:50", "throughput": 5170.07, "total_tokens": 13559488} {"current_steps": 13485, "total_steps": 42420, "loss": 0.3457, "lr": 0.0008622436506942088, "epoch": 6.357850070721358, "percentage": 31.79, "elapsed_time": "0:43:43", "remaining_time": "1:33:49", "throughput": 5170.31, "total_tokens": 13565920} {"current_steps": 13490, "total_steps": 42420, "loss": 0.3555, "lr": 0.0008621018196169582, "epoch": 6.36020744931636, "percentage": 31.8, "elapsed_time": "0:43:44", "remaining_time": "1:33:49", "throughput": 5170.71, "total_tokens": 13573088} {"current_steps": 13495, "total_steps": 42420, "loss": 0.3904, "lr": 0.0008619599272420178, "epoch": 6.3625648279113625, "percentage": 31.81, "elapsed_time": "0:43:45", "remaining_time": "1:33:48", "throughput": 5170.8, "total_tokens": 13578304} {"current_steps": 13500, "total_steps": 42420, "loss": 0.3226, "lr": 0.0008618179735934076, "epoch": 6.364922206506365, "percentage": 31.82, "elapsed_time": "0:43:46", "remaining_time": "1:33:47", "throughput": 5171.04, "total_tokens": 13584288} {"current_steps": 13505, "total_steps": 42420, "loss": 0.2818, "lr": 0.0008616759586951574, "epoch": 6.367279585101367, "percentage": 31.84, "elapsed_time": "0:43:47", "remaining_time": "1:33:46", "throughput": 5171.08, "total_tokens": 13588896} {"current_steps": 13510, "total_steps": 42420, "loss": 0.3896, "lr": 0.0008615338825713086, "epoch": 6.36963696369637, "percentage": 31.85, "elapsed_time": "0:43:48", "remaining_time": "1:33:45", "throughput": 5171.0, "total_tokens": 13592800} {"current_steps": 13515, "total_steps": 42420, "loss": 0.341, "lr": 0.000861391745245912, "epoch": 6.371994342291372, "percentage": 31.86, "elapsed_time": "0:43:49", "remaining_time": "1:33:43", "throughput": 5171.13, "total_tokens": 13597760} {"current_steps": 13520, "total_steps": 42420, "loss": 0.3323, "lr": 0.0008612495467430289, "epoch": 6.3743517208863745, "percentage": 31.87, "elapsed_time": "0:43:50", "remaining_time": "1:33:43", "throughput": 5171.43, "total_tokens": 13604032} {"current_steps": 13525, "total_steps": 42420, "loss": 0.3439, "lr": 0.0008611072870867313, "epoch": 6.376709099481377, "percentage": 31.88, "elapsed_time": "0:43:51", "remaining_time": "1:33:42", "throughput": 5171.52, "total_tokens": 13609088} {"current_steps": 13530, "total_steps": 42420, "loss": 0.3484, "lr": 0.0008609649663011011, "epoch": 6.379066478076379, "percentage": 31.9, "elapsed_time": "0:43:52", "remaining_time": "1:33:41", "throughput": 5171.73, "total_tokens": 13614816} {"current_steps": 13535, "total_steps": 42420, "loss": 0.3045, "lr": 0.0008608225844102311, "epoch": 6.381423856671382, "percentage": 31.91, "elapsed_time": "0:43:53", "remaining_time": "1:33:40", "throughput": 5171.86, "total_tokens": 13620032} {"current_steps": 13540, "total_steps": 42420, "loss": 0.3367, "lr": 0.0008606801414382239, "epoch": 6.383781235266384, "percentage": 31.92, "elapsed_time": "0:43:54", "remaining_time": "1:33:38", "throughput": 5171.9, "total_tokens": 13624512} {"current_steps": 13545, "total_steps": 42420, "loss": 0.31, "lr": 0.0008605376374091928, "epoch": 6.3861386138613865, "percentage": 31.93, "elapsed_time": "0:43:55", "remaining_time": "1:33:37", "throughput": 5172.06, "total_tokens": 13629984} {"current_steps": 13550, "total_steps": 42420, "loss": 0.3641, "lr": 0.0008603950723472611, "epoch": 6.388495992456389, "percentage": 31.94, "elapsed_time": "0:43:56", "remaining_time": "1:33:36", "throughput": 5172.2, "total_tokens": 13635520} {"current_steps": 13555, "total_steps": 42420, "loss": 0.2923, "lr": 0.0008602524462765631, "epoch": 6.39085337105139, "percentage": 31.95, "elapsed_time": "0:43:57", "remaining_time": "1:33:35", "throughput": 5172.3, "total_tokens": 13640640} {"current_steps": 13560, "total_steps": 42420, "loss": 0.4058, "lr": 0.0008601097592212425, "epoch": 6.393210749646393, "percentage": 31.97, "elapsed_time": "0:43:58", "remaining_time": "1:33:34", "throughput": 5172.43, "total_tokens": 13646048} {"current_steps": 13565, "total_steps": 42420, "loss": 0.3437, "lr": 0.000859967011205454, "epoch": 6.395568128241395, "percentage": 31.98, "elapsed_time": "0:43:59", "remaining_time": "1:33:34", "throughput": 5172.65, "total_tokens": 13652064} {"current_steps": 13570, "total_steps": 42420, "loss": 0.3609, "lr": 0.0008598242022533623, "epoch": 6.397925506836398, "percentage": 31.99, "elapsed_time": "0:44:00", "remaining_time": "1:33:33", "throughput": 5172.76, "total_tokens": 13657280} {"current_steps": 13575, "total_steps": 42420, "loss": 0.359, "lr": 0.0008596813323891428, "epoch": 6.4002828854314, "percentage": 32.0, "elapsed_time": "0:44:01", "remaining_time": "1:33:32", "throughput": 5172.9, "total_tokens": 13662528} {"current_steps": 13580, "total_steps": 42420, "loss": 0.3454, "lr": 0.0008595384016369807, "epoch": 6.402640264026402, "percentage": 32.01, "elapsed_time": "0:44:02", "remaining_time": "1:33:31", "throughput": 5173.02, "total_tokens": 13667616} {"current_steps": 13585, "total_steps": 42420, "loss": 0.3143, "lr": 0.0008593954100210717, "epoch": 6.404997642621405, "percentage": 32.02, "elapsed_time": "0:44:02", "remaining_time": "1:33:29", "throughput": 5173.02, "total_tokens": 13672000} {"current_steps": 13590, "total_steps": 42420, "loss": 0.4202, "lr": 0.0008592523575656221, "epoch": 6.407355021216407, "percentage": 32.04, "elapsed_time": "0:44:03", "remaining_time": "1:33:28", "throughput": 5173.11, "total_tokens": 13677152} {"current_steps": 13595, "total_steps": 42420, "loss": 0.3369, "lr": 0.0008591092442948482, "epoch": 6.40971239981141, "percentage": 32.05, "elapsed_time": "0:44:04", "remaining_time": "1:33:27", "throughput": 5173.28, "total_tokens": 13682720} {"current_steps": 13600, "total_steps": 42420, "loss": 0.3124, "lr": 0.0008589660702329765, "epoch": 6.412069778406412, "percentage": 32.06, "elapsed_time": "0:44:05", "remaining_time": "1:33:26", "throughput": 5173.35, "total_tokens": 13687680} {"current_steps": 13605, "total_steps": 42420, "loss": 0.3042, "lr": 0.0008588228354042439, "epoch": 6.414427157001414, "percentage": 32.07, "elapsed_time": "0:44:06", "remaining_time": "1:33:25", "throughput": 5173.43, "total_tokens": 13692448} {"current_steps": 13610, "total_steps": 42420, "loss": 0.3743, "lr": 0.0008586795398328978, "epoch": 6.416784535596417, "percentage": 32.08, "elapsed_time": "0:44:07", "remaining_time": "1:33:24", "throughput": 5173.6, "total_tokens": 13698464} {"current_steps": 13615, "total_steps": 42420, "loss": 0.3649, "lr": 0.0008585361835431956, "epoch": 6.419141914191419, "percentage": 32.1, "elapsed_time": "0:44:08", "remaining_time": "1:33:24", "throughput": 5173.87, "total_tokens": 13705248} {"current_steps": 13620, "total_steps": 42420, "loss": 0.3689, "lr": 0.0008583927665594052, "epoch": 6.421499292786422, "percentage": 32.11, "elapsed_time": "0:44:09", "remaining_time": "1:33:23", "throughput": 5174.05, "total_tokens": 13710784} {"current_steps": 13625, "total_steps": 42420, "loss": 0.3499, "lr": 0.0008582492889058046, "epoch": 6.423856671381424, "percentage": 32.12, "elapsed_time": "0:44:10", "remaining_time": "1:33:22", "throughput": 5174.05, "total_tokens": 13715104} {"current_steps": 13630, "total_steps": 42420, "loss": 0.3147, "lr": 0.000858105750606682, "epoch": 6.426214049976426, "percentage": 32.13, "elapsed_time": "0:44:11", "remaining_time": "1:33:20", "throughput": 5174.16, "total_tokens": 13720096} {"current_steps": 13635, "total_steps": 42420, "loss": 0.3587, "lr": 0.0008579621516863362, "epoch": 6.428571428571429, "percentage": 32.14, "elapsed_time": "0:44:12", "remaining_time": "1:33:19", "throughput": 5174.23, "total_tokens": 13724992} {"current_steps": 13640, "total_steps": 42420, "loss": 0.3475, "lr": 0.0008578184921690759, "epoch": 6.430928807166431, "percentage": 32.15, "elapsed_time": "0:44:13", "remaining_time": "1:33:18", "throughput": 5174.28, "total_tokens": 13729856} {"current_steps": 13645, "total_steps": 42420, "loss": 0.2915, "lr": 0.0008576747720792204, "epoch": 6.433286185761434, "percentage": 32.17, "elapsed_time": "0:44:14", "remaining_time": "1:33:17", "throughput": 5174.49, "total_tokens": 13735712} {"current_steps": 13650, "total_steps": 42420, "loss": 0.2432, "lr": 0.0008575309914410989, "epoch": 6.435643564356436, "percentage": 32.18, "elapsed_time": "0:44:15", "remaining_time": "1:33:16", "throughput": 5174.56, "total_tokens": 13740288} {"current_steps": 13655, "total_steps": 42420, "loss": 0.3131, "lr": 0.0008573871502790511, "epoch": 6.438000942951438, "percentage": 32.19, "elapsed_time": "0:44:16", "remaining_time": "1:33:15", "throughput": 5174.63, "total_tokens": 13745152} {"current_steps": 13660, "total_steps": 42420, "loss": 0.4377, "lr": 0.0008572432486174268, "epoch": 6.44035832154644, "percentage": 32.2, "elapsed_time": "0:44:17", "remaining_time": "1:33:14", "throughput": 5174.66, "total_tokens": 13749728} {"current_steps": 13665, "total_steps": 42420, "loss": 0.3106, "lr": 0.0008570992864805862, "epoch": 6.442715700141442, "percentage": 32.21, "elapsed_time": "0:44:18", "remaining_time": "1:33:13", "throughput": 5174.81, "total_tokens": 13755392} {"current_steps": 13670, "total_steps": 42420, "loss": 0.3207, "lr": 0.0008569552638928996, "epoch": 6.445073078736445, "percentage": 32.23, "elapsed_time": "0:44:19", "remaining_time": "1:33:12", "throughput": 5174.9, "total_tokens": 13760352} {"current_steps": 13675, "total_steps": 42420, "loss": 0.3353, "lr": 0.0008568111808787475, "epoch": 6.447430457331447, "percentage": 32.24, "elapsed_time": "0:44:19", "remaining_time": "1:33:11", "throughput": 5175.02, "total_tokens": 13765248} {"current_steps": 13680, "total_steps": 42420, "loss": 0.3256, "lr": 0.0008566670374625208, "epoch": 6.4497878359264496, "percentage": 32.25, "elapsed_time": "0:44:20", "remaining_time": "1:33:09", "throughput": 5175.0, "total_tokens": 13769440} {"current_steps": 13685, "total_steps": 42420, "loss": 0.2946, "lr": 0.0008565228336686207, "epoch": 6.452145214521452, "percentage": 32.26, "elapsed_time": "0:44:21", "remaining_time": "1:33:08", "throughput": 5175.01, "total_tokens": 13773888} {"current_steps": 13690, "total_steps": 42420, "loss": 0.2596, "lr": 0.0008563785695214581, "epoch": 6.454502593116454, "percentage": 32.27, "elapsed_time": "0:44:22", "remaining_time": "1:33:07", "throughput": 5175.12, "total_tokens": 13778784} {"current_steps": 13695, "total_steps": 42420, "loss": 0.3962, "lr": 0.0008562342450454545, "epoch": 6.456859971711457, "percentage": 32.28, "elapsed_time": "0:44:23", "remaining_time": "1:33:06", "throughput": 5175.26, "total_tokens": 13784096} {"current_steps": 13700, "total_steps": 42420, "loss": 0.3489, "lr": 0.000856089860265042, "epoch": 6.459217350306459, "percentage": 32.3, "elapsed_time": "0:44:24", "remaining_time": "1:33:05", "throughput": 5175.52, "total_tokens": 13790176} {"current_steps": 13705, "total_steps": 42420, "loss": 0.3693, "lr": 0.0008559454152046622, "epoch": 6.461574728901462, "percentage": 32.31, "elapsed_time": "0:44:25", "remaining_time": "1:33:04", "throughput": 5175.66, "total_tokens": 13795616} {"current_steps": 13710, "total_steps": 42420, "loss": 0.3531, "lr": 0.0008558009098887672, "epoch": 6.463932107496464, "percentage": 32.32, "elapsed_time": "0:44:26", "remaining_time": "1:33:03", "throughput": 5175.75, "total_tokens": 13800640} {"current_steps": 13715, "total_steps": 42420, "loss": 0.3632, "lr": 0.0008556563443418192, "epoch": 6.466289486091466, "percentage": 32.33, "elapsed_time": "0:44:27", "remaining_time": "1:33:02", "throughput": 5175.87, "total_tokens": 13805632} {"current_steps": 13720, "total_steps": 42420, "loss": 0.297, "lr": 0.0008555117185882907, "epoch": 6.468646864686469, "percentage": 32.34, "elapsed_time": "0:44:28", "remaining_time": "1:33:01", "throughput": 5175.91, "total_tokens": 13810272} {"current_steps": 13725, "total_steps": 42420, "loss": 0.2976, "lr": 0.0008553670326526647, "epoch": 6.471004243281471, "percentage": 32.36, "elapsed_time": "0:44:29", "remaining_time": "1:33:00", "throughput": 5176.01, "total_tokens": 13815360} {"current_steps": 13730, "total_steps": 42420, "loss": 0.478, "lr": 0.0008552222865594338, "epoch": 6.473361621876474, "percentage": 32.37, "elapsed_time": "0:44:29", "remaining_time": "1:32:59", "throughput": 5176.0, "total_tokens": 13819808} {"current_steps": 13735, "total_steps": 42420, "loss": 0.3511, "lr": 0.0008550774803331012, "epoch": 6.475719000471476, "percentage": 32.38, "elapsed_time": "0:44:30", "remaining_time": "1:32:58", "throughput": 5176.14, "total_tokens": 13824992} {"current_steps": 13740, "total_steps": 42420, "loss": 0.3263, "lr": 0.0008549326139981799, "epoch": 6.478076379066478, "percentage": 32.39, "elapsed_time": "0:44:31", "remaining_time": "1:32:56", "throughput": 5176.18, "total_tokens": 13829472} {"current_steps": 13745, "total_steps": 42420, "loss": 0.3833, "lr": 0.0008547876875791935, "epoch": 6.480433757661481, "percentage": 32.4, "elapsed_time": "0:44:32", "remaining_time": "1:32:56", "throughput": 5176.51, "total_tokens": 13836576} {"current_steps": 13750, "total_steps": 42420, "loss": 0.3676, "lr": 0.0008546427011006756, "epoch": 6.482791136256483, "percentage": 32.41, "elapsed_time": "0:44:33", "remaining_time": "1:32:55", "throughput": 5176.58, "total_tokens": 13841184} {"current_steps": 13755, "total_steps": 42420, "loss": 0.378, "lr": 0.0008544976545871699, "epoch": 6.485148514851485, "percentage": 32.43, "elapsed_time": "0:44:34", "remaining_time": "1:32:54", "throughput": 5176.69, "total_tokens": 13846304} {"current_steps": 13760, "total_steps": 42420, "loss": 0.3389, "lr": 0.0008543525480632304, "epoch": 6.487505893446487, "percentage": 32.44, "elapsed_time": "0:44:35", "remaining_time": "1:32:53", "throughput": 5176.81, "total_tokens": 13851456} {"current_steps": 13765, "total_steps": 42420, "loss": 0.3429, "lr": 0.0008542073815534209, "epoch": 6.4898632720414895, "percentage": 32.45, "elapsed_time": "0:44:36", "remaining_time": "1:32:52", "throughput": 5176.91, "total_tokens": 13856672} {"current_steps": 13770, "total_steps": 42420, "loss": 0.3897, "lr": 0.0008540621550823157, "epoch": 6.492220650636492, "percentage": 32.46, "elapsed_time": "0:44:37", "remaining_time": "1:32:51", "throughput": 5177.12, "total_tokens": 13862304} {"current_steps": 13775, "total_steps": 42420, "loss": 0.3779, "lr": 0.0008539168686744995, "epoch": 6.494578029231494, "percentage": 32.47, "elapsed_time": "0:44:38", "remaining_time": "1:32:49", "throughput": 5177.19, "total_tokens": 13867200} {"current_steps": 13780, "total_steps": 42420, "loss": 0.3728, "lr": 0.0008537715223545665, "epoch": 6.496935407826497, "percentage": 32.48, "elapsed_time": "0:44:39", "remaining_time": "1:32:48", "throughput": 5177.37, "total_tokens": 13872576} {"current_steps": 13785, "total_steps": 42420, "loss": 0.3701, "lr": 0.0008536261161471216, "epoch": 6.499292786421499, "percentage": 32.5, "elapsed_time": "0:44:40", "remaining_time": "1:32:47", "throughput": 5177.42, "total_tokens": 13877440} {"current_steps": 13790, "total_steps": 42420, "loss": 0.3572, "lr": 0.0008534806500767793, "epoch": 6.5016501650165015, "percentage": 32.51, "elapsed_time": "0:44:41", "remaining_time": "1:32:46", "throughput": 5177.63, "total_tokens": 13883200} {"current_steps": 13795, "total_steps": 42420, "loss": 0.3838, "lr": 0.0008533351241681649, "epoch": 6.504007543611504, "percentage": 32.52, "elapsed_time": "0:44:42", "remaining_time": "1:32:45", "throughput": 5177.8, "total_tokens": 13888640} {"current_steps": 13800, "total_steps": 42420, "loss": 0.4231, "lr": 0.0008531895384459129, "epoch": 6.506364922206506, "percentage": 32.53, "elapsed_time": "0:44:43", "remaining_time": "1:32:44", "throughput": 5177.82, "total_tokens": 13893280} {"current_steps": 13805, "total_steps": 42420, "loss": 0.3011, "lr": 0.0008530438929346692, "epoch": 6.508722300801509, "percentage": 32.54, "elapsed_time": "0:44:44", "remaining_time": "1:32:43", "throughput": 5178.05, "total_tokens": 13899296} {"current_steps": 13810, "total_steps": 42420, "loss": 0.3388, "lr": 0.0008528981876590886, "epoch": 6.511079679396511, "percentage": 32.56, "elapsed_time": "0:44:45", "remaining_time": "1:32:42", "throughput": 5178.13, "total_tokens": 13904416} {"current_steps": 13815, "total_steps": 42420, "loss": 0.3316, "lr": 0.0008527524226438368, "epoch": 6.5134370579915135, "percentage": 32.57, "elapsed_time": "0:44:46", "remaining_time": "1:32:41", "throughput": 5178.16, "total_tokens": 13909088} {"current_steps": 13820, "total_steps": 42420, "loss": 0.3117, "lr": 0.0008526065979135893, "epoch": 6.515794436586516, "percentage": 32.58, "elapsed_time": "0:44:47", "remaining_time": "1:32:40", "throughput": 5178.28, "total_tokens": 13914048} {"current_steps": 13825, "total_steps": 42420, "loss": 0.3927, "lr": 0.0008524607134930316, "epoch": 6.518151815181518, "percentage": 32.59, "elapsed_time": "0:44:47", "remaining_time": "1:32:39", "throughput": 5178.2, "total_tokens": 13917856} {"current_steps": 13830, "total_steps": 42420, "loss": 0.3373, "lr": 0.0008523147694068595, "epoch": 6.520509193776521, "percentage": 32.6, "elapsed_time": "0:44:48", "remaining_time": "1:32:38", "throughput": 5178.29, "total_tokens": 13922848} {"current_steps": 13835, "total_steps": 42420, "loss": 0.3492, "lr": 0.000852168765679779, "epoch": 6.522866572371523, "percentage": 32.61, "elapsed_time": "0:44:49", "remaining_time": "1:32:37", "throughput": 5178.53, "total_tokens": 13928992} {"current_steps": 13840, "total_steps": 42420, "loss": 0.3691, "lr": 0.0008520227023365058, "epoch": 6.5252239509665255, "percentage": 32.63, "elapsed_time": "0:44:50", "remaining_time": "1:32:36", "throughput": 5178.6, "total_tokens": 13933632} {"current_steps": 13845, "total_steps": 42420, "loss": 0.3109, "lr": 0.0008518765794017663, "epoch": 6.527581329561528, "percentage": 32.64, "elapsed_time": "0:44:51", "remaining_time": "1:32:34", "throughput": 5178.61, "total_tokens": 13937920} {"current_steps": 13850, "total_steps": 42420, "loss": 0.2918, "lr": 0.0008517303969002964, "epoch": 6.52993870815653, "percentage": 32.65, "elapsed_time": "0:44:52", "remaining_time": "1:32:33", "throughput": 5178.76, "total_tokens": 13943392} {"current_steps": 13855, "total_steps": 42420, "loss": 0.3372, "lr": 0.0008515841548568421, "epoch": 6.532296086751533, "percentage": 32.66, "elapsed_time": "0:44:53", "remaining_time": "1:32:33", "throughput": 5178.92, "total_tokens": 13948896} {"current_steps": 13860, "total_steps": 42420, "loss": 0.2911, "lr": 0.0008514378532961602, "epoch": 6.534653465346535, "percentage": 32.67, "elapsed_time": "0:44:54", "remaining_time": "1:32:32", "throughput": 5179.09, "total_tokens": 13954336} {"current_steps": 13865, "total_steps": 42420, "loss": 0.3981, "lr": 0.0008512914922430168, "epoch": 6.537010843941537, "percentage": 32.69, "elapsed_time": "0:44:55", "remaining_time": "1:32:31", "throughput": 5179.27, "total_tokens": 13959808} {"current_steps": 13870, "total_steps": 42420, "loss": 0.3539, "lr": 0.0008511450717221883, "epoch": 6.539368222536539, "percentage": 32.7, "elapsed_time": "0:44:56", "remaining_time": "1:32:30", "throughput": 5179.48, "total_tokens": 13965536} {"current_steps": 13875, "total_steps": 42420, "loss": 0.3643, "lr": 0.0008509985917584612, "epoch": 6.5417256011315414, "percentage": 32.71, "elapsed_time": "0:44:57", "remaining_time": "1:32:29", "throughput": 5179.66, "total_tokens": 13971104} {"current_steps": 13880, "total_steps": 42420, "loss": 0.3643, "lr": 0.0008508520523766321, "epoch": 6.544082979726544, "percentage": 32.72, "elapsed_time": "0:44:58", "remaining_time": "1:32:28", "throughput": 5179.76, "total_tokens": 13976320} {"current_steps": 13885, "total_steps": 42420, "loss": 0.3568, "lr": 0.0008507054536015078, "epoch": 6.546440358321546, "percentage": 32.73, "elapsed_time": "0:44:59", "remaining_time": "1:32:27", "throughput": 5179.85, "total_tokens": 13981216} {"current_steps": 13890, "total_steps": 42420, "loss": 0.3417, "lr": 0.0008505587954579048, "epoch": 6.548797736916549, "percentage": 32.74, "elapsed_time": "0:45:00", "remaining_time": "1:32:25", "throughput": 5179.9, "total_tokens": 13985888} {"current_steps": 13895, "total_steps": 42420, "loss": 0.3318, "lr": 0.0008504120779706499, "epoch": 6.551155115511551, "percentage": 32.76, "elapsed_time": "0:45:01", "remaining_time": "1:32:24", "throughput": 5180.08, "total_tokens": 13991456} {"current_steps": 13900, "total_steps": 42420, "loss": 0.4134, "lr": 0.0008502653011645798, "epoch": 6.5535124941065535, "percentage": 32.77, "elapsed_time": "0:45:01", "remaining_time": "1:32:23", "throughput": 5180.18, "total_tokens": 13996544} {"current_steps": 13905, "total_steps": 42420, "loss": 0.3518, "lr": 0.0008501184650645414, "epoch": 6.555869872701556, "percentage": 32.78, "elapsed_time": "0:45:02", "remaining_time": "1:32:22", "throughput": 5180.35, "total_tokens": 14002208} {"current_steps": 13910, "total_steps": 42420, "loss": 0.3344, "lr": 0.0008499715696953916, "epoch": 6.558227251296558, "percentage": 32.79, "elapsed_time": "0:45:03", "remaining_time": "1:32:21", "throughput": 5180.5, "total_tokens": 14007424} {"current_steps": 13915, "total_steps": 42420, "loss": 0.3179, "lr": 0.0008498246150819971, "epoch": 6.560584629891561, "percentage": 32.8, "elapsed_time": "0:45:04", "remaining_time": "1:32:21", "throughput": 5180.72, "total_tokens": 14013440} {"current_steps": 13920, "total_steps": 42420, "loss": 0.427, "lr": 0.000849677601249235, "epoch": 6.562942008486563, "percentage": 32.81, "elapsed_time": "0:45:05", "remaining_time": "1:32:20", "throughput": 5180.82, "total_tokens": 14018560} {"current_steps": 13925, "total_steps": 42420, "loss": 0.2913, "lr": 0.0008495305282219923, "epoch": 6.5652993870815655, "percentage": 32.83, "elapsed_time": "0:45:06", "remaining_time": "1:32:19", "throughput": 5181.11, "total_tokens": 14025152} {"current_steps": 13930, "total_steps": 42420, "loss": 0.2797, "lr": 0.0008493833960251658, "epoch": 6.567656765676568, "percentage": 32.84, "elapsed_time": "0:45:08", "remaining_time": "1:32:18", "throughput": 5181.42, "total_tokens": 14031776} {"current_steps": 13935, "total_steps": 42420, "loss": 0.3418, "lr": 0.0008492362046836623, "epoch": 6.57001414427157, "percentage": 32.85, "elapsed_time": "0:45:09", "remaining_time": "1:32:17", "throughput": 5181.61, "total_tokens": 14037440} {"current_steps": 13940, "total_steps": 42420, "loss": 0.3834, "lr": 0.0008490889542223994, "epoch": 6.572371522866573, "percentage": 32.86, "elapsed_time": "0:45:10", "remaining_time": "1:32:17", "throughput": 5181.88, "total_tokens": 14044448} {"current_steps": 13945, "total_steps": 42420, "loss": 0.2997, "lr": 0.0008489416446663036, "epoch": 6.574728901461575, "percentage": 32.87, "elapsed_time": "0:45:11", "remaining_time": "1:32:16", "throughput": 5181.93, "total_tokens": 14049120} {"current_steps": 13950, "total_steps": 42420, "loss": 0.3596, "lr": 0.000848794276040312, "epoch": 6.5770862800565775, "percentage": 32.89, "elapsed_time": "0:45:12", "remaining_time": "1:32:14", "throughput": 5182.0, "total_tokens": 14054016} {"current_steps": 13955, "total_steps": 42420, "loss": 0.3236, "lr": 0.0008486468483693717, "epoch": 6.579443658651579, "percentage": 32.9, "elapsed_time": "0:45:13", "remaining_time": "1:32:13", "throughput": 5182.12, "total_tokens": 14059136} {"current_steps": 13960, "total_steps": 42420, "loss": 0.2916, "lr": 0.0008484993616784398, "epoch": 6.581801037246581, "percentage": 32.91, "elapsed_time": "0:45:14", "remaining_time": "1:32:13", "throughput": 5182.31, "total_tokens": 14064960} {"current_steps": 13965, "total_steps": 42420, "loss": 0.3767, "lr": 0.000848351815992483, "epoch": 6.584158415841584, "percentage": 32.92, "elapsed_time": "0:45:14", "remaining_time": "1:32:11", "throughput": 5182.39, "total_tokens": 14069792} {"current_steps": 13970, "total_steps": 42420, "loss": 0.3093, "lr": 0.0008482042113364786, "epoch": 6.586515794436586, "percentage": 32.93, "elapsed_time": "0:45:15", "remaining_time": "1:32:10", "throughput": 5182.5, "total_tokens": 14074848} {"current_steps": 13975, "total_steps": 42420, "loss": 0.3597, "lr": 0.0008480565477354132, "epoch": 6.588873173031589, "percentage": 32.94, "elapsed_time": "0:45:16", "remaining_time": "1:32:09", "throughput": 5182.54, "total_tokens": 14079456} {"current_steps": 13980, "total_steps": 42420, "loss": 0.3244, "lr": 0.0008479088252142839, "epoch": 6.591230551626591, "percentage": 32.96, "elapsed_time": "0:45:17", "remaining_time": "1:32:09", "throughput": 5182.82, "total_tokens": 14086112} {"current_steps": 13985, "total_steps": 42420, "loss": 0.3573, "lr": 0.0008477610437980977, "epoch": 6.593587930221593, "percentage": 32.97, "elapsed_time": "0:45:18", "remaining_time": "1:32:07", "throughput": 5182.92, "total_tokens": 14091008} {"current_steps": 13990, "total_steps": 42420, "loss": 0.3417, "lr": 0.0008476132035118714, "epoch": 6.595945308816596, "percentage": 32.98, "elapsed_time": "0:45:19", "remaining_time": "1:32:07", "throughput": 5183.1, "total_tokens": 14096992} {"current_steps": 13995, "total_steps": 42420, "loss": 0.3455, "lr": 0.000847465304380632, "epoch": 6.598302687411598, "percentage": 32.99, "elapsed_time": "0:45:20", "remaining_time": "1:32:06", "throughput": 5183.32, "total_tokens": 14103072} {"current_steps": 14000, "total_steps": 42420, "loss": 0.3615, "lr": 0.0008473173464294162, "epoch": 6.600660066006601, "percentage": 33.0, "elapsed_time": "0:45:21", "remaining_time": "1:32:05", "throughput": 5183.32, "total_tokens": 14107552} {"current_steps": 14005, "total_steps": 42420, "loss": 0.3763, "lr": 0.0008471693296832705, "epoch": 6.603017444601603, "percentage": 33.02, "elapsed_time": "0:45:22", "remaining_time": "1:32:04", "throughput": 5183.52, "total_tokens": 14113152} {"current_steps": 14010, "total_steps": 42420, "loss": 0.3751, "lr": 0.0008470212541672521, "epoch": 6.605374823196605, "percentage": 33.03, "elapsed_time": "0:45:23", "remaining_time": "1:32:03", "throughput": 5183.7, "total_tokens": 14118752} {"current_steps": 14015, "total_steps": 42420, "loss": 0.3635, "lr": 0.0008468731199064275, "epoch": 6.607732201791608, "percentage": 33.04, "elapsed_time": "0:45:24", "remaining_time": "1:32:02", "throughput": 5183.8, "total_tokens": 14123808} {"current_steps": 14020, "total_steps": 42420, "loss": 0.332, "lr": 0.0008467249269258732, "epoch": 6.61008958038661, "percentage": 33.05, "elapsed_time": "0:45:25", "remaining_time": "1:32:01", "throughput": 5183.98, "total_tokens": 14129632} {"current_steps": 14025, "total_steps": 42420, "loss": 0.3755, "lr": 0.0008465766752506758, "epoch": 6.612446958981613, "percentage": 33.06, "elapsed_time": "0:45:26", "remaining_time": "1:31:59", "throughput": 5183.9, "total_tokens": 14133632} {"current_steps": 14030, "total_steps": 42420, "loss": 0.3271, "lr": 0.0008464283649059318, "epoch": 6.614804337576615, "percentage": 33.07, "elapsed_time": "0:45:27", "remaining_time": "1:31:59", "throughput": 5184.05, "total_tokens": 14139456} {"current_steps": 14035, "total_steps": 42420, "loss": 0.375, "lr": 0.0008462799959167477, "epoch": 6.617161716171617, "percentage": 33.09, "elapsed_time": "0:45:28", "remaining_time": "1:31:57", "throughput": 5184.05, "total_tokens": 14143904} {"current_steps": 14040, "total_steps": 42420, "loss": 0.351, "lr": 0.0008461315683082397, "epoch": 6.61951909476662, "percentage": 33.1, "elapsed_time": "0:45:29", "remaining_time": "1:31:56", "throughput": 5184.06, "total_tokens": 14148544} {"current_steps": 14045, "total_steps": 42420, "loss": 0.3559, "lr": 0.0008459830821055343, "epoch": 6.621876473361622, "percentage": 33.11, "elapsed_time": "0:45:30", "remaining_time": "1:31:55", "throughput": 5184.12, "total_tokens": 14153568} {"current_steps": 14050, "total_steps": 42420, "loss": 0.35, "lr": 0.0008458345373337675, "epoch": 6.624233851956625, "percentage": 33.12, "elapsed_time": "0:45:30", "remaining_time": "1:31:54", "throughput": 5184.05, "total_tokens": 14157408} {"current_steps": 14055, "total_steps": 42420, "loss": 0.3494, "lr": 0.0008456859340180856, "epoch": 6.626591230551627, "percentage": 33.13, "elapsed_time": "0:45:31", "remaining_time": "1:31:53", "throughput": 5184.1, "total_tokens": 14162048} {"current_steps": 14060, "total_steps": 42420, "loss": 0.2703, "lr": 0.0008455372721836443, "epoch": 6.628948609146629, "percentage": 33.14, "elapsed_time": "0:45:32", "remaining_time": "1:31:52", "throughput": 5184.21, "total_tokens": 14167200} {"current_steps": 14065, "total_steps": 42420, "loss": 0.3896, "lr": 0.0008453885518556097, "epoch": 6.631305987741631, "percentage": 33.16, "elapsed_time": "0:45:33", "remaining_time": "1:31:51", "throughput": 5184.23, "total_tokens": 14171808} {"current_steps": 14070, "total_steps": 42420, "loss": 0.3701, "lr": 0.0008452397730591578, "epoch": 6.633663366336633, "percentage": 33.17, "elapsed_time": "0:45:34", "remaining_time": "1:31:49", "throughput": 5184.28, "total_tokens": 14176384} {"current_steps": 14075, "total_steps": 42420, "loss": 0.3551, "lr": 0.000845090935819474, "epoch": 6.636020744931636, "percentage": 33.18, "elapsed_time": "0:45:35", "remaining_time": "1:31:48", "throughput": 5184.32, "total_tokens": 14180992} {"current_steps": 14080, "total_steps": 42420, "loss": 0.3661, "lr": 0.0008449420401617541, "epoch": 6.638378123526638, "percentage": 33.19, "elapsed_time": "0:45:36", "remaining_time": "1:31:47", "throughput": 5184.33, "total_tokens": 14185568} {"current_steps": 14085, "total_steps": 42420, "loss": 0.3636, "lr": 0.0008447930861112036, "epoch": 6.6407355021216405, "percentage": 33.2, "elapsed_time": "0:45:37", "remaining_time": "1:31:46", "throughput": 5184.35, "total_tokens": 14190080} {"current_steps": 14090, "total_steps": 42420, "loss": 0.3618, "lr": 0.0008446440736930379, "epoch": 6.643092880716643, "percentage": 33.22, "elapsed_time": "0:45:38", "remaining_time": "1:31:45", "throughput": 5184.55, "total_tokens": 14195936} {"current_steps": 14095, "total_steps": 42420, "loss": 0.4045, "lr": 0.0008444950029324823, "epoch": 6.645450259311645, "percentage": 33.23, "elapsed_time": "0:45:39", "remaining_time": "1:31:44", "throughput": 5184.79, "total_tokens": 14202336} {"current_steps": 14100, "total_steps": 42420, "loss": 0.3669, "lr": 0.0008443458738547718, "epoch": 6.647807637906648, "percentage": 33.24, "elapsed_time": "0:45:40", "remaining_time": "1:31:43", "throughput": 5184.87, "total_tokens": 14207360} {"current_steps": 14105, "total_steps": 42420, "loss": 0.3285, "lr": 0.0008441966864851516, "epoch": 6.65016501650165, "percentage": 33.25, "elapsed_time": "0:45:41", "remaining_time": "1:31:42", "throughput": 5184.92, "total_tokens": 14211968} {"current_steps": 14110, "total_steps": 42420, "loss": 0.3506, "lr": 0.0008440474408488763, "epoch": 6.6525223950966526, "percentage": 33.26, "elapsed_time": "0:45:41", "remaining_time": "1:31:41", "throughput": 5184.9, "total_tokens": 14216384} {"current_steps": 14115, "total_steps": 42420, "loss": 0.311, "lr": 0.0008438981369712111, "epoch": 6.654879773691655, "percentage": 33.27, "elapsed_time": "0:45:42", "remaining_time": "1:31:39", "throughput": 5184.96, "total_tokens": 14220832} {"current_steps": 14120, "total_steps": 42420, "loss": 0.3577, "lr": 0.0008437487748774302, "epoch": 6.657237152286657, "percentage": 33.29, "elapsed_time": "0:45:43", "remaining_time": "1:31:38", "throughput": 5185.05, "total_tokens": 14225760} {"current_steps": 14125, "total_steps": 42420, "loss": 0.2978, "lr": 0.0008435993545928183, "epoch": 6.65959453088166, "percentage": 33.3, "elapsed_time": "0:45:44", "remaining_time": "1:31:37", "throughput": 5185.18, "total_tokens": 14230880} {"current_steps": 14130, "total_steps": 42420, "loss": 0.3762, "lr": 0.0008434498761426696, "epoch": 6.661951909476662, "percentage": 33.31, "elapsed_time": "0:45:45", "remaining_time": "1:31:36", "throughput": 5185.21, "total_tokens": 14235296} {"current_steps": 14135, "total_steps": 42420, "loss": 0.3166, "lr": 0.0008433003395522885, "epoch": 6.664309288071665, "percentage": 33.32, "elapsed_time": "0:45:46", "remaining_time": "1:31:35", "throughput": 5185.37, "total_tokens": 14240768} {"current_steps": 14140, "total_steps": 42420, "loss": 0.2927, "lr": 0.0008431507448469886, "epoch": 6.666666666666667, "percentage": 33.33, "elapsed_time": "0:45:47", "remaining_time": "1:31:34", "throughput": 5185.51, "total_tokens": 14246240} {"current_steps": 14145, "total_steps": 42420, "loss": 0.3621, "lr": 0.000843001092052094, "epoch": 6.669024045261669, "percentage": 33.35, "elapsed_time": "0:45:48", "remaining_time": "1:31:33", "throughput": 5185.54, "total_tokens": 14250848} {"current_steps": 14150, "total_steps": 42420, "loss": 0.3359, "lr": 0.0008428513811929386, "epoch": 6.671381423856672, "percentage": 33.36, "elapsed_time": "0:45:49", "remaining_time": "1:31:32", "throughput": 5185.68, "total_tokens": 14256288} {"current_steps": 14155, "total_steps": 42420, "loss": 0.3231, "lr": 0.0008427016122948654, "epoch": 6.673738802451673, "percentage": 33.37, "elapsed_time": "0:45:50", "remaining_time": "1:31:31", "throughput": 5185.72, "total_tokens": 14260864} {"current_steps": 14160, "total_steps": 42420, "loss": 0.3321, "lr": 0.000842551785383228, "epoch": 6.676096181046676, "percentage": 33.38, "elapsed_time": "0:45:50", "remaining_time": "1:31:30", "throughput": 5185.73, "total_tokens": 14265248} {"current_steps": 14165, "total_steps": 42420, "loss": 0.3271, "lr": 0.0008424019004833896, "epoch": 6.678453559641678, "percentage": 33.39, "elapsed_time": "0:45:51", "remaining_time": "1:31:28", "throughput": 5185.8, "total_tokens": 14270144} {"current_steps": 14170, "total_steps": 42420, "loss": 0.3376, "lr": 0.0008422519576207231, "epoch": 6.6808109382366805, "percentage": 33.4, "elapsed_time": "0:45:52", "remaining_time": "1:31:27", "throughput": 5185.93, "total_tokens": 14275488} {"current_steps": 14175, "total_steps": 42420, "loss": 0.3918, "lr": 0.0008421019568206112, "epoch": 6.683168316831683, "percentage": 33.42, "elapsed_time": "0:45:53", "remaining_time": "1:31:27", "throughput": 5186.16, "total_tokens": 14281728} {"current_steps": 14180, "total_steps": 42420, "loss": 0.3442, "lr": 0.0008419518981084466, "epoch": 6.685525695426685, "percentage": 33.43, "elapsed_time": "0:45:54", "remaining_time": "1:31:26", "throughput": 5186.3, "total_tokens": 14286944} {"current_steps": 14185, "total_steps": 42420, "loss": 0.361, "lr": 0.0008418017815096317, "epoch": 6.687883074021688, "percentage": 33.44, "elapsed_time": "0:45:55", "remaining_time": "1:31:25", "throughput": 5186.57, "total_tokens": 14293312} {"current_steps": 14190, "total_steps": 42420, "loss": 0.3539, "lr": 0.0008416516070495788, "epoch": 6.69024045261669, "percentage": 33.45, "elapsed_time": "0:45:56", "remaining_time": "1:31:24", "throughput": 5186.75, "total_tokens": 14298784} {"current_steps": 14195, "total_steps": 42420, "loss": 0.3261, "lr": 0.0008415013747537094, "epoch": 6.6925978312116925, "percentage": 33.46, "elapsed_time": "0:45:57", "remaining_time": "1:31:23", "throughput": 5186.72, "total_tokens": 14302944} {"current_steps": 14200, "total_steps": 42420, "loss": 0.3394, "lr": 0.0008413510846474559, "epoch": 6.694955209806695, "percentage": 33.47, "elapsed_time": "0:45:58", "remaining_time": "1:31:21", "throughput": 5186.71, "total_tokens": 14307200} {"current_steps": 14205, "total_steps": 42420, "loss": 0.2976, "lr": 0.0008412007367562595, "epoch": 6.697312588401697, "percentage": 33.49, "elapsed_time": "0:45:59", "remaining_time": "1:31:20", "throughput": 5186.8, "total_tokens": 14312384} {"current_steps": 14210, "total_steps": 42420, "loss": 0.4346, "lr": 0.0008410503311055715, "epoch": 6.6996699669967, "percentage": 33.5, "elapsed_time": "0:46:00", "remaining_time": "1:31:19", "throughput": 5186.8, "total_tokens": 14316800} {"current_steps": 14215, "total_steps": 42420, "loss": 0.315, "lr": 0.0008408998677208534, "epoch": 6.702027345591702, "percentage": 33.51, "elapsed_time": "0:46:01", "remaining_time": "1:31:18", "throughput": 5186.87, "total_tokens": 14321888} {"current_steps": 14220, "total_steps": 42420, "loss": 0.3302, "lr": 0.0008407493466275756, "epoch": 6.7043847241867045, "percentage": 33.52, "elapsed_time": "0:46:02", "remaining_time": "1:31:17", "throughput": 5186.86, "total_tokens": 14326304} {"current_steps": 14225, "total_steps": 42420, "loss": 0.3706, "lr": 0.0008405987678512191, "epoch": 6.706742102781707, "percentage": 33.53, "elapsed_time": "0:46:02", "remaining_time": "1:31:16", "throughput": 5186.99, "total_tokens": 14331584} {"current_steps": 14230, "total_steps": 42420, "loss": 0.3666, "lr": 0.0008404481314172742, "epoch": 6.709099481376709, "percentage": 33.55, "elapsed_time": "0:46:03", "remaining_time": "1:31:15", "throughput": 5187.04, "total_tokens": 14336544} {"current_steps": 14235, "total_steps": 42420, "loss": 0.3493, "lr": 0.0008402974373512413, "epoch": 6.711456859971712, "percentage": 33.56, "elapsed_time": "0:46:04", "remaining_time": "1:31:14", "throughput": 5187.21, "total_tokens": 14342240} {"current_steps": 14240, "total_steps": 42420, "loss": 0.3605, "lr": 0.0008401466856786299, "epoch": 6.713814238566714, "percentage": 33.57, "elapsed_time": "0:46:05", "remaining_time": "1:31:13", "throughput": 5187.18, "total_tokens": 14346528} {"current_steps": 14245, "total_steps": 42420, "loss": 0.2775, "lr": 0.0008399958764249602, "epoch": 6.7161716171617165, "percentage": 33.58, "elapsed_time": "0:46:06", "remaining_time": "1:31:12", "throughput": 5187.3, "total_tokens": 14351552} {"current_steps": 14250, "total_steps": 42420, "loss": 0.3548, "lr": 0.0008398450096157615, "epoch": 6.718528995756719, "percentage": 33.59, "elapsed_time": "0:46:07", "remaining_time": "1:31:11", "throughput": 5187.47, "total_tokens": 14357184} {"current_steps": 14255, "total_steps": 42420, "loss": 0.3086, "lr": 0.0008396940852765727, "epoch": 6.720886374351721, "percentage": 33.6, "elapsed_time": "0:46:08", "remaining_time": "1:31:10", "throughput": 5187.61, "total_tokens": 14362400} {"current_steps": 14260, "total_steps": 42420, "loss": 0.3113, "lr": 0.0008395431034329431, "epoch": 6.723243752946724, "percentage": 33.62, "elapsed_time": "0:46:09", "remaining_time": "1:31:09", "throughput": 5187.64, "total_tokens": 14367008} {"current_steps": 14265, "total_steps": 42420, "loss": 0.2488, "lr": 0.000839392064110431, "epoch": 6.725601131541725, "percentage": 33.63, "elapsed_time": "0:46:10", "remaining_time": "1:31:07", "throughput": 5187.67, "total_tokens": 14371680} {"current_steps": 14270, "total_steps": 42420, "loss": 0.3895, "lr": 0.0008392409673346051, "epoch": 6.727958510136728, "percentage": 33.64, "elapsed_time": "0:46:11", "remaining_time": "1:31:06", "throughput": 5187.74, "total_tokens": 14376448} {"current_steps": 14275, "total_steps": 42420, "loss": 0.3558, "lr": 0.0008390898131310434, "epoch": 6.73031588873173, "percentage": 33.65, "elapsed_time": "0:46:12", "remaining_time": "1:31:05", "throughput": 5187.81, "total_tokens": 14381344} {"current_steps": 14280, "total_steps": 42420, "loss": 0.3088, "lr": 0.0008389386015253338, "epoch": 6.732673267326732, "percentage": 33.66, "elapsed_time": "0:46:13", "remaining_time": "1:31:04", "throughput": 5188.03, "total_tokens": 14387360} {"current_steps": 14285, "total_steps": 42420, "loss": 0.3304, "lr": 0.0008387873325430738, "epoch": 6.735030645921735, "percentage": 33.68, "elapsed_time": "0:46:14", "remaining_time": "1:31:03", "throughput": 5188.08, "total_tokens": 14392128} {"current_steps": 14290, "total_steps": 42420, "loss": 0.3515, "lr": 0.0008386360062098706, "epoch": 6.737388024516737, "percentage": 33.69, "elapsed_time": "0:46:14", "remaining_time": "1:31:02", "throughput": 5188.14, "total_tokens": 14396928} {"current_steps": 14295, "total_steps": 42420, "loss": 0.3924, "lr": 0.0008384846225513416, "epoch": 6.73974540311174, "percentage": 33.7, "elapsed_time": "0:46:15", "remaining_time": "1:31:01", "throughput": 5188.18, "total_tokens": 14401632} {"current_steps": 14300, "total_steps": 42420, "loss": 0.3491, "lr": 0.0008383331815931128, "epoch": 6.742102781706742, "percentage": 33.71, "elapsed_time": "0:46:16", "remaining_time": "1:31:00", "throughput": 5188.24, "total_tokens": 14406304} {"current_steps": 14305, "total_steps": 42420, "loss": 0.3329, "lr": 0.0008381816833608209, "epoch": 6.7444601603017444, "percentage": 33.72, "elapsed_time": "0:46:17", "remaining_time": "1:30:59", "throughput": 5188.44, "total_tokens": 14412384} {"current_steps": 14310, "total_steps": 42420, "loss": 0.3471, "lr": 0.0008380301278801121, "epoch": 6.746817538896747, "percentage": 33.73, "elapsed_time": "0:46:18", "remaining_time": "1:30:58", "throughput": 5188.43, "total_tokens": 14416576} {"current_steps": 14315, "total_steps": 42420, "loss": 0.3459, "lr": 0.0008378785151766421, "epoch": 6.749174917491749, "percentage": 33.75, "elapsed_time": "0:46:19", "remaining_time": "1:30:57", "throughput": 5188.52, "total_tokens": 14421664} {"current_steps": 14320, "total_steps": 42420, "loss": 0.3117, "lr": 0.0008377268452760764, "epoch": 6.751532296086752, "percentage": 33.76, "elapsed_time": "0:46:20", "remaining_time": "1:30:56", "throughput": 5188.57, "total_tokens": 14426432} {"current_steps": 14325, "total_steps": 42420, "loss": 0.3514, "lr": 0.00083757511820409, "epoch": 6.753889674681754, "percentage": 33.77, "elapsed_time": "0:46:21", "remaining_time": "1:30:55", "throughput": 5188.78, "total_tokens": 14432576} {"current_steps": 14330, "total_steps": 42420, "loss": 0.3277, "lr": 0.0008374233339863678, "epoch": 6.7562470532767565, "percentage": 33.78, "elapsed_time": "0:46:22", "remaining_time": "1:30:53", "throughput": 5188.69, "total_tokens": 14436288} {"current_steps": 14335, "total_steps": 42420, "loss": 0.3408, "lr": 0.0008372714926486041, "epoch": 6.758604431871759, "percentage": 33.79, "elapsed_time": "0:46:23", "remaining_time": "1:30:52", "throughput": 5188.8, "total_tokens": 14441536} {"current_steps": 14340, "total_steps": 42420, "loss": 0.3113, "lr": 0.0008371195942165034, "epoch": 6.760961810466761, "percentage": 33.8, "elapsed_time": "0:46:24", "remaining_time": "1:30:51", "throughput": 5188.75, "total_tokens": 14445536} {"current_steps": 14345, "total_steps": 42420, "loss": 0.3983, "lr": 0.0008369676387157794, "epoch": 6.763319189061764, "percentage": 33.82, "elapsed_time": "0:46:24", "remaining_time": "1:30:50", "throughput": 5188.85, "total_tokens": 14450720} {"current_steps": 14350, "total_steps": 42420, "loss": 0.3561, "lr": 0.0008368156261721554, "epoch": 6.765676567656766, "percentage": 33.83, "elapsed_time": "0:46:25", "remaining_time": "1:30:49", "throughput": 5188.81, "total_tokens": 14454720} {"current_steps": 14355, "total_steps": 42420, "loss": 0.3517, "lr": 0.0008366635566113649, "epoch": 6.768033946251768, "percentage": 33.84, "elapsed_time": "0:46:26", "remaining_time": "1:30:47", "throughput": 5188.76, "total_tokens": 14458976} {"current_steps": 14360, "total_steps": 42420, "loss": 0.35, "lr": 0.0008365114300591504, "epoch": 6.77039132484677, "percentage": 33.85, "elapsed_time": "0:46:27", "remaining_time": "1:30:46", "throughput": 5188.68, "total_tokens": 14462688} {"current_steps": 14365, "total_steps": 42420, "loss": 0.3143, "lr": 0.0008363592465412644, "epoch": 6.772748703441772, "percentage": 33.86, "elapsed_time": "0:46:28", "remaining_time": "1:30:45", "throughput": 5188.75, "total_tokens": 14467616} {"current_steps": 14370, "total_steps": 42420, "loss": 0.2783, "lr": 0.000836207006083469, "epoch": 6.775106082036775, "percentage": 33.88, "elapsed_time": "0:46:29", "remaining_time": "1:30:44", "throughput": 5188.85, "total_tokens": 14472448} {"current_steps": 14375, "total_steps": 42420, "loss": 0.4275, "lr": 0.000836054708711536, "epoch": 6.777463460631777, "percentage": 33.89, "elapsed_time": "0:46:30", "remaining_time": "1:30:43", "throughput": 5188.94, "total_tokens": 14477568} {"current_steps": 14380, "total_steps": 42420, "loss": 0.3205, "lr": 0.0008359023544512466, "epoch": 6.77982083922678, "percentage": 33.9, "elapsed_time": "0:46:31", "remaining_time": "1:30:42", "throughput": 5189.06, "total_tokens": 14482688} {"current_steps": 14385, "total_steps": 42420, "loss": 0.3037, "lr": 0.000835749943328392, "epoch": 6.782178217821782, "percentage": 33.91, "elapsed_time": "0:46:31", "remaining_time": "1:30:41", "throughput": 5189.15, "total_tokens": 14487616} {"current_steps": 14390, "total_steps": 42420, "loss": 0.3186, "lr": 0.0008355974753687727, "epoch": 6.784535596416784, "percentage": 33.92, "elapsed_time": "0:46:32", "remaining_time": "1:30:40", "throughput": 5189.3, "total_tokens": 14493376} {"current_steps": 14395, "total_steps": 42420, "loss": 0.3133, "lr": 0.0008354449505981989, "epoch": 6.786892975011787, "percentage": 33.93, "elapsed_time": "0:46:33", "remaining_time": "1:30:39", "throughput": 5189.35, "total_tokens": 14498208} {"current_steps": 14400, "total_steps": 42420, "loss": 0.3446, "lr": 0.0008352923690424904, "epoch": 6.789250353606789, "percentage": 33.95, "elapsed_time": "0:46:34", "remaining_time": "1:30:37", "throughput": 5189.32, "total_tokens": 14502464} {"current_steps": 14405, "total_steps": 42420, "loss": 0.2978, "lr": 0.0008351397307274768, "epoch": 6.791607732201792, "percentage": 33.96, "elapsed_time": "0:46:35", "remaining_time": "1:30:36", "throughput": 5189.39, "total_tokens": 14507552} {"current_steps": 14410, "total_steps": 42420, "loss": 0.303, "lr": 0.0008349870356789973, "epoch": 6.793965110796794, "percentage": 33.97, "elapsed_time": "0:46:36", "remaining_time": "1:30:35", "throughput": 5189.45, "total_tokens": 14512416} {"current_steps": 14415, "total_steps": 42420, "loss": 0.3624, "lr": 0.0008348342839229004, "epoch": 6.796322489391796, "percentage": 33.98, "elapsed_time": "0:46:37", "remaining_time": "1:30:34", "throughput": 5189.34, "total_tokens": 14516224} {"current_steps": 14420, "total_steps": 42420, "loss": 0.3717, "lr": 0.0008346814754850442, "epoch": 6.798679867986799, "percentage": 33.99, "elapsed_time": "0:46:38", "remaining_time": "1:30:33", "throughput": 5189.36, "total_tokens": 14520832} {"current_steps": 14425, "total_steps": 42420, "loss": 0.2892, "lr": 0.0008345286103912969, "epoch": 6.801037246581801, "percentage": 34.01, "elapsed_time": "0:46:39", "remaining_time": "1:30:32", "throughput": 5189.35, "total_tokens": 14525024} {"current_steps": 14430, "total_steps": 42420, "loss": 0.3479, "lr": 0.0008343756886675358, "epoch": 6.803394625176804, "percentage": 34.02, "elapsed_time": "0:46:39", "remaining_time": "1:30:30", "throughput": 5189.45, "total_tokens": 14529920} {"current_steps": 14435, "total_steps": 42420, "loss": 0.2841, "lr": 0.000834222710339648, "epoch": 6.805752003771806, "percentage": 34.03, "elapsed_time": "0:46:40", "remaining_time": "1:30:29", "throughput": 5189.54, "total_tokens": 14534784} {"current_steps": 14440, "total_steps": 42420, "loss": 0.2995, "lr": 0.0008340696754335302, "epoch": 6.808109382366808, "percentage": 34.04, "elapsed_time": "0:46:41", "remaining_time": "1:30:28", "throughput": 5189.65, "total_tokens": 14540128} {"current_steps": 14445, "total_steps": 42420, "loss": 0.4226, "lr": 0.0008339165839750884, "epoch": 6.810466760961811, "percentage": 34.05, "elapsed_time": "0:46:42", "remaining_time": "1:30:27", "throughput": 5189.72, "total_tokens": 14544960} {"current_steps": 14450, "total_steps": 42420, "loss": 0.3544, "lr": 0.0008337634359902385, "epoch": 6.812824139556813, "percentage": 34.06, "elapsed_time": "0:46:43", "remaining_time": "1:30:26", "throughput": 5189.85, "total_tokens": 14550560} {"current_steps": 14455, "total_steps": 42420, "loss": 0.3239, "lr": 0.0008336102315049059, "epoch": 6.815181518151816, "percentage": 34.08, "elapsed_time": "0:46:44", "remaining_time": "1:30:25", "throughput": 5189.92, "total_tokens": 14555520} {"current_steps": 14460, "total_steps": 42420, "loss": 0.329, "lr": 0.0008334569705450254, "epoch": 6.817538896746818, "percentage": 34.09, "elapsed_time": "0:46:45", "remaining_time": "1:30:24", "throughput": 5189.96, "total_tokens": 14560192} {"current_steps": 14465, "total_steps": 42420, "loss": 0.3487, "lr": 0.0008333036531365418, "epoch": 6.8198962753418195, "percentage": 34.1, "elapsed_time": "0:46:46", "remaining_time": "1:30:23", "throughput": 5190.1, "total_tokens": 14565632} {"current_steps": 14470, "total_steps": 42420, "loss": 0.368, "lr": 0.0008331502793054087, "epoch": 6.822253653936822, "percentage": 34.11, "elapsed_time": "0:46:47", "remaining_time": "1:30:22", "throughput": 5190.24, "total_tokens": 14571136} {"current_steps": 14475, "total_steps": 42420, "loss": 0.371, "lr": 0.0008329968490775899, "epoch": 6.824611032531824, "percentage": 34.12, "elapsed_time": "0:46:48", "remaining_time": "1:30:21", "throughput": 5190.23, "total_tokens": 14575424} {"current_steps": 14480, "total_steps": 42420, "loss": 0.3393, "lr": 0.0008328433624790587, "epoch": 6.826968411126827, "percentage": 34.13, "elapsed_time": "0:46:49", "remaining_time": "1:30:20", "throughput": 5190.19, "total_tokens": 14579584} {"current_steps": 14485, "total_steps": 42420, "loss": 0.3628, "lr": 0.0008326898195357975, "epoch": 6.829325789721829, "percentage": 34.15, "elapsed_time": "0:46:49", "remaining_time": "1:30:19", "throughput": 5190.23, "total_tokens": 14584160} {"current_steps": 14490, "total_steps": 42420, "loss": 0.3128, "lr": 0.0008325362202737985, "epoch": 6.8316831683168315, "percentage": 34.16, "elapsed_time": "0:46:50", "remaining_time": "1:30:18", "throughput": 5190.4, "total_tokens": 14589888} {"current_steps": 14495, "total_steps": 42420, "loss": 0.3182, "lr": 0.0008323825647190635, "epoch": 6.834040546911834, "percentage": 34.17, "elapsed_time": "0:46:51", "remaining_time": "1:30:16", "throughput": 5190.46, "total_tokens": 14594496} {"current_steps": 14500, "total_steps": 42420, "loss": 0.391, "lr": 0.0008322288528976041, "epoch": 6.836397925506836, "percentage": 34.18, "elapsed_time": "0:46:52", "remaining_time": "1:30:16", "throughput": 5190.61, "total_tokens": 14600032} {"current_steps": 14505, "total_steps": 42420, "loss": 0.3539, "lr": 0.0008320750848354406, "epoch": 6.838755304101839, "percentage": 34.19, "elapsed_time": "0:46:53", "remaining_time": "1:30:14", "throughput": 5190.69, "total_tokens": 14604800} {"current_steps": 14510, "total_steps": 42420, "loss": 0.3551, "lr": 0.0008319212605586037, "epoch": 6.841112682696841, "percentage": 34.21, "elapsed_time": "0:46:54", "remaining_time": "1:30:14", "throughput": 5190.95, "total_tokens": 14611488} {"current_steps": 14515, "total_steps": 42420, "loss": 0.3571, "lr": 0.0008317673800931329, "epoch": 6.8434700612918435, "percentage": 34.22, "elapsed_time": "0:46:55", "remaining_time": "1:30:13", "throughput": 5190.93, "total_tokens": 14615776} {"current_steps": 14520, "total_steps": 42420, "loss": 0.3596, "lr": 0.000831613443465078, "epoch": 6.845827439886846, "percentage": 34.23, "elapsed_time": "0:46:56", "remaining_time": "1:30:12", "throughput": 5191.13, "total_tokens": 14621536} {"current_steps": 14525, "total_steps": 42420, "loss": 0.3417, "lr": 0.0008314594507004973, "epoch": 6.848184818481848, "percentage": 34.24, "elapsed_time": "0:46:57", "remaining_time": "1:30:10", "throughput": 5191.17, "total_tokens": 14626176} {"current_steps": 14530, "total_steps": 42420, "loss": 0.3978, "lr": 0.0008313054018254597, "epoch": 6.850542197076851, "percentage": 34.25, "elapsed_time": "0:46:58", "remaining_time": "1:30:09", "throughput": 5191.12, "total_tokens": 14629952} {"current_steps": 14535, "total_steps": 42420, "loss": 0.3231, "lr": 0.0008311512968660426, "epoch": 6.852899575671853, "percentage": 34.26, "elapsed_time": "0:46:59", "remaining_time": "1:30:08", "throughput": 5191.13, "total_tokens": 14634432} {"current_steps": 14540, "total_steps": 42420, "loss": 0.3113, "lr": 0.0008309971358483337, "epoch": 6.8552569542668556, "percentage": 34.28, "elapsed_time": "0:46:59", "remaining_time": "1:30:07", "throughput": 5191.18, "total_tokens": 14639072} {"current_steps": 14545, "total_steps": 42420, "loss": 0.2878, "lr": 0.0008308429187984297, "epoch": 6.857614332861858, "percentage": 34.29, "elapsed_time": "0:47:00", "remaining_time": "1:30:06", "throughput": 5191.19, "total_tokens": 14643552} {"current_steps": 14550, "total_steps": 42420, "loss": 0.34, "lr": 0.000830688645742437, "epoch": 6.85997171145686, "percentage": 34.3, "elapsed_time": "0:47:01", "remaining_time": "1:30:04", "throughput": 5191.19, "total_tokens": 14647840} {"current_steps": 14555, "total_steps": 42420, "loss": 0.3092, "lr": 0.0008305343167064712, "epoch": 6.862329090051862, "percentage": 34.31, "elapsed_time": "0:47:02", "remaining_time": "1:30:03", "throughput": 5191.35, "total_tokens": 14653600} {"current_steps": 14560, "total_steps": 42420, "loss": 0.3186, "lr": 0.0008303799317166578, "epoch": 6.864686468646864, "percentage": 34.32, "elapsed_time": "0:47:03", "remaining_time": "1:30:03", "throughput": 5191.46, "total_tokens": 14659104} {"current_steps": 14565, "total_steps": 42420, "loss": 0.3249, "lr": 0.0008302254907991316, "epoch": 6.867043847241867, "percentage": 34.34, "elapsed_time": "0:47:04", "remaining_time": "1:30:01", "throughput": 5191.53, "total_tokens": 14664000} {"current_steps": 14570, "total_steps": 42420, "loss": 0.3963, "lr": 0.0008300709939800367, "epoch": 6.869401225836869, "percentage": 34.35, "elapsed_time": "0:47:05", "remaining_time": "1:30:00", "throughput": 5191.43, "total_tokens": 14667648} {"current_steps": 14575, "total_steps": 42420, "loss": 0.3357, "lr": 0.0008299164412855267, "epoch": 6.8717586044318715, "percentage": 34.36, "elapsed_time": "0:47:06", "remaining_time": "1:29:59", "throughput": 5191.65, "total_tokens": 14673792} {"current_steps": 14580, "total_steps": 42420, "loss": 0.3605, "lr": 0.0008297618327417649, "epoch": 6.874115983026874, "percentage": 34.37, "elapsed_time": "0:47:07", "remaining_time": "1:29:58", "throughput": 5191.86, "total_tokens": 14679712} {"current_steps": 14585, "total_steps": 42420, "loss": 0.3392, "lr": 0.0008296071683749238, "epoch": 6.876473361621876, "percentage": 34.38, "elapsed_time": "0:47:08", "remaining_time": "1:29:58", "throughput": 5191.95, "total_tokens": 14685248} {"current_steps": 14590, "total_steps": 42420, "loss": 0.3406, "lr": 0.0008294524482111856, "epoch": 6.878830740216879, "percentage": 34.39, "elapsed_time": "0:47:09", "remaining_time": "1:29:56", "throughput": 5191.95, "total_tokens": 14689664} {"current_steps": 14595, "total_steps": 42420, "loss": 0.3416, "lr": 0.0008292976722767415, "epoch": 6.881188118811881, "percentage": 34.41, "elapsed_time": "0:47:10", "remaining_time": "1:29:55", "throughput": 5192.13, "total_tokens": 14695392} {"current_steps": 14600, "total_steps": 42420, "loss": 0.2944, "lr": 0.000829142840597793, "epoch": 6.8835454974068835, "percentage": 34.42, "elapsed_time": "0:47:11", "remaining_time": "1:29:54", "throughput": 5192.22, "total_tokens": 14700608} {"current_steps": 14605, "total_steps": 42420, "loss": 0.3276, "lr": 0.0008289879532005498, "epoch": 6.885902876001886, "percentage": 34.43, "elapsed_time": "0:47:12", "remaining_time": "1:29:53", "throughput": 5192.38, "total_tokens": 14705888} {"current_steps": 14610, "total_steps": 42420, "loss": 0.3573, "lr": 0.0008288330101112322, "epoch": 6.888260254596888, "percentage": 34.44, "elapsed_time": "0:47:13", "remaining_time": "1:29:52", "throughput": 5192.48, "total_tokens": 14711040} {"current_steps": 14615, "total_steps": 42420, "loss": 0.3853, "lr": 0.0008286780113560692, "epoch": 6.890617633191891, "percentage": 34.45, "elapsed_time": "0:47:14", "remaining_time": "1:29:51", "throughput": 5192.6, "total_tokens": 14716416} {"current_steps": 14620, "total_steps": 42420, "loss": 0.3536, "lr": 0.0008285229569612994, "epoch": 6.892975011786893, "percentage": 34.46, "elapsed_time": "0:47:14", "remaining_time": "1:29:50", "throughput": 5192.61, "total_tokens": 14720896} {"current_steps": 14625, "total_steps": 42420, "loss": 0.3446, "lr": 0.0008283678469531711, "epoch": 6.8953323903818955, "percentage": 34.48, "elapsed_time": "0:47:15", "remaining_time": "1:29:49", "throughput": 5192.57, "total_tokens": 14724960} {"current_steps": 14630, "total_steps": 42420, "loss": 0.348, "lr": 0.0008282126813579417, "epoch": 6.897689768976898, "percentage": 34.49, "elapsed_time": "0:47:16", "remaining_time": "1:29:48", "throughput": 5192.61, "total_tokens": 14729760} {"current_steps": 14635, "total_steps": 42420, "loss": 0.3483, "lr": 0.0008280574602018781, "epoch": 6.9000471475719, "percentage": 34.5, "elapsed_time": "0:47:17", "remaining_time": "1:29:47", "throughput": 5192.69, "total_tokens": 14734656} {"current_steps": 14640, "total_steps": 42420, "loss": 0.3718, "lr": 0.0008279021835112566, "epoch": 6.902404526166903, "percentage": 34.51, "elapsed_time": "0:47:18", "remaining_time": "1:29:46", "throughput": 5192.75, "total_tokens": 14739488} {"current_steps": 14645, "total_steps": 42420, "loss": 0.3523, "lr": 0.0008277468513123627, "epoch": 6.904761904761905, "percentage": 34.52, "elapsed_time": "0:47:19", "remaining_time": "1:29:44", "throughput": 5192.76, "total_tokens": 14743968} {"current_steps": 14650, "total_steps": 42420, "loss": 0.3284, "lr": 0.0008275914636314918, "epoch": 6.9071192833569075, "percentage": 34.54, "elapsed_time": "0:47:20", "remaining_time": "1:29:43", "throughput": 5192.85, "total_tokens": 14749024} {"current_steps": 14655, "total_steps": 42420, "loss": 0.3289, "lr": 0.0008274360204949484, "epoch": 6.90947666195191, "percentage": 34.55, "elapsed_time": "0:47:21", "remaining_time": "1:29:42", "throughput": 5192.95, "total_tokens": 14754112} {"current_steps": 14660, "total_steps": 42420, "loss": 0.3661, "lr": 0.0008272805219290462, "epoch": 6.911834040546912, "percentage": 34.56, "elapsed_time": "0:47:22", "remaining_time": "1:29:41", "throughput": 5193.01, "total_tokens": 14759008} {"current_steps": 14665, "total_steps": 42420, "loss": 0.3482, "lr": 0.0008271249679601085, "epoch": 6.914191419141914, "percentage": 34.57, "elapsed_time": "0:47:23", "remaining_time": "1:29:40", "throughput": 5193.04, "total_tokens": 14763872} {"current_steps": 14670, "total_steps": 42420, "loss": 0.2692, "lr": 0.0008269693586144681, "epoch": 6.916548797736916, "percentage": 34.58, "elapsed_time": "0:47:23", "remaining_time": "1:29:39", "throughput": 5193.0, "total_tokens": 14767872} {"current_steps": 14675, "total_steps": 42420, "loss": 0.2969, "lr": 0.0008268136939184668, "epoch": 6.918906176331919, "percentage": 34.59, "elapsed_time": "0:47:24", "remaining_time": "1:29:38", "throughput": 5193.07, "total_tokens": 14772832} {"current_steps": 14680, "total_steps": 42420, "loss": 0.3895, "lr": 0.0008266579738984561, "epoch": 6.921263554926921, "percentage": 34.61, "elapsed_time": "0:47:25", "remaining_time": "1:29:37", "throughput": 5193.06, "total_tokens": 14777376} {"current_steps": 14685, "total_steps": 42420, "loss": 0.3487, "lr": 0.0008265021985807968, "epoch": 6.923620933521923, "percentage": 34.62, "elapsed_time": "0:47:26", "remaining_time": "1:29:36", "throughput": 5193.14, "total_tokens": 14782176} {"current_steps": 14690, "total_steps": 42420, "loss": 0.3396, "lr": 0.000826346367991859, "epoch": 6.925978312116926, "percentage": 34.63, "elapsed_time": "0:47:27", "remaining_time": "1:29:35", "throughput": 5193.22, "total_tokens": 14787392} {"current_steps": 14695, "total_steps": 42420, "loss": 0.3353, "lr": 0.0008261904821580221, "epoch": 6.928335690711928, "percentage": 34.64, "elapsed_time": "0:47:28", "remaining_time": "1:29:33", "throughput": 5193.27, "total_tokens": 14792032} {"current_steps": 14700, "total_steps": 42420, "loss": 0.3215, "lr": 0.000826034541105675, "epoch": 6.930693069306931, "percentage": 34.65, "elapsed_time": "0:47:29", "remaining_time": "1:29:32", "throughput": 5193.28, "total_tokens": 14796384} {"current_steps": 14705, "total_steps": 42420, "loss": 0.3373, "lr": 0.0008258785448612159, "epoch": 6.933050447901933, "percentage": 34.67, "elapsed_time": "0:47:30", "remaining_time": "1:29:31", "throughput": 5193.26, "total_tokens": 14800960} {"current_steps": 14710, "total_steps": 42420, "loss": 0.3274, "lr": 0.0008257224934510521, "epoch": 6.935407826496935, "percentage": 34.68, "elapsed_time": "0:47:31", "remaining_time": "1:29:30", "throughput": 5193.39, "total_tokens": 14806400} {"current_steps": 14715, "total_steps": 42420, "loss": 0.3464, "lr": 0.0008255663869016007, "epoch": 6.937765205091938, "percentage": 34.69, "elapsed_time": "0:47:31", "remaining_time": "1:29:29", "throughput": 5193.47, "total_tokens": 14811520} {"current_steps": 14720, "total_steps": 42420, "loss": 0.3648, "lr": 0.0008254102252392878, "epoch": 6.94012258368694, "percentage": 34.7, "elapsed_time": "0:47:32", "remaining_time": "1:29:28", "throughput": 5193.53, "total_tokens": 14816544} {"current_steps": 14725, "total_steps": 42420, "loss": 0.342, "lr": 0.0008252540084905489, "epoch": 6.942479962281943, "percentage": 34.71, "elapsed_time": "0:47:33", "remaining_time": "1:29:27", "throughput": 5193.68, "total_tokens": 14822240} {"current_steps": 14730, "total_steps": 42420, "loss": 0.3277, "lr": 0.0008250977366818288, "epoch": 6.944837340876945, "percentage": 34.72, "elapsed_time": "0:47:34", "remaining_time": "1:29:26", "throughput": 5193.79, "total_tokens": 14827520} {"current_steps": 14735, "total_steps": 42420, "loss": 0.381, "lr": 0.0008249414098395818, "epoch": 6.947194719471947, "percentage": 34.74, "elapsed_time": "0:47:35", "remaining_time": "1:29:25", "throughput": 5194.02, "total_tokens": 14833792} {"current_steps": 14740, "total_steps": 42420, "loss": 0.3494, "lr": 0.0008247850279902713, "epoch": 6.94955209806695, "percentage": 34.75, "elapsed_time": "0:47:36", "remaining_time": "1:29:25", "throughput": 5194.2, "total_tokens": 14839616} {"current_steps": 14745, "total_steps": 42420, "loss": 0.3247, "lr": 0.0008246285911603701, "epoch": 6.951909476661952, "percentage": 34.76, "elapsed_time": "0:47:37", "remaining_time": "1:29:24", "throughput": 5194.25, "total_tokens": 14844736} {"current_steps": 14750, "total_steps": 42420, "loss": 0.4034, "lr": 0.0008244720993763604, "epoch": 6.954266855256955, "percentage": 34.77, "elapsed_time": "0:47:38", "remaining_time": "1:29:23", "throughput": 5194.4, "total_tokens": 14850304} {"current_steps": 14755, "total_steps": 42420, "loss": 0.3486, "lr": 0.0008243155526647335, "epoch": 6.956624233851956, "percentage": 34.78, "elapsed_time": "0:47:39", "remaining_time": "1:29:22", "throughput": 5194.47, "total_tokens": 14855136} {"current_steps": 14760, "total_steps": 42420, "loss": 0.3692, "lr": 0.00082415895105199, "epoch": 6.958981612446959, "percentage": 34.79, "elapsed_time": "0:47:40", "remaining_time": "1:29:20", "throughput": 5194.58, "total_tokens": 14860352} {"current_steps": 14765, "total_steps": 42420, "loss": 0.3588, "lr": 0.0008240022945646399, "epoch": 6.961338991041961, "percentage": 34.81, "elapsed_time": "0:47:41", "remaining_time": "1:29:19", "throughput": 5194.61, "total_tokens": 14864960} {"current_steps": 14770, "total_steps": 42420, "loss": 0.3811, "lr": 0.0008238455832292027, "epoch": 6.963696369636963, "percentage": 34.82, "elapsed_time": "0:47:42", "remaining_time": "1:29:18", "throughput": 5194.68, "total_tokens": 14870080} {"current_steps": 14775, "total_steps": 42420, "loss": 0.3482, "lr": 0.0008236888170722068, "epoch": 6.966053748231966, "percentage": 34.83, "elapsed_time": "0:47:43", "remaining_time": "1:29:17", "throughput": 5194.69, "total_tokens": 14874560} {"current_steps": 14780, "total_steps": 42420, "loss": 0.38, "lr": 0.00082353199612019, "epoch": 6.968411126826968, "percentage": 34.84, "elapsed_time": "0:47:44", "remaining_time": "1:29:16", "throughput": 5194.75, "total_tokens": 14879648} {"current_steps": 14785, "total_steps": 42420, "loss": 0.3399, "lr": 0.0008233751203996996, "epoch": 6.970768505421971, "percentage": 34.85, "elapsed_time": "0:47:45", "remaining_time": "1:29:15", "throughput": 5194.84, "total_tokens": 14884672} {"current_steps": 14790, "total_steps": 42420, "loss": 0.3595, "lr": 0.0008232181899372919, "epoch": 6.973125884016973, "percentage": 34.87, "elapsed_time": "0:47:46", "remaining_time": "1:29:14", "throughput": 5194.92, "total_tokens": 14889632} {"current_steps": 14795, "total_steps": 42420, "loss": 0.3609, "lr": 0.0008230612047595327, "epoch": 6.975483262611975, "percentage": 34.88, "elapsed_time": "0:47:47", "remaining_time": "1:29:13", "throughput": 5194.91, "total_tokens": 14893824} {"current_steps": 14800, "total_steps": 42420, "loss": 0.3521, "lr": 0.0008229041648929967, "epoch": 6.977840641206978, "percentage": 34.89, "elapsed_time": "0:47:47", "remaining_time": "1:29:11", "throughput": 5194.87, "total_tokens": 14897952} {"current_steps": 14805, "total_steps": 42420, "loss": 0.3485, "lr": 0.0008227470703642682, "epoch": 6.98019801980198, "percentage": 34.9, "elapsed_time": "0:47:48", "remaining_time": "1:29:11", "throughput": 5195.07, "total_tokens": 14903712} {"current_steps": 14810, "total_steps": 42420, "loss": 0.3459, "lr": 0.0008225899211999406, "epoch": 6.982555398396983, "percentage": 34.91, "elapsed_time": "0:47:49", "remaining_time": "1:29:09", "throughput": 5195.05, "total_tokens": 14908128} {"current_steps": 14815, "total_steps": 42420, "loss": 0.3496, "lr": 0.0008224327174266166, "epoch": 6.984912776991985, "percentage": 34.92, "elapsed_time": "0:47:50", "remaining_time": "1:29:08", "throughput": 5195.05, "total_tokens": 14912576} {"current_steps": 14820, "total_steps": 42420, "loss": 0.3599, "lr": 0.0008222754590709082, "epoch": 6.987270155586987, "percentage": 34.94, "elapsed_time": "0:47:51", "remaining_time": "1:29:07", "throughput": 5195.12, "total_tokens": 14917472} {"current_steps": 14825, "total_steps": 42420, "loss": 0.3288, "lr": 0.0008221181461594365, "epoch": 6.98962753418199, "percentage": 34.95, "elapsed_time": "0:47:52", "remaining_time": "1:29:06", "throughput": 5195.23, "total_tokens": 14922592} {"current_steps": 14830, "total_steps": 42420, "loss": 0.3113, "lr": 0.0008219607787188318, "epoch": 6.991984912776992, "percentage": 34.96, "elapsed_time": "0:47:53", "remaining_time": "1:29:05", "throughput": 5195.31, "total_tokens": 14927744} {"current_steps": 14835, "total_steps": 42420, "loss": 0.3702, "lr": 0.0008218033567757339, "epoch": 6.994342291371995, "percentage": 34.97, "elapsed_time": "0:47:54", "remaining_time": "1:29:04", "throughput": 5195.42, "total_tokens": 14933024} {"current_steps": 14840, "total_steps": 42420, "loss": 0.3481, "lr": 0.0008216458803567917, "epoch": 6.996699669966997, "percentage": 34.98, "elapsed_time": "0:47:55", "remaining_time": "1:29:03", "throughput": 5195.47, "total_tokens": 14937856} {"current_steps": 14845, "total_steps": 42420, "loss": 0.3437, "lr": 0.000821488349488663, "epoch": 6.999057048561999, "percentage": 35.0, "elapsed_time": "0:47:56", "remaining_time": "1:29:02", "throughput": 5195.62, "total_tokens": 14943616} {"current_steps": 14847, "total_steps": 42420, "eval_loss": 0.3379077613353729, "epoch": 7.0, "percentage": 35.0, "elapsed_time": "0:48:15", "remaining_time": "1:29:37", "throughput": 5161.41, "total_tokens": 14945088} {"current_steps": 14850, "total_steps": 42420, "loss": 0.3421, "lr": 0.0008213307641980154, "epoch": 7.001414427157002, "percentage": 35.01, "elapsed_time": "0:48:18", "remaining_time": "1:29:40", "throughput": 5157.94, "total_tokens": 14949184} {"current_steps": 14855, "total_steps": 42420, "loss": 0.3594, "lr": 0.0008211731245115254, "epoch": 7.003771805752004, "percentage": 35.02, "elapsed_time": "0:48:19", "remaining_time": "1:29:39", "throughput": 5157.99, "total_tokens": 14953760} {"current_steps": 14860, "total_steps": 42420, "loss": 0.3512, "lr": 0.0008210154304558785, "epoch": 7.006129184347006, "percentage": 35.03, "elapsed_time": "0:48:20", "remaining_time": "1:29:38", "throughput": 5158.06, "total_tokens": 14958624} {"current_steps": 14865, "total_steps": 42420, "loss": 0.3332, "lr": 0.0008208576820577698, "epoch": 7.008486562942008, "percentage": 35.04, "elapsed_time": "0:48:20", "remaining_time": "1:29:37", "throughput": 5158.09, "total_tokens": 14963136} {"current_steps": 14870, "total_steps": 42420, "loss": 0.3198, "lr": 0.0008206998793439034, "epoch": 7.0108439415370105, "percentage": 35.05, "elapsed_time": "0:48:21", "remaining_time": "1:29:36", "throughput": 5158.14, "total_tokens": 14967840} {"current_steps": 14875, "total_steps": 42420, "loss": 0.199, "lr": 0.0008205420223409926, "epoch": 7.013201320132013, "percentage": 35.07, "elapsed_time": "0:48:22", "remaining_time": "1:29:34", "throughput": 5158.14, "total_tokens": 14972064} {"current_steps": 14880, "total_steps": 42420, "loss": 0.4588, "lr": 0.00082038411107576, "epoch": 7.015558698727015, "percentage": 35.08, "elapsed_time": "0:48:23", "remaining_time": "1:29:33", "throughput": 5158.19, "total_tokens": 14976736} {"current_steps": 14885, "total_steps": 42420, "loss": 0.3398, "lr": 0.0008202261455749369, "epoch": 7.017916077322018, "percentage": 35.09, "elapsed_time": "0:48:24", "remaining_time": "1:29:32", "throughput": 5158.13, "total_tokens": 14980640} {"current_steps": 14890, "total_steps": 42420, "loss": 0.2695, "lr": 0.0008200681258652648, "epoch": 7.02027345591702, "percentage": 35.1, "elapsed_time": "0:48:25", "remaining_time": "1:29:31", "throughput": 5158.34, "total_tokens": 14986752} {"current_steps": 14895, "total_steps": 42420, "loss": 0.3817, "lr": 0.0008199100519734932, "epoch": 7.0226308345120225, "percentage": 35.11, "elapsed_time": "0:48:26", "remaining_time": "1:29:30", "throughput": 5158.41, "total_tokens": 14991744} {"current_steps": 14900, "total_steps": 42420, "loss": 0.3619, "lr": 0.0008197519239263818, "epoch": 7.024988213107025, "percentage": 35.12, "elapsed_time": "0:48:27", "remaining_time": "1:29:29", "throughput": 5158.68, "total_tokens": 14998144} {"current_steps": 14905, "total_steps": 42420, "loss": 0.3317, "lr": 0.0008195937417506987, "epoch": 7.027345591702027, "percentage": 35.14, "elapsed_time": "0:48:28", "remaining_time": "1:29:28", "throughput": 5158.66, "total_tokens": 15002400} {"current_steps": 14910, "total_steps": 42420, "loss": 0.3367, "lr": 0.0008194355054732214, "epoch": 7.02970297029703, "percentage": 35.15, "elapsed_time": "0:48:29", "remaining_time": "1:29:27", "throughput": 5158.61, "total_tokens": 15006528} {"current_steps": 14915, "total_steps": 42420, "loss": 0.3264, "lr": 0.0008192772151207369, "epoch": 7.032060348892032, "percentage": 35.16, "elapsed_time": "0:48:29", "remaining_time": "1:29:26", "throughput": 5158.71, "total_tokens": 15011360} {"current_steps": 14920, "total_steps": 42420, "loss": 0.3435, "lr": 0.0008191188707200406, "epoch": 7.0344177274870345, "percentage": 35.17, "elapsed_time": "0:48:30", "remaining_time": "1:29:25", "throughput": 5158.8, "total_tokens": 15016448} {"current_steps": 14925, "total_steps": 42420, "loss": 0.3984, "lr": 0.000818960472297938, "epoch": 7.036775106082037, "percentage": 35.18, "elapsed_time": "0:48:31", "remaining_time": "1:29:23", "throughput": 5158.77, "total_tokens": 15020320} {"current_steps": 14930, "total_steps": 42420, "loss": 0.3849, "lr": 0.0008188020198812431, "epoch": 7.039132484677039, "percentage": 35.2, "elapsed_time": "0:48:32", "remaining_time": "1:29:22", "throughput": 5158.89, "total_tokens": 15025472} {"current_steps": 14935, "total_steps": 42420, "loss": 0.3742, "lr": 0.0008186435134967793, "epoch": 7.041489863272042, "percentage": 35.21, "elapsed_time": "0:48:33", "remaining_time": "1:29:21", "throughput": 5159.04, "total_tokens": 15031072} {"current_steps": 14940, "total_steps": 42420, "loss": 0.3528, "lr": 0.0008184849531713787, "epoch": 7.043847241867044, "percentage": 35.22, "elapsed_time": "0:48:34", "remaining_time": "1:29:20", "throughput": 5159.19, "total_tokens": 15036640} {"current_steps": 14945, "total_steps": 42420, "loss": 0.3452, "lr": 0.0008183263389318832, "epoch": 7.0462046204620465, "percentage": 35.23, "elapsed_time": "0:48:35", "remaining_time": "1:29:19", "throughput": 5159.24, "total_tokens": 15041248} {"current_steps": 14950, "total_steps": 42420, "loss": 0.3404, "lr": 0.0008181676708051433, "epoch": 7.048561999057049, "percentage": 35.24, "elapsed_time": "0:48:36", "remaining_time": "1:29:18", "throughput": 5159.26, "total_tokens": 15045632} {"current_steps": 14955, "total_steps": 42420, "loss": 0.3546, "lr": 0.0008180089488180191, "epoch": 7.050919377652051, "percentage": 35.25, "elapsed_time": "0:48:37", "remaining_time": "1:29:17", "throughput": 5159.25, "total_tokens": 15049728} {"current_steps": 14960, "total_steps": 42420, "loss": 0.3341, "lr": 0.0008178501729973793, "epoch": 7.053276756247053, "percentage": 35.27, "elapsed_time": "0:48:37", "remaining_time": "1:29:16", "throughput": 5159.38, "total_tokens": 15055040} {"current_steps": 14965, "total_steps": 42420, "loss": 0.32, "lr": 0.0008176913433701018, "epoch": 7.055634134842055, "percentage": 35.28, "elapsed_time": "0:48:39", "remaining_time": "1:29:15", "throughput": 5159.6, "total_tokens": 15061536} {"current_steps": 14970, "total_steps": 42420, "loss": 0.2758, "lr": 0.0008175324599630742, "epoch": 7.057991513437058, "percentage": 35.29, "elapsed_time": "0:48:40", "remaining_time": "1:29:14", "throughput": 5159.67, "total_tokens": 15066272} {"current_steps": 14975, "total_steps": 42420, "loss": 0.3911, "lr": 0.0008173735228031925, "epoch": 7.06034889203206, "percentage": 35.3, "elapsed_time": "0:48:40", "remaining_time": "1:29:13", "throughput": 5159.83, "total_tokens": 15071648} {"current_steps": 14980, "total_steps": 42420, "loss": 0.2946, "lr": 0.000817214531917362, "epoch": 7.0627062706270625, "percentage": 35.31, "elapsed_time": "0:48:41", "remaining_time": "1:29:12", "throughput": 5159.88, "total_tokens": 15076288} {"current_steps": 14985, "total_steps": 42420, "loss": 0.3869, "lr": 0.0008170554873324973, "epoch": 7.065063649222065, "percentage": 35.33, "elapsed_time": "0:48:42", "remaining_time": "1:29:10", "throughput": 5159.89, "total_tokens": 15080768} {"current_steps": 14990, "total_steps": 42420, "loss": 0.3195, "lr": 0.0008168963890755217, "epoch": 7.067421027817067, "percentage": 35.34, "elapsed_time": "0:48:43", "remaining_time": "1:29:09", "throughput": 5159.83, "total_tokens": 15084672} {"current_steps": 14995, "total_steps": 42420, "loss": 0.3217, "lr": 0.0008167372371733682, "epoch": 7.06977840641207, "percentage": 35.35, "elapsed_time": "0:48:44", "remaining_time": "1:29:08", "throughput": 5159.94, "total_tokens": 15089600} {"current_steps": 15000, "total_steps": 42420, "loss": 0.285, "lr": 0.0008165780316529783, "epoch": 7.072135785007072, "percentage": 35.36, "elapsed_time": "0:48:45", "remaining_time": "1:29:07", "throughput": 5160.03, "total_tokens": 15094592} {"current_steps": 15005, "total_steps": 42420, "loss": 0.3881, "lr": 0.0008164187725413028, "epoch": 7.0744931636020745, "percentage": 35.37, "elapsed_time": "0:48:46", "remaining_time": "1:29:06", "throughput": 5159.97, "total_tokens": 15098656} {"current_steps": 15010, "total_steps": 42420, "loss": 0.4019, "lr": 0.0008162594598653017, "epoch": 7.076850542197077, "percentage": 35.38, "elapsed_time": "0:48:47", "remaining_time": "1:29:05", "throughput": 5160.12, "total_tokens": 15104160} {"current_steps": 15015, "total_steps": 42420, "loss": 0.3401, "lr": 0.0008161000936519438, "epoch": 7.079207920792079, "percentage": 35.4, "elapsed_time": "0:48:47", "remaining_time": "1:29:04", "throughput": 5160.2, "total_tokens": 15108960} {"current_steps": 15020, "total_steps": 42420, "loss": 0.3536, "lr": 0.0008159406739282069, "epoch": 7.081565299387082, "percentage": 35.41, "elapsed_time": "0:48:48", "remaining_time": "1:29:03", "throughput": 5160.32, "total_tokens": 15114176} {"current_steps": 15025, "total_steps": 42420, "loss": 0.3307, "lr": 0.0008157812007210786, "epoch": 7.083922677982084, "percentage": 35.42, "elapsed_time": "0:48:49", "remaining_time": "1:29:01", "throughput": 5160.32, "total_tokens": 15118496} {"current_steps": 15030, "total_steps": 42420, "loss": 0.2733, "lr": 0.0008156216740575543, "epoch": 7.0862800565770865, "percentage": 35.43, "elapsed_time": "0:48:50", "remaining_time": "1:29:00", "throughput": 5160.44, "total_tokens": 15123776} {"current_steps": 15035, "total_steps": 42420, "loss": 0.4144, "lr": 0.00081546209396464, "epoch": 7.088637435172089, "percentage": 35.44, "elapsed_time": "0:48:51", "remaining_time": "1:29:00", "throughput": 5160.67, "total_tokens": 15130144} {"current_steps": 15040, "total_steps": 42420, "loss": 0.4162, "lr": 0.0008153024604693491, "epoch": 7.090994813767091, "percentage": 35.45, "elapsed_time": "0:48:52", "remaining_time": "1:28:58", "throughput": 5160.66, "total_tokens": 15134368} {"current_steps": 15045, "total_steps": 42420, "loss": 0.2995, "lr": 0.0008151427735987052, "epoch": 7.093352192362094, "percentage": 35.47, "elapsed_time": "0:48:53", "remaining_time": "1:28:57", "throughput": 5160.8, "total_tokens": 15139616} {"current_steps": 15050, "total_steps": 42420, "loss": 0.3356, "lr": 0.0008149830333797406, "epoch": 7.095709570957096, "percentage": 35.48, "elapsed_time": "0:48:54", "remaining_time": "1:28:56", "throughput": 5160.81, "total_tokens": 15143904} {"current_steps": 15055, "total_steps": 42420, "loss": 0.3069, "lr": 0.0008148232398394966, "epoch": 7.0980669495520985, "percentage": 35.49, "elapsed_time": "0:48:55", "remaining_time": "1:28:55", "throughput": 5161.04, "total_tokens": 15150208} {"current_steps": 15060, "total_steps": 42420, "loss": 0.4074, "lr": 0.0008146633930050236, "epoch": 7.1004243281471, "percentage": 35.5, "elapsed_time": "0:48:56", "remaining_time": "1:28:54", "throughput": 5161.07, "total_tokens": 15154656} {"current_steps": 15065, "total_steps": 42420, "loss": 0.3767, "lr": 0.0008145034929033805, "epoch": 7.102781706742102, "percentage": 35.51, "elapsed_time": "0:48:57", "remaining_time": "1:28:53", "throughput": 5161.28, "total_tokens": 15160928} {"current_steps": 15070, "total_steps": 42420, "loss": 0.337, "lr": 0.0008143435395616363, "epoch": 7.105139085337105, "percentage": 35.53, "elapsed_time": "0:48:58", "remaining_time": "1:28:52", "throughput": 5161.39, "total_tokens": 15166016} {"current_steps": 15075, "total_steps": 42420, "loss": 0.3427, "lr": 0.0008141835330068681, "epoch": 7.107496463932107, "percentage": 35.54, "elapsed_time": "0:48:59", "remaining_time": "1:28:51", "throughput": 5161.54, "total_tokens": 15171616} {"current_steps": 15080, "total_steps": 42420, "loss": 0.3591, "lr": 0.0008140234732661622, "epoch": 7.10985384252711, "percentage": 35.55, "elapsed_time": "0:49:00", "remaining_time": "1:28:50", "throughput": 5161.57, "total_tokens": 15175904} {"current_steps": 15085, "total_steps": 42420, "loss": 0.3877, "lr": 0.0008138633603666141, "epoch": 7.112211221122112, "percentage": 35.56, "elapsed_time": "0:49:01", "remaining_time": "1:28:49", "throughput": 5161.6, "total_tokens": 15180320} {"current_steps": 15090, "total_steps": 42420, "loss": 0.3483, "lr": 0.0008137031943353283, "epoch": 7.114568599717114, "percentage": 35.57, "elapsed_time": "0:49:01", "remaining_time": "1:28:48", "throughput": 5161.63, "total_tokens": 15185024} {"current_steps": 15095, "total_steps": 42420, "loss": 0.4179, "lr": 0.0008135429751994179, "epoch": 7.116925978312117, "percentage": 35.58, "elapsed_time": "0:49:02", "remaining_time": "1:28:47", "throughput": 5161.73, "total_tokens": 15190048} {"current_steps": 15100, "total_steps": 42420, "loss": 0.3241, "lr": 0.0008133827029860054, "epoch": 7.119283356907119, "percentage": 35.6, "elapsed_time": "0:49:03", "remaining_time": "1:28:45", "throughput": 5161.69, "total_tokens": 15193984} {"current_steps": 15105, "total_steps": 42420, "loss": 0.4109, "lr": 0.0008132223777222224, "epoch": 7.121640735502122, "percentage": 35.61, "elapsed_time": "0:49:04", "remaining_time": "1:28:44", "throughput": 5161.86, "total_tokens": 15199520} {"current_steps": 15110, "total_steps": 42420, "loss": 0.4475, "lr": 0.0008130619994352091, "epoch": 7.123998114097124, "percentage": 35.62, "elapsed_time": "0:49:05", "remaining_time": "1:28:43", "throughput": 5161.92, "total_tokens": 15204352} {"current_steps": 15115, "total_steps": 42420, "loss": 0.3533, "lr": 0.0008129015681521145, "epoch": 7.126355492692126, "percentage": 35.63, "elapsed_time": "0:49:06", "remaining_time": "1:28:42", "throughput": 5161.99, "total_tokens": 15209344} {"current_steps": 15120, "total_steps": 42420, "loss": 0.3789, "lr": 0.0008127410839000974, "epoch": 7.128712871287129, "percentage": 35.64, "elapsed_time": "0:49:07", "remaining_time": "1:28:41", "throughput": 5161.95, "total_tokens": 15213312} {"current_steps": 15125, "total_steps": 42420, "loss": 0.3376, "lr": 0.000812580546706325, "epoch": 7.131070249882131, "percentage": 35.66, "elapsed_time": "0:49:08", "remaining_time": "1:28:40", "throughput": 5162.2, "total_tokens": 15219744} {"current_steps": 15130, "total_steps": 42420, "loss": 0.3836, "lr": 0.000812419956597973, "epoch": 7.133427628477134, "percentage": 35.67, "elapsed_time": "0:49:09", "remaining_time": "1:28:39", "throughput": 5162.19, "total_tokens": 15224032} {"current_steps": 15135, "total_steps": 42420, "loss": 0.3457, "lr": 0.0008122593136022272, "epoch": 7.135785007072136, "percentage": 35.68, "elapsed_time": "0:49:10", "remaining_time": "1:28:38", "throughput": 5162.2, "total_tokens": 15228512} {"current_steps": 15140, "total_steps": 42420, "loss": 0.2761, "lr": 0.0008120986177462814, "epoch": 7.138142385667138, "percentage": 35.69, "elapsed_time": "0:49:11", "remaining_time": "1:28:37", "throughput": 5162.3, "total_tokens": 15233952} {"current_steps": 15145, "total_steps": 42420, "loss": 0.3388, "lr": 0.0008119378690573387, "epoch": 7.140499764262141, "percentage": 35.7, "elapsed_time": "0:49:11", "remaining_time": "1:28:35", "throughput": 5162.26, "total_tokens": 15237888} {"current_steps": 15150, "total_steps": 42420, "loss": 0.3927, "lr": 0.0008117770675626112, "epoch": 7.142857142857143, "percentage": 35.71, "elapsed_time": "0:49:12", "remaining_time": "1:28:34", "throughput": 5162.37, "total_tokens": 15242880} {"current_steps": 15155, "total_steps": 42420, "loss": 0.3903, "lr": 0.0008116162132893198, "epoch": 7.145214521452146, "percentage": 35.73, "elapsed_time": "0:49:13", "remaining_time": "1:28:33", "throughput": 5162.45, "total_tokens": 15247744} {"current_steps": 15160, "total_steps": 42420, "loss": 0.3798, "lr": 0.0008114553062646946, "epoch": 7.147571900047147, "percentage": 35.74, "elapsed_time": "0:49:14", "remaining_time": "1:28:32", "throughput": 5162.53, "total_tokens": 15252864} {"current_steps": 15165, "total_steps": 42420, "loss": 0.3682, "lr": 0.0008112943465159738, "epoch": 7.1499292786421496, "percentage": 35.75, "elapsed_time": "0:49:15", "remaining_time": "1:28:31", "throughput": 5162.65, "total_tokens": 15258304} {"current_steps": 15170, "total_steps": 42420, "loss": 0.3501, "lr": 0.0008111333340704061, "epoch": 7.152286657237152, "percentage": 35.76, "elapsed_time": "0:49:16", "remaining_time": "1:28:30", "throughput": 5162.71, "total_tokens": 15262944} {"current_steps": 15175, "total_steps": 42420, "loss": 0.3157, "lr": 0.0008109722689552475, "epoch": 7.154644035832154, "percentage": 35.77, "elapsed_time": "0:49:17", "remaining_time": "1:28:29", "throughput": 5162.73, "total_tokens": 15267456} {"current_steps": 15180, "total_steps": 42420, "loss": 0.3433, "lr": 0.0008108111511977639, "epoch": 7.157001414427157, "percentage": 35.79, "elapsed_time": "0:49:18", "remaining_time": "1:28:28", "throughput": 5162.77, "total_tokens": 15272032} {"current_steps": 15185, "total_steps": 42420, "loss": 0.2865, "lr": 0.0008106499808252294, "epoch": 7.159358793022159, "percentage": 35.8, "elapsed_time": "0:49:18", "remaining_time": "1:28:27", "throughput": 5162.81, "total_tokens": 15276672} {"current_steps": 15190, "total_steps": 42420, "loss": 0.3899, "lr": 0.0008104887578649279, "epoch": 7.161716171617162, "percentage": 35.81, "elapsed_time": "0:49:19", "remaining_time": "1:28:26", "throughput": 5162.96, "total_tokens": 15282144} {"current_steps": 15195, "total_steps": 42420, "loss": 0.3568, "lr": 0.0008103274823441515, "epoch": 7.164073550212164, "percentage": 35.82, "elapsed_time": "0:49:20", "remaining_time": "1:28:24", "throughput": 5163.05, "total_tokens": 15287008} {"current_steps": 15200, "total_steps": 42420, "loss": 0.3381, "lr": 0.0008101661542902014, "epoch": 7.166430928807166, "percentage": 35.83, "elapsed_time": "0:49:21", "remaining_time": "1:28:23", "throughput": 5163.22, "total_tokens": 15292480} {"current_steps": 15205, "total_steps": 42420, "loss": 0.3605, "lr": 0.0008100047737303876, "epoch": 7.168788307402169, "percentage": 35.84, "elapsed_time": "0:49:22", "remaining_time": "1:28:22", "throughput": 5163.34, "total_tokens": 15297600} {"current_steps": 15210, "total_steps": 42420, "loss": 0.4152, "lr": 0.0008098433406920293, "epoch": 7.171145685997171, "percentage": 35.86, "elapsed_time": "0:49:23", "remaining_time": "1:28:22", "throughput": 5163.5, "total_tokens": 15303424} {"current_steps": 15215, "total_steps": 42420, "loss": 0.3703, "lr": 0.0008096818552024541, "epoch": 7.173503064592174, "percentage": 35.87, "elapsed_time": "0:49:24", "remaining_time": "1:28:21", "throughput": 5163.87, "total_tokens": 15310784} {"current_steps": 15220, "total_steps": 42420, "loss": 0.3894, "lr": 0.0008095203172889991, "epoch": 7.175860443187176, "percentage": 35.88, "elapsed_time": "0:49:25", "remaining_time": "1:28:20", "throughput": 5164.02, "total_tokens": 15316320} {"current_steps": 15225, "total_steps": 42420, "loss": 0.4041, "lr": 0.0008093587269790096, "epoch": 7.178217821782178, "percentage": 35.89, "elapsed_time": "0:49:26", "remaining_time": "1:28:19", "throughput": 5164.17, "total_tokens": 15321760} {"current_steps": 15230, "total_steps": 42420, "loss": 0.3444, "lr": 0.0008091970842998402, "epoch": 7.180575200377181, "percentage": 35.9, "elapsed_time": "0:49:27", "remaining_time": "1:28:18", "throughput": 5164.32, "total_tokens": 15327392} {"current_steps": 15235, "total_steps": 42420, "loss": 0.3125, "lr": 0.0008090353892788544, "epoch": 7.182932578972183, "percentage": 35.91, "elapsed_time": "0:49:28", "remaining_time": "1:28:17", "throughput": 5164.38, "total_tokens": 15332320} {"current_steps": 15240, "total_steps": 42420, "loss": 0.3296, "lr": 0.0008088736419434244, "epoch": 7.185289957567186, "percentage": 35.93, "elapsed_time": "0:49:29", "remaining_time": "1:28:16", "throughput": 5164.51, "total_tokens": 15337696} {"current_steps": 15245, "total_steps": 42420, "loss": 0.3092, "lr": 0.0008087118423209309, "epoch": 7.187647336162188, "percentage": 35.94, "elapsed_time": "0:49:30", "remaining_time": "1:28:15", "throughput": 5164.62, "total_tokens": 15343008} {"current_steps": 15250, "total_steps": 42420, "loss": 0.3357, "lr": 0.0008085499904387644, "epoch": 7.19000471475719, "percentage": 35.95, "elapsed_time": "0:49:31", "remaining_time": "1:28:14", "throughput": 5164.73, "total_tokens": 15348096} {"current_steps": 15255, "total_steps": 42420, "loss": 0.2691, "lr": 0.0008083880863243234, "epoch": 7.192362093352193, "percentage": 35.96, "elapsed_time": "0:49:32", "remaining_time": "1:28:13", "throughput": 5164.79, "total_tokens": 15352864} {"current_steps": 15260, "total_steps": 42420, "loss": 0.285, "lr": 0.0008082261300050153, "epoch": 7.194719471947194, "percentage": 35.97, "elapsed_time": "0:49:33", "remaining_time": "1:28:12", "throughput": 5164.84, "total_tokens": 15357728} {"current_steps": 15265, "total_steps": 42420, "loss": 0.3243, "lr": 0.000808064121508257, "epoch": 7.197076850542197, "percentage": 35.99, "elapsed_time": "0:49:34", "remaining_time": "1:28:11", "throughput": 5164.85, "total_tokens": 15362240} {"current_steps": 15270, "total_steps": 42420, "loss": 0.4136, "lr": 0.0008079020608614733, "epoch": 7.199434229137199, "percentage": 36.0, "elapsed_time": "0:49:35", "remaining_time": "1:28:10", "throughput": 5165.03, "total_tokens": 15368128} {"current_steps": 15275, "total_steps": 42420, "loss": 0.3516, "lr": 0.0008077399480920988, "epoch": 7.2017916077322015, "percentage": 36.01, "elapsed_time": "0:49:36", "remaining_time": "1:28:08", "throughput": 5164.95, "total_tokens": 15371648} {"current_steps": 15280, "total_steps": 42420, "loss": 0.3309, "lr": 0.0008075777832275761, "epoch": 7.204148986327204, "percentage": 36.02, "elapsed_time": "0:49:37", "remaining_time": "1:28:07", "throughput": 5164.97, "total_tokens": 15376320} {"current_steps": 15285, "total_steps": 42420, "loss": 0.3564, "lr": 0.0008074155662953573, "epoch": 7.206506364922206, "percentage": 36.03, "elapsed_time": "0:49:37", "remaining_time": "1:28:06", "throughput": 5164.99, "total_tokens": 15380864} {"current_steps": 15290, "total_steps": 42420, "loss": 0.3208, "lr": 0.0008072532973229025, "epoch": 7.208863743517209, "percentage": 36.04, "elapsed_time": "0:49:38", "remaining_time": "1:28:05", "throughput": 5165.22, "total_tokens": 15387168} {"current_steps": 15295, "total_steps": 42420, "loss": 0.3754, "lr": 0.0008070909763376814, "epoch": 7.211221122112211, "percentage": 36.06, "elapsed_time": "0:49:39", "remaining_time": "1:28:04", "throughput": 5165.34, "total_tokens": 15392384} {"current_steps": 15300, "total_steps": 42420, "loss": 0.3085, "lr": 0.0008069286033671721, "epoch": 7.2135785007072135, "percentage": 36.07, "elapsed_time": "0:49:40", "remaining_time": "1:28:03", "throughput": 5165.38, "total_tokens": 15397216} {"current_steps": 15305, "total_steps": 42420, "loss": 0.3609, "lr": 0.0008067661784388616, "epoch": 7.215935879302216, "percentage": 36.08, "elapsed_time": "0:49:41", "remaining_time": "1:28:02", "throughput": 5165.41, "total_tokens": 15401920} {"current_steps": 15310, "total_steps": 42420, "loss": 0.3562, "lr": 0.0008066037015802458, "epoch": 7.218293257897218, "percentage": 36.09, "elapsed_time": "0:49:42", "remaining_time": "1:28:01", "throughput": 5165.51, "total_tokens": 15407136} {"current_steps": 15315, "total_steps": 42420, "loss": 0.3412, "lr": 0.000806441172818829, "epoch": 7.220650636492221, "percentage": 36.1, "elapsed_time": "0:49:43", "remaining_time": "1:28:00", "throughput": 5165.56, "total_tokens": 15411936} {"current_steps": 15320, "total_steps": 42420, "loss": 0.3525, "lr": 0.0008062785921821247, "epoch": 7.223008015087223, "percentage": 36.12, "elapsed_time": "0:49:44", "remaining_time": "1:27:59", "throughput": 5165.68, "total_tokens": 15417184} {"current_steps": 15325, "total_steps": 42420, "loss": 0.3418, "lr": 0.0008061159596976553, "epoch": 7.2253653936822255, "percentage": 36.13, "elapsed_time": "0:49:45", "remaining_time": "1:27:58", "throughput": 5165.69, "total_tokens": 15421632} {"current_steps": 15330, "total_steps": 42420, "loss": 0.3597, "lr": 0.0008059532753929512, "epoch": 7.227722772277228, "percentage": 36.14, "elapsed_time": "0:49:46", "remaining_time": "1:27:57", "throughput": 5165.8, "total_tokens": 15426624} {"current_steps": 15335, "total_steps": 42420, "loss": 0.3786, "lr": 0.0008057905392955525, "epoch": 7.23008015087223, "percentage": 36.15, "elapsed_time": "0:49:47", "remaining_time": "1:27:56", "throughput": 5165.9, "total_tokens": 15431520} {"current_steps": 15340, "total_steps": 42420, "loss": 0.33, "lr": 0.0008056277514330075, "epoch": 7.232437529467233, "percentage": 36.16, "elapsed_time": "0:49:48", "remaining_time": "1:27:54", "throughput": 5165.96, "total_tokens": 15436448} {"current_steps": 15345, "total_steps": 42420, "loss": 0.3319, "lr": 0.0008054649118328734, "epoch": 7.234794908062235, "percentage": 36.17, "elapsed_time": "0:49:49", "remaining_time": "1:27:54", "throughput": 5166.29, "total_tokens": 15443968} {"current_steps": 15350, "total_steps": 42420, "loss": 0.3295, "lr": 0.0008053020205227162, "epoch": 7.2371522866572375, "percentage": 36.19, "elapsed_time": "0:49:50", "remaining_time": "1:27:53", "throughput": 5166.39, "total_tokens": 15449440} {"current_steps": 15355, "total_steps": 42420, "loss": 0.3334, "lr": 0.0008051390775301109, "epoch": 7.23950966525224, "percentage": 36.2, "elapsed_time": "0:49:51", "remaining_time": "1:27:52", "throughput": 5166.41, "total_tokens": 15453888} {"current_steps": 15360, "total_steps": 42420, "loss": 0.3387, "lr": 0.0008049760828826404, "epoch": 7.2418670438472414, "percentage": 36.21, "elapsed_time": "0:49:52", "remaining_time": "1:27:51", "throughput": 5166.41, "total_tokens": 15458080} {"current_steps": 15365, "total_steps": 42420, "loss": 0.2759, "lr": 0.0008048130366078974, "epoch": 7.244224422442244, "percentage": 36.22, "elapsed_time": "0:49:52", "remaining_time": "1:27:49", "throughput": 5166.42, "total_tokens": 15462304} {"current_steps": 15370, "total_steps": 42420, "loss": 0.3543, "lr": 0.0008046499387334828, "epoch": 7.246581801037246, "percentage": 36.23, "elapsed_time": "0:49:53", "remaining_time": "1:27:48", "throughput": 5166.41, "total_tokens": 15466752} {"current_steps": 15375, "total_steps": 42420, "loss": 0.329, "lr": 0.0008044867892870063, "epoch": 7.248939179632249, "percentage": 36.24, "elapsed_time": "0:49:54", "remaining_time": "1:27:47", "throughput": 5166.56, "total_tokens": 15471936} {"current_steps": 15380, "total_steps": 42420, "loss": 0.2911, "lr": 0.0008043235882960861, "epoch": 7.251296558227251, "percentage": 36.26, "elapsed_time": "0:49:55", "remaining_time": "1:27:46", "throughput": 5166.59, "total_tokens": 15476544} {"current_steps": 15385, "total_steps": 42420, "loss": 0.362, "lr": 0.0008041603357883495, "epoch": 7.2536539368222535, "percentage": 36.27, "elapsed_time": "0:49:56", "remaining_time": "1:27:45", "throughput": 5166.6, "total_tokens": 15480960} {"current_steps": 15390, "total_steps": 42420, "loss": 0.3196, "lr": 0.0008039970317914324, "epoch": 7.256011315417256, "percentage": 36.28, "elapsed_time": "0:49:57", "remaining_time": "1:27:44", "throughput": 5166.69, "total_tokens": 15486208} {"current_steps": 15395, "total_steps": 42420, "loss": 0.2736, "lr": 0.0008038336763329796, "epoch": 7.258368694012258, "percentage": 36.29, "elapsed_time": "0:49:58", "remaining_time": "1:27:43", "throughput": 5166.79, "total_tokens": 15491328} {"current_steps": 15400, "total_steps": 42420, "loss": 0.3059, "lr": 0.0008036702694406441, "epoch": 7.260726072607261, "percentage": 36.3, "elapsed_time": "0:49:59", "remaining_time": "1:27:42", "throughput": 5166.87, "total_tokens": 15496320} {"current_steps": 15405, "total_steps": 42420, "loss": 0.3629, "lr": 0.000803506811142088, "epoch": 7.263083451202263, "percentage": 36.32, "elapsed_time": "0:50:00", "remaining_time": "1:27:41", "throughput": 5166.96, "total_tokens": 15501152} {"current_steps": 15410, "total_steps": 42420, "loss": 0.2636, "lr": 0.000803343301464982, "epoch": 7.2654408297972655, "percentage": 36.33, "elapsed_time": "0:50:00", "remaining_time": "1:27:39", "throughput": 5166.97, "total_tokens": 15505440} {"current_steps": 15415, "total_steps": 42420, "loss": 0.3781, "lr": 0.0008031797404370055, "epoch": 7.267798208392268, "percentage": 36.34, "elapsed_time": "0:50:01", "remaining_time": "1:27:38", "throughput": 5167.09, "total_tokens": 15510624} {"current_steps": 15420, "total_steps": 42420, "loss": 0.2759, "lr": 0.0008030161280858468, "epoch": 7.27015558698727, "percentage": 36.35, "elapsed_time": "0:50:02", "remaining_time": "1:27:37", "throughput": 5167.2, "total_tokens": 15515808} {"current_steps": 15425, "total_steps": 42420, "loss": 0.2662, "lr": 0.0008028524644392024, "epoch": 7.272512965582273, "percentage": 36.36, "elapsed_time": "0:50:03", "remaining_time": "1:27:36", "throughput": 5167.18, "total_tokens": 15520000} {"current_steps": 15430, "total_steps": 42420, "loss": 0.3623, "lr": 0.000802688749524778, "epoch": 7.274870344177275, "percentage": 36.37, "elapsed_time": "0:50:04", "remaining_time": "1:27:35", "throughput": 5167.32, "total_tokens": 15525184} {"current_steps": 15435, "total_steps": 42420, "loss": 0.3175, "lr": 0.0008025249833702876, "epoch": 7.2772277227722775, "percentage": 36.39, "elapsed_time": "0:50:05", "remaining_time": "1:27:34", "throughput": 5167.55, "total_tokens": 15531328} {"current_steps": 15440, "total_steps": 42420, "loss": 0.3641, "lr": 0.0008023611660034541, "epoch": 7.27958510136728, "percentage": 36.4, "elapsed_time": "0:50:06", "remaining_time": "1:27:33", "throughput": 5167.68, "total_tokens": 15536768} {"current_steps": 15445, "total_steps": 42420, "loss": 0.311, "lr": 0.0008021972974520091, "epoch": 7.281942479962282, "percentage": 36.41, "elapsed_time": "0:50:07", "remaining_time": "1:27:32", "throughput": 5167.74, "total_tokens": 15541696} {"current_steps": 15450, "total_steps": 42420, "loss": 0.2979, "lr": 0.0008020333777436925, "epoch": 7.284299858557285, "percentage": 36.42, "elapsed_time": "0:50:08", "remaining_time": "1:27:31", "throughput": 5167.76, "total_tokens": 15546400} {"current_steps": 15455, "total_steps": 42420, "loss": 0.3734, "lr": 0.0008018694069062534, "epoch": 7.286657237152287, "percentage": 36.43, "elapsed_time": "0:50:09", "remaining_time": "1:27:30", "throughput": 5167.9, "total_tokens": 15552000} {"current_steps": 15460, "total_steps": 42420, "loss": 0.3493, "lr": 0.0008017053849674491, "epoch": 7.2890146157472895, "percentage": 36.45, "elapsed_time": "0:50:10", "remaining_time": "1:27:29", "throughput": 5168.03, "total_tokens": 15557312} {"current_steps": 15465, "total_steps": 42420, "loss": 0.3414, "lr": 0.0008015413119550457, "epoch": 7.291371994342291, "percentage": 36.46, "elapsed_time": "0:50:11", "remaining_time": "1:27:28", "throughput": 5168.25, "total_tokens": 15563648} {"current_steps": 15470, "total_steps": 42420, "loss": 0.3405, "lr": 0.0008013771878968183, "epoch": 7.293729372937293, "percentage": 36.47, "elapsed_time": "0:50:12", "remaining_time": "1:27:27", "throughput": 5168.41, "total_tokens": 15569376} {"current_steps": 15475, "total_steps": 42420, "loss": 0.3733, "lr": 0.0008012130128205498, "epoch": 7.296086751532296, "percentage": 36.48, "elapsed_time": "0:50:13", "remaining_time": "1:27:26", "throughput": 5168.4, "total_tokens": 15573568} {"current_steps": 15480, "total_steps": 42420, "loss": 0.3087, "lr": 0.0008010487867540329, "epoch": 7.298444130127298, "percentage": 36.49, "elapsed_time": "0:50:14", "remaining_time": "1:27:25", "throughput": 5168.65, "total_tokens": 15579936} {"current_steps": 15485, "total_steps": 42420, "loss": 0.3507, "lr": 0.0008008845097250676, "epoch": 7.300801508722301, "percentage": 36.5, "elapsed_time": "0:50:15", "remaining_time": "1:27:25", "throughput": 5168.91, "total_tokens": 15586688} {"current_steps": 15490, "total_steps": 42420, "loss": 0.3606, "lr": 0.0008007201817614635, "epoch": 7.303158887317303, "percentage": 36.52, "elapsed_time": "0:50:16", "remaining_time": "1:27:24", "throughput": 5169.01, "total_tokens": 15592096} {"current_steps": 15495, "total_steps": 42420, "loss": 0.3408, "lr": 0.0008005558028910387, "epoch": 7.305516265912305, "percentage": 36.53, "elapsed_time": "0:50:17", "remaining_time": "1:27:23", "throughput": 5169.05, "total_tokens": 15596768} {"current_steps": 15500, "total_steps": 42420, "loss": 0.3273, "lr": 0.0008003913731416197, "epoch": 7.307873644507308, "percentage": 36.54, "elapsed_time": "0:50:18", "remaining_time": "1:27:21", "throughput": 5169.05, "total_tokens": 15600992} {"current_steps": 15505, "total_steps": 42420, "loss": 0.2966, "lr": 0.0008002268925410414, "epoch": 7.31023102310231, "percentage": 36.55, "elapsed_time": "0:50:18", "remaining_time": "1:27:20", "throughput": 5169.06, "total_tokens": 15605344} {"current_steps": 15510, "total_steps": 42420, "loss": 0.3307, "lr": 0.000800062361117148, "epoch": 7.312588401697313, "percentage": 36.56, "elapsed_time": "0:50:19", "remaining_time": "1:27:19", "throughput": 5169.25, "total_tokens": 15611072} {"current_steps": 15515, "total_steps": 42420, "loss": 0.3214, "lr": 0.0007998977788977915, "epoch": 7.314945780292315, "percentage": 36.57, "elapsed_time": "0:50:20", "remaining_time": "1:27:18", "throughput": 5169.26, "total_tokens": 15615392} {"current_steps": 15520, "total_steps": 42420, "loss": 0.3321, "lr": 0.0007997331459108331, "epoch": 7.317303158887317, "percentage": 36.59, "elapsed_time": "0:50:21", "remaining_time": "1:27:17", "throughput": 5169.3, "total_tokens": 15620160} {"current_steps": 15525, "total_steps": 42420, "loss": 0.2907, "lr": 0.0007995684621841419, "epoch": 7.31966053748232, "percentage": 36.6, "elapsed_time": "0:50:22", "remaining_time": "1:27:16", "throughput": 5169.4, "total_tokens": 15625216} {"current_steps": 15530, "total_steps": 42420, "loss": 0.252, "lr": 0.0007994037277455968, "epoch": 7.322017916077322, "percentage": 36.61, "elapsed_time": "0:50:23", "remaining_time": "1:27:15", "throughput": 5169.49, "total_tokens": 15630208} {"current_steps": 15535, "total_steps": 42420, "loss": 0.439, "lr": 0.0007992389426230842, "epoch": 7.324375294672325, "percentage": 36.62, "elapsed_time": "0:50:24", "remaining_time": "1:27:14", "throughput": 5169.53, "total_tokens": 15634784} {"current_steps": 15540, "total_steps": 42420, "loss": 0.358, "lr": 0.0007990741068444992, "epoch": 7.326732673267327, "percentage": 36.63, "elapsed_time": "0:50:25", "remaining_time": "1:27:12", "throughput": 5169.59, "total_tokens": 15639712} {"current_steps": 15545, "total_steps": 42420, "loss": 0.3054, "lr": 0.0007989092204377461, "epoch": 7.329090051862329, "percentage": 36.65, "elapsed_time": "0:50:26", "remaining_time": "1:27:12", "throughput": 5169.73, "total_tokens": 15645120} {"current_steps": 15550, "total_steps": 42420, "loss": 0.3707, "lr": 0.000798744283430737, "epoch": 7.331447430457332, "percentage": 36.66, "elapsed_time": "0:50:27", "remaining_time": "1:27:11", "throughput": 5169.9, "total_tokens": 15650816} {"current_steps": 15555, "total_steps": 42420, "loss": 0.3323, "lr": 0.0007985792958513931, "epoch": 7.333804809052333, "percentage": 36.67, "elapsed_time": "0:50:28", "remaining_time": "1:27:10", "throughput": 5170.03, "total_tokens": 15656192} {"current_steps": 15560, "total_steps": 42420, "loss": 0.3609, "lr": 0.0007984142577276441, "epoch": 7.336162187647336, "percentage": 36.68, "elapsed_time": "0:50:29", "remaining_time": "1:27:09", "throughput": 5170.19, "total_tokens": 15661664} {"current_steps": 15565, "total_steps": 42420, "loss": 0.3104, "lr": 0.0007982491690874281, "epoch": 7.338519566242338, "percentage": 36.69, "elapsed_time": "0:50:30", "remaining_time": "1:27:08", "throughput": 5170.34, "total_tokens": 15667072} {"current_steps": 15570, "total_steps": 42420, "loss": 0.339, "lr": 0.0007980840299586916, "epoch": 7.3408769448373405, "percentage": 36.7, "elapsed_time": "0:50:30", "remaining_time": "1:27:06", "throughput": 5170.34, "total_tokens": 15671296} {"current_steps": 15575, "total_steps": 42420, "loss": 0.2905, "lr": 0.00079791884036939, "epoch": 7.343234323432343, "percentage": 36.72, "elapsed_time": "0:50:31", "remaining_time": "1:27:05", "throughput": 5170.45, "total_tokens": 15676320} {"current_steps": 15580, "total_steps": 42420, "loss": 0.364, "lr": 0.0007977536003474872, "epoch": 7.345591702027345, "percentage": 36.73, "elapsed_time": "0:50:32", "remaining_time": "1:27:04", "throughput": 5170.52, "total_tokens": 15681280} {"current_steps": 15585, "total_steps": 42420, "loss": 0.3585, "lr": 0.0007975883099209554, "epoch": 7.347949080622348, "percentage": 36.74, "elapsed_time": "0:50:33", "remaining_time": "1:27:03", "throughput": 5170.51, "total_tokens": 15685536} {"current_steps": 15590, "total_steps": 42420, "loss": 0.3296, "lr": 0.0007974229691177755, "epoch": 7.35030645921735, "percentage": 36.75, "elapsed_time": "0:50:34", "remaining_time": "1:27:02", "throughput": 5170.5, "total_tokens": 15690112} {"current_steps": 15595, "total_steps": 42420, "loss": 0.3853, "lr": 0.000797257577965937, "epoch": 7.3526638378123526, "percentage": 36.76, "elapsed_time": "0:50:35", "remaining_time": "1:27:01", "throughput": 5170.63, "total_tokens": 15695488} {"current_steps": 15600, "total_steps": 42420, "loss": 0.3735, "lr": 0.0007970921364934374, "epoch": 7.355021216407355, "percentage": 36.78, "elapsed_time": "0:50:36", "remaining_time": "1:27:00", "throughput": 5170.6, "total_tokens": 15699616} {"current_steps": 15605, "total_steps": 42420, "loss": 0.3743, "lr": 0.0007969266447282837, "epoch": 7.357378595002357, "percentage": 36.79, "elapsed_time": "0:50:37", "remaining_time": "1:26:58", "throughput": 5170.53, "total_tokens": 15703360} {"current_steps": 15610, "total_steps": 42420, "loss": 0.3411, "lr": 0.0007967611026984904, "epoch": 7.35973597359736, "percentage": 36.8, "elapsed_time": "0:50:38", "remaining_time": "1:26:57", "throughput": 5170.67, "total_tokens": 15709088} {"current_steps": 15615, "total_steps": 42420, "loss": 0.346, "lr": 0.0007965955104320812, "epoch": 7.362093352192362, "percentage": 36.81, "elapsed_time": "0:50:38", "remaining_time": "1:26:56", "throughput": 5170.71, "total_tokens": 15713728} {"current_steps": 15620, "total_steps": 42420, "loss": 0.3611, "lr": 0.000796429867957088, "epoch": 7.364450730787365, "percentage": 36.82, "elapsed_time": "0:50:39", "remaining_time": "1:26:55", "throughput": 5170.69, "total_tokens": 15717824} {"current_steps": 15625, "total_steps": 42420, "loss": 0.3061, "lr": 0.0007962641753015514, "epoch": 7.366808109382367, "percentage": 36.83, "elapsed_time": "0:50:40", "remaining_time": "1:26:54", "throughput": 5170.72, "total_tokens": 15722368} {"current_steps": 15630, "total_steps": 42420, "loss": 0.3623, "lr": 0.00079609843249352, "epoch": 7.369165487977369, "percentage": 36.85, "elapsed_time": "0:50:41", "remaining_time": "1:26:53", "throughput": 5170.8, "total_tokens": 15727520} {"current_steps": 15635, "total_steps": 42420, "loss": 0.3512, "lr": 0.0007959326395610515, "epoch": 7.371522866572372, "percentage": 36.86, "elapsed_time": "0:50:42", "remaining_time": "1:26:52", "throughput": 5170.85, "total_tokens": 15732352} {"current_steps": 15640, "total_steps": 42420, "loss": 0.329, "lr": 0.0007957667965322118, "epoch": 7.373880245167374, "percentage": 36.87, "elapsed_time": "0:50:43", "remaining_time": "1:26:51", "throughput": 5171.02, "total_tokens": 15738016} {"current_steps": 15645, "total_steps": 42420, "loss": 0.3571, "lr": 0.0007956009034350753, "epoch": 7.376237623762377, "percentage": 36.88, "elapsed_time": "0:50:44", "remaining_time": "1:26:50", "throughput": 5171.06, "total_tokens": 15742688} {"current_steps": 15650, "total_steps": 42420, "loss": 0.342, "lr": 0.0007954349602977247, "epoch": 7.378595002357379, "percentage": 36.89, "elapsed_time": "0:50:45", "remaining_time": "1:26:49", "throughput": 5171.1, "total_tokens": 15747520} {"current_steps": 15655, "total_steps": 42420, "loss": 0.3719, "lr": 0.0007952689671482517, "epoch": 7.380952380952381, "percentage": 36.9, "elapsed_time": "0:50:46", "remaining_time": "1:26:48", "throughput": 5171.37, "total_tokens": 15754496} {"current_steps": 15660, "total_steps": 42420, "loss": 0.3531, "lr": 0.0007951029240147558, "epoch": 7.383309759547384, "percentage": 36.92, "elapsed_time": "0:50:47", "remaining_time": "1:26:47", "throughput": 5171.54, "total_tokens": 15760160} {"current_steps": 15665, "total_steps": 42420, "loss": 0.3232, "lr": 0.0007949368309253456, "epoch": 7.385667138142385, "percentage": 36.93, "elapsed_time": "0:50:48", "remaining_time": "1:26:46", "throughput": 5171.64, "total_tokens": 15765504} {"current_steps": 15670, "total_steps": 42420, "loss": 0.3136, "lr": 0.0007947706879081376, "epoch": 7.388024516737388, "percentage": 36.94, "elapsed_time": "0:50:49", "remaining_time": "1:26:45", "throughput": 5171.57, "total_tokens": 15769312} {"current_steps": 15675, "total_steps": 42420, "loss": 0.3382, "lr": 0.0007946044949912569, "epoch": 7.39038189533239, "percentage": 36.95, "elapsed_time": "0:50:50", "remaining_time": "1:26:44", "throughput": 5171.64, "total_tokens": 15774144} {"current_steps": 15680, "total_steps": 42420, "loss": 0.2674, "lr": 0.0007944382522028375, "epoch": 7.3927392739273925, "percentage": 36.96, "elapsed_time": "0:50:50", "remaining_time": "1:26:42", "throughput": 5171.59, "total_tokens": 15778016} {"current_steps": 15685, "total_steps": 42420, "loss": 0.3608, "lr": 0.0007942719595710213, "epoch": 7.395096652522395, "percentage": 36.98, "elapsed_time": "0:50:51", "remaining_time": "1:26:41", "throughput": 5171.61, "total_tokens": 15782720} {"current_steps": 15690, "total_steps": 42420, "loss": 0.3667, "lr": 0.0007941056171239586, "epoch": 7.397454031117397, "percentage": 36.99, "elapsed_time": "0:50:52", "remaining_time": "1:26:40", "throughput": 5171.7, "total_tokens": 15787776} {"current_steps": 15695, "total_steps": 42420, "loss": 0.3289, "lr": 0.0007939392248898087, "epoch": 7.3998114097124, "percentage": 37.0, "elapsed_time": "0:50:53", "remaining_time": "1:26:39", "throughput": 5171.8, "total_tokens": 15793024} {"current_steps": 15700, "total_steps": 42420, "loss": 0.3413, "lr": 0.0007937727828967387, "epoch": 7.402168788307402, "percentage": 37.01, "elapsed_time": "0:50:54", "remaining_time": "1:26:38", "throughput": 5171.81, "total_tokens": 15797312} {"current_steps": 15705, "total_steps": 42420, "loss": 0.3639, "lr": 0.0007936062911729247, "epoch": 7.4045261669024045, "percentage": 37.02, "elapsed_time": "0:50:55", "remaining_time": "1:26:37", "throughput": 5171.77, "total_tokens": 15801216} {"current_steps": 15710, "total_steps": 42420, "loss": 0.3527, "lr": 0.0007934397497465506, "epoch": 7.406883545497407, "percentage": 37.03, "elapsed_time": "0:50:56", "remaining_time": "1:26:36", "throughput": 5171.8, "total_tokens": 15805760} {"current_steps": 15715, "total_steps": 42420, "loss": 0.3342, "lr": 0.0007932731586458093, "epoch": 7.409240924092409, "percentage": 37.05, "elapsed_time": "0:50:57", "remaining_time": "1:26:35", "throughput": 5171.94, "total_tokens": 15811136} {"current_steps": 15720, "total_steps": 42420, "loss": 0.3348, "lr": 0.0007931065178989016, "epoch": 7.411598302687412, "percentage": 37.06, "elapsed_time": "0:50:58", "remaining_time": "1:26:33", "throughput": 5172.02, "total_tokens": 15816224} {"current_steps": 15725, "total_steps": 42420, "loss": 0.3778, "lr": 0.0007929398275340373, "epoch": 7.413955681282414, "percentage": 37.07, "elapsed_time": "0:50:58", "remaining_time": "1:26:32", "throughput": 5172.05, "total_tokens": 15820768} {"current_steps": 15730, "total_steps": 42420, "loss": 0.2599, "lr": 0.0007927730875794339, "epoch": 7.4163130598774165, "percentage": 37.08, "elapsed_time": "0:50:59", "remaining_time": "1:26:31", "throughput": 5172.04, "total_tokens": 15825184} {"current_steps": 15735, "total_steps": 42420, "loss": 0.2744, "lr": 0.0007926062980633178, "epoch": 7.418670438472419, "percentage": 37.09, "elapsed_time": "0:51:00", "remaining_time": "1:26:30", "throughput": 5172.02, "total_tokens": 15829376} {"current_steps": 15740, "total_steps": 42420, "loss": 0.3487, "lr": 0.0007924394590139237, "epoch": 7.421027817067421, "percentage": 37.11, "elapsed_time": "0:51:01", "remaining_time": "1:26:29", "throughput": 5172.21, "total_tokens": 15835264} {"current_steps": 15745, "total_steps": 42420, "loss": 0.36, "lr": 0.0007922725704594944, "epoch": 7.423385195662424, "percentage": 37.12, "elapsed_time": "0:51:02", "remaining_time": "1:26:28", "throughput": 5172.24, "total_tokens": 15839872} {"current_steps": 15750, "total_steps": 42420, "loss": 0.3413, "lr": 0.0007921056324282814, "epoch": 7.425742574257426, "percentage": 37.13, "elapsed_time": "0:51:03", "remaining_time": "1:26:27", "throughput": 5172.24, "total_tokens": 15844480} {"current_steps": 15755, "total_steps": 42420, "loss": 0.3508, "lr": 0.0007919386449485446, "epoch": 7.428099952852428, "percentage": 37.14, "elapsed_time": "0:51:04", "remaining_time": "1:26:26", "throughput": 5172.22, "total_tokens": 15848640} {"current_steps": 15760, "total_steps": 42420, "loss": 0.3425, "lr": 0.000791771608048552, "epoch": 7.43045733144743, "percentage": 37.15, "elapsed_time": "0:51:05", "remaining_time": "1:26:25", "throughput": 5172.32, "total_tokens": 15854048} {"current_steps": 15765, "total_steps": 42420, "loss": 0.3653, "lr": 0.0007916045217565801, "epoch": 7.432814710042432, "percentage": 37.16, "elapsed_time": "0:51:06", "remaining_time": "1:26:23", "throughput": 5172.34, "total_tokens": 15858432} {"current_steps": 15770, "total_steps": 42420, "loss": 0.3657, "lr": 0.000791437386100914, "epoch": 7.435172088637435, "percentage": 37.18, "elapsed_time": "0:51:06", "remaining_time": "1:26:22", "throughput": 5172.47, "total_tokens": 15863840} {"current_steps": 15775, "total_steps": 42420, "loss": 0.3158, "lr": 0.0007912702011098467, "epoch": 7.437529467232437, "percentage": 37.19, "elapsed_time": "0:51:07", "remaining_time": "1:26:21", "throughput": 5172.55, "total_tokens": 15868768} {"current_steps": 15780, "total_steps": 42420, "loss": 0.3342, "lr": 0.0007911029668116798, "epoch": 7.43988684582744, "percentage": 37.2, "elapsed_time": "0:51:08", "remaining_time": "1:26:20", "throughput": 5172.65, "total_tokens": 15873920} {"current_steps": 15785, "total_steps": 42420, "loss": 0.3889, "lr": 0.0007909356832347232, "epoch": 7.442244224422442, "percentage": 37.21, "elapsed_time": "0:51:09", "remaining_time": "1:26:19", "throughput": 5172.71, "total_tokens": 15878784} {"current_steps": 15790, "total_steps": 42420, "loss": 0.3622, "lr": 0.0007907683504072954, "epoch": 7.4446016030174444, "percentage": 37.22, "elapsed_time": "0:51:10", "remaining_time": "1:26:18", "throughput": 5172.66, "total_tokens": 15882784} {"current_steps": 15795, "total_steps": 42420, "loss": 0.3449, "lr": 0.0007906009683577227, "epoch": 7.446958981612447, "percentage": 37.23, "elapsed_time": "0:51:11", "remaining_time": "1:26:17", "throughput": 5172.9, "total_tokens": 15889312} {"current_steps": 15800, "total_steps": 42420, "loss": 0.3368, "lr": 0.0007904335371143401, "epoch": 7.449316360207449, "percentage": 37.25, "elapsed_time": "0:51:12", "remaining_time": "1:26:16", "throughput": 5172.97, "total_tokens": 15894272} {"current_steps": 15805, "total_steps": 42420, "loss": 0.344, "lr": 0.0007902660567054911, "epoch": 7.451673738802452, "percentage": 37.26, "elapsed_time": "0:51:13", "remaining_time": "1:26:15", "throughput": 5173.06, "total_tokens": 15899328} {"current_steps": 15810, "total_steps": 42420, "loss": 0.2944, "lr": 0.0007900985271595271, "epoch": 7.454031117397454, "percentage": 37.27, "elapsed_time": "0:51:14", "remaining_time": "1:26:14", "throughput": 5173.12, "total_tokens": 15904096} {"current_steps": 15815, "total_steps": 42420, "loss": 0.4262, "lr": 0.0007899309485048081, "epoch": 7.4563884959924565, "percentage": 37.28, "elapsed_time": "0:51:15", "remaining_time": "1:26:13", "throughput": 5173.17, "total_tokens": 15909088} {"current_steps": 15820, "total_steps": 42420, "loss": 0.3135, "lr": 0.0007897633207697021, "epoch": 7.458745874587459, "percentage": 37.29, "elapsed_time": "0:51:16", "remaining_time": "1:26:13", "throughput": 5173.59, "total_tokens": 15917120} {"current_steps": 15825, "total_steps": 42420, "loss": 0.3475, "lr": 0.000789595643982586, "epoch": 7.461103253182461, "percentage": 37.31, "elapsed_time": "0:51:17", "remaining_time": "1:26:11", "throughput": 5173.7, "total_tokens": 15922208} {"current_steps": 15830, "total_steps": 42420, "loss": 0.3449, "lr": 0.0007894279181718442, "epoch": 7.463460631777464, "percentage": 37.32, "elapsed_time": "0:51:18", "remaining_time": "1:26:10", "throughput": 5173.75, "total_tokens": 15926848} {"current_steps": 15835, "total_steps": 42420, "loss": 0.3066, "lr": 0.0007892601433658704, "epoch": 7.465818010372466, "percentage": 37.33, "elapsed_time": "0:51:19", "remaining_time": "1:26:09", "throughput": 5173.8, "total_tokens": 15931680} {"current_steps": 15840, "total_steps": 42420, "loss": 0.2813, "lr": 0.0007890923195930654, "epoch": 7.4681753889674685, "percentage": 37.34, "elapsed_time": "0:51:20", "remaining_time": "1:26:08", "throughput": 5173.85, "total_tokens": 15936640} {"current_steps": 15845, "total_steps": 42420, "loss": 0.3787, "lr": 0.0007889244468818392, "epoch": 7.470532767562471, "percentage": 37.35, "elapsed_time": "0:51:21", "remaining_time": "1:26:07", "throughput": 5173.84, "total_tokens": 15940768} {"current_steps": 15850, "total_steps": 42420, "loss": 0.331, "lr": 0.0007887565252606099, "epoch": 7.472890146157473, "percentage": 37.36, "elapsed_time": "0:51:21", "remaining_time": "1:26:06", "throughput": 5173.89, "total_tokens": 15945728} {"current_steps": 15855, "total_steps": 42420, "loss": 0.3969, "lr": 0.0007885885547578038, "epoch": 7.475247524752476, "percentage": 37.38, "elapsed_time": "0:51:22", "remaining_time": "1:26:05", "throughput": 5173.93, "total_tokens": 15950208} {"current_steps": 15860, "total_steps": 42420, "loss": 0.3683, "lr": 0.0007884205354018551, "epoch": 7.477604903347478, "percentage": 37.39, "elapsed_time": "0:51:23", "remaining_time": "1:26:04", "throughput": 5174.13, "total_tokens": 15956256} {"current_steps": 15865, "total_steps": 42420, "loss": 0.3872, "lr": 0.0007882524672212071, "epoch": 7.47996228194248, "percentage": 37.4, "elapsed_time": "0:51:24", "remaining_time": "1:26:03", "throughput": 5174.26, "total_tokens": 15961440} {"current_steps": 15870, "total_steps": 42420, "loss": 0.3474, "lr": 0.0007880843502443104, "epoch": 7.482319660537482, "percentage": 37.41, "elapsed_time": "0:51:25", "remaining_time": "1:26:02", "throughput": 5174.38, "total_tokens": 15966688} {"current_steps": 15875, "total_steps": 42420, "loss": 0.3769, "lr": 0.0007879161844996247, "epoch": 7.484677039132484, "percentage": 37.42, "elapsed_time": "0:51:26", "remaining_time": "1:26:01", "throughput": 5174.43, "total_tokens": 15971328} {"current_steps": 15880, "total_steps": 42420, "loss": 0.3477, "lr": 0.0007877479700156176, "epoch": 7.487034417727487, "percentage": 37.44, "elapsed_time": "0:51:27", "remaining_time": "1:26:00", "throughput": 5174.56, "total_tokens": 15976512} {"current_steps": 15885, "total_steps": 42420, "loss": 0.2912, "lr": 0.0007875797068207646, "epoch": 7.489391796322489, "percentage": 37.45, "elapsed_time": "0:51:28", "remaining_time": "1:25:59", "throughput": 5174.65, "total_tokens": 15981600} {"current_steps": 15890, "total_steps": 42420, "loss": 0.4134, "lr": 0.0007874113949435503, "epoch": 7.491749174917492, "percentage": 37.46, "elapsed_time": "0:51:29", "remaining_time": "1:25:58", "throughput": 5174.79, "total_tokens": 15987264} {"current_steps": 15895, "total_steps": 42420, "loss": 0.3433, "lr": 0.0007872430344124667, "epoch": 7.494106553512494, "percentage": 37.47, "elapsed_time": "0:51:30", "remaining_time": "1:25:57", "throughput": 5174.85, "total_tokens": 15992288} {"current_steps": 15900, "total_steps": 42420, "loss": 0.3581, "lr": 0.0007870746252560146, "epoch": 7.496463932107496, "percentage": 37.48, "elapsed_time": "0:51:31", "remaining_time": "1:25:56", "throughput": 5174.88, "total_tokens": 15996992} {"current_steps": 15905, "total_steps": 42420, "loss": 0.3654, "lr": 0.0007869061675027025, "epoch": 7.498821310702499, "percentage": 37.49, "elapsed_time": "0:51:32", "remaining_time": "1:25:54", "throughput": 5174.91, "total_tokens": 16001728} {"current_steps": 15910, "total_steps": 42420, "loss": 0.3414, "lr": 0.0007867376611810478, "epoch": 7.501178689297501, "percentage": 37.51, "elapsed_time": "0:51:33", "remaining_time": "1:25:53", "throughput": 5175.0, "total_tokens": 16007072} {"current_steps": 15915, "total_steps": 42420, "loss": 0.3236, "lr": 0.0007865691063195754, "epoch": 7.503536067892504, "percentage": 37.52, "elapsed_time": "0:51:33", "remaining_time": "1:25:52", "throughput": 5175.01, "total_tokens": 16011488} {"current_steps": 15920, "total_steps": 42420, "loss": 0.3536, "lr": 0.0007864005029468192, "epoch": 7.505893446487506, "percentage": 37.53, "elapsed_time": "0:51:35", "remaining_time": "1:25:51", "throughput": 5175.16, "total_tokens": 16017184} {"current_steps": 15925, "total_steps": 42420, "loss": 0.3634, "lr": 0.0007862318510913204, "epoch": 7.508250825082508, "percentage": 37.54, "elapsed_time": "0:51:36", "remaining_time": "1:25:51", "throughput": 5175.39, "total_tokens": 16023776} {"current_steps": 15930, "total_steps": 42420, "loss": 0.3405, "lr": 0.0007860631507816294, "epoch": 7.510608203677511, "percentage": 37.55, "elapsed_time": "0:51:37", "remaining_time": "1:25:50", "throughput": 5175.61, "total_tokens": 16030112} {"current_steps": 15935, "total_steps": 42420, "loss": 0.3587, "lr": 0.0007858944020463039, "epoch": 7.512965582272513, "percentage": 37.56, "elapsed_time": "0:51:38", "remaining_time": "1:25:49", "throughput": 5175.74, "total_tokens": 16035680} {"current_steps": 15940, "total_steps": 42420, "loss": 0.3632, "lr": 0.0007857256049139104, "epoch": 7.515322960867516, "percentage": 37.58, "elapsed_time": "0:51:39", "remaining_time": "1:25:48", "throughput": 5175.86, "total_tokens": 16041056} {"current_steps": 15945, "total_steps": 42420, "loss": 0.3384, "lr": 0.0007855567594130233, "epoch": 7.517680339462518, "percentage": 37.59, "elapsed_time": "0:51:40", "remaining_time": "1:25:47", "throughput": 5175.98, "total_tokens": 16046176} {"current_steps": 15950, "total_steps": 42420, "loss": 0.2815, "lr": 0.0007853878655722255, "epoch": 7.52003771805752, "percentage": 37.6, "elapsed_time": "0:51:41", "remaining_time": "1:25:46", "throughput": 5176.07, "total_tokens": 16051232} {"current_steps": 15955, "total_steps": 42420, "loss": 0.4778, "lr": 0.0007852189234201074, "epoch": 7.522395096652522, "percentage": 37.61, "elapsed_time": "0:51:41", "remaining_time": "1:25:45", "throughput": 5176.06, "total_tokens": 16055648} {"current_steps": 15960, "total_steps": 42420, "loss": 0.3953, "lr": 0.0007850499329852684, "epoch": 7.524752475247524, "percentage": 37.62, "elapsed_time": "0:51:42", "remaining_time": "1:25:44", "throughput": 5176.14, "total_tokens": 16060672} {"current_steps": 15965, "total_steps": 42420, "loss": 0.352, "lr": 0.0007848808942963156, "epoch": 7.527109853842527, "percentage": 37.64, "elapsed_time": "0:51:43", "remaining_time": "1:25:43", "throughput": 5176.16, "total_tokens": 16065216} {"current_steps": 15970, "total_steps": 42420, "loss": 0.3535, "lr": 0.0007847118073818644, "epoch": 7.529467232437529, "percentage": 37.65, "elapsed_time": "0:51:44", "remaining_time": "1:25:41", "throughput": 5176.25, "total_tokens": 16070400} {"current_steps": 15975, "total_steps": 42420, "loss": 0.3521, "lr": 0.0007845426722705384, "epoch": 7.5318246110325315, "percentage": 37.66, "elapsed_time": "0:51:45", "remaining_time": "1:25:40", "throughput": 5176.36, "total_tokens": 16075648} {"current_steps": 15980, "total_steps": 42420, "loss": 0.3076, "lr": 0.0007843734889909692, "epoch": 7.534181989627534, "percentage": 37.67, "elapsed_time": "0:51:46", "remaining_time": "1:25:40", "throughput": 5176.55, "total_tokens": 16081472} {"current_steps": 15985, "total_steps": 42420, "loss": 0.414, "lr": 0.0007842042575717968, "epoch": 7.536539368222536, "percentage": 37.68, "elapsed_time": "0:51:47", "remaining_time": "1:25:39", "throughput": 5176.77, "total_tokens": 16087808} {"current_steps": 15990, "total_steps": 42420, "loss": 0.3328, "lr": 0.0007840349780416689, "epoch": 7.538896746817539, "percentage": 37.69, "elapsed_time": "0:51:48", "remaining_time": "1:25:38", "throughput": 5176.89, "total_tokens": 16093312} {"current_steps": 15995, "total_steps": 42420, "loss": 0.3218, "lr": 0.0007838656504292421, "epoch": 7.541254125412541, "percentage": 37.71, "elapsed_time": "0:51:49", "remaining_time": "1:25:37", "throughput": 5176.97, "total_tokens": 16098144} {"current_steps": 16000, "total_steps": 42420, "loss": 0.3481, "lr": 0.0007836962747631803, "epoch": 7.5436115040075435, "percentage": 37.72, "elapsed_time": "0:51:50", "remaining_time": "1:25:36", "throughput": 5177.27, "total_tokens": 16105248} {"current_steps": 16005, "total_steps": 42420, "loss": 0.349, "lr": 0.000783526851072156, "epoch": 7.545968882602546, "percentage": 37.73, "elapsed_time": "0:51:51", "remaining_time": "1:25:35", "throughput": 5177.35, "total_tokens": 16110240} {"current_steps": 16010, "total_steps": 42420, "loss": 0.3545, "lr": 0.00078335737938485, "epoch": 7.548326261197548, "percentage": 37.74, "elapsed_time": "0:51:52", "remaining_time": "1:25:34", "throughput": 5177.66, "total_tokens": 16116960} {"current_steps": 16015, "total_steps": 42420, "loss": 0.2652, "lr": 0.0007831878597299506, "epoch": 7.550683639792551, "percentage": 37.75, "elapsed_time": "0:51:53", "remaining_time": "1:25:33", "throughput": 5177.71, "total_tokens": 16121568} {"current_steps": 16020, "total_steps": 42420, "loss": 0.406, "lr": 0.000783018292136155, "epoch": 7.553041018387553, "percentage": 37.77, "elapsed_time": "0:51:54", "remaining_time": "1:25:32", "throughput": 5177.82, "total_tokens": 16126880} {"current_steps": 16025, "total_steps": 42420, "loss": 0.3325, "lr": 0.0007828486766321679, "epoch": 7.5553983969825556, "percentage": 37.78, "elapsed_time": "0:51:55", "remaining_time": "1:25:31", "throughput": 5178.06, "total_tokens": 16133152} {"current_steps": 16030, "total_steps": 42420, "loss": 0.3253, "lr": 0.0007826790132467022, "epoch": 7.557755775577558, "percentage": 37.79, "elapsed_time": "0:51:56", "remaining_time": "1:25:30", "throughput": 5178.11, "total_tokens": 16137888} {"current_steps": 16035, "total_steps": 42420, "loss": 0.3484, "lr": 0.0007825093020084792, "epoch": 7.56011315417256, "percentage": 37.8, "elapsed_time": "0:51:57", "remaining_time": "1:25:29", "throughput": 5178.19, "total_tokens": 16143104} {"current_steps": 16040, "total_steps": 42420, "loss": 0.3266, "lr": 0.0007823395429462282, "epoch": 7.562470532767563, "percentage": 37.81, "elapsed_time": "0:51:58", "remaining_time": "1:25:28", "throughput": 5178.1, "total_tokens": 16146688} {"current_steps": 16045, "total_steps": 42420, "loss": 0.3528, "lr": 0.0007821697360886865, "epoch": 7.564827911362565, "percentage": 37.82, "elapsed_time": "0:51:59", "remaining_time": "1:25:27", "throughput": 5178.19, "total_tokens": 16151648} {"current_steps": 16050, "total_steps": 42420, "loss": 0.3188, "lr": 0.0007819998814645992, "epoch": 7.567185289957568, "percentage": 37.84, "elapsed_time": "0:52:00", "remaining_time": "1:25:26", "throughput": 5178.37, "total_tokens": 16157472} {"current_steps": 16055, "total_steps": 42420, "loss": 0.347, "lr": 0.0007818299791027204, "epoch": 7.56954266855257, "percentage": 37.85, "elapsed_time": "0:52:01", "remaining_time": "1:25:25", "throughput": 5178.4, "total_tokens": 16161920} {"current_steps": 16060, "total_steps": 42420, "loss": 0.2751, "lr": 0.000781660029031811, "epoch": 7.571900047147572, "percentage": 37.86, "elapsed_time": "0:52:01", "remaining_time": "1:25:24", "throughput": 5178.39, "total_tokens": 16166336} {"current_steps": 16065, "total_steps": 42420, "loss": 0.4063, "lr": 0.0007814900312806412, "epoch": 7.574257425742574, "percentage": 37.87, "elapsed_time": "0:52:02", "remaining_time": "1:25:22", "throughput": 5178.38, "total_tokens": 16170592} {"current_steps": 16070, "total_steps": 42420, "loss": 0.3158, "lr": 0.0007813199858779885, "epoch": 7.576614804337576, "percentage": 37.88, "elapsed_time": "0:52:03", "remaining_time": "1:25:21", "throughput": 5178.44, "total_tokens": 16175776} {"current_steps": 16075, "total_steps": 42420, "loss": 0.3413, "lr": 0.0007811498928526386, "epoch": 7.578972182932579, "percentage": 37.89, "elapsed_time": "0:52:04", "remaining_time": "1:25:20", "throughput": 5178.52, "total_tokens": 16180608} {"current_steps": 16080, "total_steps": 42420, "loss": 0.3357, "lr": 0.0007809797522333857, "epoch": 7.581329561527581, "percentage": 37.91, "elapsed_time": "0:52:05", "remaining_time": "1:25:19", "throughput": 5178.62, "total_tokens": 16185664} {"current_steps": 16085, "total_steps": 42420, "loss": 0.3571, "lr": 0.0007808095640490314, "epoch": 7.5836869401225835, "percentage": 37.92, "elapsed_time": "0:52:06", "remaining_time": "1:25:18", "throughput": 5178.8, "total_tokens": 16191584} {"current_steps": 16090, "total_steps": 42420, "loss": 0.3481, "lr": 0.0007806393283283857, "epoch": 7.586044318717586, "percentage": 37.93, "elapsed_time": "0:52:07", "remaining_time": "1:25:17", "throughput": 5178.94, "total_tokens": 16197280} {"current_steps": 16095, "total_steps": 42420, "loss": 0.3504, "lr": 0.0007804690451002668, "epoch": 7.588401697312588, "percentage": 37.94, "elapsed_time": "0:52:08", "remaining_time": "1:25:16", "throughput": 5179.04, "total_tokens": 16202528} {"current_steps": 16100, "total_steps": 42420, "loss": 0.3148, "lr": 0.0007802987143935008, "epoch": 7.590759075907591, "percentage": 37.95, "elapsed_time": "0:52:09", "remaining_time": "1:25:16", "throughput": 5179.18, "total_tokens": 16208224} {"current_steps": 16105, "total_steps": 42420, "loss": 0.2828, "lr": 0.0007801283362369214, "epoch": 7.593116454502593, "percentage": 37.97, "elapsed_time": "0:52:10", "remaining_time": "1:25:14", "throughput": 5179.18, "total_tokens": 16212512} {"current_steps": 16110, "total_steps": 42420, "loss": 0.439, "lr": 0.000779957910659371, "epoch": 7.5954738330975955, "percentage": 37.98, "elapsed_time": "0:52:11", "remaining_time": "1:25:13", "throughput": 5179.22, "total_tokens": 16217344} {"current_steps": 16115, "total_steps": 42420, "loss": 0.3337, "lr": 0.0007797874376896998, "epoch": 7.597831211692598, "percentage": 37.99, "elapsed_time": "0:52:12", "remaining_time": "1:25:12", "throughput": 5179.28, "total_tokens": 16222112} {"current_steps": 16120, "total_steps": 42420, "loss": 0.2931, "lr": 0.0007796169173567661, "epoch": 7.6001885902876, "percentage": 38.0, "elapsed_time": "0:52:13", "remaining_time": "1:25:11", "throughput": 5179.36, "total_tokens": 16227392} {"current_steps": 16125, "total_steps": 42420, "loss": 0.3438, "lr": 0.0007794463496894358, "epoch": 7.602545968882603, "percentage": 38.01, "elapsed_time": "0:52:14", "remaining_time": "1:25:10", "throughput": 5179.53, "total_tokens": 16233376} {"current_steps": 16130, "total_steps": 42420, "loss": 0.3349, "lr": 0.0007792757347165831, "epoch": 7.604903347477605, "percentage": 38.02, "elapsed_time": "0:52:14", "remaining_time": "1:25:09", "throughput": 5179.55, "total_tokens": 16237888} {"current_steps": 16135, "total_steps": 42420, "loss": 0.3517, "lr": 0.0007791050724670904, "epoch": 7.6072607260726075, "percentage": 38.04, "elapsed_time": "0:52:16", "remaining_time": "1:25:08", "throughput": 5179.71, "total_tokens": 16243840} {"current_steps": 16140, "total_steps": 42420, "loss": 0.2939, "lr": 0.000778934362969848, "epoch": 7.60961810466761, "percentage": 38.05, "elapsed_time": "0:52:16", "remaining_time": "1:25:07", "throughput": 5179.73, "total_tokens": 16248416} {"current_steps": 16145, "total_steps": 42420, "loss": 0.3164, "lr": 0.0007787636062537537, "epoch": 7.611975483262612, "percentage": 38.06, "elapsed_time": "0:52:17", "remaining_time": "1:25:06", "throughput": 5179.77, "total_tokens": 16253088} {"current_steps": 16150, "total_steps": 42420, "loss": 0.4031, "lr": 0.0007785928023477142, "epoch": 7.614332861857615, "percentage": 38.07, "elapsed_time": "0:52:18", "remaining_time": "1:25:05", "throughput": 5179.97, "total_tokens": 16259104} {"current_steps": 16155, "total_steps": 42420, "loss": 0.2989, "lr": 0.0007784219512806435, "epoch": 7.616690240452616, "percentage": 38.08, "elapsed_time": "0:52:19", "remaining_time": "1:25:04", "throughput": 5180.02, "total_tokens": 16263936} {"current_steps": 16160, "total_steps": 42420, "loss": 0.3313, "lr": 0.0007782510530814638, "epoch": 7.619047619047619, "percentage": 38.1, "elapsed_time": "0:52:20", "remaining_time": "1:25:03", "throughput": 5180.07, "total_tokens": 16268800} {"current_steps": 16165, "total_steps": 42420, "loss": 0.3071, "lr": 0.0007780801077791052, "epoch": 7.621404997642621, "percentage": 38.11, "elapsed_time": "0:52:21", "remaining_time": "1:25:02", "throughput": 5180.2, "total_tokens": 16274272} {"current_steps": 16170, "total_steps": 42420, "loss": 0.4008, "lr": 0.0007779091154025059, "epoch": 7.623762376237623, "percentage": 38.12, "elapsed_time": "0:52:22", "remaining_time": "1:25:01", "throughput": 5180.26, "total_tokens": 16279328} {"current_steps": 16175, "total_steps": 42420, "loss": 0.3421, "lr": 0.0007777380759806119, "epoch": 7.626119754832626, "percentage": 38.13, "elapsed_time": "0:52:23", "remaining_time": "1:25:00", "throughput": 5180.53, "total_tokens": 16286272} {"current_steps": 16180, "total_steps": 42420, "loss": 0.3091, "lr": 0.0007775669895423776, "epoch": 7.628477133427628, "percentage": 38.14, "elapsed_time": "0:52:24", "remaining_time": "1:25:00", "throughput": 5180.66, "total_tokens": 16291872} {"current_steps": 16185, "total_steps": 42420, "loss": 0.3657, "lr": 0.0007773958561167646, "epoch": 7.630834512022631, "percentage": 38.15, "elapsed_time": "0:52:25", "remaining_time": "1:24:58", "throughput": 5180.73, "total_tokens": 16296832} {"current_steps": 16190, "total_steps": 42420, "loss": 0.3861, "lr": 0.0007772246757327431, "epoch": 7.633191890617633, "percentage": 38.17, "elapsed_time": "0:52:26", "remaining_time": "1:24:57", "throughput": 5180.87, "total_tokens": 16302368} {"current_steps": 16195, "total_steps": 42420, "loss": 0.3404, "lr": 0.000777053448419291, "epoch": 7.635549269212635, "percentage": 38.18, "elapsed_time": "0:52:27", "remaining_time": "1:24:56", "throughput": 5180.88, "total_tokens": 16306752} {"current_steps": 16200, "total_steps": 42420, "loss": 0.3075, "lr": 0.0007768821742053941, "epoch": 7.637906647807638, "percentage": 38.19, "elapsed_time": "0:52:28", "remaining_time": "1:24:55", "throughput": 5180.89, "total_tokens": 16311232} {"current_steps": 16205, "total_steps": 42420, "loss": 0.3577, "lr": 0.0007767108531200464, "epoch": 7.64026402640264, "percentage": 38.2, "elapsed_time": "0:52:29", "remaining_time": "1:24:54", "throughput": 5180.99, "total_tokens": 16316448} {"current_steps": 16210, "total_steps": 42420, "loss": 0.3793, "lr": 0.0007765394851922496, "epoch": 7.642621404997643, "percentage": 38.21, "elapsed_time": "0:52:30", "remaining_time": "1:24:53", "throughput": 5181.06, "total_tokens": 16321472} {"current_steps": 16215, "total_steps": 42420, "loss": 0.3063, "lr": 0.0007763680704510134, "epoch": 7.644978783592645, "percentage": 38.22, "elapsed_time": "0:52:31", "remaining_time": "1:24:52", "throughput": 5181.09, "total_tokens": 16326176} {"current_steps": 16220, "total_steps": 42420, "loss": 0.3226, "lr": 0.0007761966089253552, "epoch": 7.6473361621876474, "percentage": 38.24, "elapsed_time": "0:52:32", "remaining_time": "1:24:51", "throughput": 5181.22, "total_tokens": 16331616} {"current_steps": 16225, "total_steps": 42420, "loss": 0.3587, "lr": 0.0007760251006443007, "epoch": 7.64969354078265, "percentage": 38.25, "elapsed_time": "0:52:33", "remaining_time": "1:24:50", "throughput": 5181.31, "total_tokens": 16336864} {"current_steps": 16230, "total_steps": 42420, "loss": 0.3419, "lr": 0.0007758535456368832, "epoch": 7.652050919377652, "percentage": 38.26, "elapsed_time": "0:52:34", "remaining_time": "1:24:49", "throughput": 5181.46, "total_tokens": 16342432} {"current_steps": 16235, "total_steps": 42420, "loss": 0.341, "lr": 0.0007756819439321443, "epoch": 7.654408297972655, "percentage": 38.27, "elapsed_time": "0:52:34", "remaining_time": "1:24:48", "throughput": 5181.47, "total_tokens": 16346784} {"current_steps": 16240, "total_steps": 42420, "loss": 0.321, "lr": 0.0007755102955591329, "epoch": 7.656765676567657, "percentage": 38.28, "elapsed_time": "0:52:35", "remaining_time": "1:24:47", "throughput": 5181.58, "total_tokens": 16352160} {"current_steps": 16245, "total_steps": 42420, "loss": 0.3489, "lr": 0.0007753386005469066, "epoch": 7.6591230551626595, "percentage": 38.3, "elapsed_time": "0:52:36", "remaining_time": "1:24:46", "throughput": 5181.65, "total_tokens": 16356992} {"current_steps": 16250, "total_steps": 42420, "loss": 0.3514, "lr": 0.00077516685892453, "epoch": 7.661480433757662, "percentage": 38.31, "elapsed_time": "0:52:37", "remaining_time": "1:24:45", "throughput": 5181.74, "total_tokens": 16362304} {"current_steps": 16255, "total_steps": 42420, "loss": 0.2723, "lr": 0.0007749950707210765, "epoch": 7.663837812352664, "percentage": 38.32, "elapsed_time": "0:52:38", "remaining_time": "1:24:44", "throughput": 5181.78, "total_tokens": 16366944} {"current_steps": 16260, "total_steps": 42420, "loss": 0.3597, "lr": 0.0007748232359656264, "epoch": 7.666195190947667, "percentage": 38.33, "elapsed_time": "0:52:39", "remaining_time": "1:24:43", "throughput": 5181.89, "total_tokens": 16372160} {"current_steps": 16265, "total_steps": 42420, "loss": 0.3051, "lr": 0.0007746513546872689, "epoch": 7.668552569542668, "percentage": 38.34, "elapsed_time": "0:52:40", "remaining_time": "1:24:41", "throughput": 5181.91, "total_tokens": 16376576} {"current_steps": 16270, "total_steps": 42420, "loss": 0.3565, "lr": 0.0007744794269151001, "epoch": 7.670909948137671, "percentage": 38.35, "elapsed_time": "0:52:41", "remaining_time": "1:24:40", "throughput": 5181.9, "total_tokens": 16381056} {"current_steps": 16275, "total_steps": 42420, "loss": 0.343, "lr": 0.000774307452678225, "epoch": 7.673267326732673, "percentage": 38.37, "elapsed_time": "0:52:42", "remaining_time": "1:24:39", "throughput": 5181.96, "total_tokens": 16385920} {"current_steps": 16280, "total_steps": 42420, "loss": 0.3254, "lr": 0.0007741354320057554, "epoch": 7.675624705327675, "percentage": 38.38, "elapsed_time": "0:52:42", "remaining_time": "1:24:38", "throughput": 5181.9, "total_tokens": 16389824} {"current_steps": 16285, "total_steps": 42420, "loss": 0.3282, "lr": 0.000773963364926812, "epoch": 7.677982083922678, "percentage": 38.39, "elapsed_time": "0:52:43", "remaining_time": "1:24:37", "throughput": 5181.95, "total_tokens": 16394432} {"current_steps": 16290, "total_steps": 42420, "loss": 0.3272, "lr": 0.0007737912514705223, "epoch": 7.68033946251768, "percentage": 38.4, "elapsed_time": "0:52:44", "remaining_time": "1:24:36", "throughput": 5182.02, "total_tokens": 16399296} {"current_steps": 16295, "total_steps": 42420, "loss": 0.3685, "lr": 0.0007736190916660227, "epoch": 7.682696841112683, "percentage": 38.41, "elapsed_time": "0:52:45", "remaining_time": "1:24:35", "throughput": 5182.04, "total_tokens": 16403776} {"current_steps": 16300, "total_steps": 42420, "loss": 0.3378, "lr": 0.0007734468855424563, "epoch": 7.685054219707685, "percentage": 38.43, "elapsed_time": "0:52:46", "remaining_time": "1:24:33", "throughput": 5182.0, "total_tokens": 16407840} {"current_steps": 16305, "total_steps": 42420, "loss": 0.3238, "lr": 0.0007732746331289753, "epoch": 7.687411598302687, "percentage": 38.44, "elapsed_time": "0:52:47", "remaining_time": "1:24:32", "throughput": 5182.05, "total_tokens": 16412608} {"current_steps": 16310, "total_steps": 42420, "loss": 0.3011, "lr": 0.0007731023344547388, "epoch": 7.68976897689769, "percentage": 38.45, "elapsed_time": "0:52:48", "remaining_time": "1:24:31", "throughput": 5182.05, "total_tokens": 16417024} {"current_steps": 16315, "total_steps": 42420, "loss": 0.4138, "lr": 0.0007729299895489142, "epoch": 7.692126355492692, "percentage": 38.46, "elapsed_time": "0:52:48", "remaining_time": "1:24:30", "throughput": 5182.12, "total_tokens": 16421856} {"current_steps": 16320, "total_steps": 42420, "loss": 0.3009, "lr": 0.0007727575984406762, "epoch": 7.694483734087695, "percentage": 38.47, "elapsed_time": "0:52:49", "remaining_time": "1:24:29", "throughput": 5182.18, "total_tokens": 16426848} {"current_steps": 16325, "total_steps": 42420, "loss": 0.3747, "lr": 0.0007725851611592083, "epoch": 7.696841112682697, "percentage": 38.48, "elapsed_time": "0:52:50", "remaining_time": "1:24:28", "throughput": 5182.21, "total_tokens": 16431488} {"current_steps": 16330, "total_steps": 42420, "loss": 0.3465, "lr": 0.0007724126777337006, "epoch": 7.699198491277699, "percentage": 38.5, "elapsed_time": "0:52:51", "remaining_time": "1:24:27", "throughput": 5182.3, "total_tokens": 16436384} {"current_steps": 16335, "total_steps": 42420, "loss": 0.3425, "lr": 0.0007722401481933519, "epoch": 7.701555869872702, "percentage": 38.51, "elapsed_time": "0:52:52", "remaining_time": "1:24:26", "throughput": 5182.42, "total_tokens": 16441856} {"current_steps": 16340, "total_steps": 42420, "loss": 0.2971, "lr": 0.0007720675725673685, "epoch": 7.703913248467704, "percentage": 38.52, "elapsed_time": "0:52:53", "remaining_time": "1:24:25", "throughput": 5182.44, "total_tokens": 16446528} {"current_steps": 16345, "total_steps": 42420, "loss": 0.355, "lr": 0.0007718949508849644, "epoch": 7.706270627062707, "percentage": 38.53, "elapsed_time": "0:52:54", "remaining_time": "1:24:24", "throughput": 5182.58, "total_tokens": 16451744} {"current_steps": 16350, "total_steps": 42420, "loss": 0.4207, "lr": 0.0007717222831753617, "epoch": 7.708628005657709, "percentage": 38.54, "elapsed_time": "0:52:55", "remaining_time": "1:24:23", "throughput": 5182.69, "total_tokens": 16456960} {"current_steps": 16355, "total_steps": 42420, "loss": 0.3677, "lr": 0.00077154956946779, "epoch": 7.7109853842527105, "percentage": 38.55, "elapsed_time": "0:52:56", "remaining_time": "1:24:22", "throughput": 5182.92, "total_tokens": 16463168} {"current_steps": 16360, "total_steps": 42420, "loss": 0.3337, "lr": 0.0007713768097914871, "epoch": 7.713342762847713, "percentage": 38.57, "elapsed_time": "0:52:57", "remaining_time": "1:24:21", "throughput": 5182.97, "total_tokens": 16467776} {"current_steps": 16365, "total_steps": 42420, "loss": 0.3842, "lr": 0.0007712040041756978, "epoch": 7.715700141442715, "percentage": 38.58, "elapsed_time": "0:52:58", "remaining_time": "1:24:20", "throughput": 5183.04, "total_tokens": 16472672} {"current_steps": 16370, "total_steps": 42420, "loss": 0.3335, "lr": 0.0007710311526496754, "epoch": 7.718057520037718, "percentage": 38.59, "elapsed_time": "0:52:58", "remaining_time": "1:24:18", "throughput": 5183.02, "total_tokens": 16476800} {"current_steps": 16375, "total_steps": 42420, "loss": 0.3022, "lr": 0.0007708582552426806, "epoch": 7.72041489863272, "percentage": 38.6, "elapsed_time": "0:52:59", "remaining_time": "1:24:17", "throughput": 5183.05, "total_tokens": 16481632} {"current_steps": 16380, "total_steps": 42420, "loss": 0.4421, "lr": 0.0007706853119839822, "epoch": 7.7227722772277225, "percentage": 38.61, "elapsed_time": "0:53:00", "remaining_time": "1:24:16", "throughput": 5183.01, "total_tokens": 16485536} {"current_steps": 16385, "total_steps": 42420, "loss": 0.3723, "lr": 0.0007705123229028565, "epoch": 7.725129655822725, "percentage": 38.63, "elapsed_time": "0:53:01", "remaining_time": "1:24:15", "throughput": 5183.08, "total_tokens": 16490528} {"current_steps": 16390, "total_steps": 42420, "loss": 0.3344, "lr": 0.0007703392880285874, "epoch": 7.727487034417727, "percentage": 38.64, "elapsed_time": "0:53:02", "remaining_time": "1:24:14", "throughput": 5183.12, "total_tokens": 16495200} {"current_steps": 16395, "total_steps": 42420, "loss": 0.305, "lr": 0.0007701662073904672, "epoch": 7.72984441301273, "percentage": 38.65, "elapsed_time": "0:53:03", "remaining_time": "1:24:13", "throughput": 5183.15, "total_tokens": 16500032} {"current_steps": 16400, "total_steps": 42420, "loss": 0.3202, "lr": 0.0007699930810177951, "epoch": 7.732201791607732, "percentage": 38.66, "elapsed_time": "0:53:04", "remaining_time": "1:24:12", "throughput": 5183.19, "total_tokens": 16504672} {"current_steps": 16405, "total_steps": 42420, "loss": 0.2458, "lr": 0.0007698199089398785, "epoch": 7.7345591702027345, "percentage": 38.67, "elapsed_time": "0:53:05", "remaining_time": "1:24:10", "throughput": 5183.17, "total_tokens": 16508896} {"current_steps": 16410, "total_steps": 42420, "loss": 0.4733, "lr": 0.0007696466911860327, "epoch": 7.736916548797737, "percentage": 38.68, "elapsed_time": "0:53:06", "remaining_time": "1:24:09", "throughput": 5183.29, "total_tokens": 16514080} {"current_steps": 16415, "total_steps": 42420, "loss": 0.3912, "lr": 0.0007694734277855804, "epoch": 7.739273927392739, "percentage": 38.7, "elapsed_time": "0:53:06", "remaining_time": "1:24:08", "throughput": 5183.44, "total_tokens": 16519584} {"current_steps": 16420, "total_steps": 42420, "loss": 0.3213, "lr": 0.0007693001187678523, "epoch": 7.741631305987742, "percentage": 38.71, "elapsed_time": "0:53:07", "remaining_time": "1:24:07", "throughput": 5183.47, "total_tokens": 16524192} {"current_steps": 16425, "total_steps": 42420, "loss": 0.3579, "lr": 0.0007691267641621864, "epoch": 7.743988684582744, "percentage": 38.72, "elapsed_time": "0:53:08", "remaining_time": "1:24:06", "throughput": 5183.47, "total_tokens": 16528512} {"current_steps": 16430, "total_steps": 42420, "loss": 0.3617, "lr": 0.000768953363997929, "epoch": 7.7463460631777465, "percentage": 38.73, "elapsed_time": "0:53:09", "remaining_time": "1:24:05", "throughput": 5183.55, "total_tokens": 16533376} {"current_steps": 16435, "total_steps": 42420, "loss": 0.3523, "lr": 0.0007687799183044335, "epoch": 7.748703441772749, "percentage": 38.74, "elapsed_time": "0:53:10", "remaining_time": "1:24:04", "throughput": 5183.55, "total_tokens": 16537824} {"current_steps": 16440, "total_steps": 42420, "loss": 0.3626, "lr": 0.0007686064271110616, "epoch": 7.751060820367751, "percentage": 38.76, "elapsed_time": "0:53:11", "remaining_time": "1:24:03", "throughput": 5183.57, "total_tokens": 16542336} {"current_steps": 16445, "total_steps": 42420, "loss": 0.3252, "lr": 0.0007684328904471821, "epoch": 7.753418198962754, "percentage": 38.77, "elapsed_time": "0:53:12", "remaining_time": "1:24:02", "throughput": 5183.85, "total_tokens": 16549408} {"current_steps": 16450, "total_steps": 42420, "loss": 0.3446, "lr": 0.0007682593083421721, "epoch": 7.755775577557756, "percentage": 38.78, "elapsed_time": "0:53:13", "remaining_time": "1:24:01", "throughput": 5183.89, "total_tokens": 16553856} {"current_steps": 16455, "total_steps": 42420, "loss": 0.3492, "lr": 0.0007680856808254158, "epoch": 7.7581329561527586, "percentage": 38.79, "elapsed_time": "0:53:14", "remaining_time": "1:24:00", "throughput": 5183.87, "total_tokens": 16558464} {"current_steps": 16460, "total_steps": 42420, "loss": 0.3074, "lr": 0.0007679120079263058, "epoch": 7.760490334747761, "percentage": 38.8, "elapsed_time": "0:53:15", "remaining_time": "1:23:59", "throughput": 5184.05, "total_tokens": 16564320} {"current_steps": 16465, "total_steps": 42420, "loss": 0.251, "lr": 0.0007677382896742416, "epoch": 7.7628477133427625, "percentage": 38.81, "elapsed_time": "0:53:16", "remaining_time": "1:23:58", "throughput": 5184.17, "total_tokens": 16569952} {"current_steps": 16470, "total_steps": 42420, "loss": 0.3735, "lr": 0.0007675645260986309, "epoch": 7.765205091937765, "percentage": 38.83, "elapsed_time": "0:53:17", "remaining_time": "1:23:57", "throughput": 5184.36, "total_tokens": 16575680} {"current_steps": 16475, "total_steps": 42420, "loss": 0.4055, "lr": 0.000767390717228889, "epoch": 7.767562470532767, "percentage": 38.84, "elapsed_time": "0:53:18", "remaining_time": "1:23:56", "throughput": 5184.38, "total_tokens": 16580256} {"current_steps": 16480, "total_steps": 42420, "loss": 0.346, "lr": 0.0007672168630944386, "epoch": 7.76991984912777, "percentage": 38.85, "elapsed_time": "0:53:19", "remaining_time": "1:23:55", "throughput": 5184.53, "total_tokens": 16585696} {"current_steps": 16485, "total_steps": 42420, "loss": 0.3422, "lr": 0.0007670429637247105, "epoch": 7.772277227722772, "percentage": 38.86, "elapsed_time": "0:53:20", "remaining_time": "1:23:54", "throughput": 5184.65, "total_tokens": 16591040} {"current_steps": 16490, "total_steps": 42420, "loss": 0.3755, "lr": 0.0007668690191491428, "epoch": 7.7746346063177745, "percentage": 38.87, "elapsed_time": "0:53:21", "remaining_time": "1:23:53", "throughput": 5184.77, "total_tokens": 16596736} {"current_steps": 16495, "total_steps": 42420, "loss": 0.3473, "lr": 0.000766695029397181, "epoch": 7.776991984912777, "percentage": 38.88, "elapsed_time": "0:53:21", "remaining_time": "1:23:52", "throughput": 5184.78, "total_tokens": 16601248} {"current_steps": 16500, "total_steps": 42420, "loss": 0.352, "lr": 0.0007665209944982792, "epoch": 7.779349363507779, "percentage": 38.9, "elapsed_time": "0:53:22", "remaining_time": "1:23:51", "throughput": 5184.85, "total_tokens": 16605984} {"current_steps": 16505, "total_steps": 42420, "loss": 0.3624, "lr": 0.0007663469144818984, "epoch": 7.781706742102782, "percentage": 38.91, "elapsed_time": "0:53:23", "remaining_time": "1:23:50", "throughput": 5184.88, "total_tokens": 16610592} {"current_steps": 16510, "total_steps": 42420, "loss": 0.3494, "lr": 0.0007661727893775071, "epoch": 7.784064120697784, "percentage": 38.92, "elapsed_time": "0:53:24", "remaining_time": "1:23:48", "throughput": 5184.86, "total_tokens": 16614688} {"current_steps": 16515, "total_steps": 42420, "loss": 0.3141, "lr": 0.000765998619214582, "epoch": 7.7864214992927865, "percentage": 38.93, "elapsed_time": "0:53:25", "remaining_time": "1:23:48", "throughput": 5185.12, "total_tokens": 16621632} {"current_steps": 16520, "total_steps": 42420, "loss": 0.2809, "lr": 0.0007658244040226071, "epoch": 7.788778877887789, "percentage": 38.94, "elapsed_time": "0:53:26", "remaining_time": "1:23:47", "throughput": 5185.22, "total_tokens": 16626976} {"current_steps": 16525, "total_steps": 42420, "loss": 0.277, "lr": 0.000765650143831074, "epoch": 7.791136256482791, "percentage": 38.96, "elapsed_time": "0:53:27", "remaining_time": "1:23:46", "throughput": 5185.31, "total_tokens": 16631840} {"current_steps": 16530, "total_steps": 42420, "loss": 0.2843, "lr": 0.000765475838669482, "epoch": 7.793493635077794, "percentage": 38.97, "elapsed_time": "0:53:28", "remaining_time": "1:23:45", "throughput": 5185.45, "total_tokens": 16637440} {"current_steps": 16535, "total_steps": 42420, "loss": 0.4402, "lr": 0.000765301488567338, "epoch": 7.795851013672796, "percentage": 38.98, "elapsed_time": "0:53:29", "remaining_time": "1:23:44", "throughput": 5185.52, "total_tokens": 16642400} {"current_steps": 16540, "total_steps": 42420, "loss": 0.3718, "lr": 0.0007651270935541566, "epoch": 7.7982083922677985, "percentage": 38.99, "elapsed_time": "0:53:30", "remaining_time": "1:23:43", "throughput": 5185.52, "total_tokens": 16646848} {"current_steps": 16545, "total_steps": 42420, "loss": 0.3728, "lr": 0.00076495265365946, "epoch": 7.800565770862801, "percentage": 39.0, "elapsed_time": "0:53:31", "remaining_time": "1:23:42", "throughput": 5185.62, "total_tokens": 16652160} {"current_steps": 16550, "total_steps": 42420, "loss": 0.355, "lr": 0.0007647781689127778, "epoch": 7.802923149457803, "percentage": 39.01, "elapsed_time": "0:53:32", "remaining_time": "1:23:41", "throughput": 5185.85, "total_tokens": 16658624} {"current_steps": 16555, "total_steps": 42420, "loss": 0.3622, "lr": 0.0007646036393436469, "epoch": 7.805280528052805, "percentage": 39.03, "elapsed_time": "0:53:33", "remaining_time": "1:23:40", "throughput": 5185.95, "total_tokens": 16663840} {"current_steps": 16560, "total_steps": 42420, "loss": 0.359, "lr": 0.0007644290649816128, "epoch": 7.807637906647807, "percentage": 39.04, "elapsed_time": "0:53:34", "remaining_time": "1:23:39", "throughput": 5186.05, "total_tokens": 16669024} {"current_steps": 16565, "total_steps": 42420, "loss": 0.3368, "lr": 0.0007642544458562277, "epoch": 7.80999528524281, "percentage": 39.05, "elapsed_time": "0:53:35", "remaining_time": "1:23:38", "throughput": 5186.11, "total_tokens": 16674080} {"current_steps": 16570, "total_steps": 42420, "loss": 0.3178, "lr": 0.0007640797819970519, "epoch": 7.812352663837812, "percentage": 39.06, "elapsed_time": "0:53:36", "remaining_time": "1:23:37", "throughput": 5186.24, "total_tokens": 16679680} {"current_steps": 16575, "total_steps": 42420, "loss": 0.269, "lr": 0.0007639050734336527, "epoch": 7.814710042432814, "percentage": 39.07, "elapsed_time": "0:53:37", "remaining_time": "1:23:36", "throughput": 5186.45, "total_tokens": 16685984} {"current_steps": 16580, "total_steps": 42420, "loss": 0.3658, "lr": 0.0007637303201956054, "epoch": 7.817067421027817, "percentage": 39.09, "elapsed_time": "0:53:38", "remaining_time": "1:23:35", "throughput": 5186.63, "total_tokens": 16692064} {"current_steps": 16585, "total_steps": 42420, "loss": 0.2233, "lr": 0.0007635555223124927, "epoch": 7.819424799622819, "percentage": 39.1, "elapsed_time": "0:53:39", "remaining_time": "1:23:34", "throughput": 5186.62, "total_tokens": 16696256} {"current_steps": 16590, "total_steps": 42420, "loss": 0.4752, "lr": 0.0007633806798139052, "epoch": 7.821782178217822, "percentage": 39.11, "elapsed_time": "0:53:40", "remaining_time": "1:23:33", "throughput": 5186.75, "total_tokens": 16701504} {"current_steps": 16595, "total_steps": 42420, "loss": 0.3603, "lr": 0.0007632057927294402, "epoch": 7.824139556812824, "percentage": 39.12, "elapsed_time": "0:53:40", "remaining_time": "1:23:32", "throughput": 5186.8, "total_tokens": 16706432} {"current_steps": 16600, "total_steps": 42420, "loss": 0.3347, "lr": 0.0007630308610887037, "epoch": 7.826496935407826, "percentage": 39.13, "elapsed_time": "0:53:41", "remaining_time": "1:23:31", "throughput": 5186.92, "total_tokens": 16711648} {"current_steps": 16605, "total_steps": 42420, "loss": 0.3431, "lr": 0.0007628558849213085, "epoch": 7.828854314002829, "percentage": 39.14, "elapsed_time": "0:53:42", "remaining_time": "1:23:30", "throughput": 5187.07, "total_tokens": 16717568} {"current_steps": 16610, "total_steps": 42420, "loss": 0.3321, "lr": 0.000762680864256875, "epoch": 7.831211692597831, "percentage": 39.16, "elapsed_time": "0:53:43", "remaining_time": "1:23:29", "throughput": 5187.08, "total_tokens": 16722048} {"current_steps": 16615, "total_steps": 42420, "loss": 0.3414, "lr": 0.0007625057991250311, "epoch": 7.833569071192834, "percentage": 39.17, "elapsed_time": "0:53:44", "remaining_time": "1:23:28", "throughput": 5187.22, "total_tokens": 16727840} {"current_steps": 16620, "total_steps": 42420, "loss": 0.3975, "lr": 0.0007623306895554126, "epoch": 7.835926449787836, "percentage": 39.18, "elapsed_time": "0:53:46", "remaining_time": "1:23:28", "throughput": 5187.42, "total_tokens": 16735104} {"current_steps": 16625, "total_steps": 42420, "loss": 0.3872, "lr": 0.0007621555355776624, "epoch": 7.838283828382838, "percentage": 39.19, "elapsed_time": "0:53:47", "remaining_time": "1:23:26", "throughput": 5187.5, "total_tokens": 16740160} {"current_steps": 16630, "total_steps": 42420, "loss": 0.3705, "lr": 0.0007619803372214313, "epoch": 7.840641206977841, "percentage": 39.2, "elapsed_time": "0:53:48", "remaining_time": "1:23:26", "throughput": 5187.74, "total_tokens": 16746816} {"current_steps": 16635, "total_steps": 42420, "loss": 0.3751, "lr": 0.0007618050945163772, "epoch": 7.842998585572843, "percentage": 39.21, "elapsed_time": "0:53:49", "remaining_time": "1:23:25", "throughput": 5187.87, "total_tokens": 16752096} {"current_steps": 16640, "total_steps": 42420, "loss": 0.3563, "lr": 0.0007616298074921659, "epoch": 7.845355964167846, "percentage": 39.23, "elapsed_time": "0:53:49", "remaining_time": "1:23:24", "throughput": 5187.94, "total_tokens": 16756896} {"current_steps": 16645, "total_steps": 42420, "loss": 0.3336, "lr": 0.0007614544761784703, "epoch": 7.847713342762848, "percentage": 39.24, "elapsed_time": "0:53:50", "remaining_time": "1:23:22", "throughput": 5187.96, "total_tokens": 16761472} {"current_steps": 16650, "total_steps": 42420, "loss": 0.2916, "lr": 0.0007612791006049713, "epoch": 7.8500707213578504, "percentage": 39.25, "elapsed_time": "0:53:51", "remaining_time": "1:23:21", "throughput": 5188.07, "total_tokens": 16766688} {"current_steps": 16655, "total_steps": 42420, "loss": 0.4746, "lr": 0.0007611036808013568, "epoch": 7.852428099952853, "percentage": 39.26, "elapsed_time": "0:53:52", "remaining_time": "1:23:20", "throughput": 5188.06, "total_tokens": 16771072} {"current_steps": 16660, "total_steps": 42420, "loss": 0.3334, "lr": 0.0007609282167973223, "epoch": 7.854785478547855, "percentage": 39.27, "elapsed_time": "0:53:53", "remaining_time": "1:23:20", "throughput": 5188.26, "total_tokens": 16777696} {"current_steps": 16665, "total_steps": 42420, "loss": 0.3416, "lr": 0.000760752708622571, "epoch": 7.857142857142857, "percentage": 39.29, "elapsed_time": "0:53:54", "remaining_time": "1:23:19", "throughput": 5188.35, "total_tokens": 16782848} {"current_steps": 16670, "total_steps": 42420, "loss": 0.312, "lr": 0.0007605771563068134, "epoch": 7.859500235737859, "percentage": 39.3, "elapsed_time": "0:53:55", "remaining_time": "1:23:17", "throughput": 5188.3, "total_tokens": 16786816} {"current_steps": 16675, "total_steps": 42420, "loss": 0.3198, "lr": 0.0007604015598797678, "epoch": 7.861857614332862, "percentage": 39.31, "elapsed_time": "0:53:56", "remaining_time": "1:23:16", "throughput": 5188.39, "total_tokens": 16791648} {"current_steps": 16680, "total_steps": 42420, "loss": 0.3609, "lr": 0.0007602259193711594, "epoch": 7.864214992927864, "percentage": 39.32, "elapsed_time": "0:53:57", "remaining_time": "1:23:15", "throughput": 5188.56, "total_tokens": 16797536} {"current_steps": 16685, "total_steps": 42420, "loss": 0.4182, "lr": 0.0007600502348107211, "epoch": 7.866572371522866, "percentage": 39.33, "elapsed_time": "0:53:58", "remaining_time": "1:23:14", "throughput": 5188.72, "total_tokens": 16803296} {"current_steps": 16690, "total_steps": 42420, "loss": 0.275, "lr": 0.0007598745062281935, "epoch": 7.868929750117869, "percentage": 39.34, "elapsed_time": "0:53:59", "remaining_time": "1:23:13", "throughput": 5188.67, "total_tokens": 16807232} {"current_steps": 16695, "total_steps": 42420, "loss": 0.3712, "lr": 0.0007596987336533246, "epoch": 7.871287128712871, "percentage": 39.36, "elapsed_time": "0:54:00", "remaining_time": "1:23:12", "throughput": 5188.87, "total_tokens": 16813120} {"current_steps": 16700, "total_steps": 42420, "loss": 0.3757, "lr": 0.0007595229171158692, "epoch": 7.873644507307874, "percentage": 39.37, "elapsed_time": "0:54:01", "remaining_time": "1:23:12", "throughput": 5189.09, "total_tokens": 16819552} {"current_steps": 16705, "total_steps": 42420, "loss": 0.3609, "lr": 0.0007593470566455906, "epoch": 7.876001885902876, "percentage": 39.38, "elapsed_time": "0:54:02", "remaining_time": "1:23:11", "throughput": 5189.26, "total_tokens": 16825664} {"current_steps": 16710, "total_steps": 42420, "loss": 0.3363, "lr": 0.0007591711522722587, "epoch": 7.878359264497878, "percentage": 39.39, "elapsed_time": "0:54:03", "remaining_time": "1:23:10", "throughput": 5189.31, "total_tokens": 16830400} {"current_steps": 16715, "total_steps": 42420, "loss": 0.3652, "lr": 0.0007589952040256513, "epoch": 7.880716643092881, "percentage": 39.4, "elapsed_time": "0:54:04", "remaining_time": "1:23:09", "throughput": 5189.43, "total_tokens": 16836160} {"current_steps": 16720, "total_steps": 42420, "loss": 0.3017, "lr": 0.0007588192119355533, "epoch": 7.883074021687883, "percentage": 39.42, "elapsed_time": "0:54:05", "remaining_time": "1:23:08", "throughput": 5189.46, "total_tokens": 16840736} {"current_steps": 16725, "total_steps": 42420, "loss": 0.3689, "lr": 0.0007586431760317571, "epoch": 7.885431400282886, "percentage": 39.43, "elapsed_time": "0:54:06", "remaining_time": "1:23:07", "throughput": 5189.51, "total_tokens": 16845632} {"current_steps": 16730, "total_steps": 42420, "loss": 0.2353, "lr": 0.0007584670963440629, "epoch": 7.887788778877888, "percentage": 39.44, "elapsed_time": "0:54:06", "remaining_time": "1:23:05", "throughput": 5189.5, "total_tokens": 16849760} {"current_steps": 16735, "total_steps": 42420, "loss": 0.4102, "lr": 0.0007582909729022777, "epoch": 7.89014615747289, "percentage": 39.45, "elapsed_time": "0:54:07", "remaining_time": "1:23:04", "throughput": 5189.49, "total_tokens": 16854368} {"current_steps": 16740, "total_steps": 42420, "loss": 0.281, "lr": 0.0007581148057362164, "epoch": 7.892503536067893, "percentage": 39.46, "elapsed_time": "0:54:08", "remaining_time": "1:23:03", "throughput": 5189.48, "total_tokens": 16858528} {"current_steps": 16745, "total_steps": 42420, "loss": 0.4264, "lr": 0.0007579385948757009, "epoch": 7.894860914662895, "percentage": 39.47, "elapsed_time": "0:54:09", "remaining_time": "1:23:02", "throughput": 5189.65, "total_tokens": 16864640} {"current_steps": 16750, "total_steps": 42420, "loss": 0.3477, "lr": 0.0007577623403505609, "epoch": 7.897218293257898, "percentage": 39.49, "elapsed_time": "0:54:10", "remaining_time": "1:23:01", "throughput": 5189.71, "total_tokens": 16869600} {"current_steps": 16755, "total_steps": 42420, "loss": 0.3771, "lr": 0.0007575860421906333, "epoch": 7.899575671852899, "percentage": 39.5, "elapsed_time": "0:54:11", "remaining_time": "1:23:00", "throughput": 5189.76, "total_tokens": 16874176} {"current_steps": 16760, "total_steps": 42420, "loss": 0.3487, "lr": 0.0007574097004257622, "epoch": 7.9019330504479015, "percentage": 39.51, "elapsed_time": "0:54:12", "remaining_time": "1:22:59", "throughput": 5189.87, "total_tokens": 16879712} {"current_steps": 16765, "total_steps": 42420, "loss": 0.3132, "lr": 0.0007572333150857994, "epoch": 7.904290429042904, "percentage": 39.52, "elapsed_time": "0:54:13", "remaining_time": "1:22:58", "throughput": 5189.87, "total_tokens": 16883904} {"current_steps": 16770, "total_steps": 42420, "loss": 0.3587, "lr": 0.000757056886200604, "epoch": 7.906647807637906, "percentage": 39.53, "elapsed_time": "0:54:14", "remaining_time": "1:22:57", "throughput": 5189.94, "total_tokens": 16889216} {"current_steps": 16775, "total_steps": 42420, "loss": 0.3905, "lr": 0.0007568804138000422, "epoch": 7.909005186232909, "percentage": 39.55, "elapsed_time": "0:54:15", "remaining_time": "1:22:56", "throughput": 5189.91, "total_tokens": 16893536} {"current_steps": 16780, "total_steps": 42420, "loss": 0.3153, "lr": 0.0007567038979139882, "epoch": 7.911362564827911, "percentage": 39.56, "elapsed_time": "0:54:16", "remaining_time": "1:22:55", "throughput": 5189.98, "total_tokens": 16898592} {"current_steps": 16785, "total_steps": 42420, "loss": 0.3553, "lr": 0.0007565273385723227, "epoch": 7.9137199434229135, "percentage": 39.57, "elapsed_time": "0:54:16", "remaining_time": "1:22:54", "throughput": 5190.03, "total_tokens": 16903456} {"current_steps": 16790, "total_steps": 42420, "loss": 0.3316, "lr": 0.0007563507358049344, "epoch": 7.916077322017916, "percentage": 39.58, "elapsed_time": "0:54:17", "remaining_time": "1:22:53", "throughput": 5190.17, "total_tokens": 16909120} {"current_steps": 16795, "total_steps": 42420, "loss": 0.3713, "lr": 0.000756174089641719, "epoch": 7.918434700612918, "percentage": 39.59, "elapsed_time": "0:54:18", "remaining_time": "1:22:52", "throughput": 5190.33, "total_tokens": 16914816} {"current_steps": 16800, "total_steps": 42420, "loss": 0.367, "lr": 0.00075599740011258, "epoch": 7.920792079207921, "percentage": 39.6, "elapsed_time": "0:54:19", "remaining_time": "1:22:51", "throughput": 5190.46, "total_tokens": 16920192} {"current_steps": 16805, "total_steps": 42420, "loss": 0.3495, "lr": 0.0007558206672474276, "epoch": 7.923149457802923, "percentage": 39.62, "elapsed_time": "0:54:20", "remaining_time": "1:22:50", "throughput": 5190.55, "total_tokens": 16925344} {"current_steps": 16810, "total_steps": 42420, "loss": 0.3358, "lr": 0.0007556438910761798, "epoch": 7.9255068363979255, "percentage": 39.63, "elapsed_time": "0:54:21", "remaining_time": "1:22:49", "throughput": 5190.62, "total_tokens": 16930432} {"current_steps": 16815, "total_steps": 42420, "loss": 0.2897, "lr": 0.000755467071628762, "epoch": 7.927864214992928, "percentage": 39.64, "elapsed_time": "0:54:22", "remaining_time": "1:22:48", "throughput": 5190.73, "total_tokens": 16935712} {"current_steps": 16820, "total_steps": 42420, "loss": 0.4465, "lr": 0.0007552902089351063, "epoch": 7.93022159358793, "percentage": 39.65, "elapsed_time": "0:54:23", "remaining_time": "1:22:47", "throughput": 5190.85, "total_tokens": 16941088} {"current_steps": 16825, "total_steps": 42420, "loss": 0.3012, "lr": 0.0007551133030251531, "epoch": 7.932578972182933, "percentage": 39.66, "elapsed_time": "0:54:24", "remaining_time": "1:22:46", "throughput": 5190.87, "total_tokens": 16945664} {"current_steps": 16830, "total_steps": 42420, "loss": 0.3418, "lr": 0.0007549363539288491, "epoch": 7.934936350777935, "percentage": 39.67, "elapsed_time": "0:54:25", "remaining_time": "1:22:45", "throughput": 5190.95, "total_tokens": 16950624} {"current_steps": 16835, "total_steps": 42420, "loss": 0.3458, "lr": 0.000754759361676149, "epoch": 7.9372937293729375, "percentage": 39.69, "elapsed_time": "0:54:26", "remaining_time": "1:22:44", "throughput": 5191.07, "total_tokens": 16956064} {"current_steps": 16840, "total_steps": 42420, "loss": 0.2849, "lr": 0.0007545823262970146, "epoch": 7.93965110796794, "percentage": 39.7, "elapsed_time": "0:54:27", "remaining_time": "1:22:42", "throughput": 5191.11, "total_tokens": 16960544} {"current_steps": 16845, "total_steps": 42420, "loss": 0.2635, "lr": 0.0007544052478214147, "epoch": 7.942008486562942, "percentage": 39.71, "elapsed_time": "0:54:28", "remaining_time": "1:22:41", "throughput": 5191.22, "total_tokens": 16965888} {"current_steps": 16850, "total_steps": 42420, "loss": 0.4798, "lr": 0.000754228126279326, "epoch": 7.944365865157945, "percentage": 39.72, "elapsed_time": "0:54:29", "remaining_time": "1:22:40", "throughput": 5191.32, "total_tokens": 16970944} {"current_steps": 16855, "total_steps": 42420, "loss": 0.3385, "lr": 0.0007540509617007322, "epoch": 7.946723243752947, "percentage": 39.73, "elapsed_time": "0:54:30", "remaining_time": "1:22:39", "throughput": 5191.42, "total_tokens": 16976000} {"current_steps": 16860, "total_steps": 42420, "loss": 0.3187, "lr": 0.0007538737541156241, "epoch": 7.9490806223479495, "percentage": 39.75, "elapsed_time": "0:54:30", "remaining_time": "1:22:38", "throughput": 5191.46, "total_tokens": 16980992} {"current_steps": 16865, "total_steps": 42420, "loss": 0.3341, "lr": 0.0007536965035539999, "epoch": 7.951438000942951, "percentage": 39.76, "elapsed_time": "0:54:31", "remaining_time": "1:22:37", "throughput": 5191.51, "total_tokens": 16985568} {"current_steps": 16870, "total_steps": 42420, "loss": 0.3629, "lr": 0.000753519210045865, "epoch": 7.9537953795379535, "percentage": 39.77, "elapsed_time": "0:54:32", "remaining_time": "1:22:36", "throughput": 5191.51, "total_tokens": 16990016} {"current_steps": 16875, "total_steps": 42420, "loss": 0.3499, "lr": 0.0007533418736212325, "epoch": 7.956152758132956, "percentage": 39.78, "elapsed_time": "0:54:33", "remaining_time": "1:22:35", "throughput": 5191.77, "total_tokens": 16996928} {"current_steps": 16880, "total_steps": 42420, "loss": 0.3426, "lr": 0.0007531644943101224, "epoch": 7.958510136727958, "percentage": 39.79, "elapsed_time": "0:54:34", "remaining_time": "1:22:34", "throughput": 5191.83, "total_tokens": 17001728} {"current_steps": 16885, "total_steps": 42420, "loss": 0.3442, "lr": 0.000752987072142562, "epoch": 7.960867515322961, "percentage": 39.8, "elapsed_time": "0:54:35", "remaining_time": "1:22:33", "throughput": 5191.85, "total_tokens": 17006400} {"current_steps": 16890, "total_steps": 42420, "loss": 0.3362, "lr": 0.0007528096071485854, "epoch": 7.963224893917963, "percentage": 39.82, "elapsed_time": "0:54:36", "remaining_time": "1:22:32", "throughput": 5191.98, "total_tokens": 17012096} {"current_steps": 16895, "total_steps": 42420, "loss": 0.3078, "lr": 0.000752632099358235, "epoch": 7.9655822725129655, "percentage": 39.83, "elapsed_time": "0:54:37", "remaining_time": "1:22:31", "throughput": 5192.01, "total_tokens": 17016576} {"current_steps": 16900, "total_steps": 42420, "loss": 0.3541, "lr": 0.0007524545488015596, "epoch": 7.967939651107968, "percentage": 39.84, "elapsed_time": "0:54:38", "remaining_time": "1:22:30", "throughput": 5192.15, "total_tokens": 17022048} {"current_steps": 16905, "total_steps": 42420, "loss": 0.3451, "lr": 0.0007522769555086156, "epoch": 7.97029702970297, "percentage": 39.85, "elapsed_time": "0:54:39", "remaining_time": "1:22:29", "throughput": 5192.24, "total_tokens": 17027296} {"current_steps": 16910, "total_steps": 42420, "loss": 0.4038, "lr": 0.0007520993195094663, "epoch": 7.972654408297973, "percentage": 39.86, "elapsed_time": "0:54:40", "remaining_time": "1:22:28", "throughput": 5192.24, "total_tokens": 17031648} {"current_steps": 16915, "total_steps": 42420, "loss": 0.3426, "lr": 0.0007519216408341827, "epoch": 7.975011786892975, "percentage": 39.88, "elapsed_time": "0:54:41", "remaining_time": "1:22:27", "throughput": 5192.29, "total_tokens": 17036352} {"current_steps": 16920, "total_steps": 42420, "loss": 0.3552, "lr": 0.0007517439195128426, "epoch": 7.9773691654879775, "percentage": 39.89, "elapsed_time": "0:54:42", "remaining_time": "1:22:26", "throughput": 5192.39, "total_tokens": 17041952} {"current_steps": 16925, "total_steps": 42420, "loss": 0.3668, "lr": 0.0007515661555755314, "epoch": 7.97972654408298, "percentage": 39.9, "elapsed_time": "0:54:42", "remaining_time": "1:22:25", "throughput": 5192.45, "total_tokens": 17046784} {"current_steps": 16930, "total_steps": 42420, "loss": 0.3561, "lr": 0.0007513883490523414, "epoch": 7.982083922677982, "percentage": 39.91, "elapsed_time": "0:54:43", "remaining_time": "1:22:24", "throughput": 5192.56, "total_tokens": 17052288} {"current_steps": 16935, "total_steps": 42420, "loss": 0.356, "lr": 0.0007512104999733719, "epoch": 7.984441301272985, "percentage": 39.92, "elapsed_time": "0:54:44", "remaining_time": "1:22:23", "throughput": 5192.71, "total_tokens": 17057792} {"current_steps": 16940, "total_steps": 42420, "loss": 0.423, "lr": 0.0007510326083687304, "epoch": 7.986798679867987, "percentage": 39.93, "elapsed_time": "0:54:45", "remaining_time": "1:22:22", "throughput": 5192.85, "total_tokens": 17063552} {"current_steps": 16945, "total_steps": 42420, "loss": 0.3648, "lr": 0.0007508546742685307, "epoch": 7.9891560584629895, "percentage": 39.95, "elapsed_time": "0:54:46", "remaining_time": "1:22:21", "throughput": 5192.99, "total_tokens": 17069248} {"current_steps": 16950, "total_steps": 42420, "loss": 0.3312, "lr": 0.0007506766977028939, "epoch": 7.991513437057992, "percentage": 39.96, "elapsed_time": "0:54:47", "remaining_time": "1:22:20", "throughput": 5192.99, "total_tokens": 17073600} {"current_steps": 16955, "total_steps": 42420, "loss": 0.3268, "lr": 0.0007504986787019481, "epoch": 7.993870815652993, "percentage": 39.97, "elapsed_time": "0:54:48", "remaining_time": "1:22:19", "throughput": 5193.02, "total_tokens": 17078400} {"current_steps": 16960, "total_steps": 42420, "loss": 0.3574, "lr": 0.0007503206172958294, "epoch": 7.996228194247996, "percentage": 39.98, "elapsed_time": "0:54:49", "remaining_time": "1:22:18", "throughput": 5193.18, "total_tokens": 17084128} {"current_steps": 16965, "total_steps": 42420, "loss": 0.3653, "lr": 0.0007501425135146805, "epoch": 7.998585572842998, "percentage": 39.99, "elapsed_time": "0:54:50", "remaining_time": "1:22:17", "throughput": 5193.31, "total_tokens": 17089472} {"current_steps": 16968, "total_steps": 42420, "eval_loss": 0.3673172891139984, "epoch": 8.0, "percentage": 40.0, "elapsed_time": "0:55:10", "remaining_time": "1:22:45", "throughput": 5163.45, "total_tokens": 17092000} {"current_steps": 16970, "total_steps": 42420, "loss": 0.3675, "lr": 0.0007499643673886512, "epoch": 8.000942951438, "percentage": 40.0, "elapsed_time": "0:55:12", "remaining_time": "1:22:47", "throughput": 5160.6, "total_tokens": 17094208} {"current_steps": 16975, "total_steps": 42420, "loss": 0.3678, "lr": 0.0007497861789478986, "epoch": 8.003300330033003, "percentage": 40.02, "elapsed_time": "0:55:13", "remaining_time": "1:22:46", "throughput": 5160.76, "total_tokens": 17099872} {"current_steps": 16980, "total_steps": 42420, "loss": 0.3676, "lr": 0.0007496079482225869, "epoch": 8.005657708628005, "percentage": 40.03, "elapsed_time": "0:55:14", "remaining_time": "1:22:45", "throughput": 5160.82, "total_tokens": 17104768} {"current_steps": 16985, "total_steps": 42420, "loss": 0.3109, "lr": 0.000749429675242888, "epoch": 8.008015087223008, "percentage": 40.04, "elapsed_time": "0:55:15", "remaining_time": "1:22:44", "throughput": 5160.89, "total_tokens": 17109824} {"current_steps": 16990, "total_steps": 42420, "loss": 0.3829, "lr": 0.0007492513600389799, "epoch": 8.01037246581801, "percentage": 40.05, "elapsed_time": "0:55:16", "remaining_time": "1:22:43", "throughput": 5161.02, "total_tokens": 17115360} {"current_steps": 16995, "total_steps": 42420, "loss": 0.3864, "lr": 0.0007490730026410486, "epoch": 8.012729844413013, "percentage": 40.06, "elapsed_time": "0:55:17", "remaining_time": "1:22:42", "throughput": 5161.06, "total_tokens": 17120160} {"current_steps": 17000, "total_steps": 42420, "loss": 0.3293, "lr": 0.0007488946030792871, "epoch": 8.015087223008015, "percentage": 40.08, "elapsed_time": "0:55:18", "remaining_time": "1:22:41", "throughput": 5161.11, "total_tokens": 17124928} {"current_steps": 17005, "total_steps": 42420, "loss": 0.3675, "lr": 0.0007487161613838953, "epoch": 8.017444601603017, "percentage": 40.09, "elapsed_time": "0:55:18", "remaining_time": "1:22:40", "throughput": 5161.19, "total_tokens": 17129888} {"current_steps": 17010, "total_steps": 42420, "loss": 0.3468, "lr": 0.0007485376775850802, "epoch": 8.01980198019802, "percentage": 40.1, "elapsed_time": "0:55:19", "remaining_time": "1:22:39", "throughput": 5161.26, "total_tokens": 17134848} {"current_steps": 17015, "total_steps": 42420, "loss": 0.3599, "lr": 0.0007483591517130563, "epoch": 8.022159358793022, "percentage": 40.11, "elapsed_time": "0:55:20", "remaining_time": "1:22:38", "throughput": 5161.26, "total_tokens": 17139296} {"current_steps": 17020, "total_steps": 42420, "loss": 0.3548, "lr": 0.0007481805837980448, "epoch": 8.024516737388025, "percentage": 40.12, "elapsed_time": "0:55:21", "remaining_time": "1:22:36", "throughput": 5161.28, "total_tokens": 17143584} {"current_steps": 17025, "total_steps": 42420, "loss": 0.3828, "lr": 0.0007480019738702743, "epoch": 8.026874115983027, "percentage": 40.13, "elapsed_time": "0:55:22", "remaining_time": "1:22:35", "throughput": 5161.36, "total_tokens": 17148576} {"current_steps": 17030, "total_steps": 42420, "loss": 0.3358, "lr": 0.0007478233219599808, "epoch": 8.02923149457803, "percentage": 40.15, "elapsed_time": "0:55:23", "remaining_time": "1:22:34", "throughput": 5161.39, "total_tokens": 17153216} {"current_steps": 17035, "total_steps": 42420, "loss": 0.3355, "lr": 0.0007476446280974064, "epoch": 8.031588873173032, "percentage": 40.16, "elapsed_time": "0:55:24", "remaining_time": "1:22:33", "throughput": 5161.54, "total_tokens": 17158688} {"current_steps": 17040, "total_steps": 42420, "loss": 0.2932, "lr": 0.0007474658923128012, "epoch": 8.033946251768034, "percentage": 40.17, "elapsed_time": "0:55:25", "remaining_time": "1:22:32", "throughput": 5161.61, "total_tokens": 17163776} {"current_steps": 17045, "total_steps": 42420, "loss": 0.3031, "lr": 0.0007472871146364223, "epoch": 8.036303630363037, "percentage": 40.18, "elapsed_time": "0:55:26", "remaining_time": "1:22:31", "throughput": 5161.62, "total_tokens": 17168256} {"current_steps": 17050, "total_steps": 42420, "loss": 0.3839, "lr": 0.0007471082950985336, "epoch": 8.038661008958039, "percentage": 40.19, "elapsed_time": "0:55:27", "remaining_time": "1:22:30", "throughput": 5161.75, "total_tokens": 17173792} {"current_steps": 17055, "total_steps": 42420, "loss": 0.4567, "lr": 0.0007469294337294061, "epoch": 8.041018387553041, "percentage": 40.21, "elapsed_time": "0:55:28", "remaining_time": "1:22:29", "throughput": 5161.99, "total_tokens": 17180256} {"current_steps": 17060, "total_steps": 42420, "loss": 0.3539, "lr": 0.0007467505305593181, "epoch": 8.043375766148044, "percentage": 40.22, "elapsed_time": "0:55:29", "remaining_time": "1:22:28", "throughput": 5162.12, "total_tokens": 17185632} {"current_steps": 17065, "total_steps": 42420, "loss": 0.3682, "lr": 0.0007465715856185549, "epoch": 8.045733144743046, "percentage": 40.23, "elapsed_time": "0:55:30", "remaining_time": "1:22:27", "throughput": 5162.16, "total_tokens": 17190496} {"current_steps": 17070, "total_steps": 42420, "loss": 0.3457, "lr": 0.0007463925989374088, "epoch": 8.048090523338049, "percentage": 40.24, "elapsed_time": "0:55:30", "remaining_time": "1:22:26", "throughput": 5162.2, "total_tokens": 17195264} {"current_steps": 17075, "total_steps": 42420, "loss": 0.3287, "lr": 0.0007462135705461792, "epoch": 8.050447901933051, "percentage": 40.25, "elapsed_time": "0:55:31", "remaining_time": "1:22:25", "throughput": 5162.32, "total_tokens": 17200640} {"current_steps": 17080, "total_steps": 42420, "loss": 0.2935, "lr": 0.0007460345004751726, "epoch": 8.052805280528053, "percentage": 40.26, "elapsed_time": "0:55:33", "remaining_time": "1:22:24", "throughput": 5162.45, "total_tokens": 17206496} {"current_steps": 17085, "total_steps": 42420, "loss": 0.3199, "lr": 0.0007458553887547025, "epoch": 8.055162659123056, "percentage": 40.28, "elapsed_time": "0:55:33", "remaining_time": "1:22:23", "throughput": 5162.47, "total_tokens": 17210880} {"current_steps": 17090, "total_steps": 42420, "loss": 0.386, "lr": 0.0007456762354150895, "epoch": 8.057520037718058, "percentage": 40.29, "elapsed_time": "0:55:34", "remaining_time": "1:22:22", "throughput": 5162.57, "total_tokens": 17216032} {"current_steps": 17095, "total_steps": 42420, "loss": 0.3877, "lr": 0.0007454970404866611, "epoch": 8.05987741631306, "percentage": 40.3, "elapsed_time": "0:55:35", "remaining_time": "1:22:21", "throughput": 5162.57, "total_tokens": 17220224} {"current_steps": 17100, "total_steps": 42420, "loss": 0.3316, "lr": 0.0007453178039997521, "epoch": 8.062234794908063, "percentage": 40.31, "elapsed_time": "0:55:36", "remaining_time": "1:22:20", "throughput": 5162.67, "total_tokens": 17225568} {"current_steps": 17105, "total_steps": 42420, "loss": 0.3583, "lr": 0.0007451385259847041, "epoch": 8.064592173503065, "percentage": 40.32, "elapsed_time": "0:55:37", "remaining_time": "1:22:19", "throughput": 5162.7, "total_tokens": 17230144} {"current_steps": 17110, "total_steps": 42420, "loss": 0.3297, "lr": 0.000744959206471866, "epoch": 8.066949552098066, "percentage": 40.33, "elapsed_time": "0:55:38", "remaining_time": "1:22:18", "throughput": 5162.81, "total_tokens": 17235456} {"current_steps": 17115, "total_steps": 42420, "loss": 0.3213, "lr": 0.0007447798454915933, "epoch": 8.069306930693068, "percentage": 40.35, "elapsed_time": "0:55:39", "remaining_time": "1:22:17", "throughput": 5163.04, "total_tokens": 17241888} {"current_steps": 17120, "total_steps": 42420, "loss": 0.2948, "lr": 0.0007446004430742489, "epoch": 8.07166430928807, "percentage": 40.36, "elapsed_time": "0:55:40", "remaining_time": "1:22:16", "throughput": 5163.04, "total_tokens": 17246272} {"current_steps": 17125, "total_steps": 42420, "loss": 0.4513, "lr": 0.0007444209992502025, "epoch": 8.074021687883073, "percentage": 40.37, "elapsed_time": "0:55:41", "remaining_time": "1:22:15", "throughput": 5163.04, "total_tokens": 17250784} {"current_steps": 17130, "total_steps": 42420, "loss": 0.3276, "lr": 0.0007442415140498311, "epoch": 8.076379066478076, "percentage": 40.38, "elapsed_time": "0:55:42", "remaining_time": "1:22:14", "throughput": 5163.08, "total_tokens": 17255456} {"current_steps": 17135, "total_steps": 42420, "loss": 0.2875, "lr": 0.0007440619875035183, "epoch": 8.078736445073078, "percentage": 40.39, "elapsed_time": "0:55:43", "remaining_time": "1:22:13", "throughput": 5163.23, "total_tokens": 17261472} {"current_steps": 17140, "total_steps": 42420, "loss": 0.3312, "lr": 0.0007438824196416549, "epoch": 8.08109382366808, "percentage": 40.41, "elapsed_time": "0:55:43", "remaining_time": "1:22:12", "throughput": 5163.25, "total_tokens": 17265888} {"current_steps": 17145, "total_steps": 42420, "loss": 0.3148, "lr": 0.0007437028104946387, "epoch": 8.083451202263083, "percentage": 40.42, "elapsed_time": "0:55:44", "remaining_time": "1:22:11", "throughput": 5163.33, "total_tokens": 17270912} {"current_steps": 17150, "total_steps": 42420, "loss": 0.3815, "lr": 0.0007435231600928746, "epoch": 8.085808580858085, "percentage": 40.43, "elapsed_time": "0:55:45", "remaining_time": "1:22:09", "throughput": 5163.4, "total_tokens": 17275872} {"current_steps": 17155, "total_steps": 42420, "loss": 0.3157, "lr": 0.0007433434684667743, "epoch": 8.088165959453088, "percentage": 40.44, "elapsed_time": "0:55:46", "remaining_time": "1:22:08", "throughput": 5163.35, "total_tokens": 17279712} {"current_steps": 17160, "total_steps": 42420, "loss": 0.3504, "lr": 0.0007431637356467565, "epoch": 8.09052333804809, "percentage": 40.45, "elapsed_time": "0:55:47", "remaining_time": "1:22:07", "throughput": 5163.37, "total_tokens": 17284256} {"current_steps": 17165, "total_steps": 42420, "loss": 0.329, "lr": 0.000742983961663247, "epoch": 8.092880716643092, "percentage": 40.46, "elapsed_time": "0:55:48", "remaining_time": "1:22:06", "throughput": 5163.38, "total_tokens": 17288736} {"current_steps": 17170, "total_steps": 42420, "loss": 0.3228, "lr": 0.0007428041465466782, "epoch": 8.095238095238095, "percentage": 40.48, "elapsed_time": "0:55:49", "remaining_time": "1:22:05", "throughput": 5163.37, "total_tokens": 17292960} {"current_steps": 17175, "total_steps": 42420, "loss": 0.3322, "lr": 0.0007426242903274902, "epoch": 8.097595473833097, "percentage": 40.49, "elapsed_time": "0:55:50", "remaining_time": "1:22:04", "throughput": 5163.5, "total_tokens": 17298304} {"current_steps": 17180, "total_steps": 42420, "loss": 0.3643, "lr": 0.0007424443930361293, "epoch": 8.0999528524281, "percentage": 40.5, "elapsed_time": "0:55:51", "remaining_time": "1:22:03", "throughput": 5163.56, "total_tokens": 17303200} {"current_steps": 17185, "total_steps": 42420, "loss": 0.2948, "lr": 0.0007422644547030493, "epoch": 8.102310231023102, "percentage": 40.51, "elapsed_time": "0:55:51", "remaining_time": "1:22:02", "throughput": 5163.66, "total_tokens": 17308544} {"current_steps": 17190, "total_steps": 42420, "loss": 0.3713, "lr": 0.0007420844753587102, "epoch": 8.104667609618105, "percentage": 40.52, "elapsed_time": "0:55:52", "remaining_time": "1:22:01", "throughput": 5163.68, "total_tokens": 17313280} {"current_steps": 17195, "total_steps": 42420, "loss": 0.363, "lr": 0.0007419044550335801, "epoch": 8.107024988213107, "percentage": 40.54, "elapsed_time": "0:55:53", "remaining_time": "1:22:00", "throughput": 5163.81, "total_tokens": 17318784} {"current_steps": 17200, "total_steps": 42420, "loss": 0.3373, "lr": 0.0007417243937581328, "epoch": 8.10938236680811, "percentage": 40.55, "elapsed_time": "0:55:54", "remaining_time": "1:21:59", "throughput": 5163.92, "total_tokens": 17324096} {"current_steps": 17205, "total_steps": 42420, "loss": 0.3425, "lr": 0.00074154429156285, "epoch": 8.111739745403112, "percentage": 40.56, "elapsed_time": "0:55:55", "remaining_time": "1:21:58", "throughput": 5164.07, "total_tokens": 17329760} {"current_steps": 17210, "total_steps": 42420, "loss": 0.3214, "lr": 0.0007413641484782198, "epoch": 8.114097123998114, "percentage": 40.57, "elapsed_time": "0:55:56", "remaining_time": "1:21:57", "throughput": 5164.19, "total_tokens": 17335200} {"current_steps": 17215, "total_steps": 42420, "loss": 0.3292, "lr": 0.0007411839645347375, "epoch": 8.116454502593117, "percentage": 40.58, "elapsed_time": "0:55:57", "remaining_time": "1:21:56", "throughput": 5164.25, "total_tokens": 17339872} {"current_steps": 17220, "total_steps": 42420, "loss": 0.3722, "lr": 0.0007410037397629051, "epoch": 8.118811881188119, "percentage": 40.59, "elapsed_time": "0:55:58", "remaining_time": "1:21:55", "throughput": 5164.38, "total_tokens": 17345472} {"current_steps": 17225, "total_steps": 42420, "loss": 0.3591, "lr": 0.0007408234741932315, "epoch": 8.121169259783121, "percentage": 40.61, "elapsed_time": "0:55:59", "remaining_time": "1:21:54", "throughput": 5164.51, "total_tokens": 17350816} {"current_steps": 17230, "total_steps": 42420, "loss": 0.3561, "lr": 0.0007406431678562326, "epoch": 8.123526638378124, "percentage": 40.62, "elapsed_time": "0:56:00", "remaining_time": "1:21:53", "throughput": 5164.53, "total_tokens": 17355584} {"current_steps": 17235, "total_steps": 42420, "loss": 0.3608, "lr": 0.0007404628207824313, "epoch": 8.125884016973126, "percentage": 40.63, "elapsed_time": "0:56:01", "remaining_time": "1:21:51", "throughput": 5164.62, "total_tokens": 17360608} {"current_steps": 17240, "total_steps": 42420, "loss": 0.3482, "lr": 0.0007402824330023573, "epoch": 8.128241395568129, "percentage": 40.64, "elapsed_time": "0:56:02", "remaining_time": "1:21:51", "throughput": 5164.84, "total_tokens": 17367200} {"current_steps": 17245, "total_steps": 42420, "loss": 0.3612, "lr": 0.0007401020045465472, "epoch": 8.130598774163131, "percentage": 40.65, "elapsed_time": "0:56:03", "remaining_time": "1:21:49", "throughput": 5164.82, "total_tokens": 17371104} {"current_steps": 17250, "total_steps": 42420, "loss": 0.3278, "lr": 0.0007399215354455443, "epoch": 8.132956152758133, "percentage": 40.66, "elapsed_time": "0:56:04", "remaining_time": "1:21:48", "throughput": 5164.8, "total_tokens": 17375328} {"current_steps": 17255, "total_steps": 42420, "loss": 0.298, "lr": 0.0007397410257298991, "epoch": 8.135313531353136, "percentage": 40.68, "elapsed_time": "0:56:05", "remaining_time": "1:21:47", "throughput": 5164.86, "total_tokens": 17380160} {"current_steps": 17260, "total_steps": 42420, "loss": 0.3517, "lr": 0.0007395604754301689, "epoch": 8.137670909948138, "percentage": 40.69, "elapsed_time": "0:56:05", "remaining_time": "1:21:46", "throughput": 5164.93, "total_tokens": 17384992} {"current_steps": 17265, "total_steps": 42420, "loss": 0.3916, "lr": 0.0007393798845769178, "epoch": 8.14002828854314, "percentage": 40.7, "elapsed_time": "0:56:06", "remaining_time": "1:21:45", "throughput": 5165.0, "total_tokens": 17389888} {"current_steps": 17270, "total_steps": 42420, "loss": 0.3302, "lr": 0.0007391992532007164, "epoch": 8.142385667138143, "percentage": 40.71, "elapsed_time": "0:56:07", "remaining_time": "1:21:44", "throughput": 5165.1, "total_tokens": 17395072} {"current_steps": 17275, "total_steps": 42420, "loss": 0.3434, "lr": 0.0007390185813321429, "epoch": 8.144743045733145, "percentage": 40.72, "elapsed_time": "0:56:08", "remaining_time": "1:21:43", "throughput": 5165.25, "total_tokens": 17400608} {"current_steps": 17280, "total_steps": 42420, "loss": 0.3611, "lr": 0.0007388378690017821, "epoch": 8.147100424328148, "percentage": 40.74, "elapsed_time": "0:56:09", "remaining_time": "1:21:42", "throughput": 5165.24, "total_tokens": 17404960} {"current_steps": 17285, "total_steps": 42420, "loss": 0.3683, "lr": 0.0007386571162402251, "epoch": 8.14945780292315, "percentage": 40.75, "elapsed_time": "0:56:10", "remaining_time": "1:21:41", "throughput": 5165.34, "total_tokens": 17410400} {"current_steps": 17290, "total_steps": 42420, "loss": 0.3539, "lr": 0.0007384763230780705, "epoch": 8.151815181518153, "percentage": 40.76, "elapsed_time": "0:56:11", "remaining_time": "1:21:40", "throughput": 5165.32, "total_tokens": 17414560} {"current_steps": 17295, "total_steps": 42420, "loss": 0.3534, "lr": 0.0007382954895459235, "epoch": 8.154172560113155, "percentage": 40.77, "elapsed_time": "0:56:12", "remaining_time": "1:21:39", "throughput": 5165.36, "total_tokens": 17419456} {"current_steps": 17300, "total_steps": 42420, "loss": 0.3395, "lr": 0.0007381146156743963, "epoch": 8.156529938708157, "percentage": 40.78, "elapsed_time": "0:56:13", "remaining_time": "1:21:38", "throughput": 5165.45, "total_tokens": 17424672} {"current_steps": 17305, "total_steps": 42420, "loss": 0.2898, "lr": 0.0007379337014941076, "epoch": 8.15888731730316, "percentage": 40.79, "elapsed_time": "0:56:14", "remaining_time": "1:21:37", "throughput": 5165.51, "total_tokens": 17429408} {"current_steps": 17310, "total_steps": 42420, "loss": 0.3543, "lr": 0.0007377527470356829, "epoch": 8.16124469589816, "percentage": 40.81, "elapsed_time": "0:56:15", "remaining_time": "1:21:35", "throughput": 5165.62, "total_tokens": 17434528} {"current_steps": 17315, "total_steps": 42420, "loss": 0.2826, "lr": 0.0007375717523297552, "epoch": 8.163602074493163, "percentage": 40.82, "elapsed_time": "0:56:15", "remaining_time": "1:21:34", "throughput": 5165.67, "total_tokens": 17439200} {"current_steps": 17320, "total_steps": 42420, "loss": 0.3589, "lr": 0.0007373907174069634, "epoch": 8.165959453088165, "percentage": 40.83, "elapsed_time": "0:56:16", "remaining_time": "1:21:33", "throughput": 5165.69, "total_tokens": 17443616} {"current_steps": 17325, "total_steps": 42420, "loss": 0.3534, "lr": 0.0007372096422979538, "epoch": 8.168316831683168, "percentage": 40.84, "elapsed_time": "0:56:17", "remaining_time": "1:21:32", "throughput": 5165.7, "total_tokens": 17447936} {"current_steps": 17330, "total_steps": 42420, "loss": 0.3088, "lr": 0.0007370285270333794, "epoch": 8.17067421027817, "percentage": 40.85, "elapsed_time": "0:56:18", "remaining_time": "1:21:31", "throughput": 5165.78, "total_tokens": 17453504} {"current_steps": 17335, "total_steps": 42420, "loss": 0.3058, "lr": 0.0007368473716439, "epoch": 8.173031588873172, "percentage": 40.87, "elapsed_time": "0:56:19", "remaining_time": "1:21:30", "throughput": 5165.92, "total_tokens": 17459008} {"current_steps": 17340, "total_steps": 42420, "loss": 0.297, "lr": 0.000736666176160182, "epoch": 8.175388967468175, "percentage": 40.88, "elapsed_time": "0:56:20", "remaining_time": "1:21:29", "throughput": 5166.04, "total_tokens": 17464704} {"current_steps": 17345, "total_steps": 42420, "loss": 0.3412, "lr": 0.0007364849406128986, "epoch": 8.177746346063177, "percentage": 40.89, "elapsed_time": "0:56:21", "remaining_time": "1:21:28", "throughput": 5166.07, "total_tokens": 17469440} {"current_steps": 17350, "total_steps": 42420, "loss": 0.3227, "lr": 0.0007363036650327302, "epoch": 8.18010372465818, "percentage": 40.9, "elapsed_time": "0:56:22", "remaining_time": "1:21:27", "throughput": 5166.04, "total_tokens": 17473408} {"current_steps": 17355, "total_steps": 42420, "loss": 0.3484, "lr": 0.0007361223494503633, "epoch": 8.182461103253182, "percentage": 40.91, "elapsed_time": "0:56:23", "remaining_time": "1:21:26", "throughput": 5166.39, "total_tokens": 17481504} {"current_steps": 17360, "total_steps": 42420, "loss": 0.3381, "lr": 0.0007359409938964919, "epoch": 8.184818481848184, "percentage": 40.92, "elapsed_time": "0:56:24", "remaining_time": "1:21:26", "throughput": 5166.55, "total_tokens": 17487328} {"current_steps": 17365, "total_steps": 42420, "loss": 0.363, "lr": 0.0007357595984018163, "epoch": 8.187175860443187, "percentage": 40.94, "elapsed_time": "0:56:25", "remaining_time": "1:21:24", "throughput": 5166.6, "total_tokens": 17492000} {"current_steps": 17370, "total_steps": 42420, "loss": 0.2959, "lr": 0.0007355781629970436, "epoch": 8.18953323903819, "percentage": 40.95, "elapsed_time": "0:56:26", "remaining_time": "1:21:23", "throughput": 5166.66, "total_tokens": 17497024} {"current_steps": 17375, "total_steps": 42420, "loss": 0.3586, "lr": 0.0007353966877128878, "epoch": 8.191890617633192, "percentage": 40.96, "elapsed_time": "0:56:27", "remaining_time": "1:21:22", "throughput": 5166.85, "total_tokens": 17502944} {"current_steps": 17380, "total_steps": 42420, "loss": 0.3759, "lr": 0.0007352151725800696, "epoch": 8.194247996228194, "percentage": 40.97, "elapsed_time": "0:56:28", "remaining_time": "1:21:21", "throughput": 5166.88, "total_tokens": 17507360} {"current_steps": 17385, "total_steps": 42420, "loss": 0.3518, "lr": 0.0007350336176293165, "epoch": 8.196605374823196, "percentage": 40.98, "elapsed_time": "0:56:29", "remaining_time": "1:21:20", "throughput": 5166.87, "total_tokens": 17511616} {"current_steps": 17390, "total_steps": 42420, "loss": 0.337, "lr": 0.0007348520228913624, "epoch": 8.198962753418199, "percentage": 40.99, "elapsed_time": "0:56:30", "remaining_time": "1:21:19", "throughput": 5166.9, "total_tokens": 17516160} {"current_steps": 17395, "total_steps": 42420, "loss": 0.3448, "lr": 0.0007346703883969485, "epoch": 8.201320132013201, "percentage": 41.01, "elapsed_time": "0:56:30", "remaining_time": "1:21:18", "throughput": 5166.98, "total_tokens": 17521056} {"current_steps": 17400, "total_steps": 42420, "loss": 0.2887, "lr": 0.0007344887141768223, "epoch": 8.203677510608204, "percentage": 41.02, "elapsed_time": "0:56:31", "remaining_time": "1:21:17", "throughput": 5167.07, "total_tokens": 17526272} {"current_steps": 17405, "total_steps": 42420, "loss": 0.4335, "lr": 0.0007343070002617381, "epoch": 8.206034889203206, "percentage": 41.03, "elapsed_time": "0:56:32", "remaining_time": "1:21:16", "throughput": 5167.09, "total_tokens": 17531104} {"current_steps": 17410, "total_steps": 42420, "loss": 0.3756, "lr": 0.0007341252466824572, "epoch": 8.208392267798208, "percentage": 41.04, "elapsed_time": "0:56:33", "remaining_time": "1:21:15", "throughput": 5167.09, "total_tokens": 17535296} {"current_steps": 17415, "total_steps": 42420, "loss": 0.2849, "lr": 0.0007339434534697471, "epoch": 8.21074964639321, "percentage": 41.05, "elapsed_time": "0:56:34", "remaining_time": "1:21:14", "throughput": 5167.17, "total_tokens": 17540448} {"current_steps": 17420, "total_steps": 42420, "loss": 0.3341, "lr": 0.0007337616206543827, "epoch": 8.213107024988213, "percentage": 41.07, "elapsed_time": "0:56:35", "remaining_time": "1:21:13", "throughput": 5167.29, "total_tokens": 17546080} {"current_steps": 17425, "total_steps": 42420, "loss": 0.3283, "lr": 0.0007335797482671447, "epoch": 8.215464403583216, "percentage": 41.08, "elapsed_time": "0:56:36", "remaining_time": "1:21:12", "throughput": 5167.32, "total_tokens": 17550688} {"current_steps": 17430, "total_steps": 42420, "loss": 0.2788, "lr": 0.0007333978363388214, "epoch": 8.217821782178218, "percentage": 41.09, "elapsed_time": "0:56:37", "remaining_time": "1:21:11", "throughput": 5167.42, "total_tokens": 17555936} {"current_steps": 17435, "total_steps": 42420, "loss": 0.3316, "lr": 0.0007332158849002073, "epoch": 8.22017916077322, "percentage": 41.1, "elapsed_time": "0:56:38", "remaining_time": "1:21:10", "throughput": 5167.61, "total_tokens": 17562208} {"current_steps": 17440, "total_steps": 42420, "loss": 0.2638, "lr": 0.0007330338939821036, "epoch": 8.222536539368223, "percentage": 41.11, "elapsed_time": "0:56:39", "remaining_time": "1:21:09", "throughput": 5167.72, "total_tokens": 17567616} {"current_steps": 17445, "total_steps": 42420, "loss": 0.3728, "lr": 0.0007328518636153184, "epoch": 8.224893917963225, "percentage": 41.12, "elapsed_time": "0:56:40", "remaining_time": "1:21:08", "throughput": 5167.78, "total_tokens": 17572352} {"current_steps": 17450, "total_steps": 42420, "loss": 0.3815, "lr": 0.0007326697938306664, "epoch": 8.227251296558228, "percentage": 41.14, "elapsed_time": "0:56:41", "remaining_time": "1:21:06", "throughput": 5167.74, "total_tokens": 17576384} {"current_steps": 17455, "total_steps": 42420, "loss": 0.3469, "lr": 0.0007324876846589686, "epoch": 8.22960867515323, "percentage": 41.15, "elapsed_time": "0:56:42", "remaining_time": "1:21:05", "throughput": 5167.74, "total_tokens": 17580960} {"current_steps": 17460, "total_steps": 42420, "loss": 0.3473, "lr": 0.0007323055361310533, "epoch": 8.231966053748232, "percentage": 41.16, "elapsed_time": "0:56:43", "remaining_time": "1:21:04", "throughput": 5167.86, "total_tokens": 17586304} {"current_steps": 17465, "total_steps": 42420, "loss": 0.3388, "lr": 0.000732123348277755, "epoch": 8.234323432343235, "percentage": 41.17, "elapsed_time": "0:56:43", "remaining_time": "1:21:03", "throughput": 5167.88, "total_tokens": 17590848} {"current_steps": 17470, "total_steps": 42420, "loss": 0.354, "lr": 0.0007319411211299152, "epoch": 8.236680810938237, "percentage": 41.18, "elapsed_time": "0:56:44", "remaining_time": "1:21:02", "throughput": 5167.97, "total_tokens": 17596192} {"current_steps": 17475, "total_steps": 42420, "loss": 0.3456, "lr": 0.0007317588547183816, "epoch": 8.23903818953324, "percentage": 41.2, "elapsed_time": "0:56:45", "remaining_time": "1:21:01", "throughput": 5168.13, "total_tokens": 17602016} {"current_steps": 17480, "total_steps": 42420, "loss": 0.3379, "lr": 0.000731576549074009, "epoch": 8.241395568128242, "percentage": 41.21, "elapsed_time": "0:56:46", "remaining_time": "1:21:00", "throughput": 5168.22, "total_tokens": 17607424} {"current_steps": 17485, "total_steps": 42420, "loss": 0.3595, "lr": 0.0007313942042276586, "epoch": 8.243752946723244, "percentage": 41.22, "elapsed_time": "0:56:47", "remaining_time": "1:20:59", "throughput": 5168.16, "total_tokens": 17611136} {"current_steps": 17490, "total_steps": 42420, "loss": 0.3252, "lr": 0.0007312118202101982, "epoch": 8.246110325318247, "percentage": 41.23, "elapsed_time": "0:56:48", "remaining_time": "1:20:58", "throughput": 5168.21, "total_tokens": 17615904} {"current_steps": 17495, "total_steps": 42420, "loss": 0.2917, "lr": 0.0007310293970525023, "epoch": 8.24846770391325, "percentage": 41.24, "elapsed_time": "0:56:49", "remaining_time": "1:20:57", "throughput": 5168.3, "total_tokens": 17621088} {"current_steps": 17500, "total_steps": 42420, "loss": 0.3536, "lr": 0.0007308469347854522, "epoch": 8.250825082508252, "percentage": 41.25, "elapsed_time": "0:56:50", "remaining_time": "1:20:56", "throughput": 5168.35, "total_tokens": 17625696} {"current_steps": 17505, "total_steps": 42420, "loss": 0.3507, "lr": 0.000730664433439936, "epoch": 8.253182461103254, "percentage": 41.27, "elapsed_time": "0:56:51", "remaining_time": "1:20:55", "throughput": 5168.49, "total_tokens": 17631168} {"current_steps": 17510, "total_steps": 42420, "loss": 0.3641, "lr": 0.0007304818930468472, "epoch": 8.255539839698255, "percentage": 41.28, "elapsed_time": "0:56:52", "remaining_time": "1:20:54", "throughput": 5168.76, "total_tokens": 17638880} {"current_steps": 17515, "total_steps": 42420, "loss": 0.273, "lr": 0.0007302993136370873, "epoch": 8.257897218293257, "percentage": 41.29, "elapsed_time": "0:56:53", "remaining_time": "1:20:53", "throughput": 5168.79, "total_tokens": 17643392} {"current_steps": 17520, "total_steps": 42420, "loss": 0.4053, "lr": 0.000730116695241564, "epoch": 8.26025459688826, "percentage": 41.3, "elapsed_time": "0:56:54", "remaining_time": "1:20:52", "throughput": 5168.93, "total_tokens": 17648864} {"current_steps": 17525, "total_steps": 42420, "loss": 0.2932, "lr": 0.0007299340378911914, "epoch": 8.262611975483262, "percentage": 41.31, "elapsed_time": "0:56:55", "remaining_time": "1:20:51", "throughput": 5169.04, "total_tokens": 17654432} {"current_steps": 17530, "total_steps": 42420, "loss": 0.3683, "lr": 0.0007297513416168902, "epoch": 8.264969354078264, "percentage": 41.32, "elapsed_time": "0:56:56", "remaining_time": "1:20:50", "throughput": 5169.04, "total_tokens": 17658624} {"current_steps": 17535, "total_steps": 42420, "loss": 0.3567, "lr": 0.0007295686064495877, "epoch": 8.267326732673267, "percentage": 41.34, "elapsed_time": "0:56:57", "remaining_time": "1:20:49", "throughput": 5169.25, "total_tokens": 17664800} {"current_steps": 17540, "total_steps": 42420, "loss": 0.3536, "lr": 0.0007293858324202181, "epoch": 8.269684111268269, "percentage": 41.35, "elapsed_time": "0:56:58", "remaining_time": "1:20:48", "throughput": 5169.33, "total_tokens": 17669664} {"current_steps": 17545, "total_steps": 42420, "loss": 0.3306, "lr": 0.0007292030195597218, "epoch": 8.272041489863271, "percentage": 41.36, "elapsed_time": "0:56:59", "remaining_time": "1:20:47", "throughput": 5169.49, "total_tokens": 17675392} {"current_steps": 17550, "total_steps": 42420, "loss": 0.3634, "lr": 0.0007290201678990459, "epoch": 8.274398868458274, "percentage": 41.37, "elapsed_time": "0:56:59", "remaining_time": "1:20:46", "throughput": 5169.43, "total_tokens": 17679360} {"current_steps": 17555, "total_steps": 42420, "loss": 0.2986, "lr": 0.0007288372774691441, "epoch": 8.276756247053276, "percentage": 41.38, "elapsed_time": "0:57:00", "remaining_time": "1:20:45", "throughput": 5169.59, "total_tokens": 17684992} {"current_steps": 17560, "total_steps": 42420, "loss": 0.3513, "lr": 0.0007286543483009767, "epoch": 8.279113625648279, "percentage": 41.4, "elapsed_time": "0:57:01", "remaining_time": "1:20:44", "throughput": 5169.61, "total_tokens": 17689472} {"current_steps": 17565, "total_steps": 42420, "loss": 0.3594, "lr": 0.0007284713804255105, "epoch": 8.281471004243281, "percentage": 41.41, "elapsed_time": "0:57:02", "remaining_time": "1:20:43", "throughput": 5169.7, "total_tokens": 17694368} {"current_steps": 17570, "total_steps": 42420, "loss": 0.3589, "lr": 0.0007282883738737189, "epoch": 8.283828382838283, "percentage": 41.42, "elapsed_time": "0:57:03", "remaining_time": "1:20:41", "throughput": 5169.68, "total_tokens": 17698272} {"current_steps": 17575, "total_steps": 42420, "loss": 0.349, "lr": 0.0007281053286765816, "epoch": 8.286185761433286, "percentage": 41.43, "elapsed_time": "0:57:04", "remaining_time": "1:20:40", "throughput": 5169.71, "total_tokens": 17703072} {"current_steps": 17580, "total_steps": 42420, "loss": 0.3552, "lr": 0.0007279222448650851, "epoch": 8.288543140028288, "percentage": 41.44, "elapsed_time": "0:57:05", "remaining_time": "1:20:39", "throughput": 5169.78, "total_tokens": 17708032} {"current_steps": 17585, "total_steps": 42420, "loss": 0.359, "lr": 0.0007277391224702225, "epoch": 8.29090051862329, "percentage": 41.45, "elapsed_time": "0:57:06", "remaining_time": "1:20:38", "throughput": 5169.88, "total_tokens": 17713440} {"current_steps": 17590, "total_steps": 42420, "loss": 0.3634, "lr": 0.0007275559615229933, "epoch": 8.293257897218293, "percentage": 41.47, "elapsed_time": "0:57:07", "remaining_time": "1:20:37", "throughput": 5169.91, "total_tokens": 17718176} {"current_steps": 17595, "total_steps": 42420, "loss": 0.3238, "lr": 0.0007273727620544035, "epoch": 8.295615275813295, "percentage": 41.48, "elapsed_time": "0:57:07", "remaining_time": "1:20:36", "throughput": 5169.87, "total_tokens": 17722208} {"current_steps": 17600, "total_steps": 42420, "loss": 0.2944, "lr": 0.0007271895240954655, "epoch": 8.297972654408298, "percentage": 41.49, "elapsed_time": "0:57:08", "remaining_time": "1:20:35", "throughput": 5169.89, "total_tokens": 17726848} {"current_steps": 17605, "total_steps": 42420, "loss": 0.2689, "lr": 0.0007270062476771986, "epoch": 8.3003300330033, "percentage": 41.5, "elapsed_time": "0:57:09", "remaining_time": "1:20:34", "throughput": 5169.88, "total_tokens": 17731296} {"current_steps": 17610, "total_steps": 42420, "loss": 0.3213, "lr": 0.0007268229328306285, "epoch": 8.302687411598303, "percentage": 41.51, "elapsed_time": "0:57:10", "remaining_time": "1:20:33", "throughput": 5169.96, "total_tokens": 17736416} {"current_steps": 17615, "total_steps": 42420, "loss": 0.3669, "lr": 0.0007266395795867868, "epoch": 8.305044790193305, "percentage": 41.53, "elapsed_time": "0:57:11", "remaining_time": "1:20:32", "throughput": 5170.1, "total_tokens": 17742208} {"current_steps": 17620, "total_steps": 42420, "loss": 0.3349, "lr": 0.0007264561879767126, "epoch": 8.307402168788308, "percentage": 41.54, "elapsed_time": "0:57:12", "remaining_time": "1:20:31", "throughput": 5170.09, "total_tokens": 17746336} {"current_steps": 17625, "total_steps": 42420, "loss": 0.3323, "lr": 0.0007262727580314508, "epoch": 8.30975954738331, "percentage": 41.55, "elapsed_time": "0:57:13", "remaining_time": "1:20:30", "throughput": 5170.13, "total_tokens": 17750848} {"current_steps": 17630, "total_steps": 42420, "loss": 0.3265, "lr": 0.0007260892897820529, "epoch": 8.312116925978312, "percentage": 41.56, "elapsed_time": "0:57:14", "remaining_time": "1:20:28", "throughput": 5170.05, "total_tokens": 17754496} {"current_steps": 17635, "total_steps": 42420, "loss": 0.3599, "lr": 0.000725905783259577, "epoch": 8.314474304573315, "percentage": 41.57, "elapsed_time": "0:57:14", "remaining_time": "1:20:27", "throughput": 5170.05, "total_tokens": 17758912} {"current_steps": 17640, "total_steps": 42420, "loss": 0.3474, "lr": 0.0007257222384950877, "epoch": 8.316831683168317, "percentage": 41.58, "elapsed_time": "0:57:15", "remaining_time": "1:20:26", "throughput": 5170.25, "total_tokens": 17764928} {"current_steps": 17645, "total_steps": 42420, "loss": 0.3449, "lr": 0.0007255386555196559, "epoch": 8.31918906176332, "percentage": 41.6, "elapsed_time": "0:57:16", "remaining_time": "1:20:25", "throughput": 5170.26, "total_tokens": 17769248} {"current_steps": 17650, "total_steps": 42420, "loss": 0.3559, "lr": 0.0007253550343643592, "epoch": 8.321546440358322, "percentage": 41.61, "elapsed_time": "0:57:17", "remaining_time": "1:20:24", "throughput": 5170.24, "total_tokens": 17773312} {"current_steps": 17655, "total_steps": 42420, "loss": 0.3452, "lr": 0.0007251713750602815, "epoch": 8.323903818953324, "percentage": 41.62, "elapsed_time": "0:57:18", "remaining_time": "1:20:23", "throughput": 5170.37, "total_tokens": 17778784} {"current_steps": 17660, "total_steps": 42420, "loss": 0.3112, "lr": 0.0007249876776385132, "epoch": 8.326261197548327, "percentage": 41.63, "elapsed_time": "0:57:19", "remaining_time": "1:20:22", "throughput": 5170.49, "total_tokens": 17784224} {"current_steps": 17665, "total_steps": 42420, "loss": 0.3421, "lr": 0.000724803942130151, "epoch": 8.32861857614333, "percentage": 41.64, "elapsed_time": "0:57:20", "remaining_time": "1:20:21", "throughput": 5170.65, "total_tokens": 17789952} {"current_steps": 17670, "total_steps": 42420, "loss": 0.3557, "lr": 0.0007246201685662985, "epoch": 8.330975954738332, "percentage": 41.65, "elapsed_time": "0:57:21", "remaining_time": "1:20:20", "throughput": 5170.74, "total_tokens": 17795232} {"current_steps": 17675, "total_steps": 42420, "loss": 0.357, "lr": 0.0007244363569780654, "epoch": 8.333333333333334, "percentage": 41.67, "elapsed_time": "0:57:22", "remaining_time": "1:20:19", "throughput": 5170.78, "total_tokens": 17800224} {"current_steps": 17680, "total_steps": 42420, "loss": 0.3332, "lr": 0.0007242525073965677, "epoch": 8.335690711928336, "percentage": 41.68, "elapsed_time": "0:57:23", "remaining_time": "1:20:18", "throughput": 5170.9, "total_tokens": 17805376} {"current_steps": 17685, "total_steps": 42420, "loss": 0.3389, "lr": 0.0007240686198529279, "epoch": 8.338048090523339, "percentage": 41.69, "elapsed_time": "0:57:24", "remaining_time": "1:20:17", "throughput": 5171.06, "total_tokens": 17811104} {"current_steps": 17690, "total_steps": 42420, "loss": 0.3401, "lr": 0.0007238846943782754, "epoch": 8.340405469118341, "percentage": 41.7, "elapsed_time": "0:57:25", "remaining_time": "1:20:16", "throughput": 5171.04, "total_tokens": 17815232} {"current_steps": 17695, "total_steps": 42420, "loss": 0.3488, "lr": 0.0007237007310037456, "epoch": 8.342762847713344, "percentage": 41.71, "elapsed_time": "0:57:26", "remaining_time": "1:20:15", "throughput": 5171.12, "total_tokens": 17820288} {"current_steps": 17700, "total_steps": 42420, "loss": 0.3231, "lr": 0.00072351672976048, "epoch": 8.345120226308346, "percentage": 41.73, "elapsed_time": "0:57:27", "remaining_time": "1:20:14", "throughput": 5171.24, "total_tokens": 17825728} {"current_steps": 17705, "total_steps": 42420, "loss": 0.3526, "lr": 0.0007233326906796272, "epoch": 8.347477604903348, "percentage": 41.74, "elapsed_time": "0:57:28", "remaining_time": "1:20:13", "throughput": 5171.28, "total_tokens": 17830624} {"current_steps": 17710, "total_steps": 42420, "loss": 0.3644, "lr": 0.0007231486137923419, "epoch": 8.34983498349835, "percentage": 41.75, "elapsed_time": "0:57:28", "remaining_time": "1:20:12", "throughput": 5171.37, "total_tokens": 17836064} {"current_steps": 17715, "total_steps": 42420, "loss": 0.3747, "lr": 0.0007229644991297852, "epoch": 8.352192362093351, "percentage": 41.76, "elapsed_time": "0:57:29", "remaining_time": "1:20:11", "throughput": 5171.51, "total_tokens": 17841472} {"current_steps": 17720, "total_steps": 42420, "loss": 0.3428, "lr": 0.0007227803467231244, "epoch": 8.354549740688354, "percentage": 41.77, "elapsed_time": "0:57:30", "remaining_time": "1:20:10", "throughput": 5171.46, "total_tokens": 17845536} {"current_steps": 17725, "total_steps": 42420, "loss": 0.3413, "lr": 0.0007225961566035335, "epoch": 8.356907119283356, "percentage": 41.78, "elapsed_time": "0:57:31", "remaining_time": "1:20:08", "throughput": 5171.43, "total_tokens": 17849664} {"current_steps": 17730, "total_steps": 42420, "loss": 0.3177, "lr": 0.0007224119288021929, "epoch": 8.359264497878359, "percentage": 41.8, "elapsed_time": "0:57:32", "remaining_time": "1:20:07", "throughput": 5171.59, "total_tokens": 17855392} {"current_steps": 17735, "total_steps": 42420, "loss": 0.2826, "lr": 0.000722227663350289, "epoch": 8.361621876473361, "percentage": 41.81, "elapsed_time": "0:57:33", "remaining_time": "1:20:06", "throughput": 5171.64, "total_tokens": 17860224} {"current_steps": 17740, "total_steps": 42420, "loss": 0.4879, "lr": 0.0007220433602790149, "epoch": 8.363979255068363, "percentage": 41.82, "elapsed_time": "0:57:34", "remaining_time": "1:20:05", "throughput": 5171.74, "total_tokens": 17865888} {"current_steps": 17745, "total_steps": 42420, "loss": 0.3834, "lr": 0.00072185901961957, "epoch": 8.366336633663366, "percentage": 41.83, "elapsed_time": "0:57:35", "remaining_time": "1:20:04", "throughput": 5171.7, "total_tokens": 17869920} {"current_steps": 17750, "total_steps": 42420, "loss": 0.3519, "lr": 0.0007216746414031599, "epoch": 8.368694012258368, "percentage": 41.84, "elapsed_time": "0:57:36", "remaining_time": "1:20:03", "throughput": 5171.72, "total_tokens": 17874464} {"current_steps": 17755, "total_steps": 42420, "loss": 0.3274, "lr": 0.0007214902256609971, "epoch": 8.37105139085337, "percentage": 41.86, "elapsed_time": "0:57:37", "remaining_time": "1:20:02", "throughput": 5171.77, "total_tokens": 17879264} {"current_steps": 17760, "total_steps": 42420, "loss": 0.3157, "lr": 0.0007213057724242996, "epoch": 8.373408769448373, "percentage": 41.87, "elapsed_time": "0:57:38", "remaining_time": "1:20:01", "throughput": 5171.95, "total_tokens": 17885152} {"current_steps": 17765, "total_steps": 42420, "loss": 0.3382, "lr": 0.0007211212817242924, "epoch": 8.375766148043375, "percentage": 41.88, "elapsed_time": "0:57:39", "remaining_time": "1:20:00", "throughput": 5172.06, "total_tokens": 17890304} {"current_steps": 17770, "total_steps": 42420, "loss": 0.3088, "lr": 0.0007209367535922067, "epoch": 8.378123526638378, "percentage": 41.89, "elapsed_time": "0:57:40", "remaining_time": "1:19:59", "throughput": 5172.19, "total_tokens": 17896160} {"current_steps": 17775, "total_steps": 42420, "loss": 0.4062, "lr": 0.0007207521880592799, "epoch": 8.38048090523338, "percentage": 41.9, "elapsed_time": "0:57:41", "remaining_time": "1:19:58", "throughput": 5172.33, "total_tokens": 17901888} {"current_steps": 17780, "total_steps": 42420, "loss": 0.3169, "lr": 0.0007205675851567557, "epoch": 8.382838283828383, "percentage": 41.91, "elapsed_time": "0:57:42", "remaining_time": "1:19:57", "throughput": 5172.45, "total_tokens": 17907328} {"current_steps": 17785, "total_steps": 42420, "loss": 0.3666, "lr": 0.0007203829449158846, "epoch": 8.385195662423385, "percentage": 41.93, "elapsed_time": "0:57:43", "remaining_time": "1:19:56", "throughput": 5172.56, "total_tokens": 17912736} {"current_steps": 17790, "total_steps": 42420, "loss": 0.3727, "lr": 0.0007201982673679226, "epoch": 8.387553041018387, "percentage": 41.94, "elapsed_time": "0:57:43", "remaining_time": "1:19:55", "throughput": 5172.49, "total_tokens": 17916704} {"current_steps": 17795, "total_steps": 42420, "loss": 0.3686, "lr": 0.0007200135525441327, "epoch": 8.38991041961339, "percentage": 41.95, "elapsed_time": "0:57:44", "remaining_time": "1:19:54", "throughput": 5172.57, "total_tokens": 17921568} {"current_steps": 17800, "total_steps": 42420, "loss": 0.3493, "lr": 0.000719828800475784, "epoch": 8.392267798208392, "percentage": 41.96, "elapsed_time": "0:57:45", "remaining_time": "1:19:53", "throughput": 5172.71, "total_tokens": 17927616} {"current_steps": 17805, "total_steps": 42420, "loss": 0.3573, "lr": 0.0007196440111941519, "epoch": 8.394625176803395, "percentage": 41.97, "elapsed_time": "0:57:46", "remaining_time": "1:19:52", "throughput": 5172.83, "total_tokens": 17933248} {"current_steps": 17810, "total_steps": 42420, "loss": 0.3261, "lr": 0.0007194591847305181, "epoch": 8.396982555398397, "percentage": 41.98, "elapsed_time": "0:57:47", "remaining_time": "1:19:51", "throughput": 5172.89, "total_tokens": 17938176} {"current_steps": 17815, "total_steps": 42420, "loss": 0.4108, "lr": 0.0007192743211161703, "epoch": 8.3993399339934, "percentage": 42.0, "elapsed_time": "0:57:48", "remaining_time": "1:19:50", "throughput": 5172.84, "total_tokens": 17942240} {"current_steps": 17820, "total_steps": 42420, "loss": 0.3482, "lr": 0.0007190894203824033, "epoch": 8.401697312588402, "percentage": 42.01, "elapsed_time": "0:57:49", "remaining_time": "1:19:49", "throughput": 5172.94, "total_tokens": 17947264} {"current_steps": 17825, "total_steps": 42420, "loss": 0.3874, "lr": 0.0007189044825605172, "epoch": 8.404054691183404, "percentage": 42.02, "elapsed_time": "0:57:50", "remaining_time": "1:19:48", "throughput": 5172.93, "total_tokens": 17951488} {"current_steps": 17830, "total_steps": 42420, "loss": 0.354, "lr": 0.0007187195076818189, "epoch": 8.406412069778407, "percentage": 42.03, "elapsed_time": "0:57:51", "remaining_time": "1:19:47", "throughput": 5172.94, "total_tokens": 17955904} {"current_steps": 17835, "total_steps": 42420, "loss": 0.3476, "lr": 0.0007185344957776217, "epoch": 8.408769448373409, "percentage": 42.04, "elapsed_time": "0:57:52", "remaining_time": "1:19:46", "throughput": 5173.02, "total_tokens": 17960960} {"current_steps": 17840, "total_steps": 42420, "loss": 0.3392, "lr": 0.0007183494468792449, "epoch": 8.411126826968411, "percentage": 42.06, "elapsed_time": "0:57:52", "remaining_time": "1:19:45", "throughput": 5173.07, "total_tokens": 17965952} {"current_steps": 17845, "total_steps": 42420, "loss": 0.3331, "lr": 0.0007181643610180141, "epoch": 8.413484205563414, "percentage": 42.07, "elapsed_time": "0:57:53", "remaining_time": "1:19:44", "throughput": 5173.17, "total_tokens": 17971136} {"current_steps": 17850, "total_steps": 42420, "loss": 0.2485, "lr": 0.0007179792382252611, "epoch": 8.415841584158416, "percentage": 42.08, "elapsed_time": "0:57:54", "remaining_time": "1:19:42", "throughput": 5173.25, "total_tokens": 17976096} {"current_steps": 17855, "total_steps": 42420, "loss": 0.3626, "lr": 0.0007177940785323242, "epoch": 8.418198962753419, "percentage": 42.09, "elapsed_time": "0:57:55", "remaining_time": "1:19:41", "throughput": 5173.3, "total_tokens": 17980928} {"current_steps": 17860, "total_steps": 42420, "loss": 0.4188, "lr": 0.0007176088819705478, "epoch": 8.420556341348421, "percentage": 42.1, "elapsed_time": "0:57:56", "remaining_time": "1:19:41", "throughput": 5173.47, "total_tokens": 17986848} {"current_steps": 17865, "total_steps": 42420, "loss": 0.337, "lr": 0.0007174236485712825, "epoch": 8.422913719943423, "percentage": 42.11, "elapsed_time": "0:57:57", "remaining_time": "1:19:39", "throughput": 5173.53, "total_tokens": 17991328} {"current_steps": 17870, "total_steps": 42420, "loss": 0.3593, "lr": 0.000717238378365885, "epoch": 8.425271098538426, "percentage": 42.13, "elapsed_time": "0:57:58", "remaining_time": "1:19:38", "throughput": 5173.58, "total_tokens": 17996000} {"current_steps": 17875, "total_steps": 42420, "loss": 0.3425, "lr": 0.0007170530713857187, "epoch": 8.427628477133428, "percentage": 42.14, "elapsed_time": "0:57:59", "remaining_time": "1:19:37", "throughput": 5173.68, "total_tokens": 18001376} {"current_steps": 17880, "total_steps": 42420, "loss": 0.3189, "lr": 0.0007168677276621527, "epoch": 8.42998585572843, "percentage": 42.15, "elapsed_time": "0:58:00", "remaining_time": "1:19:36", "throughput": 5173.82, "total_tokens": 18006816} {"current_steps": 17885, "total_steps": 42420, "loss": 0.3839, "lr": 0.0007166823472265625, "epoch": 8.432343234323433, "percentage": 42.16, "elapsed_time": "0:58:01", "remaining_time": "1:19:35", "throughput": 5174.03, "total_tokens": 18013248} {"current_steps": 17890, "total_steps": 42420, "loss": 0.3125, "lr": 0.00071649693011033, "epoch": 8.434700612918435, "percentage": 42.17, "elapsed_time": "0:58:02", "remaining_time": "1:19:34", "throughput": 5174.05, "total_tokens": 18017664} {"current_steps": 17895, "total_steps": 42420, "loss": 0.3436, "lr": 0.0007163114763448432, "epoch": 8.437057991513438, "percentage": 42.19, "elapsed_time": "0:58:03", "remaining_time": "1:19:33", "throughput": 5174.08, "total_tokens": 18022432} {"current_steps": 17900, "total_steps": 42420, "loss": 0.311, "lr": 0.0007161259859614962, "epoch": 8.43941537010844, "percentage": 42.2, "elapsed_time": "0:58:04", "remaining_time": "1:19:32", "throughput": 5174.14, "total_tokens": 18027296} {"current_steps": 17905, "total_steps": 42420, "loss": 0.4162, "lr": 0.0007159404589916896, "epoch": 8.441772748703443, "percentage": 42.21, "elapsed_time": "0:58:05", "remaining_time": "1:19:31", "throughput": 5174.23, "total_tokens": 18032288} {"current_steps": 17910, "total_steps": 42420, "loss": 0.3415, "lr": 0.0007157548954668295, "epoch": 8.444130127298443, "percentage": 42.22, "elapsed_time": "0:58:05", "remaining_time": "1:19:30", "throughput": 5174.24, "total_tokens": 18036608} {"current_steps": 17915, "total_steps": 42420, "loss": 0.335, "lr": 0.0007155692954183288, "epoch": 8.446487505893446, "percentage": 42.23, "elapsed_time": "0:58:06", "remaining_time": "1:19:29", "throughput": 5174.25, "total_tokens": 18041056} {"current_steps": 17920, "total_steps": 42420, "loss": 0.3328, "lr": 0.0007153836588776068, "epoch": 8.448844884488448, "percentage": 42.24, "elapsed_time": "0:58:07", "remaining_time": "1:19:28", "throughput": 5174.35, "total_tokens": 18046368} {"current_steps": 17925, "total_steps": 42420, "loss": 0.3565, "lr": 0.0007151979858760883, "epoch": 8.45120226308345, "percentage": 42.26, "elapsed_time": "0:58:08", "remaining_time": "1:19:27", "throughput": 5174.48, "total_tokens": 18051584} {"current_steps": 17930, "total_steps": 42420, "loss": 0.3182, "lr": 0.0007150122764452046, "epoch": 8.453559641678453, "percentage": 42.27, "elapsed_time": "0:58:09", "remaining_time": "1:19:26", "throughput": 5174.53, "total_tokens": 18056608} {"current_steps": 17935, "total_steps": 42420, "loss": 0.3296, "lr": 0.0007148265306163932, "epoch": 8.455917020273455, "percentage": 42.28, "elapsed_time": "0:58:10", "remaining_time": "1:19:25", "throughput": 5174.6, "total_tokens": 18061728} {"current_steps": 17940, "total_steps": 42420, "loss": 0.3919, "lr": 0.0007146407484210976, "epoch": 8.458274398868458, "percentage": 42.29, "elapsed_time": "0:58:11", "remaining_time": "1:19:24", "throughput": 5174.73, "total_tokens": 18067392} {"current_steps": 17945, "total_steps": 42420, "loss": 0.3084, "lr": 0.0007144549298907679, "epoch": 8.46063177746346, "percentage": 42.3, "elapsed_time": "0:58:12", "remaining_time": "1:19:23", "throughput": 5174.82, "total_tokens": 18072896} {"current_steps": 17950, "total_steps": 42420, "loss": 0.3564, "lr": 0.0007142690750568597, "epoch": 8.462989156058462, "percentage": 42.31, "elapsed_time": "0:58:13", "remaining_time": "1:19:22", "throughput": 5174.94, "total_tokens": 18078432} {"current_steps": 17955, "total_steps": 42420, "loss": 0.3262, "lr": 0.0007140831839508352, "epoch": 8.465346534653465, "percentage": 42.33, "elapsed_time": "0:58:14", "remaining_time": "1:19:21", "throughput": 5175.03, "total_tokens": 18083712} {"current_steps": 17960, "total_steps": 42420, "loss": 0.3734, "lr": 0.0007138972566041625, "epoch": 8.467703913248467, "percentage": 42.34, "elapsed_time": "0:58:15", "remaining_time": "1:19:20", "throughput": 5175.17, "total_tokens": 18089152} {"current_steps": 17965, "total_steps": 42420, "loss": 0.3278, "lr": 0.0007137112930483162, "epoch": 8.47006129184347, "percentage": 42.35, "elapsed_time": "0:58:16", "remaining_time": "1:19:19", "throughput": 5175.25, "total_tokens": 18094336} {"current_steps": 17970, "total_steps": 42420, "loss": 0.3357, "lr": 0.0007135252933147764, "epoch": 8.472418670438472, "percentage": 42.36, "elapsed_time": "0:58:17", "remaining_time": "1:19:18", "throughput": 5175.37, "total_tokens": 18099936} {"current_steps": 17975, "total_steps": 42420, "loss": 0.3824, "lr": 0.0007133392574350299, "epoch": 8.474776049033474, "percentage": 42.37, "elapsed_time": "0:58:18", "remaining_time": "1:19:17", "throughput": 5175.37, "total_tokens": 18104256} {"current_steps": 17980, "total_steps": 42420, "loss": 0.3471, "lr": 0.0007131531854405695, "epoch": 8.477133427628477, "percentage": 42.39, "elapsed_time": "0:58:18", "remaining_time": "1:19:16", "throughput": 5175.31, "total_tokens": 18108096} {"current_steps": 17985, "total_steps": 42420, "loss": 0.3301, "lr": 0.000712967077362894, "epoch": 8.47949080622348, "percentage": 42.4, "elapsed_time": "0:58:19", "remaining_time": "1:19:14", "throughput": 5175.31, "total_tokens": 18112608} {"current_steps": 17990, "total_steps": 42420, "loss": 0.3462, "lr": 0.0007127809332335083, "epoch": 8.481848184818482, "percentage": 42.41, "elapsed_time": "0:58:20", "remaining_time": "1:19:13", "throughput": 5175.26, "total_tokens": 18116448} {"current_steps": 17995, "total_steps": 42420, "loss": 0.3371, "lr": 0.0007125947530839232, "epoch": 8.484205563413484, "percentage": 42.42, "elapsed_time": "0:58:21", "remaining_time": "1:19:12", "throughput": 5175.35, "total_tokens": 18121376} {"current_steps": 18000, "total_steps": 42420, "loss": 0.2684, "lr": 0.0007124085369456562, "epoch": 8.486562942008486, "percentage": 42.43, "elapsed_time": "0:58:22", "remaining_time": "1:19:11", "throughput": 5175.35, "total_tokens": 18125504} {"current_steps": 18005, "total_steps": 42420, "loss": 0.3513, "lr": 0.0007122222848502304, "epoch": 8.488920320603489, "percentage": 42.44, "elapsed_time": "0:58:23", "remaining_time": "1:19:10", "throughput": 5175.49, "total_tokens": 18130944} {"current_steps": 18010, "total_steps": 42420, "loss": 0.3824, "lr": 0.0007120359968291752, "epoch": 8.491277699198491, "percentage": 42.46, "elapsed_time": "0:58:24", "remaining_time": "1:19:09", "throughput": 5175.53, "total_tokens": 18135584} {"current_steps": 18015, "total_steps": 42420, "loss": 0.3331, "lr": 0.0007118496729140258, "epoch": 8.493635077793494, "percentage": 42.47, "elapsed_time": "0:58:25", "remaining_time": "1:19:08", "throughput": 5175.61, "total_tokens": 18140640} {"current_steps": 18020, "total_steps": 42420, "loss": 0.3307, "lr": 0.0007116633131363237, "epoch": 8.495992456388496, "percentage": 42.48, "elapsed_time": "0:58:25", "remaining_time": "1:19:07", "throughput": 5175.61, "total_tokens": 18145024} {"current_steps": 18025, "total_steps": 42420, "loss": 0.3624, "lr": 0.0007114769175276165, "epoch": 8.498349834983498, "percentage": 42.49, "elapsed_time": "0:58:26", "remaining_time": "1:19:06", "throughput": 5175.7, "total_tokens": 18150176} {"current_steps": 18030, "total_steps": 42420, "loss": 0.3032, "lr": 0.000711290486119458, "epoch": 8.500707213578501, "percentage": 42.5, "elapsed_time": "0:58:27", "remaining_time": "1:19:05", "throughput": 5175.91, "total_tokens": 18156480} {"current_steps": 18035, "total_steps": 42420, "loss": 0.3207, "lr": 0.0007111040189434077, "epoch": 8.503064592173503, "percentage": 42.52, "elapsed_time": "0:58:28", "remaining_time": "1:19:04", "throughput": 5176.01, "total_tokens": 18161696} {"current_steps": 18040, "total_steps": 42420, "loss": 0.3525, "lr": 0.0007109175160310312, "epoch": 8.505421970768506, "percentage": 42.53, "elapsed_time": "0:58:29", "remaining_time": "1:19:03", "throughput": 5176.14, "total_tokens": 18167072} {"current_steps": 18045, "total_steps": 42420, "loss": 0.3325, "lr": 0.0007107309774139004, "epoch": 8.507779349363508, "percentage": 42.54, "elapsed_time": "0:58:30", "remaining_time": "1:19:02", "throughput": 5176.13, "total_tokens": 18171360} {"current_steps": 18050, "total_steps": 42420, "loss": 0.3517, "lr": 0.0007105444031235932, "epoch": 8.51013672795851, "percentage": 42.55, "elapsed_time": "0:58:31", "remaining_time": "1:19:01", "throughput": 5176.27, "total_tokens": 18176800} {"current_steps": 18055, "total_steps": 42420, "loss": 0.3309, "lr": 0.0007103577931916934, "epoch": 8.512494106553513, "percentage": 42.56, "elapsed_time": "0:58:32", "remaining_time": "1:19:00", "throughput": 5176.36, "total_tokens": 18181760} {"current_steps": 18060, "total_steps": 42420, "loss": 0.3767, "lr": 0.0007101711476497909, "epoch": 8.514851485148515, "percentage": 42.57, "elapsed_time": "0:58:33", "remaining_time": "1:18:59", "throughput": 5176.58, "total_tokens": 18188576} {"current_steps": 18065, "total_steps": 42420, "loss": 0.3464, "lr": 0.0007099844665294816, "epoch": 8.517208863743518, "percentage": 42.59, "elapsed_time": "0:58:34", "remaining_time": "1:18:58", "throughput": 5176.61, "total_tokens": 18192928} {"current_steps": 18070, "total_steps": 42420, "loss": 0.3493, "lr": 0.0007097977498623676, "epoch": 8.51956624233852, "percentage": 42.6, "elapsed_time": "0:58:35", "remaining_time": "1:18:57", "throughput": 5176.76, "total_tokens": 18198528} {"current_steps": 18075, "total_steps": 42420, "loss": 0.362, "lr": 0.0007096109976800565, "epoch": 8.521923620933523, "percentage": 42.61, "elapsed_time": "0:58:36", "remaining_time": "1:18:56", "throughput": 5176.78, "total_tokens": 18203072} {"current_steps": 18080, "total_steps": 42420, "loss": 0.3473, "lr": 0.0007094242100141625, "epoch": 8.524280999528525, "percentage": 42.62, "elapsed_time": "0:58:37", "remaining_time": "1:18:54", "throughput": 5176.83, "total_tokens": 18207776} {"current_steps": 18085, "total_steps": 42420, "loss": 0.2914, "lr": 0.0007092373868963056, "epoch": 8.526638378123527, "percentage": 42.63, "elapsed_time": "0:58:38", "remaining_time": "1:18:53", "throughput": 5176.89, "total_tokens": 18212640} {"current_steps": 18090, "total_steps": 42420, "loss": 0.3778, "lr": 0.0007090505283581116, "epoch": 8.52899575671853, "percentage": 42.64, "elapsed_time": "0:58:39", "remaining_time": "1:18:52", "throughput": 5177.01, "total_tokens": 18218048} {"current_steps": 18095, "total_steps": 42420, "loss": 0.4252, "lr": 0.000708863634431213, "epoch": 8.531353135313532, "percentage": 42.66, "elapsed_time": "0:58:39", "remaining_time": "1:18:51", "throughput": 5177.11, "total_tokens": 18223264} {"current_steps": 18100, "total_steps": 42420, "loss": 0.3525, "lr": 0.000708676705147247, "epoch": 8.533710513908535, "percentage": 42.67, "elapsed_time": "0:58:41", "remaining_time": "1:18:50", "throughput": 5177.24, "total_tokens": 18229088} {"current_steps": 18105, "total_steps": 42420, "loss": 0.3514, "lr": 0.0007084897405378581, "epoch": 8.536067892503535, "percentage": 42.68, "elapsed_time": "0:58:42", "remaining_time": "1:18:50", "throughput": 5177.48, "total_tokens": 18235616} {"current_steps": 18110, "total_steps": 42420, "loss": 0.3368, "lr": 0.0007083027406346962, "epoch": 8.53842527109854, "percentage": 42.69, "elapsed_time": "0:58:43", "remaining_time": "1:18:49", "throughput": 5177.56, "total_tokens": 18240608} {"current_steps": 18115, "total_steps": 42420, "loss": 0.3184, "lr": 0.000708115705469417, "epoch": 8.54078264969354, "percentage": 42.7, "elapsed_time": "0:58:43", "remaining_time": "1:18:47", "throughput": 5177.6, "total_tokens": 18245152} {"current_steps": 18120, "total_steps": 42420, "loss": 0.3548, "lr": 0.0007079286350736824, "epoch": 8.543140028288542, "percentage": 42.72, "elapsed_time": "0:58:44", "remaining_time": "1:18:47", "throughput": 5177.72, "total_tokens": 18250624} {"current_steps": 18125, "total_steps": 42420, "loss": 0.3665, "lr": 0.0007077415294791604, "epoch": 8.545497406883545, "percentage": 42.73, "elapsed_time": "0:58:45", "remaining_time": "1:18:45", "throughput": 5177.78, "total_tokens": 18255712} {"current_steps": 18130, "total_steps": 42420, "loss": 0.3212, "lr": 0.0007075543887175246, "epoch": 8.547854785478547, "percentage": 42.74, "elapsed_time": "0:58:46", "remaining_time": "1:18:45", "throughput": 5177.88, "total_tokens": 18261152} {"current_steps": 18135, "total_steps": 42420, "loss": 0.3182, "lr": 0.000707367212820455, "epoch": 8.55021216407355, "percentage": 42.75, "elapsed_time": "0:58:47", "remaining_time": "1:18:43", "throughput": 5177.85, "total_tokens": 18265184} {"current_steps": 18140, "total_steps": 42420, "loss": 0.3374, "lr": 0.0007071800018196371, "epoch": 8.552569542668552, "percentage": 42.76, "elapsed_time": "0:58:48", "remaining_time": "1:18:42", "throughput": 5177.9, "total_tokens": 18269824} {"current_steps": 18145, "total_steps": 42420, "loss": 0.3347, "lr": 0.0007069927557467626, "epoch": 8.554926921263554, "percentage": 42.77, "elapsed_time": "0:58:49", "remaining_time": "1:18:41", "throughput": 5177.95, "total_tokens": 18274592} {"current_steps": 18150, "total_steps": 42420, "loss": 0.2945, "lr": 0.0007068054746335291, "epoch": 8.557284299858557, "percentage": 42.79, "elapsed_time": "0:58:50", "remaining_time": "1:18:40", "throughput": 5178.01, "total_tokens": 18279488} {"current_steps": 18155, "total_steps": 42420, "loss": 0.4085, "lr": 0.0007066181585116401, "epoch": 8.55964167845356, "percentage": 42.8, "elapsed_time": "0:58:51", "remaining_time": "1:18:39", "throughput": 5178.01, "total_tokens": 18283840} {"current_steps": 18160, "total_steps": 42420, "loss": 0.3246, "lr": 0.0007064308074128051, "epoch": 8.561999057048562, "percentage": 42.81, "elapsed_time": "0:58:51", "remaining_time": "1:18:38", "throughput": 5177.95, "total_tokens": 18287712} {"current_steps": 18165, "total_steps": 42420, "loss": 0.323, "lr": 0.0007062434213687392, "epoch": 8.564356435643564, "percentage": 42.82, "elapsed_time": "0:58:52", "remaining_time": "1:18:37", "throughput": 5178.04, "total_tokens": 18292928} {"current_steps": 18170, "total_steps": 42420, "loss": 0.3643, "lr": 0.000706056000411164, "epoch": 8.566713814238566, "percentage": 42.83, "elapsed_time": "0:58:53", "remaining_time": "1:18:36", "throughput": 5178.14, "total_tokens": 18298048} {"current_steps": 18175, "total_steps": 42420, "loss": 0.3506, "lr": 0.0007058685445718067, "epoch": 8.569071192833569, "percentage": 42.85, "elapsed_time": "0:58:54", "remaining_time": "1:18:35", "throughput": 5178.14, "total_tokens": 18302464} {"current_steps": 18180, "total_steps": 42420, "loss": 0.3209, "lr": 0.0007056810538824002, "epoch": 8.571428571428571, "percentage": 42.86, "elapsed_time": "0:58:55", "remaining_time": "1:18:33", "throughput": 5178.19, "total_tokens": 18307360} {"current_steps": 18185, "total_steps": 42420, "loss": 0.3271, "lr": 0.0007054935283746833, "epoch": 8.573785950023574, "percentage": 42.87, "elapsed_time": "0:58:56", "remaining_time": "1:18:32", "throughput": 5178.1, "total_tokens": 18310848} {"current_steps": 18190, "total_steps": 42420, "loss": 0.3639, "lr": 0.0007053059680804013, "epoch": 8.576143328618576, "percentage": 42.88, "elapsed_time": "0:58:57", "remaining_time": "1:18:31", "throughput": 5178.22, "total_tokens": 18316480} {"current_steps": 18195, "total_steps": 42420, "loss": 0.3575, "lr": 0.0007051183730313049, "epoch": 8.578500707213578, "percentage": 42.89, "elapsed_time": "0:58:58", "remaining_time": "1:18:30", "throughput": 5178.3, "total_tokens": 18321568} {"current_steps": 18200, "total_steps": 42420, "loss": 0.3631, "lr": 0.0007049307432591505, "epoch": 8.58085808580858, "percentage": 42.9, "elapsed_time": "0:58:59", "remaining_time": "1:18:29", "throughput": 5178.41, "total_tokens": 18326784} {"current_steps": 18205, "total_steps": 42420, "loss": 0.3411, "lr": 0.0007047430787957008, "epoch": 8.583215464403583, "percentage": 42.92, "elapsed_time": "0:58:59", "remaining_time": "1:18:28", "throughput": 5178.42, "total_tokens": 18331392} {"current_steps": 18210, "total_steps": 42420, "loss": 0.3468, "lr": 0.0007045553796727242, "epoch": 8.585572842998586, "percentage": 42.93, "elapsed_time": "0:59:00", "remaining_time": "1:18:27", "throughput": 5178.51, "total_tokens": 18336480} {"current_steps": 18215, "total_steps": 42420, "loss": 0.324, "lr": 0.0007043676459219952, "epoch": 8.587930221593588, "percentage": 42.94, "elapsed_time": "0:59:01", "remaining_time": "1:18:26", "throughput": 5178.73, "total_tokens": 18342912} {"current_steps": 18220, "total_steps": 42420, "loss": 0.3828, "lr": 0.0007041798775752934, "epoch": 8.59028760018859, "percentage": 42.95, "elapsed_time": "0:59:02", "remaining_time": "1:18:25", "throughput": 5178.77, "total_tokens": 18347808} {"current_steps": 18225, "total_steps": 42420, "loss": 0.3229, "lr": 0.0007039920746644053, "epoch": 8.592644978783593, "percentage": 42.96, "elapsed_time": "0:59:03", "remaining_time": "1:18:24", "throughput": 5178.88, "total_tokens": 18353376} {"current_steps": 18230, "total_steps": 42420, "loss": 0.3607, "lr": 0.0007038042372211225, "epoch": 8.595002357378595, "percentage": 42.98, "elapsed_time": "0:59:04", "remaining_time": "1:18:23", "throughput": 5178.97, "total_tokens": 18358592} {"current_steps": 18235, "total_steps": 42420, "loss": 0.3644, "lr": 0.0007036163652772427, "epoch": 8.597359735973598, "percentage": 42.99, "elapsed_time": "0:59:05", "remaining_time": "1:18:22", "throughput": 5179.21, "total_tokens": 18365280} {"current_steps": 18240, "total_steps": 42420, "loss": 0.3562, "lr": 0.0007034284588645696, "epoch": 8.5997171145686, "percentage": 43.0, "elapsed_time": "0:59:06", "remaining_time": "1:18:21", "throughput": 5179.17, "total_tokens": 18369088} {"current_steps": 18245, "total_steps": 42420, "loss": 0.3463, "lr": 0.0007032405180149123, "epoch": 8.602074493163602, "percentage": 43.01, "elapsed_time": "0:59:07", "remaining_time": "1:18:20", "throughput": 5179.17, "total_tokens": 18373440} {"current_steps": 18250, "total_steps": 42420, "loss": 0.3817, "lr": 0.0007030525427600861, "epoch": 8.604431871758605, "percentage": 43.02, "elapsed_time": "0:59:08", "remaining_time": "1:18:19", "throughput": 5179.28, "total_tokens": 18378880} {"current_steps": 18255, "total_steps": 42420, "loss": 0.3316, "lr": 0.0007028645331319121, "epoch": 8.606789250353607, "percentage": 43.03, "elapsed_time": "0:59:09", "remaining_time": "1:18:18", "throughput": 5179.32, "total_tokens": 18383584} {"current_steps": 18260, "total_steps": 42420, "loss": 0.3634, "lr": 0.0007026764891622172, "epoch": 8.60914662894861, "percentage": 43.05, "elapsed_time": "0:59:10", "remaining_time": "1:18:17", "throughput": 5179.34, "total_tokens": 18388224} {"current_steps": 18265, "total_steps": 42420, "loss": 0.3408, "lr": 0.0007024884108828338, "epoch": 8.611504007543612, "percentage": 43.06, "elapsed_time": "0:59:11", "remaining_time": "1:18:16", "throughput": 5179.41, "total_tokens": 18393280} {"current_steps": 18270, "total_steps": 42420, "loss": 0.3458, "lr": 0.0007023002983256004, "epoch": 8.613861386138614, "percentage": 43.07, "elapsed_time": "0:59:12", "remaining_time": "1:18:15", "throughput": 5179.58, "total_tokens": 18399488} {"current_steps": 18275, "total_steps": 42420, "loss": 0.3293, "lr": 0.0007021121515223615, "epoch": 8.616218764733617, "percentage": 43.08, "elapsed_time": "0:59:13", "remaining_time": "1:18:14", "throughput": 5179.68, "total_tokens": 18404864} {"current_steps": 18280, "total_steps": 42420, "loss": 0.3072, "lr": 0.000701923970504967, "epoch": 8.61857614332862, "percentage": 43.09, "elapsed_time": "0:59:14", "remaining_time": "1:18:13", "throughput": 5179.75, "total_tokens": 18409696} {"current_steps": 18285, "total_steps": 42420, "loss": 0.3489, "lr": 0.0007017357553052727, "epoch": 8.620933521923622, "percentage": 43.1, "elapsed_time": "0:59:15", "remaining_time": "1:18:12", "throughput": 5179.81, "total_tokens": 18414496} {"current_steps": 18290, "total_steps": 42420, "loss": 0.393, "lr": 0.0007015475059551404, "epoch": 8.623290900518624, "percentage": 43.12, "elapsed_time": "0:59:15", "remaining_time": "1:18:11", "throughput": 5179.75, "total_tokens": 18418528} {"current_steps": 18295, "total_steps": 42420, "loss": 0.3758, "lr": 0.0007013592224864374, "epoch": 8.625648279113626, "percentage": 43.13, "elapsed_time": "0:59:16", "remaining_time": "1:18:10", "throughput": 5179.83, "total_tokens": 18423744} {"current_steps": 18300, "total_steps": 42420, "loss": 0.3148, "lr": 0.0007011709049310369, "epoch": 8.628005657708629, "percentage": 43.14, "elapsed_time": "0:59:17", "remaining_time": "1:18:09", "throughput": 5179.82, "total_tokens": 18428224} {"current_steps": 18305, "total_steps": 42420, "loss": 0.3406, "lr": 0.0007009825533208178, "epoch": 8.630363036303631, "percentage": 43.15, "elapsed_time": "0:59:18", "remaining_time": "1:18:08", "throughput": 5179.86, "total_tokens": 18433024} {"current_steps": 18310, "total_steps": 42420, "loss": 0.3121, "lr": 0.000700794167687665, "epoch": 8.632720414898632, "percentage": 43.16, "elapsed_time": "0:59:19", "remaining_time": "1:18:07", "throughput": 5179.96, "total_tokens": 18438144} {"current_steps": 18315, "total_steps": 42420, "loss": 0.3578, "lr": 0.0007006057480634689, "epoch": 8.635077793493634, "percentage": 43.18, "elapsed_time": "0:59:20", "remaining_time": "1:18:06", "throughput": 5180.05, "total_tokens": 18443200} {"current_steps": 18320, "total_steps": 42420, "loss": 0.3506, "lr": 0.0007004172944801258, "epoch": 8.637435172088637, "percentage": 43.19, "elapsed_time": "0:59:21", "remaining_time": "1:18:05", "throughput": 5180.22, "total_tokens": 18448960} {"current_steps": 18325, "total_steps": 42420, "loss": 0.3863, "lr": 0.0007002288069695376, "epoch": 8.639792550683639, "percentage": 43.2, "elapsed_time": "0:59:22", "remaining_time": "1:18:03", "throughput": 5180.23, "total_tokens": 18453408} {"current_steps": 18330, "total_steps": 42420, "loss": 0.3566, "lr": 0.000700040285563612, "epoch": 8.642149929278641, "percentage": 43.21, "elapsed_time": "0:59:23", "remaining_time": "1:18:03", "throughput": 5180.41, "total_tokens": 18459616} {"current_steps": 18335, "total_steps": 42420, "loss": 0.3408, "lr": 0.0006998517302942626, "epoch": 8.644507307873644, "percentage": 43.22, "elapsed_time": "0:59:24", "remaining_time": "1:18:02", "throughput": 5180.41, "total_tokens": 18464224} {"current_steps": 18340, "total_steps": 42420, "loss": 0.3413, "lr": 0.0006996631411934085, "epoch": 8.646864686468646, "percentage": 43.23, "elapsed_time": "0:59:25", "remaining_time": "1:18:00", "throughput": 5180.47, "total_tokens": 18469088} {"current_steps": 18345, "total_steps": 42420, "loss": 0.3508, "lr": 0.0006994745182929748, "epoch": 8.649222065063649, "percentage": 43.25, "elapsed_time": "0:59:25", "remaining_time": "1:17:59", "throughput": 5180.46, "total_tokens": 18473440} {"current_steps": 18350, "total_steps": 42420, "loss": 0.2663, "lr": 0.000699285861624892, "epoch": 8.651579443658651, "percentage": 43.26, "elapsed_time": "0:59:26", "remaining_time": "1:17:58", "throughput": 5180.48, "total_tokens": 18478080} {"current_steps": 18355, "total_steps": 42420, "loss": 0.4481, "lr": 0.0006990971712210965, "epoch": 8.653936822253653, "percentage": 43.27, "elapsed_time": "0:59:27", "remaining_time": "1:17:57", "throughput": 5180.51, "total_tokens": 18482592} {"current_steps": 18360, "total_steps": 42420, "loss": 0.3171, "lr": 0.0006989084471135304, "epoch": 8.656294200848656, "percentage": 43.28, "elapsed_time": "0:59:28", "remaining_time": "1:17:56", "throughput": 5180.47, "total_tokens": 18486592} {"current_steps": 18365, "total_steps": 42420, "loss": 0.3485, "lr": 0.0006987196893341414, "epoch": 8.658651579443658, "percentage": 43.29, "elapsed_time": "0:59:29", "remaining_time": "1:17:55", "throughput": 5180.56, "total_tokens": 18491616} {"current_steps": 18370, "total_steps": 42420, "loss": 0.345, "lr": 0.0006985308979148831, "epoch": 8.66100895803866, "percentage": 43.31, "elapsed_time": "0:59:30", "remaining_time": "1:17:54", "throughput": 5180.66, "total_tokens": 18497152} {"current_steps": 18375, "total_steps": 42420, "loss": 0.3596, "lr": 0.0006983420728877145, "epoch": 8.663366336633663, "percentage": 43.32, "elapsed_time": "0:59:31", "remaining_time": "1:17:53", "throughput": 5180.69, "total_tokens": 18501792} {"current_steps": 18380, "total_steps": 42420, "loss": 0.3563, "lr": 0.000698153214284601, "epoch": 8.665723715228665, "percentage": 43.33, "elapsed_time": "0:59:32", "remaining_time": "1:17:52", "throughput": 5180.84, "total_tokens": 18507776} {"current_steps": 18385, "total_steps": 42420, "loss": 0.3691, "lr": 0.0006979643221375124, "epoch": 8.668081093823668, "percentage": 43.34, "elapsed_time": "0:59:33", "remaining_time": "1:17:51", "throughput": 5180.93, "total_tokens": 18513056} {"current_steps": 18390, "total_steps": 42420, "loss": 0.3351, "lr": 0.0006977753964784255, "epoch": 8.67043847241867, "percentage": 43.35, "elapsed_time": "0:59:34", "remaining_time": "1:17:50", "throughput": 5180.92, "total_tokens": 18517376} {"current_steps": 18395, "total_steps": 42420, "loss": 0.3376, "lr": 0.0006975864373393219, "epoch": 8.672795851013673, "percentage": 43.36, "elapsed_time": "0:59:35", "remaining_time": "1:17:49", "throughput": 5181.07, "total_tokens": 18523040} {"current_steps": 18400, "total_steps": 42420, "loss": 0.4346, "lr": 0.0006973974447521895, "epoch": 8.675153229608675, "percentage": 43.38, "elapsed_time": "0:59:35", "remaining_time": "1:17:48", "throughput": 5181.03, "total_tokens": 18526816} {"current_steps": 18405, "total_steps": 42420, "loss": 0.4017, "lr": 0.0006972084187490211, "epoch": 8.677510608203677, "percentage": 43.39, "elapsed_time": "0:59:36", "remaining_time": "1:17:47", "throughput": 5181.08, "total_tokens": 18531552} {"current_steps": 18410, "total_steps": 42420, "loss": 0.3519, "lr": 0.0006970193593618158, "epoch": 8.67986798679868, "percentage": 43.4, "elapsed_time": "0:59:37", "remaining_time": "1:17:46", "throughput": 5181.25, "total_tokens": 18537792} {"current_steps": 18415, "total_steps": 42420, "loss": 0.3437, "lr": 0.0006968302666225782, "epoch": 8.682225365393682, "percentage": 43.41, "elapsed_time": "0:59:38", "remaining_time": "1:17:45", "throughput": 5181.27, "total_tokens": 18542336} {"current_steps": 18420, "total_steps": 42420, "loss": 0.3466, "lr": 0.0006966411405633186, "epoch": 8.684582743988685, "percentage": 43.42, "elapsed_time": "0:59:39", "remaining_time": "1:17:43", "throughput": 5181.34, "total_tokens": 18547168} {"current_steps": 18425, "total_steps": 42420, "loss": 0.3356, "lr": 0.0006964519812160526, "epoch": 8.686940122583687, "percentage": 43.43, "elapsed_time": "0:59:40", "remaining_time": "1:17:42", "throughput": 5181.46, "total_tokens": 18552544} {"current_steps": 18430, "total_steps": 42420, "loss": 0.3515, "lr": 0.0006962627886128016, "epoch": 8.68929750117869, "percentage": 43.45, "elapsed_time": "0:59:41", "remaining_time": "1:17:41", "throughput": 5181.5, "total_tokens": 18557024} {"current_steps": 18435, "total_steps": 42420, "loss": 0.3047, "lr": 0.0006960735627855929, "epoch": 8.691654879773692, "percentage": 43.46, "elapsed_time": "0:59:42", "remaining_time": "1:17:40", "throughput": 5181.47, "total_tokens": 18561248} {"current_steps": 18440, "total_steps": 42420, "loss": 0.3801, "lr": 0.0006958843037664593, "epoch": 8.694012258368694, "percentage": 43.47, "elapsed_time": "0:59:43", "remaining_time": "1:17:39", "throughput": 5181.47, "total_tokens": 18565920} {"current_steps": 18445, "total_steps": 42420, "loss": 0.3667, "lr": 0.0006956950115874389, "epoch": 8.696369636963697, "percentage": 43.48, "elapsed_time": "0:59:44", "remaining_time": "1:17:38", "throughput": 5181.56, "total_tokens": 18570976} {"current_steps": 18450, "total_steps": 42420, "loss": 0.3401, "lr": 0.0006955056862805758, "epoch": 8.698727015558699, "percentage": 43.49, "elapsed_time": "0:59:44", "remaining_time": "1:17:37", "throughput": 5181.57, "total_tokens": 18575360} {"current_steps": 18455, "total_steps": 42420, "loss": 0.3516, "lr": 0.0006953163278779193, "epoch": 8.701084394153701, "percentage": 43.51, "elapsed_time": "0:59:45", "remaining_time": "1:17:36", "throughput": 5181.62, "total_tokens": 18579936} {"current_steps": 18460, "total_steps": 42420, "loss": 0.3567, "lr": 0.000695126936411525, "epoch": 8.703441772748704, "percentage": 43.52, "elapsed_time": "0:59:46", "remaining_time": "1:17:35", "throughput": 5181.78, "total_tokens": 18586016} {"current_steps": 18465, "total_steps": 42420, "loss": 0.3399, "lr": 0.0006949375119134535, "epoch": 8.705799151343706, "percentage": 43.53, "elapsed_time": "0:59:47", "remaining_time": "1:17:34", "throughput": 5181.83, "total_tokens": 18590816} {"current_steps": 18470, "total_steps": 42420, "loss": 0.3078, "lr": 0.000694748054415771, "epoch": 8.708156529938709, "percentage": 43.54, "elapsed_time": "0:59:48", "remaining_time": "1:17:33", "throughput": 5181.96, "total_tokens": 18596800} {"current_steps": 18475, "total_steps": 42420, "loss": 0.2537, "lr": 0.0006945585639505494, "epoch": 8.710513908533711, "percentage": 43.55, "elapsed_time": "0:59:49", "remaining_time": "1:17:32", "throughput": 5182.05, "total_tokens": 18602080} {"current_steps": 18480, "total_steps": 42420, "loss": 0.2556, "lr": 0.0006943690405498664, "epoch": 8.712871287128714, "percentage": 43.56, "elapsed_time": "0:59:50", "remaining_time": "1:17:31", "throughput": 5182.08, "total_tokens": 18606688} {"current_steps": 18485, "total_steps": 42420, "loss": 0.4581, "lr": 0.0006941794842458051, "epoch": 8.715228665723716, "percentage": 43.58, "elapsed_time": "0:59:51", "remaining_time": "1:17:30", "throughput": 5182.13, "total_tokens": 18611520} {"current_steps": 18490, "total_steps": 42420, "loss": 0.3079, "lr": 0.000693989895070454, "epoch": 8.717586044318718, "percentage": 43.59, "elapsed_time": "0:59:52", "remaining_time": "1:17:29", "throughput": 5182.35, "total_tokens": 18618016} {"current_steps": 18495, "total_steps": 42420, "loss": 0.2807, "lr": 0.0006938002730559075, "epoch": 8.71994342291372, "percentage": 43.6, "elapsed_time": "0:59:53", "remaining_time": "1:17:28", "throughput": 5182.42, "total_tokens": 18622976} {"current_steps": 18500, "total_steps": 42420, "loss": 0.405, "lr": 0.0006936106182342652, "epoch": 8.722300801508723, "percentage": 43.61, "elapsed_time": "0:59:54", "remaining_time": "1:17:27", "throughput": 5182.43, "total_tokens": 18627424} {"current_steps": 18505, "total_steps": 42420, "loss": 0.3422, "lr": 0.0006934209306376326, "epoch": 8.724658180103724, "percentage": 43.62, "elapsed_time": "0:59:55", "remaining_time": "1:17:26", "throughput": 5182.52, "total_tokens": 18632800} {"current_steps": 18510, "total_steps": 42420, "loss": 0.3462, "lr": 0.0006932312102981206, "epoch": 8.727015558698728, "percentage": 43.64, "elapsed_time": "0:59:56", "remaining_time": "1:17:25", "throughput": 5182.7, "total_tokens": 18639008} {"current_steps": 18515, "total_steps": 42420, "loss": 0.3502, "lr": 0.0006930414572478456, "epoch": 8.729372937293729, "percentage": 43.65, "elapsed_time": "0:59:57", "remaining_time": "1:17:24", "throughput": 5182.72, "total_tokens": 18643648} {"current_steps": 18520, "total_steps": 42420, "loss": 0.3405, "lr": 0.0006928516715189295, "epoch": 8.731730315888731, "percentage": 43.66, "elapsed_time": "0:59:58", "remaining_time": "1:17:23", "throughput": 5182.74, "total_tokens": 18648256} {"current_steps": 18525, "total_steps": 42420, "loss": 0.3542, "lr": 0.0006926618531435, "epoch": 8.734087694483733, "percentage": 43.67, "elapsed_time": "0:59:59", "remaining_time": "1:17:22", "throughput": 5182.78, "total_tokens": 18652864} {"current_steps": 18530, "total_steps": 42420, "loss": 0.2824, "lr": 0.0006924720021536896, "epoch": 8.736445073078736, "percentage": 43.68, "elapsed_time": "0:59:59", "remaining_time": "1:17:21", "throughput": 5182.81, "total_tokens": 18657920} {"current_steps": 18535, "total_steps": 42420, "loss": 0.3279, "lr": 0.0006922821185816375, "epoch": 8.738802451673738, "percentage": 43.69, "elapsed_time": "1:00:01", "remaining_time": "1:17:20", "throughput": 5182.99, "total_tokens": 18664192} {"current_steps": 18540, "total_steps": 42420, "loss": 0.3971, "lr": 0.0006920922024594875, "epoch": 8.74115983026874, "percentage": 43.71, "elapsed_time": "1:00:01", "remaining_time": "1:17:19", "throughput": 5183.05, "total_tokens": 18669312} {"current_steps": 18545, "total_steps": 42420, "loss": 0.2812, "lr": 0.0006919022538193891, "epoch": 8.743517208863743, "percentage": 43.72, "elapsed_time": "1:00:02", "remaining_time": "1:17:18", "throughput": 5183.09, "total_tokens": 18673920} {"current_steps": 18550, "total_steps": 42420, "loss": 0.2902, "lr": 0.0006917122726934975, "epoch": 8.745874587458745, "percentage": 43.73, "elapsed_time": "1:00:03", "remaining_time": "1:17:17", "throughput": 5183.12, "total_tokens": 18678464} {"current_steps": 18555, "total_steps": 42420, "loss": 0.3943, "lr": 0.0006915222591139731, "epoch": 8.748231966053748, "percentage": 43.74, "elapsed_time": "1:00:04", "remaining_time": "1:17:16", "throughput": 5183.17, "total_tokens": 18683552} {"current_steps": 18560, "total_steps": 42420, "loss": 0.3102, "lr": 0.0006913322131129821, "epoch": 8.75058934464875, "percentage": 43.75, "elapsed_time": "1:00:05", "remaining_time": "1:17:15", "throughput": 5183.24, "total_tokens": 18688416} {"current_steps": 18565, "total_steps": 42420, "loss": 0.2998, "lr": 0.0006911421347226962, "epoch": 8.752946723243753, "percentage": 43.76, "elapsed_time": "1:00:06", "remaining_time": "1:17:14", "throughput": 5183.26, "total_tokens": 18692992} {"current_steps": 18570, "total_steps": 42420, "loss": 0.3637, "lr": 0.0006909520239752922, "epoch": 8.755304101838755, "percentage": 43.78, "elapsed_time": "1:00:07", "remaining_time": "1:17:12", "throughput": 5183.27, "total_tokens": 18697312} {"current_steps": 18575, "total_steps": 42420, "loss": 0.3343, "lr": 0.0006907618809029528, "epoch": 8.757661480433757, "percentage": 43.79, "elapsed_time": "1:00:08", "remaining_time": "1:17:11", "throughput": 5183.28, "total_tokens": 18701792} {"current_steps": 18580, "total_steps": 42420, "loss": 0.3692, "lr": 0.0006905717055378658, "epoch": 8.76001885902876, "percentage": 43.8, "elapsed_time": "1:00:09", "remaining_time": "1:17:10", "throughput": 5183.42, "total_tokens": 18707488} {"current_steps": 18585, "total_steps": 42420, "loss": 0.3098, "lr": 0.0006903814979122248, "epoch": 8.762376237623762, "percentage": 43.81, "elapsed_time": "1:00:10", "remaining_time": "1:17:09", "throughput": 5183.5, "total_tokens": 18712544} {"current_steps": 18590, "total_steps": 42420, "loss": 0.3287, "lr": 0.0006901912580582288, "epoch": 8.764733616218765, "percentage": 43.82, "elapsed_time": "1:00:10", "remaining_time": "1:17:08", "throughput": 5183.6, "total_tokens": 18717984} {"current_steps": 18595, "total_steps": 42420, "loss": 0.3493, "lr": 0.000690000986008082, "epoch": 8.767090994813767, "percentage": 43.84, "elapsed_time": "1:00:11", "remaining_time": "1:17:07", "throughput": 5183.65, "total_tokens": 18722624} {"current_steps": 18600, "total_steps": 42420, "loss": 0.3267, "lr": 0.0006898106817939943, "epoch": 8.76944837340877, "percentage": 43.85, "elapsed_time": "1:00:12", "remaining_time": "1:17:06", "throughput": 5183.84, "total_tokens": 18729024} {"current_steps": 18605, "total_steps": 42420, "loss": 0.3403, "lr": 0.000689620345448181, "epoch": 8.771805752003772, "percentage": 43.86, "elapsed_time": "1:00:13", "remaining_time": "1:17:05", "throughput": 5183.92, "total_tokens": 18734112} {"current_steps": 18610, "total_steps": 42420, "loss": 0.3337, "lr": 0.0006894299770028628, "epoch": 8.774163130598774, "percentage": 43.87, "elapsed_time": "1:00:14", "remaining_time": "1:17:04", "throughput": 5183.88, "total_tokens": 18737888} {"current_steps": 18615, "total_steps": 42420, "loss": 0.3414, "lr": 0.0006892395764902657, "epoch": 8.776520509193777, "percentage": 43.88, "elapsed_time": "1:00:15", "remaining_time": "1:17:03", "throughput": 5183.9, "total_tokens": 18742336} {"current_steps": 18620, "total_steps": 42420, "loss": 0.3331, "lr": 0.0006890491439426216, "epoch": 8.778877887788779, "percentage": 43.89, "elapsed_time": "1:00:16", "remaining_time": "1:17:02", "throughput": 5183.98, "total_tokens": 18747456} {"current_steps": 18625, "total_steps": 42420, "loss": 0.3277, "lr": 0.0006888586793921671, "epoch": 8.781235266383781, "percentage": 43.91, "elapsed_time": "1:00:17", "remaining_time": "1:17:01", "throughput": 5184.07, "total_tokens": 18752704} {"current_steps": 18630, "total_steps": 42420, "loss": 0.316, "lr": 0.000688668182871145, "epoch": 8.783592644978784, "percentage": 43.92, "elapsed_time": "1:00:18", "remaining_time": "1:17:00", "throughput": 5184.11, "total_tokens": 18757824} {"current_steps": 18635, "total_steps": 42420, "loss": 0.3372, "lr": 0.0006884776544118028, "epoch": 8.785950023573786, "percentage": 43.93, "elapsed_time": "1:00:19", "remaining_time": "1:16:59", "throughput": 5184.19, "total_tokens": 18762688} {"current_steps": 18640, "total_steps": 42420, "loss": 0.3589, "lr": 0.0006882870940463939, "epoch": 8.788307402168789, "percentage": 43.94, "elapsed_time": "1:00:20", "remaining_time": "1:16:58", "throughput": 5184.31, "total_tokens": 18768416} {"current_steps": 18645, "total_steps": 42420, "loss": 0.3051, "lr": 0.000688096501807177, "epoch": 8.790664780763791, "percentage": 43.95, "elapsed_time": "1:00:21", "remaining_time": "1:16:57", "throughput": 5184.41, "total_tokens": 18773600} {"current_steps": 18650, "total_steps": 42420, "loss": 0.3012, "lr": 0.000687905877726416, "epoch": 8.793022159358793, "percentage": 43.97, "elapsed_time": "1:00:21", "remaining_time": "1:16:56", "throughput": 5184.4, "total_tokens": 18777824} {"current_steps": 18655, "total_steps": 42420, "loss": 0.3188, "lr": 0.0006877152218363803, "epoch": 8.795379537953796, "percentage": 43.98, "elapsed_time": "1:00:22", "remaining_time": "1:16:55", "throughput": 5184.5, "total_tokens": 18782976} {"current_steps": 18660, "total_steps": 42420, "loss": 0.3568, "lr": 0.0006875245341693448, "epoch": 8.797736916548798, "percentage": 43.99, "elapsed_time": "1:00:23", "remaining_time": "1:16:54", "throughput": 5184.62, "total_tokens": 18788640} {"current_steps": 18665, "total_steps": 42420, "loss": 0.3743, "lr": 0.0006873338147575897, "epoch": 8.8000942951438, "percentage": 44.0, "elapsed_time": "1:00:24", "remaining_time": "1:16:53", "throughput": 5184.79, "total_tokens": 18794624} {"current_steps": 18670, "total_steps": 42420, "loss": 0.3366, "lr": 0.0006871430636334005, "epoch": 8.802451673738803, "percentage": 44.01, "elapsed_time": "1:00:25", "remaining_time": "1:16:52", "throughput": 5184.9, "total_tokens": 18799872} {"current_steps": 18675, "total_steps": 42420, "loss": 0.3686, "lr": 0.0006869522808290681, "epoch": 8.804809052333805, "percentage": 44.02, "elapsed_time": "1:00:26", "remaining_time": "1:16:51", "throughput": 5185.02, "total_tokens": 18805312} {"current_steps": 18680, "total_steps": 42420, "loss": 0.3404, "lr": 0.0006867614663768889, "epoch": 8.807166430928808, "percentage": 44.04, "elapsed_time": "1:00:27", "remaining_time": "1:16:50", "throughput": 5185.03, "total_tokens": 18809984} {"current_steps": 18685, "total_steps": 42420, "loss": 0.33, "lr": 0.0006865706203091644, "epoch": 8.80952380952381, "percentage": 44.05, "elapsed_time": "1:00:28", "remaining_time": "1:16:49", "throughput": 5185.18, "total_tokens": 18815904} {"current_steps": 18690, "total_steps": 42420, "loss": 0.3523, "lr": 0.0006863797426582017, "epoch": 8.811881188118813, "percentage": 44.06, "elapsed_time": "1:00:29", "remaining_time": "1:16:48", "throughput": 5185.18, "total_tokens": 18820288} {"current_steps": 18695, "total_steps": 42420, "loss": 0.3512, "lr": 0.0006861888334563131, "epoch": 8.814238566713815, "percentage": 44.07, "elapsed_time": "1:00:30", "remaining_time": "1:16:47", "throughput": 5185.25, "total_tokens": 18825120} {"current_steps": 18700, "total_steps": 42420, "loss": 0.344, "lr": 0.0006859978927358164, "epoch": 8.816595945308817, "percentage": 44.08, "elapsed_time": "1:00:31", "remaining_time": "1:16:46", "throughput": 5185.27, "total_tokens": 18829664} {"current_steps": 18705, "total_steps": 42420, "loss": 0.2957, "lr": 0.0006858069205290343, "epoch": 8.81895332390382, "percentage": 44.09, "elapsed_time": "1:00:32", "remaining_time": "1:16:45", "throughput": 5185.28, "total_tokens": 18834144} {"current_steps": 18710, "total_steps": 42420, "loss": 0.3686, "lr": 0.0006856159168682955, "epoch": 8.82131070249882, "percentage": 44.11, "elapsed_time": "1:00:33", "remaining_time": "1:16:44", "throughput": 5185.3, "total_tokens": 18838784} {"current_steps": 18715, "total_steps": 42420, "loss": 0.3811, "lr": 0.0006854248817859337, "epoch": 8.823668081093825, "percentage": 44.12, "elapsed_time": "1:00:34", "remaining_time": "1:16:43", "throughput": 5185.45, "total_tokens": 18844352} {"current_steps": 18720, "total_steps": 42420, "loss": 0.3539, "lr": 0.0006852338153142874, "epoch": 8.826025459688825, "percentage": 44.13, "elapsed_time": "1:00:35", "remaining_time": "1:16:42", "throughput": 5185.56, "total_tokens": 18849760} {"current_steps": 18725, "total_steps": 42420, "loss": 0.3803, "lr": 0.0006850427174857015, "epoch": 8.828382838283828, "percentage": 44.14, "elapsed_time": "1:00:35", "remaining_time": "1:16:40", "throughput": 5185.57, "total_tokens": 18854272} {"current_steps": 18730, "total_steps": 42420, "loss": 0.341, "lr": 0.0006848515883325253, "epoch": 8.83074021687883, "percentage": 44.15, "elapsed_time": "1:00:36", "remaining_time": "1:16:39", "throughput": 5185.66, "total_tokens": 18859392} {"current_steps": 18735, "total_steps": 42420, "loss": 0.3117, "lr": 0.0006846604278871139, "epoch": 8.833097595473832, "percentage": 44.17, "elapsed_time": "1:00:37", "remaining_time": "1:16:38", "throughput": 5185.74, "total_tokens": 18864256} {"current_steps": 18740, "total_steps": 42420, "loss": 0.2919, "lr": 0.0006844692361818273, "epoch": 8.835454974068835, "percentage": 44.18, "elapsed_time": "1:00:38", "remaining_time": "1:16:37", "throughput": 5185.8, "total_tokens": 18869216} {"current_steps": 18745, "total_steps": 42420, "loss": 0.258, "lr": 0.0006842780132490312, "epoch": 8.837812352663837, "percentage": 44.19, "elapsed_time": "1:00:39", "remaining_time": "1:16:36", "throughput": 5185.87, "total_tokens": 18874272} {"current_steps": 18750, "total_steps": 42420, "loss": 0.3962, "lr": 0.0006840867591210962, "epoch": 8.84016973125884, "percentage": 44.2, "elapsed_time": "1:00:40", "remaining_time": "1:16:35", "throughput": 5185.84, "total_tokens": 18878272} {"current_steps": 18755, "total_steps": 42420, "loss": 0.3154, "lr": 0.0006838954738303986, "epoch": 8.842527109853842, "percentage": 44.21, "elapsed_time": "1:00:41", "remaining_time": "1:16:34", "throughput": 5185.89, "total_tokens": 18882880} {"current_steps": 18760, "total_steps": 42420, "loss": 0.3677, "lr": 0.0006837041574093197, "epoch": 8.844884488448844, "percentage": 44.22, "elapsed_time": "1:00:42", "remaining_time": "1:16:33", "throughput": 5186.01, "total_tokens": 18888512} {"current_steps": 18765, "total_steps": 42420, "loss": 0.3224, "lr": 0.000683512809890246, "epoch": 8.847241867043847, "percentage": 44.24, "elapsed_time": "1:00:43", "remaining_time": "1:16:32", "throughput": 5186.05, "total_tokens": 18893312} {"current_steps": 18770, "total_steps": 42420, "loss": 0.3395, "lr": 0.0006833214313055694, "epoch": 8.84959924563885, "percentage": 44.25, "elapsed_time": "1:00:44", "remaining_time": "1:16:31", "throughput": 5186.1, "total_tokens": 18898272} {"current_steps": 18775, "total_steps": 42420, "loss": 0.3557, "lr": 0.0006831300216876872, "epoch": 8.851956624233852, "percentage": 44.26, "elapsed_time": "1:00:44", "remaining_time": "1:16:30", "throughput": 5186.11, "total_tokens": 18902592} {"current_steps": 18780, "total_steps": 42420, "loss": 0.3456, "lr": 0.0006829385810690016, "epoch": 8.854314002828854, "percentage": 44.27, "elapsed_time": "1:00:45", "remaining_time": "1:16:29", "throughput": 5186.19, "total_tokens": 18907520} {"current_steps": 18785, "total_steps": 42420, "loss": 0.3189, "lr": 0.0006827471094819204, "epoch": 8.856671381423856, "percentage": 44.28, "elapsed_time": "1:00:46", "remaining_time": "1:16:28", "throughput": 5186.29, "total_tokens": 18912736} {"current_steps": 18790, "total_steps": 42420, "loss": 0.2825, "lr": 0.0006825556069588564, "epoch": 8.859028760018859, "percentage": 44.3, "elapsed_time": "1:00:47", "remaining_time": "1:16:27", "throughput": 5186.4, "total_tokens": 18918272} {"current_steps": 18795, "total_steps": 42420, "loss": 0.37, "lr": 0.0006823640735322279, "epoch": 8.861386138613861, "percentage": 44.31, "elapsed_time": "1:00:48", "remaining_time": "1:16:26", "throughput": 5186.47, "total_tokens": 18923264} {"current_steps": 18800, "total_steps": 42420, "loss": 0.3066, "lr": 0.0006821725092344582, "epoch": 8.863743517208864, "percentage": 44.32, "elapsed_time": "1:00:49", "remaining_time": "1:16:25", "throughput": 5186.52, "total_tokens": 18928288} {"current_steps": 18805, "total_steps": 42420, "loss": 0.3165, "lr": 0.0006819809140979755, "epoch": 8.866100895803866, "percentage": 44.33, "elapsed_time": "1:00:50", "remaining_time": "1:16:24", "throughput": 5186.54, "total_tokens": 18932800} {"current_steps": 18810, "total_steps": 42420, "loss": 0.4513, "lr": 0.000681789288155214, "epoch": 8.868458274398868, "percentage": 44.34, "elapsed_time": "1:00:51", "remaining_time": "1:16:23", "throughput": 5186.58, "total_tokens": 18937568} {"current_steps": 18815, "total_steps": 42420, "loss": 0.3444, "lr": 0.0006815976314386126, "epoch": 8.87081565299387, "percentage": 44.35, "elapsed_time": "1:00:52", "remaining_time": "1:16:22", "throughput": 5186.8, "total_tokens": 18944128} {"current_steps": 18820, "total_steps": 42420, "loss": 0.3605, "lr": 0.0006814059439806157, "epoch": 8.873173031588873, "percentage": 44.37, "elapsed_time": "1:00:53", "remaining_time": "1:16:21", "throughput": 5186.78, "total_tokens": 18948288} {"current_steps": 18825, "total_steps": 42420, "loss": 0.4349, "lr": 0.0006812142258136724, "epoch": 8.875530410183876, "percentage": 44.38, "elapsed_time": "1:00:54", "remaining_time": "1:16:20", "throughput": 5186.89, "total_tokens": 18953536} {"current_steps": 18830, "total_steps": 42420, "loss": 0.3133, "lr": 0.0006810224769702376, "epoch": 8.877887788778878, "percentage": 44.39, "elapsed_time": "1:00:55", "remaining_time": "1:16:19", "throughput": 5187.0, "total_tokens": 18958880} {"current_steps": 18835, "total_steps": 42420, "loss": 0.3467, "lr": 0.000680830697482771, "epoch": 8.88024516737388, "percentage": 44.4, "elapsed_time": "1:00:56", "remaining_time": "1:16:18", "throughput": 5187.16, "total_tokens": 18964704} {"current_steps": 18840, "total_steps": 42420, "loss": 0.3459, "lr": 0.0006806388873837377, "epoch": 8.882602545968883, "percentage": 44.41, "elapsed_time": "1:00:57", "remaining_time": "1:16:17", "throughput": 5187.33, "total_tokens": 18970816} {"current_steps": 18845, "total_steps": 42420, "loss": 0.3998, "lr": 0.0006804470467056077, "epoch": 8.884959924563885, "percentage": 44.42, "elapsed_time": "1:00:58", "remaining_time": "1:16:16", "throughput": 5187.39, "total_tokens": 18975840} {"current_steps": 18850, "total_steps": 42420, "loss": 0.261, "lr": 0.0006802551754808567, "epoch": 8.887317303158888, "percentage": 44.44, "elapsed_time": "1:00:58", "remaining_time": "1:16:15", "throughput": 5187.43, "total_tokens": 18980672} {"current_steps": 18855, "total_steps": 42420, "loss": 0.4051, "lr": 0.0006800632737419648, "epoch": 8.88967468175389, "percentage": 44.45, "elapsed_time": "1:00:59", "remaining_time": "1:16:14", "throughput": 5187.45, "total_tokens": 18985152} {"current_steps": 18860, "total_steps": 42420, "loss": 0.3312, "lr": 0.0006798713415214183, "epoch": 8.892032060348892, "percentage": 44.46, "elapsed_time": "1:01:00", "remaining_time": "1:16:13", "throughput": 5187.51, "total_tokens": 18990464} {"current_steps": 18865, "total_steps": 42420, "loss": 0.3075, "lr": 0.0006796793788517075, "epoch": 8.894389438943895, "percentage": 44.47, "elapsed_time": "1:01:01", "remaining_time": "1:16:12", "throughput": 5187.59, "total_tokens": 18995392} {"current_steps": 18870, "total_steps": 42420, "loss": 0.3494, "lr": 0.0006794873857653286, "epoch": 8.896746817538897, "percentage": 44.48, "elapsed_time": "1:01:02", "remaining_time": "1:16:11", "throughput": 5187.82, "total_tokens": 19002144} {"current_steps": 18875, "total_steps": 42420, "loss": 0.331, "lr": 0.0006792953622947829, "epoch": 8.8991041961339, "percentage": 44.5, "elapsed_time": "1:01:03", "remaining_time": "1:16:10", "throughput": 5187.79, "total_tokens": 19006336} {"current_steps": 18880, "total_steps": 42420, "loss": 0.3561, "lr": 0.0006791033084725767, "epoch": 8.901461574728902, "percentage": 44.51, "elapsed_time": "1:01:04", "remaining_time": "1:16:09", "throughput": 5187.81, "total_tokens": 19011072} {"current_steps": 18885, "total_steps": 42420, "loss": 0.3393, "lr": 0.0006789112243312213, "epoch": 8.903818953323904, "percentage": 44.52, "elapsed_time": "1:01:05", "remaining_time": "1:16:08", "throughput": 5187.97, "total_tokens": 19016736} {"current_steps": 18890, "total_steps": 42420, "loss": 0.2966, "lr": 0.0006787191099032333, "epoch": 8.906176331918907, "percentage": 44.53, "elapsed_time": "1:01:06", "remaining_time": "1:16:06", "throughput": 5187.94, "total_tokens": 19021088} {"current_steps": 18895, "total_steps": 42420, "loss": 0.3359, "lr": 0.0006785269652211345, "epoch": 8.90853371051391, "percentage": 44.54, "elapsed_time": "1:01:07", "remaining_time": "1:16:05", "throughput": 5187.88, "total_tokens": 19024992} {"current_steps": 18900, "total_steps": 42420, "loss": 0.3207, "lr": 0.0006783347903174518, "epoch": 8.910891089108912, "percentage": 44.55, "elapsed_time": "1:01:08", "remaining_time": "1:16:04", "throughput": 5188.03, "total_tokens": 19030720} {"current_steps": 18905, "total_steps": 42420, "loss": 0.3416, "lr": 0.000678142585224717, "epoch": 8.913248467703912, "percentage": 44.57, "elapsed_time": "1:01:09", "remaining_time": "1:16:03", "throughput": 5188.18, "total_tokens": 19036416} {"current_steps": 18910, "total_steps": 42420, "loss": 0.3222, "lr": 0.0006779503499754672, "epoch": 8.915605846298917, "percentage": 44.58, "elapsed_time": "1:01:09", "remaining_time": "1:16:02", "throughput": 5188.09, "total_tokens": 19040064} {"current_steps": 18915, "total_steps": 42420, "loss": 0.3752, "lr": 0.0006777580846022447, "epoch": 8.917963224893917, "percentage": 44.59, "elapsed_time": "1:01:10", "remaining_time": "1:16:01", "throughput": 5188.18, "total_tokens": 19045184} {"current_steps": 18920, "total_steps": 42420, "loss": 0.3153, "lr": 0.0006775657891375964, "epoch": 8.92032060348892, "percentage": 44.6, "elapsed_time": "1:01:11", "remaining_time": "1:16:00", "throughput": 5188.22, "total_tokens": 19049920} {"current_steps": 18925, "total_steps": 42420, "loss": 0.3135, "lr": 0.0006773734636140751, "epoch": 8.922677982083922, "percentage": 44.61, "elapsed_time": "1:01:12", "remaining_time": "1:15:59", "throughput": 5188.21, "total_tokens": 19054112} {"current_steps": 18930, "total_steps": 42420, "loss": 0.3177, "lr": 0.0006771811080642378, "epoch": 8.925035360678924, "percentage": 44.63, "elapsed_time": "1:01:13", "remaining_time": "1:15:58", "throughput": 5188.35, "total_tokens": 19059808} {"current_steps": 18935, "total_steps": 42420, "loss": 0.2573, "lr": 0.0006769887225206475, "epoch": 8.927392739273927, "percentage": 44.64, "elapsed_time": "1:01:14", "remaining_time": "1:15:57", "throughput": 5188.34, "total_tokens": 19063968} {"current_steps": 18940, "total_steps": 42420, "loss": 0.399, "lr": 0.0006767963070158714, "epoch": 8.92975011786893, "percentage": 44.65, "elapsed_time": "1:01:15", "remaining_time": "1:15:56", "throughput": 5188.48, "total_tokens": 19069600} {"current_steps": 18945, "total_steps": 42420, "loss": 0.3976, "lr": 0.0006766038615824826, "epoch": 8.932107496463932, "percentage": 44.66, "elapsed_time": "1:01:16", "remaining_time": "1:15:55", "throughput": 5188.5, "total_tokens": 19074272} {"current_steps": 18950, "total_steps": 42420, "loss": 0.3191, "lr": 0.0006764113862530583, "epoch": 8.934464875058934, "percentage": 44.67, "elapsed_time": "1:01:17", "remaining_time": "1:15:54", "throughput": 5188.58, "total_tokens": 19079744} {"current_steps": 18955, "total_steps": 42420, "loss": 0.3688, "lr": 0.0006762188810601816, "epoch": 8.936822253653936, "percentage": 44.68, "elapsed_time": "1:01:18", "remaining_time": "1:15:53", "throughput": 5188.71, "total_tokens": 19085504} {"current_steps": 18960, "total_steps": 42420, "loss": 0.36, "lr": 0.0006760263460364406, "epoch": 8.939179632248939, "percentage": 44.7, "elapsed_time": "1:01:19", "remaining_time": "1:15:52", "throughput": 5188.73, "total_tokens": 19090272} {"current_steps": 18965, "total_steps": 42420, "loss": 0.3727, "lr": 0.0006758337812144279, "epoch": 8.941537010843941, "percentage": 44.71, "elapsed_time": "1:01:20", "remaining_time": "1:15:51", "throughput": 5188.84, "total_tokens": 19095520} {"current_steps": 18970, "total_steps": 42420, "loss": 0.3299, "lr": 0.0006756411866267416, "epoch": 8.943894389438944, "percentage": 44.72, "elapsed_time": "1:01:20", "remaining_time": "1:15:50", "throughput": 5188.86, "total_tokens": 19099936} {"current_steps": 18975, "total_steps": 42420, "loss": 0.346, "lr": 0.0006754485623059844, "epoch": 8.946251768033946, "percentage": 44.73, "elapsed_time": "1:01:21", "remaining_time": "1:15:49", "throughput": 5188.91, "total_tokens": 19104672} {"current_steps": 18980, "total_steps": 42420, "loss": 0.353, "lr": 0.0006752559082847647, "epoch": 8.948609146628948, "percentage": 44.74, "elapsed_time": "1:01:22", "remaining_time": "1:15:48", "throughput": 5188.96, "total_tokens": 19109792} {"current_steps": 18985, "total_steps": 42420, "loss": 0.2522, "lr": 0.0006750632245956954, "epoch": 8.95096652522395, "percentage": 44.75, "elapsed_time": "1:01:23", "remaining_time": "1:15:47", "throughput": 5189.05, "total_tokens": 19115264} {"current_steps": 18990, "total_steps": 42420, "loss": 0.4122, "lr": 0.0006748705112713945, "epoch": 8.953323903818953, "percentage": 44.77, "elapsed_time": "1:01:24", "remaining_time": "1:15:46", "throughput": 5189.04, "total_tokens": 19119712} {"current_steps": 18995, "total_steps": 42420, "loss": 0.3911, "lr": 0.0006746777683444852, "epoch": 8.955681282413956, "percentage": 44.78, "elapsed_time": "1:01:25", "remaining_time": "1:15:45", "throughput": 5189.23, "total_tokens": 19126144} {"current_steps": 19000, "total_steps": 42420, "loss": 0.3084, "lr": 0.0006744849958475956, "epoch": 8.958038661008958, "percentage": 44.79, "elapsed_time": "1:01:26", "remaining_time": "1:15:44", "throughput": 5189.21, "total_tokens": 19130368} {"current_steps": 19005, "total_steps": 42420, "loss": 0.3334, "lr": 0.0006742921938133589, "epoch": 8.96039603960396, "percentage": 44.8, "elapsed_time": "1:01:27", "remaining_time": "1:15:43", "throughput": 5189.22, "total_tokens": 19134752} {"current_steps": 19010, "total_steps": 42420, "loss": 0.3279, "lr": 0.0006740993622744129, "epoch": 8.962753418198963, "percentage": 44.81, "elapsed_time": "1:01:28", "remaining_time": "1:15:42", "throughput": 5189.34, "total_tokens": 19140128} {"current_steps": 19015, "total_steps": 42420, "loss": 0.3193, "lr": 0.0006739065012634011, "epoch": 8.965110796793965, "percentage": 44.83, "elapsed_time": "1:01:29", "remaining_time": "1:15:41", "throughput": 5189.42, "total_tokens": 19145408} {"current_steps": 19020, "total_steps": 42420, "loss": 0.3683, "lr": 0.0006737136108129712, "epoch": 8.967468175388968, "percentage": 44.84, "elapsed_time": "1:01:30", "remaining_time": "1:15:40", "throughput": 5189.51, "total_tokens": 19150624} {"current_steps": 19025, "total_steps": 42420, "loss": 0.334, "lr": 0.0006735206909557767, "epoch": 8.96982555398397, "percentage": 44.85, "elapsed_time": "1:01:31", "remaining_time": "1:15:38", "throughput": 5189.53, "total_tokens": 19155040} {"current_steps": 19030, "total_steps": 42420, "loss": 0.3473, "lr": 0.0006733277417244754, "epoch": 8.972182932578972, "percentage": 44.86, "elapsed_time": "1:01:31", "remaining_time": "1:15:37", "throughput": 5189.53, "total_tokens": 19159424} {"current_steps": 19035, "total_steps": 42420, "loss": 0.3657, "lr": 0.0006731347631517302, "epoch": 8.974540311173975, "percentage": 44.87, "elapsed_time": "1:01:33", "remaining_time": "1:15:37", "throughput": 5189.79, "total_tokens": 19166496} {"current_steps": 19040, "total_steps": 42420, "loss": 0.3274, "lr": 0.0006729417552702093, "epoch": 8.976897689768977, "percentage": 44.88, "elapsed_time": "1:01:34", "remaining_time": "1:15:36", "throughput": 5189.86, "total_tokens": 19171424} {"current_steps": 19045, "total_steps": 42420, "loss": 0.2922, "lr": 0.0006727487181125856, "epoch": 8.97925506836398, "percentage": 44.9, "elapsed_time": "1:01:34", "remaining_time": "1:15:34", "throughput": 5189.88, "total_tokens": 19175936} {"current_steps": 19050, "total_steps": 42420, "loss": 0.3836, "lr": 0.0006725556517115372, "epoch": 8.981612446958982, "percentage": 44.91, "elapsed_time": "1:01:35", "remaining_time": "1:15:34", "throughput": 5189.98, "total_tokens": 19181536} {"current_steps": 19055, "total_steps": 42420, "loss": 0.2328, "lr": 0.0006723625560997466, "epoch": 8.983969825553984, "percentage": 44.92, "elapsed_time": "1:01:36", "remaining_time": "1:15:32", "throughput": 5189.96, "total_tokens": 19185792} {"current_steps": 19060, "total_steps": 42420, "loss": 0.4366, "lr": 0.000672169431309902, "epoch": 8.986327204148987, "percentage": 44.93, "elapsed_time": "1:01:37", "remaining_time": "1:15:31", "throughput": 5190.12, "total_tokens": 19191584} {"current_steps": 19065, "total_steps": 42420, "loss": 0.3602, "lr": 0.0006719762773746957, "epoch": 8.98868458274399, "percentage": 44.94, "elapsed_time": "1:01:38", "remaining_time": "1:15:30", "throughput": 5190.15, "total_tokens": 19196000} {"current_steps": 19070, "total_steps": 42420, "loss": 0.3265, "lr": 0.0006717830943268259, "epoch": 8.991041961338992, "percentage": 44.96, "elapsed_time": "1:01:39", "remaining_time": "1:15:29", "throughput": 5190.27, "total_tokens": 19201728} {"current_steps": 19075, "total_steps": 42420, "loss": 0.3414, "lr": 0.0006715898821989947, "epoch": 8.993399339933994, "percentage": 44.97, "elapsed_time": "1:01:40", "remaining_time": "1:15:28", "throughput": 5190.38, "total_tokens": 19206624} {"current_steps": 19080, "total_steps": 42420, "loss": 0.3142, "lr": 0.0006713966410239099, "epoch": 8.995756718528996, "percentage": 44.98, "elapsed_time": "1:01:41", "remaining_time": "1:15:27", "throughput": 5190.5, "total_tokens": 19212288} {"current_steps": 19085, "total_steps": 42420, "loss": 0.2817, "lr": 0.0006712033708342841, "epoch": 8.998114097123999, "percentage": 44.99, "elapsed_time": "1:01:42", "remaining_time": "1:15:26", "throughput": 5190.57, "total_tokens": 19217472} {"current_steps": 19089, "total_steps": 42420, "eval_loss": 0.35525232553482056, "epoch": 9.0, "percentage": 45.0, "elapsed_time": "1:02:02", "remaining_time": "1:15:49", "throughput": 5164.07, "total_tokens": 19221568} {"current_steps": 19090, "total_steps": 42420, "loss": 0.3851, "lr": 0.0006710100716628344, "epoch": 9.000471475719001, "percentage": 45.0, "elapsed_time": "1:02:04", "remaining_time": "1:15:51", "throughput": 5161.13, "total_tokens": 19222272} {"current_steps": 19095, "total_steps": 42420, "loss": 0.3077, "lr": 0.0006708167435422832, "epoch": 9.002828854314004, "percentage": 45.01, "elapsed_time": "1:02:05", "remaining_time": "1:15:50", "throughput": 5161.2, "total_tokens": 19227136} {"current_steps": 19100, "total_steps": 42420, "loss": 0.3342, "lr": 0.0006706233865053575, "epoch": 9.005186232909006, "percentage": 45.03, "elapsed_time": "1:02:06", "remaining_time": "1:15:49", "throughput": 5161.22, "total_tokens": 19231584} {"current_steps": 19105, "total_steps": 42420, "loss": 0.3653, "lr": 0.0006704300005847894, "epoch": 9.007543611504008, "percentage": 45.04, "elapsed_time": "1:02:07", "remaining_time": "1:15:48", "throughput": 5161.37, "total_tokens": 19237632} {"current_steps": 19110, "total_steps": 42420, "loss": 0.3639, "lr": 0.0006702365858133163, "epoch": 9.009900990099009, "percentage": 45.05, "elapsed_time": "1:02:08", "remaining_time": "1:15:47", "throughput": 5161.35, "total_tokens": 19241760} {"current_steps": 19115, "total_steps": 42420, "loss": 0.3563, "lr": 0.0006700431422236794, "epoch": 9.012258368694011, "percentage": 45.06, "elapsed_time": "1:02:09", "remaining_time": "1:15:46", "throughput": 5161.49, "total_tokens": 19247424} {"current_steps": 19120, "total_steps": 42420, "loss": 0.3747, "lr": 0.0006698496698486257, "epoch": 9.014615747289014, "percentage": 45.07, "elapsed_time": "1:02:09", "remaining_time": "1:15:45", "throughput": 5161.5, "total_tokens": 19251872} {"current_steps": 19125, "total_steps": 42420, "loss": 0.3632, "lr": 0.0006696561687209068, "epoch": 9.016973125884016, "percentage": 45.08, "elapsed_time": "1:02:10", "remaining_time": "1:15:44", "throughput": 5161.53, "total_tokens": 19256608} {"current_steps": 19130, "total_steps": 42420, "loss": 0.3494, "lr": 0.0006694626388732792, "epoch": 9.019330504479019, "percentage": 45.1, "elapsed_time": "1:02:11", "remaining_time": "1:15:43", "throughput": 5161.61, "total_tokens": 19261856} {"current_steps": 19135, "total_steps": 42420, "loss": 0.348, "lr": 0.0006692690803385042, "epoch": 9.021687883074021, "percentage": 45.11, "elapsed_time": "1:02:12", "remaining_time": "1:15:42", "throughput": 5161.61, "total_tokens": 19266080} {"current_steps": 19140, "total_steps": 42420, "loss": 0.313, "lr": 0.000669075493149348, "epoch": 9.024045261669023, "percentage": 45.12, "elapsed_time": "1:02:13", "remaining_time": "1:15:40", "throughput": 5161.63, "total_tokens": 19270464} {"current_steps": 19145, "total_steps": 42420, "loss": 0.3046, "lr": 0.0006688818773385812, "epoch": 9.026402640264026, "percentage": 45.13, "elapsed_time": "1:02:14", "remaining_time": "1:15:39", "throughput": 5161.62, "total_tokens": 19274880} {"current_steps": 19150, "total_steps": 42420, "loss": 0.3432, "lr": 0.0006686882329389801, "epoch": 9.028760018859028, "percentage": 45.14, "elapsed_time": "1:02:15", "remaining_time": "1:15:38", "throughput": 5161.64, "total_tokens": 19279424} {"current_steps": 19155, "total_steps": 42420, "loss": 0.372, "lr": 0.0006684945599833255, "epoch": 9.03111739745403, "percentage": 45.16, "elapsed_time": "1:02:16", "remaining_time": "1:15:37", "throughput": 5161.66, "total_tokens": 19284128} {"current_steps": 19160, "total_steps": 42420, "loss": 0.3732, "lr": 0.0006683008585044027, "epoch": 9.033474776049033, "percentage": 45.17, "elapsed_time": "1:02:16", "remaining_time": "1:15:36", "throughput": 5161.69, "total_tokens": 19288960} {"current_steps": 19165, "total_steps": 42420, "loss": 0.351, "lr": 0.0006681071285350021, "epoch": 9.035832154644035, "percentage": 45.18, "elapsed_time": "1:02:18", "remaining_time": "1:15:35", "throughput": 5161.95, "total_tokens": 19295904} {"current_steps": 19170, "total_steps": 42420, "loss": 0.3488, "lr": 0.0006679133701079189, "epoch": 9.038189533239038, "percentage": 45.19, "elapsed_time": "1:02:19", "remaining_time": "1:15:34", "throughput": 5162.04, "total_tokens": 19301120} {"current_steps": 19175, "total_steps": 42420, "loss": 0.3629, "lr": 0.0006677195832559534, "epoch": 9.04054691183404, "percentage": 45.2, "elapsed_time": "1:02:19", "remaining_time": "1:15:33", "throughput": 5162.09, "total_tokens": 19305856} {"current_steps": 19180, "total_steps": 42420, "loss": 0.349, "lr": 0.00066752576801191, "epoch": 9.042904290429043, "percentage": 45.21, "elapsed_time": "1:02:20", "remaining_time": "1:15:32", "throughput": 5162.13, "total_tokens": 19310592} {"current_steps": 19185, "total_steps": 42420, "loss": 0.3603, "lr": 0.0006673319244085985, "epoch": 9.045261669024045, "percentage": 45.23, "elapsed_time": "1:02:21", "remaining_time": "1:15:31", "throughput": 5162.14, "total_tokens": 19315040} {"current_steps": 19190, "total_steps": 42420, "loss": 0.3409, "lr": 0.0006671380524788332, "epoch": 9.047619047619047, "percentage": 45.24, "elapsed_time": "1:02:22", "remaining_time": "1:15:30", "throughput": 5162.13, "total_tokens": 19319040} {"current_steps": 19195, "total_steps": 42420, "loss": 0.3447, "lr": 0.0006669441522554337, "epoch": 9.04997642621405, "percentage": 45.25, "elapsed_time": "1:02:23", "remaining_time": "1:15:29", "throughput": 5162.22, "total_tokens": 19324576} {"current_steps": 19200, "total_steps": 42420, "loss": 0.3309, "lr": 0.0006667502237712236, "epoch": 9.052333804809052, "percentage": 45.26, "elapsed_time": "1:02:24", "remaining_time": "1:15:28", "throughput": 5162.24, "total_tokens": 19328992} {"current_steps": 19205, "total_steps": 42420, "loss": 0.3085, "lr": 0.0006665562670590319, "epoch": 9.054691183404055, "percentage": 45.27, "elapsed_time": "1:02:25", "remaining_time": "1:15:27", "throughput": 5162.27, "total_tokens": 19333696} {"current_steps": 19210, "total_steps": 42420, "loss": 0.2905, "lr": 0.0006663622821516922, "epoch": 9.057048561999057, "percentage": 45.29, "elapsed_time": "1:02:26", "remaining_time": "1:15:26", "throughput": 5162.43, "total_tokens": 19339488} {"current_steps": 19215, "total_steps": 42420, "loss": 0.3472, "lr": 0.0006661682690820427, "epoch": 9.05940594059406, "percentage": 45.3, "elapsed_time": "1:02:27", "remaining_time": "1:15:25", "throughput": 5162.45, "total_tokens": 19344160} {"current_steps": 19220, "total_steps": 42420, "loss": 0.3141, "lr": 0.0006659742278829268, "epoch": 9.061763319189062, "percentage": 45.31, "elapsed_time": "1:02:27", "remaining_time": "1:15:24", "throughput": 5162.44, "total_tokens": 19348576} {"current_steps": 19225, "total_steps": 42420, "loss": 0.4185, "lr": 0.0006657801585871921, "epoch": 9.064120697784064, "percentage": 45.32, "elapsed_time": "1:02:28", "remaining_time": "1:15:22", "throughput": 5162.41, "total_tokens": 19352704} {"current_steps": 19230, "total_steps": 42420, "loss": 0.3206, "lr": 0.0006655860612276912, "epoch": 9.066478076379067, "percentage": 45.33, "elapsed_time": "1:02:29", "remaining_time": "1:15:21", "throughput": 5162.54, "total_tokens": 19358272} {"current_steps": 19235, "total_steps": 42420, "loss": 0.3374, "lr": 0.000665391935837282, "epoch": 9.068835454974069, "percentage": 45.34, "elapsed_time": "1:02:30", "remaining_time": "1:15:20", "throughput": 5162.56, "total_tokens": 19362944} {"current_steps": 19240, "total_steps": 42420, "loss": 0.3179, "lr": 0.000665197782448826, "epoch": 9.071192833569071, "percentage": 45.36, "elapsed_time": "1:02:31", "remaining_time": "1:15:19", "throughput": 5162.71, "total_tokens": 19368800} {"current_steps": 19245, "total_steps": 42420, "loss": 0.3832, "lr": 0.0006650036010951903, "epoch": 9.073550212164074, "percentage": 45.37, "elapsed_time": "1:02:32", "remaining_time": "1:15:18", "throughput": 5162.78, "total_tokens": 19373696} {"current_steps": 19250, "total_steps": 42420, "loss": 0.3483, "lr": 0.0006648093918092467, "epoch": 9.075907590759076, "percentage": 45.38, "elapsed_time": "1:02:33", "remaining_time": "1:15:17", "throughput": 5162.8, "total_tokens": 19378016} {"current_steps": 19255, "total_steps": 42420, "loss": 0.3566, "lr": 0.0006646151546238712, "epoch": 9.078264969354079, "percentage": 45.39, "elapsed_time": "1:02:34", "remaining_time": "1:15:16", "throughput": 5162.89, "total_tokens": 19382944} {"current_steps": 19260, "total_steps": 42420, "loss": 0.3883, "lr": 0.0006644208895719452, "epoch": 9.080622347949081, "percentage": 45.4, "elapsed_time": "1:02:35", "remaining_time": "1:15:15", "throughput": 5162.92, "total_tokens": 19387744} {"current_steps": 19265, "total_steps": 42420, "loss": 0.3474, "lr": 0.0006642265966863542, "epoch": 9.082979726544083, "percentage": 45.41, "elapsed_time": "1:02:36", "remaining_time": "1:15:14", "throughput": 5162.96, "total_tokens": 19392544} {"current_steps": 19270, "total_steps": 42420, "loss": 0.345, "lr": 0.0006640322759999887, "epoch": 9.085337105139086, "percentage": 45.43, "elapsed_time": "1:02:36", "remaining_time": "1:15:13", "throughput": 5162.93, "total_tokens": 19396448} {"current_steps": 19275, "total_steps": 42420, "loss": 0.3828, "lr": 0.000663837927545744, "epoch": 9.087694483734088, "percentage": 45.44, "elapsed_time": "1:02:37", "remaining_time": "1:15:12", "throughput": 5163.04, "total_tokens": 19401600} {"current_steps": 19280, "total_steps": 42420, "loss": 0.2997, "lr": 0.0006636435513565201, "epoch": 9.09005186232909, "percentage": 45.45, "elapsed_time": "1:02:38", "remaining_time": "1:15:11", "throughput": 5163.0, "total_tokens": 19405536} {"current_steps": 19285, "total_steps": 42420, "loss": 0.3051, "lr": 0.0006634491474652214, "epoch": 9.092409240924093, "percentage": 45.46, "elapsed_time": "1:02:39", "remaining_time": "1:15:10", "throughput": 5163.11, "total_tokens": 19410944} {"current_steps": 19290, "total_steps": 42420, "loss": 0.2906, "lr": 0.0006632547159047573, "epoch": 9.094766619519095, "percentage": 45.47, "elapsed_time": "1:02:40", "remaining_time": "1:15:09", "throughput": 5163.16, "total_tokens": 19415904} {"current_steps": 19295, "total_steps": 42420, "loss": 0.3735, "lr": 0.0006630602567080416, "epoch": 9.097123998114098, "percentage": 45.49, "elapsed_time": "1:02:41", "remaining_time": "1:15:07", "throughput": 5163.14, "total_tokens": 19419904} {"current_steps": 19300, "total_steps": 42420, "loss": 0.334, "lr": 0.0006628657699079932, "epoch": 9.0994813767091, "percentage": 45.5, "elapsed_time": "1:02:42", "remaining_time": "1:15:06", "throughput": 5163.24, "total_tokens": 19425248} {"current_steps": 19305, "total_steps": 42420, "loss": 0.2928, "lr": 0.000662671255537535, "epoch": 9.101838755304103, "percentage": 45.51, "elapsed_time": "1:02:43", "remaining_time": "1:15:06", "throughput": 5163.46, "total_tokens": 19432256} {"current_steps": 19310, "total_steps": 42420, "loss": 0.3776, "lr": 0.0006624767136295955, "epoch": 9.104196133899103, "percentage": 45.52, "elapsed_time": "1:02:44", "remaining_time": "1:15:05", "throughput": 5163.63, "total_tokens": 19438080} {"current_steps": 19315, "total_steps": 42420, "loss": 0.3268, "lr": 0.0006622821442171068, "epoch": 9.106553512494106, "percentage": 45.53, "elapsed_time": "1:02:45", "remaining_time": "1:15:04", "throughput": 5163.6, "total_tokens": 19442080} {"current_steps": 19320, "total_steps": 42420, "loss": 0.3227, "lr": 0.0006620875473330068, "epoch": 9.108910891089108, "percentage": 45.54, "elapsed_time": "1:02:46", "remaining_time": "1:15:03", "throughput": 5163.74, "total_tokens": 19448064} {"current_steps": 19325, "total_steps": 42420, "loss": 0.3303, "lr": 0.000661892923010237, "epoch": 9.11126826968411, "percentage": 45.56, "elapsed_time": "1:02:47", "remaining_time": "1:15:02", "throughput": 5163.76, "total_tokens": 19452864} {"current_steps": 19330, "total_steps": 42420, "loss": 0.285, "lr": 0.0006616982712817441, "epoch": 9.113625648279113, "percentage": 45.57, "elapsed_time": "1:02:48", "remaining_time": "1:15:01", "throughput": 5163.87, "total_tokens": 19458144} {"current_steps": 19335, "total_steps": 42420, "loss": 0.338, "lr": 0.0006615035921804793, "epoch": 9.115983026874115, "percentage": 45.58, "elapsed_time": "1:02:49", "remaining_time": "1:15:00", "throughput": 5163.91, "total_tokens": 19462976} {"current_steps": 19340, "total_steps": 42420, "loss": 0.3397, "lr": 0.0006613088857393986, "epoch": 9.118340405469118, "percentage": 45.59, "elapsed_time": "1:02:49", "remaining_time": "1:14:58", "throughput": 5163.92, "total_tokens": 19467328} {"current_steps": 19345, "total_steps": 42420, "loss": 0.3754, "lr": 0.0006611141519914624, "epoch": 9.12069778406412, "percentage": 45.6, "elapsed_time": "1:02:50", "remaining_time": "1:14:57", "throughput": 5164.05, "total_tokens": 19472800} {"current_steps": 19350, "total_steps": 42420, "loss": 0.3912, "lr": 0.0006609193909696358, "epoch": 9.123055162659123, "percentage": 45.62, "elapsed_time": "1:02:51", "remaining_time": "1:14:56", "throughput": 5164.13, "total_tokens": 19478016} {"current_steps": 19355, "total_steps": 42420, "loss": 0.3754, "lr": 0.0006607246027068885, "epoch": 9.125412541254125, "percentage": 45.63, "elapsed_time": "1:02:52", "remaining_time": "1:14:55", "throughput": 5164.16, "total_tokens": 19482656} {"current_steps": 19360, "total_steps": 42420, "loss": 0.447, "lr": 0.0006605297872361953, "epoch": 9.127769919849127, "percentage": 45.64, "elapsed_time": "1:02:53", "remaining_time": "1:14:54", "throughput": 5164.23, "total_tokens": 19487648} {"current_steps": 19365, "total_steps": 42420, "loss": 0.3248, "lr": 0.0006603349445905343, "epoch": 9.13012729844413, "percentage": 45.65, "elapsed_time": "1:02:54", "remaining_time": "1:14:53", "throughput": 5164.32, "total_tokens": 19492896} {"current_steps": 19370, "total_steps": 42420, "loss": 0.3465, "lr": 0.0006601400748028896, "epoch": 9.132484677039132, "percentage": 45.66, "elapsed_time": "1:02:55", "remaining_time": "1:14:52", "throughput": 5164.43, "total_tokens": 19498400} {"current_steps": 19375, "total_steps": 42420, "loss": 0.2923, "lr": 0.0006599451779062494, "epoch": 9.134842055634135, "percentage": 45.67, "elapsed_time": "1:02:56", "remaining_time": "1:14:51", "throughput": 5164.45, "total_tokens": 19502784} {"current_steps": 19380, "total_steps": 42420, "loss": 0.3928, "lr": 0.0006597502539336062, "epoch": 9.137199434229137, "percentage": 45.69, "elapsed_time": "1:02:57", "remaining_time": "1:14:50", "throughput": 5164.49, "total_tokens": 19507424} {"current_steps": 19385, "total_steps": 42420, "loss": 0.3678, "lr": 0.0006595553029179575, "epoch": 9.13955681282414, "percentage": 45.7, "elapsed_time": "1:02:58", "remaining_time": "1:14:49", "throughput": 5164.54, "total_tokens": 19512480} {"current_steps": 19390, "total_steps": 42420, "loss": 0.34, "lr": 0.000659360324892305, "epoch": 9.141914191419142, "percentage": 45.71, "elapsed_time": "1:02:59", "remaining_time": "1:14:48", "throughput": 5164.55, "total_tokens": 19516928} {"current_steps": 19395, "total_steps": 42420, "loss": 0.3132, "lr": 0.0006591653198896553, "epoch": 9.144271570014144, "percentage": 45.72, "elapsed_time": "1:02:59", "remaining_time": "1:14:47", "throughput": 5164.61, "total_tokens": 19521824} {"current_steps": 19400, "total_steps": 42420, "loss": 0.3606, "lr": 0.0006589702879430195, "epoch": 9.146628948609147, "percentage": 45.73, "elapsed_time": "1:03:00", "remaining_time": "1:14:46", "throughput": 5164.73, "total_tokens": 19527008} {"current_steps": 19405, "total_steps": 42420, "loss": 0.3429, "lr": 0.000658775229085413, "epoch": 9.148986327204149, "percentage": 45.74, "elapsed_time": "1:03:01", "remaining_time": "1:14:45", "throughput": 5164.99, "total_tokens": 19533920} {"current_steps": 19410, "total_steps": 42420, "loss": 0.3391, "lr": 0.0006585801433498562, "epoch": 9.151343705799151, "percentage": 45.76, "elapsed_time": "1:03:02", "remaining_time": "1:14:44", "throughput": 5165.0, "total_tokens": 19538208} {"current_steps": 19415, "total_steps": 42420, "loss": 0.3458, "lr": 0.0006583850307693735, "epoch": 9.153701084394154, "percentage": 45.77, "elapsed_time": "1:03:03", "remaining_time": "1:14:43", "throughput": 5165.1, "total_tokens": 19543584} {"current_steps": 19420, "total_steps": 42420, "loss": 0.2989, "lr": 0.0006581898913769945, "epoch": 9.156058462989156, "percentage": 45.78, "elapsed_time": "1:03:04", "remaining_time": "1:14:42", "throughput": 5165.13, "total_tokens": 19548352} {"current_steps": 19425, "total_steps": 42420, "loss": 0.3764, "lr": 0.0006579947252057526, "epoch": 9.158415841584159, "percentage": 45.79, "elapsed_time": "1:03:05", "remaining_time": "1:14:41", "throughput": 5165.2, "total_tokens": 19553280} {"current_steps": 19430, "total_steps": 42420, "loss": 0.3217, "lr": 0.0006577995322886865, "epoch": 9.160773220179161, "percentage": 45.8, "elapsed_time": "1:03:06", "remaining_time": "1:14:40", "throughput": 5165.24, "total_tokens": 19557920} {"current_steps": 19435, "total_steps": 42420, "loss": 0.3374, "lr": 0.0006576043126588387, "epoch": 9.163130598774163, "percentage": 45.82, "elapsed_time": "1:03:07", "remaining_time": "1:14:39", "throughput": 5165.25, "total_tokens": 19562304} {"current_steps": 19440, "total_steps": 42420, "loss": 0.3438, "lr": 0.0006574090663492569, "epoch": 9.165487977369166, "percentage": 45.83, "elapsed_time": "1:03:08", "remaining_time": "1:14:38", "throughput": 5165.3, "total_tokens": 19567424} {"current_steps": 19445, "total_steps": 42420, "loss": 0.3378, "lr": 0.0006572137933929927, "epoch": 9.167845355964168, "percentage": 45.84, "elapsed_time": "1:03:09", "remaining_time": "1:14:36", "throughput": 5165.37, "total_tokens": 19572160} {"current_steps": 19450, "total_steps": 42420, "loss": 0.3093, "lr": 0.0006570184938231027, "epoch": 9.17020273455917, "percentage": 45.85, "elapsed_time": "1:03:10", "remaining_time": "1:14:35", "throughput": 5165.46, "total_tokens": 19577408} {"current_steps": 19455, "total_steps": 42420, "loss": 0.3598, "lr": 0.0006568231676726477, "epoch": 9.172560113154173, "percentage": 45.86, "elapsed_time": "1:03:11", "remaining_time": "1:14:35", "throughput": 5165.64, "total_tokens": 19583456} {"current_steps": 19460, "total_steps": 42420, "loss": 0.3643, "lr": 0.0006566278149746931, "epoch": 9.174917491749175, "percentage": 45.87, "elapsed_time": "1:03:12", "remaining_time": "1:14:34", "throughput": 5165.77, "total_tokens": 19589184} {"current_steps": 19465, "total_steps": 42420, "loss": 0.3236, "lr": 0.0006564324357623088, "epoch": 9.177274870344178, "percentage": 45.89, "elapsed_time": "1:03:12", "remaining_time": "1:14:33", "throughput": 5165.79, "total_tokens": 19593792} {"current_steps": 19470, "total_steps": 42420, "loss": 0.3829, "lr": 0.0006562370300685692, "epoch": 9.17963224893918, "percentage": 45.9, "elapsed_time": "1:03:14", "remaining_time": "1:14:32", "throughput": 5166.0, "total_tokens": 19600480} {"current_steps": 19475, "total_steps": 42420, "loss": 0.3551, "lr": 0.0006560415979265532, "epoch": 9.181989627534183, "percentage": 45.91, "elapsed_time": "1:03:15", "remaining_time": "1:14:31", "throughput": 5166.02, "total_tokens": 19605152} {"current_steps": 19480, "total_steps": 42420, "loss": 0.3446, "lr": 0.0006558461393693442, "epoch": 9.184347006129185, "percentage": 45.92, "elapsed_time": "1:03:15", "remaining_time": "1:14:30", "throughput": 5166.03, "total_tokens": 19609632} {"current_steps": 19485, "total_steps": 42420, "loss": 0.3627, "lr": 0.0006556506544300298, "epoch": 9.186704384724187, "percentage": 45.93, "elapsed_time": "1:03:16", "remaining_time": "1:14:28", "throughput": 5166.01, "total_tokens": 19613696} {"current_steps": 19490, "total_steps": 42420, "loss": 0.3426, "lr": 0.0006554551431417024, "epoch": 9.18906176331919, "percentage": 45.95, "elapsed_time": "1:03:17", "remaining_time": "1:14:27", "throughput": 5166.05, "total_tokens": 19618368} {"current_steps": 19495, "total_steps": 42420, "loss": 0.3294, "lr": 0.0006552596055374587, "epoch": 9.191419141914192, "percentage": 45.96, "elapsed_time": "1:03:18", "remaining_time": "1:14:26", "throughput": 5166.17, "total_tokens": 19623808} {"current_steps": 19500, "total_steps": 42420, "loss": 0.2947, "lr": 0.0006550640416504, "epoch": 9.193776520509195, "percentage": 45.97, "elapsed_time": "1:03:19", "remaining_time": "1:14:25", "throughput": 5166.29, "total_tokens": 19628992} {"current_steps": 19505, "total_steps": 42420, "loss": 0.3267, "lr": 0.0006548684515136319, "epoch": 9.196133899104197, "percentage": 45.98, "elapsed_time": "1:03:20", "remaining_time": "1:14:24", "throughput": 5166.41, "total_tokens": 19634816} {"current_steps": 19510, "total_steps": 42420, "loss": 0.3045, "lr": 0.0006546728351602644, "epoch": 9.198491277699198, "percentage": 45.99, "elapsed_time": "1:03:21", "remaining_time": "1:14:23", "throughput": 5166.45, "total_tokens": 19639680} {"current_steps": 19515, "total_steps": 42420, "loss": 0.4101, "lr": 0.0006544771926234122, "epoch": 9.2008486562942, "percentage": 46.0, "elapsed_time": "1:03:22", "remaining_time": "1:14:22", "throughput": 5166.53, "total_tokens": 19644608} {"current_steps": 19520, "total_steps": 42420, "loss": 0.3736, "lr": 0.0006542815239361941, "epoch": 9.203206034889202, "percentage": 46.02, "elapsed_time": "1:03:23", "remaining_time": "1:14:21", "throughput": 5166.72, "total_tokens": 19650944} {"current_steps": 19525, "total_steps": 42420, "loss": 0.3625, "lr": 0.0006540858291317336, "epoch": 9.205563413484205, "percentage": 46.03, "elapsed_time": "1:03:24", "remaining_time": "1:14:20", "throughput": 5166.81, "total_tokens": 19656000} {"current_steps": 19530, "total_steps": 42420, "loss": 0.3603, "lr": 0.0006538901082431584, "epoch": 9.207920792079207, "percentage": 46.04, "elapsed_time": "1:03:25", "remaining_time": "1:14:19", "throughput": 5166.78, "total_tokens": 19660096} {"current_steps": 19535, "total_steps": 42420, "loss": 0.3708, "lr": 0.0006536943613036008, "epoch": 9.21027817067421, "percentage": 46.05, "elapsed_time": "1:03:26", "remaining_time": "1:14:18", "throughput": 5166.91, "total_tokens": 19665696} {"current_steps": 19540, "total_steps": 42420, "loss": 0.321, "lr": 0.0006534985883461975, "epoch": 9.212635549269212, "percentage": 46.06, "elapsed_time": "1:03:26", "remaining_time": "1:14:17", "throughput": 5166.98, "total_tokens": 19670496} {"current_steps": 19545, "total_steps": 42420, "loss": 0.3367, "lr": 0.0006533027894040894, "epoch": 9.214992927864214, "percentage": 46.07, "elapsed_time": "1:03:27", "remaining_time": "1:14:16", "throughput": 5167.14, "total_tokens": 19676416} {"current_steps": 19550, "total_steps": 42420, "loss": 0.3151, "lr": 0.0006531069645104218, "epoch": 9.217350306459217, "percentage": 46.09, "elapsed_time": "1:03:28", "remaining_time": "1:14:15", "throughput": 5167.19, "total_tokens": 19681344} {"current_steps": 19555, "total_steps": 42420, "loss": 0.4429, "lr": 0.0006529111136983448, "epoch": 9.21970768505422, "percentage": 46.1, "elapsed_time": "1:03:29", "remaining_time": "1:14:14", "throughput": 5167.23, "total_tokens": 19686176} {"current_steps": 19560, "total_steps": 42420, "loss": 0.3491, "lr": 0.0006527152370010127, "epoch": 9.222065063649222, "percentage": 46.11, "elapsed_time": "1:03:30", "remaining_time": "1:14:13", "throughput": 5167.39, "total_tokens": 19692256} {"current_steps": 19565, "total_steps": 42420, "loss": 0.3353, "lr": 0.0006525193344515839, "epoch": 9.224422442244224, "percentage": 46.12, "elapsed_time": "1:03:31", "remaining_time": "1:14:12", "throughput": 5167.48, "total_tokens": 19697504} {"current_steps": 19570, "total_steps": 42420, "loss": 0.3402, "lr": 0.0006523234060832213, "epoch": 9.226779820839226, "percentage": 46.13, "elapsed_time": "1:03:32", "remaining_time": "1:14:11", "throughput": 5167.53, "total_tokens": 19702240} {"current_steps": 19575, "total_steps": 42420, "loss": 0.3073, "lr": 0.0006521274519290924, "epoch": 9.229137199434229, "percentage": 46.15, "elapsed_time": "1:03:33", "remaining_time": "1:14:10", "throughput": 5167.55, "total_tokens": 19706688} {"current_steps": 19580, "total_steps": 42420, "loss": 0.3677, "lr": 0.0006519314720223688, "epoch": 9.231494578029231, "percentage": 46.16, "elapsed_time": "1:03:34", "remaining_time": "1:14:09", "throughput": 5167.63, "total_tokens": 19711904} {"current_steps": 19585, "total_steps": 42420, "loss": 0.3524, "lr": 0.0006517354663962268, "epoch": 9.233851956624234, "percentage": 46.17, "elapsed_time": "1:03:35", "remaining_time": "1:14:08", "throughput": 5167.7, "total_tokens": 19716704} {"current_steps": 19590, "total_steps": 42420, "loss": 0.3791, "lr": 0.0006515394350838466, "epoch": 9.236209335219236, "percentage": 46.18, "elapsed_time": "1:03:36", "remaining_time": "1:14:07", "throughput": 5167.82, "total_tokens": 19722176} {"current_steps": 19595, "total_steps": 42420, "loss": 0.3393, "lr": 0.0006513433781184131, "epoch": 9.238566713814238, "percentage": 46.19, "elapsed_time": "1:03:37", "remaining_time": "1:14:06", "throughput": 5167.9, "total_tokens": 19727168} {"current_steps": 19600, "total_steps": 42420, "loss": 0.337, "lr": 0.0006511472955331152, "epoch": 9.24092409240924, "percentage": 46.2, "elapsed_time": "1:03:38", "remaining_time": "1:14:05", "throughput": 5167.87, "total_tokens": 19731488} {"current_steps": 19605, "total_steps": 42420, "loss": 0.347, "lr": 0.0006509511873611464, "epoch": 9.243281471004243, "percentage": 46.22, "elapsed_time": "1:03:38", "remaining_time": "1:14:04", "throughput": 5167.88, "total_tokens": 19735936} {"current_steps": 19610, "total_steps": 42420, "loss": 0.3322, "lr": 0.0006507550536357049, "epoch": 9.245638849599246, "percentage": 46.23, "elapsed_time": "1:03:40", "remaining_time": "1:14:03", "throughput": 5168.04, "total_tokens": 19742208} {"current_steps": 19615, "total_steps": 42420, "loss": 0.3542, "lr": 0.0006505588943899923, "epoch": 9.247996228194248, "percentage": 46.24, "elapsed_time": "1:03:40", "remaining_time": "1:14:02", "throughput": 5168.11, "total_tokens": 19747296} {"current_steps": 19620, "total_steps": 42420, "loss": 0.3694, "lr": 0.0006503627096572152, "epoch": 9.25035360678925, "percentage": 46.25, "elapsed_time": "1:03:42", "remaining_time": "1:14:01", "throughput": 5168.26, "total_tokens": 19753216} {"current_steps": 19625, "total_steps": 42420, "loss": 0.3028, "lr": 0.0006501664994705843, "epoch": 9.252710985384253, "percentage": 46.26, "elapsed_time": "1:03:42", "remaining_time": "1:14:00", "throughput": 5168.32, "total_tokens": 19758272} {"current_steps": 19630, "total_steps": 42420, "loss": 0.3325, "lr": 0.000649970263863315, "epoch": 9.255068363979255, "percentage": 46.28, "elapsed_time": "1:03:43", "remaining_time": "1:13:59", "throughput": 5168.45, "total_tokens": 19763936} {"current_steps": 19635, "total_steps": 42420, "loss": 0.3025, "lr": 0.0006497740028686262, "epoch": 9.257425742574258, "percentage": 46.29, "elapsed_time": "1:03:44", "remaining_time": "1:13:58", "throughput": 5168.55, "total_tokens": 19769344} {"current_steps": 19640, "total_steps": 42420, "loss": 0.4005, "lr": 0.0006495777165197417, "epoch": 9.25978312116926, "percentage": 46.3, "elapsed_time": "1:03:45", "remaining_time": "1:13:57", "throughput": 5168.68, "total_tokens": 19775200} {"current_steps": 19645, "total_steps": 42420, "loss": 0.338, "lr": 0.0006493814048498895, "epoch": 9.262140499764262, "percentage": 46.31, "elapsed_time": "1:03:46", "remaining_time": "1:13:56", "throughput": 5168.74, "total_tokens": 19779936} {"current_steps": 19650, "total_steps": 42420, "loss": 0.3533, "lr": 0.000649185067892302, "epoch": 9.264497878359265, "percentage": 46.32, "elapsed_time": "1:03:47", "remaining_time": "1:13:55", "throughput": 5168.83, "total_tokens": 19785216} {"current_steps": 19655, "total_steps": 42420, "loss": 0.3539, "lr": 0.0006489887056802153, "epoch": 9.266855256954267, "percentage": 46.33, "elapsed_time": "1:03:48", "remaining_time": "1:13:54", "throughput": 5168.85, "total_tokens": 19789920} {"current_steps": 19660, "total_steps": 42420, "loss": 0.3478, "lr": 0.0006487923182468705, "epoch": 9.26921263554927, "percentage": 46.35, "elapsed_time": "1:03:49", "remaining_time": "1:13:53", "throughput": 5168.9, "total_tokens": 19794816} {"current_steps": 19665, "total_steps": 42420, "loss": 0.3646, "lr": 0.0006485959056255124, "epoch": 9.271570014144272, "percentage": 46.36, "elapsed_time": "1:03:50", "remaining_time": "1:13:52", "throughput": 5168.97, "total_tokens": 19799840} {"current_steps": 19670, "total_steps": 42420, "loss": 0.3149, "lr": 0.0006483994678493908, "epoch": 9.273927392739274, "percentage": 46.37, "elapsed_time": "1:03:51", "remaining_time": "1:13:51", "throughput": 5168.94, "total_tokens": 19803904} {"current_steps": 19675, "total_steps": 42420, "loss": 0.317, "lr": 0.0006482030049517588, "epoch": 9.276284771334277, "percentage": 46.38, "elapsed_time": "1:03:52", "remaining_time": "1:13:50", "throughput": 5168.99, "total_tokens": 19808864} {"current_steps": 19680, "total_steps": 42420, "loss": 0.2259, "lr": 0.0006480065169658743, "epoch": 9.27864214992928, "percentage": 46.39, "elapsed_time": "1:03:53", "remaining_time": "1:13:49", "throughput": 5169.1, "total_tokens": 19814368} {"current_steps": 19685, "total_steps": 42420, "loss": 0.3492, "lr": 0.0006478100039249997, "epoch": 9.280999528524282, "percentage": 46.4, "elapsed_time": "1:03:54", "remaining_time": "1:13:48", "throughput": 5169.22, "total_tokens": 19819936} {"current_steps": 19690, "total_steps": 42420, "loss": 0.3041, "lr": 0.000647613465862401, "epoch": 9.283356907119284, "percentage": 46.42, "elapsed_time": "1:03:55", "remaining_time": "1:13:47", "throughput": 5169.5, "total_tokens": 19827360} {"current_steps": 19695, "total_steps": 42420, "loss": 0.3491, "lr": 0.0006474169028113488, "epoch": 9.285714285714286, "percentage": 46.43, "elapsed_time": "1:03:56", "remaining_time": "1:13:46", "throughput": 5169.58, "total_tokens": 19832576} {"current_steps": 19700, "total_steps": 42420, "loss": 0.3433, "lr": 0.0006472203148051181, "epoch": 9.288071664309289, "percentage": 46.44, "elapsed_time": "1:03:57", "remaining_time": "1:13:45", "throughput": 5169.66, "total_tokens": 19837632} {"current_steps": 19705, "total_steps": 42420, "loss": 0.3243, "lr": 0.0006470237018769878, "epoch": 9.290429042904291, "percentage": 46.45, "elapsed_time": "1:03:58", "remaining_time": "1:13:44", "throughput": 5169.74, "total_tokens": 19842720} {"current_steps": 19710, "total_steps": 42420, "loss": 0.3248, "lr": 0.0006468270640602408, "epoch": 9.292786421499294, "percentage": 46.46, "elapsed_time": "1:03:59", "remaining_time": "1:13:43", "throughput": 5169.79, "total_tokens": 19847424} {"current_steps": 19715, "total_steps": 42420, "loss": 0.3286, "lr": 0.0006466304013881653, "epoch": 9.295143800094294, "percentage": 46.48, "elapsed_time": "1:03:59", "remaining_time": "1:13:42", "throughput": 5169.78, "total_tokens": 19851712} {"current_steps": 19720, "total_steps": 42420, "loss": 0.3211, "lr": 0.000646433713894052, "epoch": 9.297501178689297, "percentage": 46.49, "elapsed_time": "1:04:01", "remaining_time": "1:13:41", "throughput": 5169.95, "total_tokens": 19857856} {"current_steps": 19725, "total_steps": 42420, "loss": 0.2904, "lr": 0.0006462370016111974, "epoch": 9.299858557284299, "percentage": 46.5, "elapsed_time": "1:04:02", "remaining_time": "1:13:40", "throughput": 5170.08, "total_tokens": 19863808} {"current_steps": 19730, "total_steps": 42420, "loss": 0.2914, "lr": 0.0006460402645729014, "epoch": 9.302215935879302, "percentage": 46.51, "elapsed_time": "1:04:02", "remaining_time": "1:13:39", "throughput": 5170.1, "total_tokens": 19868160} {"current_steps": 19735, "total_steps": 42420, "loss": 0.2884, "lr": 0.0006458435028124684, "epoch": 9.304573314474304, "percentage": 46.52, "elapsed_time": "1:04:03", "remaining_time": "1:13:38", "throughput": 5170.16, "total_tokens": 19873024} {"current_steps": 19740, "total_steps": 42420, "loss": 0.2896, "lr": 0.0006456467163632065, "epoch": 9.306930693069306, "percentage": 46.53, "elapsed_time": "1:04:04", "remaining_time": "1:13:37", "throughput": 5170.13, "total_tokens": 19876928} {"current_steps": 19745, "total_steps": 42420, "loss": 0.3513, "lr": 0.0006454499052584284, "epoch": 9.309288071664309, "percentage": 46.55, "elapsed_time": "1:04:05", "remaining_time": "1:13:35", "throughput": 5170.1, "total_tokens": 19880960} {"current_steps": 19750, "total_steps": 42420, "loss": 0.2548, "lr": 0.0006452530695314511, "epoch": 9.311645450259311, "percentage": 46.56, "elapsed_time": "1:04:06", "remaining_time": "1:13:34", "throughput": 5170.15, "total_tokens": 19885824} {"current_steps": 19755, "total_steps": 42420, "loss": 0.2682, "lr": 0.0006450562092155952, "epoch": 9.314002828854314, "percentage": 46.57, "elapsed_time": "1:04:07", "remaining_time": "1:13:34", "throughput": 5170.38, "total_tokens": 19892672} {"current_steps": 19760, "total_steps": 42420, "loss": 0.2915, "lr": 0.0006448593243441862, "epoch": 9.316360207449316, "percentage": 46.58, "elapsed_time": "1:04:08", "remaining_time": "1:13:33", "throughput": 5170.5, "total_tokens": 19898080} {"current_steps": 19765, "total_steps": 42420, "loss": 0.2519, "lr": 0.000644662414950553, "epoch": 9.318717586044318, "percentage": 46.59, "elapsed_time": "1:04:09", "remaining_time": "1:13:32", "throughput": 5170.6, "total_tokens": 19903712} {"current_steps": 19770, "total_steps": 42420, "loss": 0.3088, "lr": 0.0006444654810680293, "epoch": 9.32107496463932, "percentage": 46.61, "elapsed_time": "1:04:10", "remaining_time": "1:13:31", "throughput": 5170.69, "total_tokens": 19908960} {"current_steps": 19775, "total_steps": 42420, "loss": 0.2514, "lr": 0.0006442685227299526, "epoch": 9.323432343234323, "percentage": 46.62, "elapsed_time": "1:04:11", "remaining_time": "1:13:30", "throughput": 5170.74, "total_tokens": 19913920} {"current_steps": 19780, "total_steps": 42420, "loss": 0.2804, "lr": 0.0006440715399696643, "epoch": 9.325789721829326, "percentage": 46.63, "elapsed_time": "1:04:12", "remaining_time": "1:13:29", "throughput": 5170.74, "total_tokens": 19918176} {"current_steps": 19785, "total_steps": 42420, "loss": 0.4481, "lr": 0.0006438745328205105, "epoch": 9.328147100424328, "percentage": 46.64, "elapsed_time": "1:04:12", "remaining_time": "1:13:27", "throughput": 5170.74, "total_tokens": 19922560} {"current_steps": 19790, "total_steps": 42420, "loss": 0.3544, "lr": 0.0006436775013158412, "epoch": 9.33050447901933, "percentage": 46.65, "elapsed_time": "1:04:13", "remaining_time": "1:13:26", "throughput": 5170.77, "total_tokens": 19927008} {"current_steps": 19795, "total_steps": 42420, "loss": 0.3519, "lr": 0.0006434804454890107, "epoch": 9.332861857614333, "percentage": 46.66, "elapsed_time": "1:04:14", "remaining_time": "1:13:25", "throughput": 5170.81, "total_tokens": 19931520} {"current_steps": 19800, "total_steps": 42420, "loss": 0.3494, "lr": 0.0006432833653733765, "epoch": 9.335219236209335, "percentage": 46.68, "elapsed_time": "1:04:15", "remaining_time": "1:13:24", "throughput": 5170.9, "total_tokens": 19937088} {"current_steps": 19805, "total_steps": 42420, "loss": 0.3563, "lr": 0.0006430862610023017, "epoch": 9.337576614804338, "percentage": 46.69, "elapsed_time": "1:04:16", "remaining_time": "1:13:23", "throughput": 5170.9, "total_tokens": 19941376} {"current_steps": 19810, "total_steps": 42420, "loss": 0.3222, "lr": 0.0006428891324091521, "epoch": 9.33993399339934, "percentage": 46.7, "elapsed_time": "1:04:17", "remaining_time": "1:13:22", "throughput": 5170.99, "total_tokens": 19946496} {"current_steps": 19815, "total_steps": 42420, "loss": 0.3266, "lr": 0.0006426919796272988, "epoch": 9.342291371994342, "percentage": 46.71, "elapsed_time": "1:04:18", "remaining_time": "1:13:21", "throughput": 5171.05, "total_tokens": 19951456} {"current_steps": 19820, "total_steps": 42420, "loss": 0.4259, "lr": 0.0006424948026901159, "epoch": 9.344648750589345, "percentage": 46.72, "elapsed_time": "1:04:19", "remaining_time": "1:13:20", "throughput": 5171.06, "total_tokens": 19955744} {"current_steps": 19825, "total_steps": 42420, "loss": 0.3486, "lr": 0.0006422976016309825, "epoch": 9.347006129184347, "percentage": 46.74, "elapsed_time": "1:04:19", "remaining_time": "1:13:19", "throughput": 5171.07, "total_tokens": 19960128} {"current_steps": 19830, "total_steps": 42420, "loss": 0.3379, "lr": 0.000642100376483281, "epoch": 9.34936350777935, "percentage": 46.75, "elapsed_time": "1:04:20", "remaining_time": "1:13:18", "throughput": 5171.19, "total_tokens": 19965888} {"current_steps": 19835, "total_steps": 42420, "loss": 0.3581, "lr": 0.0006419031272803986, "epoch": 9.351720886374352, "percentage": 46.76, "elapsed_time": "1:04:21", "remaining_time": "1:13:17", "throughput": 5171.23, "total_tokens": 19970336} {"current_steps": 19840, "total_steps": 42420, "loss": 0.3485, "lr": 0.0006417058540557262, "epoch": 9.354078264969354, "percentage": 46.77, "elapsed_time": "1:04:22", "remaining_time": "1:13:16", "throughput": 5171.29, "total_tokens": 19975328} {"current_steps": 19845, "total_steps": 42420, "loss": 0.3296, "lr": 0.0006415085568426586, "epoch": 9.356435643564357, "percentage": 46.78, "elapsed_time": "1:04:23", "remaining_time": "1:13:15", "throughput": 5171.27, "total_tokens": 19979424} {"current_steps": 19850, "total_steps": 42420, "loss": 0.3544, "lr": 0.0006413112356745952, "epoch": 9.35879302215936, "percentage": 46.79, "elapsed_time": "1:04:24", "remaining_time": "1:13:13", "throughput": 5171.23, "total_tokens": 19983456} {"current_steps": 19855, "total_steps": 42420, "loss": 0.3157, "lr": 0.0006411138905849386, "epoch": 9.361150400754362, "percentage": 46.81, "elapsed_time": "1:04:25", "remaining_time": "1:13:12", "throughput": 5171.33, "total_tokens": 19988992} {"current_steps": 19860, "total_steps": 42420, "loss": 0.3095, "lr": 0.0006409165216070967, "epoch": 9.363507779349364, "percentage": 46.82, "elapsed_time": "1:04:26", "remaining_time": "1:13:11", "throughput": 5171.39, "total_tokens": 19993824} {"current_steps": 19865, "total_steps": 42420, "loss": 0.3958, "lr": 0.00064071912877448, "epoch": 9.365865157944366, "percentage": 46.83, "elapsed_time": "1:04:27", "remaining_time": "1:13:10", "throughput": 5171.35, "total_tokens": 19998112} {"current_steps": 19870, "total_steps": 42420, "loss": 0.3739, "lr": 0.0006405217121205043, "epoch": 9.368222536539369, "percentage": 46.84, "elapsed_time": "1:04:28", "remaining_time": "1:13:09", "throughput": 5171.43, "total_tokens": 20003136} {"current_steps": 19875, "total_steps": 42420, "loss": 0.3511, "lr": 0.0006403242716785884, "epoch": 9.370579915134371, "percentage": 46.85, "elapsed_time": "1:04:28", "remaining_time": "1:13:08", "throughput": 5171.5, "total_tokens": 20008416} {"current_steps": 19880, "total_steps": 42420, "loss": 0.3461, "lr": 0.0006401268074821561, "epoch": 9.372937293729374, "percentage": 46.86, "elapsed_time": "1:04:29", "remaining_time": "1:13:07", "throughput": 5171.59, "total_tokens": 20013920} {"current_steps": 19885, "total_steps": 42420, "loss": 0.333, "lr": 0.0006399293195646345, "epoch": 9.375294672324376, "percentage": 46.88, "elapsed_time": "1:04:30", "remaining_time": "1:13:06", "throughput": 5171.7, "total_tokens": 20019360} {"current_steps": 19890, "total_steps": 42420, "loss": 0.3406, "lr": 0.0006397318079594548, "epoch": 9.377652050919378, "percentage": 46.89, "elapsed_time": "1:04:32", "remaining_time": "1:13:05", "throughput": 5171.88, "total_tokens": 20025696} {"current_steps": 19895, "total_steps": 42420, "loss": 0.3771, "lr": 0.0006395342727000526, "epoch": 9.38000942951438, "percentage": 46.9, "elapsed_time": "1:04:32", "remaining_time": "1:13:04", "throughput": 5171.84, "total_tokens": 20029632} {"current_steps": 19900, "total_steps": 42420, "loss": 0.3758, "lr": 0.0006393367138198674, "epoch": 9.382366808109383, "percentage": 46.91, "elapsed_time": "1:04:33", "remaining_time": "1:13:03", "throughput": 5171.88, "total_tokens": 20034272} {"current_steps": 19905, "total_steps": 42420, "loss": 0.3612, "lr": 0.000639139131352342, "epoch": 9.384724186704386, "percentage": 46.92, "elapsed_time": "1:04:34", "remaining_time": "1:13:02", "throughput": 5171.84, "total_tokens": 20038272} {"current_steps": 19910, "total_steps": 42420, "loss": 0.3471, "lr": 0.0006389415253309243, "epoch": 9.387081565299386, "percentage": 46.94, "elapsed_time": "1:04:35", "remaining_time": "1:13:01", "throughput": 5171.93, "total_tokens": 20043392} {"current_steps": 19915, "total_steps": 42420, "loss": 0.3534, "lr": 0.0006387438957890654, "epoch": 9.389438943894389, "percentage": 46.95, "elapsed_time": "1:04:36", "remaining_time": "1:13:00", "throughput": 5172.16, "total_tokens": 20050624} {"current_steps": 19920, "total_steps": 42420, "loss": 0.3858, "lr": 0.0006385462427602209, "epoch": 9.391796322489391, "percentage": 46.96, "elapsed_time": "1:04:37", "remaining_time": "1:12:59", "throughput": 5172.22, "total_tokens": 20055616} {"current_steps": 19925, "total_steps": 42420, "loss": 0.3301, "lr": 0.0006383485662778497, "epoch": 9.394153701084393, "percentage": 46.97, "elapsed_time": "1:04:38", "remaining_time": "1:12:58", "throughput": 5172.2, "total_tokens": 20059936} {"current_steps": 19930, "total_steps": 42420, "loss": 0.3064, "lr": 0.0006381508663754152, "epoch": 9.396511079679396, "percentage": 46.98, "elapsed_time": "1:04:39", "remaining_time": "1:12:57", "throughput": 5172.16, "total_tokens": 20063872} {"current_steps": 19935, "total_steps": 42420, "loss": 0.3016, "lr": 0.0006379531430863848, "epoch": 9.398868458274398, "percentage": 46.99, "elapsed_time": "1:04:40", "remaining_time": "1:12:56", "throughput": 5172.3, "total_tokens": 20069472} {"current_steps": 19940, "total_steps": 42420, "loss": 0.365, "lr": 0.0006377553964442294, "epoch": 9.4012258368694, "percentage": 47.01, "elapsed_time": "1:04:41", "remaining_time": "1:12:55", "throughput": 5172.35, "total_tokens": 20074368} {"current_steps": 19945, "total_steps": 42420, "loss": 0.3031, "lr": 0.0006375576264824245, "epoch": 9.403583215464403, "percentage": 47.02, "elapsed_time": "1:04:41", "remaining_time": "1:12:54", "throughput": 5172.34, "total_tokens": 20078624} {"current_steps": 19950, "total_steps": 42420, "loss": 0.344, "lr": 0.000637359833234449, "epoch": 9.405940594059405, "percentage": 47.03, "elapsed_time": "1:04:42", "remaining_time": "1:12:53", "throughput": 5172.39, "total_tokens": 20083328} {"current_steps": 19955, "total_steps": 42420, "loss": 0.3345, "lr": 0.0006371620167337859, "epoch": 9.408297972654408, "percentage": 47.04, "elapsed_time": "1:04:43", "remaining_time": "1:12:52", "throughput": 5172.55, "total_tokens": 20089248} {"current_steps": 19960, "total_steps": 42420, "loss": 0.3491, "lr": 0.0006369641770139224, "epoch": 9.41065535124941, "percentage": 47.05, "elapsed_time": "1:04:44", "remaining_time": "1:12:51", "throughput": 5172.52, "total_tokens": 20093408} {"current_steps": 19965, "total_steps": 42420, "loss": 0.3241, "lr": 0.0006367663141083492, "epoch": 9.413012729844413, "percentage": 47.07, "elapsed_time": "1:04:45", "remaining_time": "1:12:50", "throughput": 5172.55, "total_tokens": 20097952} {"current_steps": 19970, "total_steps": 42420, "loss": 0.3547, "lr": 0.0006365684280505612, "epoch": 9.415370108439415, "percentage": 47.08, "elapsed_time": "1:04:46", "remaining_time": "1:12:49", "throughput": 5172.57, "total_tokens": 20102560} {"current_steps": 19975, "total_steps": 42420, "loss": 0.3277, "lr": 0.0006363705188740569, "epoch": 9.417727487034417, "percentage": 47.09, "elapsed_time": "1:04:47", "remaining_time": "1:12:47", "throughput": 5172.66, "total_tokens": 20107584} {"current_steps": 19980, "total_steps": 42420, "loss": 0.3411, "lr": 0.0006361725866123395, "epoch": 9.42008486562942, "percentage": 47.1, "elapsed_time": "1:04:48", "remaining_time": "1:12:46", "throughput": 5172.67, "total_tokens": 20112416} {"current_steps": 19985, "total_steps": 42420, "loss": 0.3418, "lr": 0.0006359746312989152, "epoch": 9.422442244224422, "percentage": 47.11, "elapsed_time": "1:04:49", "remaining_time": "1:12:45", "throughput": 5172.81, "total_tokens": 20118048} {"current_steps": 19990, "total_steps": 42420, "loss": 0.3296, "lr": 0.0006357766529672944, "epoch": 9.424799622819425, "percentage": 47.12, "elapsed_time": "1:04:49", "remaining_time": "1:12:44", "throughput": 5172.8, "total_tokens": 20122080} {"current_steps": 19995, "total_steps": 42420, "loss": 0.369, "lr": 0.0006355786516509916, "epoch": 9.427157001414427, "percentage": 47.14, "elapsed_time": "1:04:50", "remaining_time": "1:12:43", "throughput": 5172.87, "total_tokens": 20127136} {"current_steps": 20000, "total_steps": 42420, "loss": 0.3481, "lr": 0.0006353806273835252, "epoch": 9.42951438000943, "percentage": 47.15, "elapsed_time": "1:04:51", "remaining_time": "1:12:42", "throughput": 5172.99, "total_tokens": 20132544} {"current_steps": 20005, "total_steps": 42420, "loss": 0.3244, "lr": 0.0006351825801984173, "epoch": 9.431871758604432, "percentage": 47.16, "elapsed_time": "1:04:52", "remaining_time": "1:12:41", "throughput": 5173.04, "total_tokens": 20137504} {"current_steps": 20010, "total_steps": 42420, "loss": 0.3126, "lr": 0.0006349845101291936, "epoch": 9.434229137199434, "percentage": 47.17, "elapsed_time": "1:04:53", "remaining_time": "1:12:40", "throughput": 5173.04, "total_tokens": 20141888} {"current_steps": 20015, "total_steps": 42420, "loss": 0.3308, "lr": 0.0006347864172093844, "epoch": 9.436586515794437, "percentage": 47.18, "elapsed_time": "1:04:54", "remaining_time": "1:12:39", "throughput": 5173.11, "total_tokens": 20147104} {"current_steps": 20020, "total_steps": 42420, "loss": 0.2922, "lr": 0.0006345883014725231, "epoch": 9.438943894389439, "percentage": 47.19, "elapsed_time": "1:04:55", "remaining_time": "1:12:38", "throughput": 5173.12, "total_tokens": 20151456} {"current_steps": 20025, "total_steps": 42420, "loss": 0.2893, "lr": 0.0006343901629521478, "epoch": 9.441301272984441, "percentage": 47.21, "elapsed_time": "1:04:56", "remaining_time": "1:12:37", "throughput": 5173.13, "total_tokens": 20155712} {"current_steps": 20030, "total_steps": 42420, "loss": 0.372, "lr": 0.0006341920016817997, "epoch": 9.443658651579444, "percentage": 47.22, "elapsed_time": "1:04:57", "remaining_time": "1:12:36", "throughput": 5173.2, "total_tokens": 20160672} {"current_steps": 20035, "total_steps": 42420, "loss": 0.353, "lr": 0.0006339938176950241, "epoch": 9.446016030174446, "percentage": 47.23, "elapsed_time": "1:04:58", "remaining_time": "1:12:35", "throughput": 5173.27, "total_tokens": 20165760} {"current_steps": 20040, "total_steps": 42420, "loss": 0.3721, "lr": 0.0006337956110253702, "epoch": 9.448373408769449, "percentage": 47.24, "elapsed_time": "1:04:58", "remaining_time": "1:12:34", "throughput": 5173.29, "total_tokens": 20170304} {"current_steps": 20045, "total_steps": 42420, "loss": 0.3381, "lr": 0.0006335973817063912, "epoch": 9.450730787364451, "percentage": 47.25, "elapsed_time": "1:04:59", "remaining_time": "1:12:33", "throughput": 5173.36, "total_tokens": 20175264} {"current_steps": 20050, "total_steps": 42420, "loss": 0.3983, "lr": 0.0006333991297716437, "epoch": 9.453088165959453, "percentage": 47.27, "elapsed_time": "1:05:00", "remaining_time": "1:12:32", "throughput": 5173.42, "total_tokens": 20180288} {"current_steps": 20055, "total_steps": 42420, "loss": 0.3363, "lr": 0.0006332008552546886, "epoch": 9.455445544554456, "percentage": 47.28, "elapsed_time": "1:05:01", "remaining_time": "1:12:31", "throughput": 5173.48, "total_tokens": 20185248} {"current_steps": 20060, "total_steps": 42420, "loss": 0.3256, "lr": 0.0006330025581890902, "epoch": 9.457802923149458, "percentage": 47.29, "elapsed_time": "1:05:02", "remaining_time": "1:12:29", "throughput": 5173.49, "total_tokens": 20189632} {"current_steps": 20065, "total_steps": 42420, "loss": 0.3927, "lr": 0.000632804238608417, "epoch": 9.46016030174446, "percentage": 47.3, "elapsed_time": "1:05:03", "remaining_time": "1:12:28", "throughput": 5173.57, "total_tokens": 20194880} {"current_steps": 20070, "total_steps": 42420, "loss": 0.3228, "lr": 0.0006326058965462414, "epoch": 9.462517680339463, "percentage": 47.31, "elapsed_time": "1:05:04", "remaining_time": "1:12:27", "throughput": 5173.61, "total_tokens": 20199712} {"current_steps": 20075, "total_steps": 42420, "loss": 0.3356, "lr": 0.0006324075320361386, "epoch": 9.464875058934465, "percentage": 47.32, "elapsed_time": "1:05:05", "remaining_time": "1:12:26", "throughput": 5173.67, "total_tokens": 20204608} {"current_steps": 20080, "total_steps": 42420, "loss": 0.3232, "lr": 0.0006322091451116889, "epoch": 9.467232437529468, "percentage": 47.34, "elapsed_time": "1:05:06", "remaining_time": "1:12:25", "throughput": 5173.81, "total_tokens": 20210496} {"current_steps": 20085, "total_steps": 42420, "loss": 0.3378, "lr": 0.0006320107358064758, "epoch": 9.46958981612447, "percentage": 47.35, "elapsed_time": "1:05:07", "remaining_time": "1:12:24", "throughput": 5173.94, "total_tokens": 20215968} {"current_steps": 20090, "total_steps": 42420, "loss": 0.3052, "lr": 0.0006318123041540866, "epoch": 9.471947194719473, "percentage": 47.36, "elapsed_time": "1:05:08", "remaining_time": "1:12:24", "throughput": 5174.06, "total_tokens": 20221600} {"current_steps": 20095, "total_steps": 42420, "loss": 0.2946, "lr": 0.0006316138501881122, "epoch": 9.474304573314475, "percentage": 47.37, "elapsed_time": "1:05:09", "remaining_time": "1:12:22", "throughput": 5174.09, "total_tokens": 20226112} {"current_steps": 20100, "total_steps": 42420, "loss": 0.3784, "lr": 0.0006314153739421476, "epoch": 9.476661951909477, "percentage": 47.38, "elapsed_time": "1:05:10", "remaining_time": "1:12:21", "throughput": 5174.17, "total_tokens": 20231296} {"current_steps": 20105, "total_steps": 42420, "loss": 0.3519, "lr": 0.0006312168754497916, "epoch": 9.47901933050448, "percentage": 47.4, "elapsed_time": "1:05:10", "remaining_time": "1:12:20", "throughput": 5174.24, "total_tokens": 20236384} {"current_steps": 20110, "total_steps": 42420, "loss": 0.3154, "lr": 0.0006310183547446466, "epoch": 9.481376709099482, "percentage": 47.41, "elapsed_time": "1:05:11", "remaining_time": "1:12:19", "throughput": 5174.24, "total_tokens": 20240896} {"current_steps": 20115, "total_steps": 42420, "loss": 0.3474, "lr": 0.0006308198118603187, "epoch": 9.483734087694483, "percentage": 47.42, "elapsed_time": "1:05:12", "remaining_time": "1:12:18", "throughput": 5174.24, "total_tokens": 20245440} {"current_steps": 20120, "total_steps": 42420, "loss": 0.3075, "lr": 0.0006306212468304178, "epoch": 9.486091466289485, "percentage": 47.43, "elapsed_time": "1:05:13", "remaining_time": "1:12:17", "throughput": 5174.24, "total_tokens": 20249856} {"current_steps": 20125, "total_steps": 42420, "loss": 0.3187, "lr": 0.0006304226596885575, "epoch": 9.488448844884488, "percentage": 47.44, "elapsed_time": "1:05:14", "remaining_time": "1:12:16", "throughput": 5174.25, "total_tokens": 20254336} {"current_steps": 20130, "total_steps": 42420, "loss": 0.3345, "lr": 0.0006302240504683555, "epoch": 9.49080622347949, "percentage": 47.45, "elapsed_time": "1:05:15", "remaining_time": "1:12:15", "throughput": 5174.29, "total_tokens": 20259360} {"current_steps": 20135, "total_steps": 42420, "loss": 0.3699, "lr": 0.0006300254192034328, "epoch": 9.493163602074493, "percentage": 47.47, "elapsed_time": "1:05:16", "remaining_time": "1:12:14", "throughput": 5174.34, "total_tokens": 20264288} {"current_steps": 20140, "total_steps": 42420, "loss": 0.3218, "lr": 0.0006298267659274144, "epoch": 9.495520980669495, "percentage": 47.48, "elapsed_time": "1:05:17", "remaining_time": "1:12:13", "throughput": 5174.44, "total_tokens": 20269728} {"current_steps": 20145, "total_steps": 42420, "loss": 0.3518, "lr": 0.0006296280906739287, "epoch": 9.497878359264497, "percentage": 47.49, "elapsed_time": "1:05:18", "remaining_time": "1:12:12", "throughput": 5174.57, "total_tokens": 20275488} {"current_steps": 20150, "total_steps": 42420, "loss": 0.3335, "lr": 0.0006294293934766084, "epoch": 9.5002357378595, "percentage": 47.5, "elapsed_time": "1:05:19", "remaining_time": "1:12:11", "throughput": 5174.64, "total_tokens": 20280608} {"current_steps": 20155, "total_steps": 42420, "loss": 0.3243, "lr": 0.0006292306743690895, "epoch": 9.502593116454502, "percentage": 47.51, "elapsed_time": "1:05:20", "remaining_time": "1:12:10", "throughput": 5174.81, "total_tokens": 20286560} {"current_steps": 20160, "total_steps": 42420, "loss": 0.3615, "lr": 0.0006290319333850113, "epoch": 9.504950495049505, "percentage": 47.52, "elapsed_time": "1:05:21", "remaining_time": "1:12:09", "throughput": 5174.95, "total_tokens": 20292480} {"current_steps": 20165, "total_steps": 42420, "loss": 0.32, "lr": 0.0006288331705580177, "epoch": 9.507307873644507, "percentage": 47.54, "elapsed_time": "1:05:22", "remaining_time": "1:12:08", "throughput": 5174.97, "total_tokens": 20296992} {"current_steps": 20170, "total_steps": 42420, "loss": 0.3779, "lr": 0.0006286343859217558, "epoch": 9.50966525223951, "percentage": 47.55, "elapsed_time": "1:05:22", "remaining_time": "1:12:07", "throughput": 5174.99, "total_tokens": 20301344} {"current_steps": 20175, "total_steps": 42420, "loss": 0.3365, "lr": 0.0006284355795098765, "epoch": 9.512022630834512, "percentage": 47.56, "elapsed_time": "1:05:23", "remaining_time": "1:12:06", "throughput": 5175.07, "total_tokens": 20306400} {"current_steps": 20180, "total_steps": 42420, "loss": 0.3564, "lr": 0.0006282367513560343, "epoch": 9.514380009429514, "percentage": 47.57, "elapsed_time": "1:05:24", "remaining_time": "1:12:05", "throughput": 5175.23, "total_tokens": 20312384} {"current_steps": 20185, "total_steps": 42420, "loss": 0.3713, "lr": 0.0006280379014938874, "epoch": 9.516737388024517, "percentage": 47.58, "elapsed_time": "1:05:25", "remaining_time": "1:12:04", "throughput": 5175.3, "total_tokens": 20317536} {"current_steps": 20190, "total_steps": 42420, "loss": 0.3686, "lr": 0.0006278390299570977, "epoch": 9.519094766619519, "percentage": 47.6, "elapsed_time": "1:05:26", "remaining_time": "1:12:03", "throughput": 5175.35, "total_tokens": 20322848} {"current_steps": 20195, "total_steps": 42420, "loss": 0.3168, "lr": 0.000627640136779331, "epoch": 9.521452145214521, "percentage": 47.61, "elapsed_time": "1:05:27", "remaining_time": "1:12:02", "throughput": 5175.38, "total_tokens": 20327584} {"current_steps": 20200, "total_steps": 42420, "loss": 0.3012, "lr": 0.0006274412219942562, "epoch": 9.523809523809524, "percentage": 47.62, "elapsed_time": "1:05:28", "remaining_time": "1:12:01", "throughput": 5175.38, "total_tokens": 20332000} {"current_steps": 20205, "total_steps": 42420, "loss": 0.3049, "lr": 0.0006272422856355462, "epoch": 9.526166902404526, "percentage": 47.63, "elapsed_time": "1:05:29", "remaining_time": "1:12:00", "throughput": 5175.47, "total_tokens": 20337184} {"current_steps": 20210, "total_steps": 42420, "loss": 0.354, "lr": 0.0006270433277368779, "epoch": 9.528524280999529, "percentage": 47.64, "elapsed_time": "1:05:30", "remaining_time": "1:11:59", "throughput": 5175.47, "total_tokens": 20341408} {"current_steps": 20215, "total_steps": 42420, "loss": 0.3941, "lr": 0.0006268443483319312, "epoch": 9.530881659594531, "percentage": 47.65, "elapsed_time": "1:05:31", "remaining_time": "1:11:58", "throughput": 5175.58, "total_tokens": 20346976} {"current_steps": 20220, "total_steps": 42420, "loss": 0.3605, "lr": 0.00062664534745439, "epoch": 9.533239038189533, "percentage": 47.67, "elapsed_time": "1:05:32", "remaining_time": "1:11:57", "throughput": 5175.71, "total_tokens": 20352768} {"current_steps": 20225, "total_steps": 42420, "loss": 0.3332, "lr": 0.000626446325137942, "epoch": 9.535596416784536, "percentage": 47.68, "elapsed_time": "1:05:33", "remaining_time": "1:11:56", "throughput": 5175.8, "total_tokens": 20358208} {"current_steps": 20230, "total_steps": 42420, "loss": 0.3386, "lr": 0.0006262472814162779, "epoch": 9.537953795379538, "percentage": 47.69, "elapsed_time": "1:05:34", "remaining_time": "1:11:55", "throughput": 5175.85, "total_tokens": 20363232} {"current_steps": 20235, "total_steps": 42420, "loss": 0.3376, "lr": 0.0006260482163230927, "epoch": 9.54031117397454, "percentage": 47.7, "elapsed_time": "1:05:35", "remaining_time": "1:11:54", "throughput": 5175.91, "total_tokens": 20368192} {"current_steps": 20240, "total_steps": 42420, "loss": 0.2999, "lr": 0.0006258491298920845, "epoch": 9.542668552569543, "percentage": 47.71, "elapsed_time": "1:05:36", "remaining_time": "1:11:53", "throughput": 5176.13, "total_tokens": 20375072} {"current_steps": 20245, "total_steps": 42420, "loss": 0.3016, "lr": 0.0006256500221569556, "epoch": 9.545025931164545, "percentage": 47.73, "elapsed_time": "1:05:37", "remaining_time": "1:11:52", "throughput": 5176.35, "total_tokens": 20381760} {"current_steps": 20250, "total_steps": 42420, "loss": 0.3602, "lr": 0.0006254508931514112, "epoch": 9.547383309759548, "percentage": 47.74, "elapsed_time": "1:05:38", "remaining_time": "1:11:51", "throughput": 5176.35, "total_tokens": 20386112} {"current_steps": 20255, "total_steps": 42420, "loss": 0.3309, "lr": 0.0006252517429091608, "epoch": 9.54974068835455, "percentage": 47.75, "elapsed_time": "1:05:39", "remaining_time": "1:11:50", "throughput": 5176.36, "total_tokens": 20390624} {"current_steps": 20260, "total_steps": 42420, "loss": 0.3445, "lr": 0.000625052571463917, "epoch": 9.552098066949553, "percentage": 47.76, "elapsed_time": "1:05:40", "remaining_time": "1:11:49", "throughput": 5176.4, "total_tokens": 20395520} {"current_steps": 20265, "total_steps": 42420, "loss": 0.3169, "lr": 0.0006248533788493958, "epoch": 9.554455445544555, "percentage": 47.77, "elapsed_time": "1:05:40", "remaining_time": "1:11:48", "throughput": 5176.38, "total_tokens": 20399616} {"current_steps": 20270, "total_steps": 42420, "loss": 0.3243, "lr": 0.0006246541650993177, "epoch": 9.556812824139557, "percentage": 47.78, "elapsed_time": "1:05:41", "remaining_time": "1:11:47", "throughput": 5176.55, "total_tokens": 20405632} {"current_steps": 20275, "total_steps": 42420, "loss": 0.3411, "lr": 0.000624454930247406, "epoch": 9.55917020273456, "percentage": 47.8, "elapsed_time": "1:05:42", "remaining_time": "1:11:46", "throughput": 5176.59, "total_tokens": 20410560} {"current_steps": 20280, "total_steps": 42420, "loss": 0.315, "lr": 0.0006242556743273876, "epoch": 9.561527581329562, "percentage": 47.81, "elapsed_time": "1:05:43", "remaining_time": "1:11:45", "throughput": 5176.59, "total_tokens": 20414784} {"current_steps": 20285, "total_steps": 42420, "loss": 0.3382, "lr": 0.0006240563973729932, "epoch": 9.563884959924565, "percentage": 47.82, "elapsed_time": "1:05:44", "remaining_time": "1:11:44", "throughput": 5176.69, "total_tokens": 20420640} {"current_steps": 20290, "total_steps": 42420, "loss": 0.3359, "lr": 0.0006238570994179573, "epoch": 9.566242338519567, "percentage": 47.83, "elapsed_time": "1:05:45", "remaining_time": "1:11:43", "throughput": 5176.69, "total_tokens": 20425120} {"current_steps": 20295, "total_steps": 42420, "loss": 0.3038, "lr": 0.0006236577804960174, "epoch": 9.56859971711457, "percentage": 47.84, "elapsed_time": "1:05:46", "remaining_time": "1:11:42", "throughput": 5176.81, "total_tokens": 20430688} {"current_steps": 20300, "total_steps": 42420, "loss": 0.3188, "lr": 0.0006234584406409147, "epoch": 9.570957095709572, "percentage": 47.85, "elapsed_time": "1:05:47", "remaining_time": "1:11:41", "throughput": 5176.88, "total_tokens": 20435936} {"current_steps": 20305, "total_steps": 42420, "loss": 0.3083, "lr": 0.0006232590798863943, "epoch": 9.573314474304574, "percentage": 47.87, "elapsed_time": "1:05:48", "remaining_time": "1:11:40", "throughput": 5177.0, "total_tokens": 20441600} {"current_steps": 20310, "total_steps": 42420, "loss": 0.4537, "lr": 0.0006230596982662044, "epoch": 9.575671852899575, "percentage": 47.88, "elapsed_time": "1:05:49", "remaining_time": "1:11:39", "throughput": 5177.09, "total_tokens": 20447104} {"current_steps": 20315, "total_steps": 42420, "loss": 0.3407, "lr": 0.0006228602958140972, "epoch": 9.578029231494579, "percentage": 47.89, "elapsed_time": "1:05:50", "remaining_time": "1:11:38", "throughput": 5177.18, "total_tokens": 20452608} {"current_steps": 20320, "total_steps": 42420, "loss": 0.363, "lr": 0.000622660872563828, "epoch": 9.58038661008958, "percentage": 47.9, "elapsed_time": "1:05:51", "remaining_time": "1:11:37", "throughput": 5177.24, "total_tokens": 20457472} {"current_steps": 20325, "total_steps": 42420, "loss": 0.3714, "lr": 0.0006224614285491555, "epoch": 9.582743988684582, "percentage": 47.91, "elapsed_time": "1:05:52", "remaining_time": "1:11:36", "throughput": 5177.36, "total_tokens": 20463520} {"current_steps": 20330, "total_steps": 42420, "loss": 0.3615, "lr": 0.0006222619638038424, "epoch": 9.585101367279584, "percentage": 47.93, "elapsed_time": "1:05:53", "remaining_time": "1:11:35", "throughput": 5177.41, "total_tokens": 20468544} {"current_steps": 20335, "total_steps": 42420, "loss": 0.3298, "lr": 0.000622062478361655, "epoch": 9.587458745874587, "percentage": 47.94, "elapsed_time": "1:05:54", "remaining_time": "1:11:34", "throughput": 5177.37, "total_tokens": 20472480} {"current_steps": 20340, "total_steps": 42420, "loss": 0.2733, "lr": 0.0006218629722563625, "epoch": 9.58981612446959, "percentage": 47.95, "elapsed_time": "1:05:55", "remaining_time": "1:11:33", "throughput": 5177.34, "total_tokens": 20476576} {"current_steps": 20345, "total_steps": 42420, "loss": 0.3565, "lr": 0.0006216634455217377, "epoch": 9.592173503064592, "percentage": 47.96, "elapsed_time": "1:05:55", "remaining_time": "1:11:32", "throughput": 5177.32, "total_tokens": 20480832} {"current_steps": 20350, "total_steps": 42420, "loss": 0.4367, "lr": 0.0006214638981915573, "epoch": 9.594530881659594, "percentage": 47.97, "elapsed_time": "1:05:56", "remaining_time": "1:11:31", "throughput": 5177.41, "total_tokens": 20486144} {"current_steps": 20355, "total_steps": 42420, "loss": 0.2877, "lr": 0.0006212643302996014, "epoch": 9.596888260254596, "percentage": 47.98, "elapsed_time": "1:05:57", "remaining_time": "1:11:30", "throughput": 5177.43, "total_tokens": 20490784} {"current_steps": 20360, "total_steps": 42420, "loss": 0.3494, "lr": 0.0006210647418796534, "epoch": 9.599245638849599, "percentage": 48.0, "elapsed_time": "1:05:58", "remaining_time": "1:11:29", "throughput": 5177.49, "total_tokens": 20495680} {"current_steps": 20365, "total_steps": 42420, "loss": 0.3225, "lr": 0.0006208651329655, "epoch": 9.601603017444601, "percentage": 48.01, "elapsed_time": "1:05:59", "remaining_time": "1:11:28", "throughput": 5177.46, "total_tokens": 20499872} {"current_steps": 20370, "total_steps": 42420, "loss": 0.365, "lr": 0.0006206655035909318, "epoch": 9.603960396039604, "percentage": 48.02, "elapsed_time": "1:06:00", "remaining_time": "1:11:27", "throughput": 5177.55, "total_tokens": 20505184} {"current_steps": 20375, "total_steps": 42420, "loss": 0.3199, "lr": 0.0006204658537897427, "epoch": 9.606317774634606, "percentage": 48.03, "elapsed_time": "1:06:01", "remaining_time": "1:11:26", "throughput": 5177.61, "total_tokens": 20510272} {"current_steps": 20380, "total_steps": 42420, "loss": 0.3388, "lr": 0.0006202661835957299, "epoch": 9.608675153229608, "percentage": 48.04, "elapsed_time": "1:06:02", "remaining_time": "1:11:25", "throughput": 5177.73, "total_tokens": 20516160} {"current_steps": 20385, "total_steps": 42420, "loss": 0.3745, "lr": 0.0006200664930426943, "epoch": 9.61103253182461, "percentage": 48.06, "elapsed_time": "1:06:03", "remaining_time": "1:11:24", "throughput": 5177.8, "total_tokens": 20521152} {"current_steps": 20390, "total_steps": 42420, "loss": 0.3215, "lr": 0.0006198667821644399, "epoch": 9.613389910419613, "percentage": 48.07, "elapsed_time": "1:06:04", "remaining_time": "1:11:22", "throughput": 5177.83, "total_tokens": 20525664} {"current_steps": 20395, "total_steps": 42420, "loss": 0.3821, "lr": 0.0006196670509947744, "epoch": 9.615747289014616, "percentage": 48.08, "elapsed_time": "1:06:05", "remaining_time": "1:11:21", "throughput": 5177.88, "total_tokens": 20530560} {"current_steps": 20400, "total_steps": 42420, "loss": 0.3333, "lr": 0.0006194672995675093, "epoch": 9.618104667609618, "percentage": 48.09, "elapsed_time": "1:06:05", "remaining_time": "1:11:20", "throughput": 5177.89, "total_tokens": 20535136} {"current_steps": 20405, "total_steps": 42420, "loss": 0.34, "lr": 0.0006192675279164587, "epoch": 9.62046204620462, "percentage": 48.1, "elapsed_time": "1:06:06", "remaining_time": "1:11:19", "throughput": 5177.9, "total_tokens": 20539520} {"current_steps": 20410, "total_steps": 42420, "loss": 0.343, "lr": 0.0006190677360754407, "epoch": 9.622819424799623, "percentage": 48.11, "elapsed_time": "1:06:07", "remaining_time": "1:11:18", "throughput": 5177.92, "total_tokens": 20544128} {"current_steps": 20415, "total_steps": 42420, "loss": 0.3583, "lr": 0.0006188679240782765, "epoch": 9.625176803394625, "percentage": 48.13, "elapsed_time": "1:06:08", "remaining_time": "1:11:17", "throughput": 5177.98, "total_tokens": 20549088} {"current_steps": 20420, "total_steps": 42420, "loss": 0.2929, "lr": 0.0006186680919587911, "epoch": 9.627534181989628, "percentage": 48.14, "elapsed_time": "1:06:09", "remaining_time": "1:11:16", "throughput": 5178.15, "total_tokens": 20555360} {"current_steps": 20425, "total_steps": 42420, "loss": 0.3529, "lr": 0.0006184682397508127, "epoch": 9.62989156058463, "percentage": 48.15, "elapsed_time": "1:06:10", "remaining_time": "1:11:15", "throughput": 5178.18, "total_tokens": 20560256} {"current_steps": 20430, "total_steps": 42420, "loss": 0.3761, "lr": 0.0006182683674881726, "epoch": 9.632248939179632, "percentage": 48.16, "elapsed_time": "1:06:11", "remaining_time": "1:11:14", "throughput": 5178.17, "total_tokens": 20564480} {"current_steps": 20435, "total_steps": 42420, "loss": 0.3231, "lr": 0.0006180684752047061, "epoch": 9.634606317774635, "percentage": 48.17, "elapsed_time": "1:06:12", "remaining_time": "1:11:13", "throughput": 5178.21, "total_tokens": 20569504} {"current_steps": 20440, "total_steps": 42420, "loss": 0.3339, "lr": 0.0006178685629342514, "epoch": 9.636963696369637, "percentage": 48.18, "elapsed_time": "1:06:13", "remaining_time": "1:11:12", "throughput": 5178.23, "total_tokens": 20573984} {"current_steps": 20445, "total_steps": 42420, "loss": 0.3336, "lr": 0.0006176686307106503, "epoch": 9.63932107496464, "percentage": 48.2, "elapsed_time": "1:06:14", "remaining_time": "1:11:11", "throughput": 5178.27, "total_tokens": 20578880} {"current_steps": 20450, "total_steps": 42420, "loss": 0.3004, "lr": 0.0006174686785677479, "epoch": 9.641678453559642, "percentage": 48.21, "elapsed_time": "1:06:14", "remaining_time": "1:11:10", "throughput": 5178.3, "total_tokens": 20583680} {"current_steps": 20455, "total_steps": 42420, "loss": 0.3899, "lr": 0.0006172687065393927, "epoch": 9.644035832154644, "percentage": 48.22, "elapsed_time": "1:06:15", "remaining_time": "1:11:09", "throughput": 5178.36, "total_tokens": 20588608} {"current_steps": 20460, "total_steps": 42420, "loss": 0.3678, "lr": 0.0006170687146594367, "epoch": 9.646393210749647, "percentage": 48.23, "elapsed_time": "1:06:16", "remaining_time": "1:11:08", "throughput": 5178.37, "total_tokens": 20592960} {"current_steps": 20465, "total_steps": 42420, "loss": 0.313, "lr": 0.0006168687029617348, "epoch": 9.64875058934465, "percentage": 48.24, "elapsed_time": "1:06:17", "remaining_time": "1:11:07", "throughput": 5178.43, "total_tokens": 20598048} {"current_steps": 20470, "total_steps": 42420, "loss": 0.3373, "lr": 0.0006166686714801459, "epoch": 9.651107967939652, "percentage": 48.26, "elapsed_time": "1:06:18", "remaining_time": "1:11:06", "throughput": 5178.62, "total_tokens": 20604384} {"current_steps": 20475, "total_steps": 42420, "loss": 0.3292, "lr": 0.0006164686202485319, "epoch": 9.653465346534654, "percentage": 48.27, "elapsed_time": "1:06:19", "remaining_time": "1:11:05", "throughput": 5178.75, "total_tokens": 20609920} {"current_steps": 20480, "total_steps": 42420, "loss": 0.3519, "lr": 0.0006162685493007582, "epoch": 9.655822725129656, "percentage": 48.28, "elapsed_time": "1:06:20", "remaining_time": "1:11:04", "throughput": 5178.78, "total_tokens": 20614624} {"current_steps": 20485, "total_steps": 42420, "loss": 0.3522, "lr": 0.000616068458670693, "epoch": 9.658180103724659, "percentage": 48.29, "elapsed_time": "1:06:21", "remaining_time": "1:11:03", "throughput": 5178.86, "total_tokens": 20619968} {"current_steps": 20490, "total_steps": 42420, "loss": 0.3494, "lr": 0.0006158683483922085, "epoch": 9.660537482319661, "percentage": 48.3, "elapsed_time": "1:06:22", "remaining_time": "1:11:02", "throughput": 5178.96, "total_tokens": 20625504} {"current_steps": 20495, "total_steps": 42420, "loss": 0.3669, "lr": 0.00061566821849918, "epoch": 9.662894860914664, "percentage": 48.31, "elapsed_time": "1:06:23", "remaining_time": "1:11:01", "throughput": 5179.14, "total_tokens": 20631904} {"current_steps": 20500, "total_steps": 42420, "loss": 0.3585, "lr": 0.0006154680690254859, "epoch": 9.665252239509666, "percentage": 48.33, "elapsed_time": "1:06:24", "remaining_time": "1:11:00", "throughput": 5179.23, "total_tokens": 20637024} {"current_steps": 20505, "total_steps": 42420, "loss": 0.3545, "lr": 0.0006152679000050084, "epoch": 9.667609618104667, "percentage": 48.34, "elapsed_time": "1:06:25", "remaining_time": "1:10:59", "throughput": 5179.31, "total_tokens": 20642368} {"current_steps": 20510, "total_steps": 42420, "loss": 0.3224, "lr": 0.0006150677114716325, "epoch": 9.66996699669967, "percentage": 48.35, "elapsed_time": "1:06:26", "remaining_time": "1:10:58", "throughput": 5179.46, "total_tokens": 20648320} {"current_steps": 20515, "total_steps": 42420, "loss": 0.3441, "lr": 0.0006148675034592468, "epoch": 9.672324375294671, "percentage": 48.36, "elapsed_time": "1:06:27", "remaining_time": "1:10:57", "throughput": 5179.59, "total_tokens": 20654048} {"current_steps": 20520, "total_steps": 42420, "loss": 0.2821, "lr": 0.0006146672760017431, "epoch": 9.674681753889674, "percentage": 48.37, "elapsed_time": "1:06:28", "remaining_time": "1:10:56", "throughput": 5179.68, "total_tokens": 20659200} {"current_steps": 20525, "total_steps": 42420, "loss": 0.2955, "lr": 0.0006144670291330166, "epoch": 9.677039132484676, "percentage": 48.39, "elapsed_time": "1:06:29", "remaining_time": "1:10:55", "throughput": 5179.71, "total_tokens": 20663936} {"current_steps": 20530, "total_steps": 42420, "loss": 0.4694, "lr": 0.0006142667628869654, "epoch": 9.679396511079679, "percentage": 48.4, "elapsed_time": "1:06:30", "remaining_time": "1:10:54", "throughput": 5179.76, "total_tokens": 20668992} {"current_steps": 20535, "total_steps": 42420, "loss": 0.3971, "lr": 0.0006140664772974916, "epoch": 9.681753889674681, "percentage": 48.41, "elapsed_time": "1:06:31", "remaining_time": "1:10:53", "throughput": 5179.88, "total_tokens": 20674656} {"current_steps": 20540, "total_steps": 42420, "loss": 0.3446, "lr": 0.0006138661723984998, "epoch": 9.684111268269683, "percentage": 48.42, "elapsed_time": "1:06:32", "remaining_time": "1:10:52", "throughput": 5180.06, "total_tokens": 20681184} {"current_steps": 20545, "total_steps": 42420, "loss": 0.3493, "lr": 0.0006136658482238985, "epoch": 9.686468646864686, "percentage": 48.43, "elapsed_time": "1:06:33", "remaining_time": "1:10:52", "throughput": 5180.23, "total_tokens": 20687360} {"current_steps": 20550, "total_steps": 42420, "loss": 0.3315, "lr": 0.0006134655048075989, "epoch": 9.688826025459688, "percentage": 48.44, "elapsed_time": "1:06:34", "remaining_time": "1:10:51", "throughput": 5180.34, "total_tokens": 20692896} {"current_steps": 20555, "total_steps": 42420, "loss": 0.3673, "lr": 0.0006132651421835158, "epoch": 9.69118340405469, "percentage": 48.46, "elapsed_time": "1:06:35", "remaining_time": "1:10:50", "throughput": 5180.41, "total_tokens": 20697856} {"current_steps": 20560, "total_steps": 42420, "loss": 0.3274, "lr": 0.0006130647603855673, "epoch": 9.693540782649693, "percentage": 48.47, "elapsed_time": "1:06:36", "remaining_time": "1:10:49", "throughput": 5180.49, "total_tokens": 20702880} {"current_steps": 20565, "total_steps": 42420, "loss": 0.3068, "lr": 0.0006128643594476748, "epoch": 9.695898161244696, "percentage": 48.48, "elapsed_time": "1:06:37", "remaining_time": "1:10:47", "throughput": 5180.56, "total_tokens": 20708000} {"current_steps": 20570, "total_steps": 42420, "loss": 0.3162, "lr": 0.0006126639394037624, "epoch": 9.698255539839698, "percentage": 48.49, "elapsed_time": "1:06:38", "remaining_time": "1:10:46", "throughput": 5180.57, "total_tokens": 20712608} {"current_steps": 20575, "total_steps": 42420, "loss": 0.2889, "lr": 0.000612463500287758, "epoch": 9.7006129184347, "percentage": 48.5, "elapsed_time": "1:06:39", "remaining_time": "1:10:45", "throughput": 5180.61, "total_tokens": 20717408} {"current_steps": 20580, "total_steps": 42420, "loss": 0.3521, "lr": 0.0006122630421335926, "epoch": 9.702970297029703, "percentage": 48.51, "elapsed_time": "1:06:39", "remaining_time": "1:10:44", "throughput": 5180.64, "total_tokens": 20722336} {"current_steps": 20585, "total_steps": 42420, "loss": 0.3956, "lr": 0.0006120625649752002, "epoch": 9.705327675624705, "percentage": 48.53, "elapsed_time": "1:06:41", "remaining_time": "1:10:43", "throughput": 5180.82, "total_tokens": 20728608} {"current_steps": 20590, "total_steps": 42420, "loss": 0.2885, "lr": 0.0006118620688465183, "epoch": 9.707685054219708, "percentage": 48.54, "elapsed_time": "1:06:41", "remaining_time": "1:10:42", "throughput": 5180.87, "total_tokens": 20733696} {"current_steps": 20595, "total_steps": 42420, "loss": 0.3299, "lr": 0.0006116615537814874, "epoch": 9.71004243281471, "percentage": 48.55, "elapsed_time": "1:06:42", "remaining_time": "1:10:42", "throughput": 5180.97, "total_tokens": 20739072} {"current_steps": 20600, "total_steps": 42420, "loss": 0.344, "lr": 0.0006114610198140514, "epoch": 9.712399811409712, "percentage": 48.56, "elapsed_time": "1:06:43", "remaining_time": "1:10:41", "throughput": 5181.08, "total_tokens": 20744736} {"current_steps": 20605, "total_steps": 42420, "loss": 0.3248, "lr": 0.0006112604669781572, "epoch": 9.714757190004715, "percentage": 48.57, "elapsed_time": "1:06:44", "remaining_time": "1:10:40", "throughput": 5181.16, "total_tokens": 20750016} {"current_steps": 20610, "total_steps": 42420, "loss": 0.3006, "lr": 0.0006110598953077552, "epoch": 9.717114568599717, "percentage": 48.59, "elapsed_time": "1:06:45", "remaining_time": "1:10:39", "throughput": 5181.22, "total_tokens": 20755008} {"current_steps": 20615, "total_steps": 42420, "loss": 0.2992, "lr": 0.0006108593048367984, "epoch": 9.71947194719472, "percentage": 48.6, "elapsed_time": "1:06:46", "remaining_time": "1:10:38", "throughput": 5181.3, "total_tokens": 20760320} {"current_steps": 20620, "total_steps": 42420, "loss": 0.3262, "lr": 0.0006106586955992437, "epoch": 9.721829325789722, "percentage": 48.61, "elapsed_time": "1:06:47", "remaining_time": "1:10:37", "throughput": 5181.39, "total_tokens": 20765984} {"current_steps": 20625, "total_steps": 42420, "loss": 0.328, "lr": 0.0006104580676290506, "epoch": 9.724186704384724, "percentage": 48.62, "elapsed_time": "1:06:48", "remaining_time": "1:10:36", "throughput": 5181.39, "total_tokens": 20770528} {"current_steps": 20630, "total_steps": 42420, "loss": 0.3993, "lr": 0.0006102574209601823, "epoch": 9.726544082979727, "percentage": 48.63, "elapsed_time": "1:06:49", "remaining_time": "1:10:35", "throughput": 5181.53, "total_tokens": 20776416} {"current_steps": 20635, "total_steps": 42420, "loss": 0.3062, "lr": 0.0006100567556266046, "epoch": 9.72890146157473, "percentage": 48.64, "elapsed_time": "1:06:50", "remaining_time": "1:10:34", "throughput": 5181.52, "total_tokens": 20780608} {"current_steps": 20640, "total_steps": 42420, "loss": 0.3423, "lr": 0.0006098560716622868, "epoch": 9.731258840169732, "percentage": 48.66, "elapsed_time": "1:06:51", "remaining_time": "1:10:32", "throughput": 5181.53, "total_tokens": 20785152} {"current_steps": 20645, "total_steps": 42420, "loss": 0.3416, "lr": 0.0006096553691012013, "epoch": 9.733616218764734, "percentage": 48.67, "elapsed_time": "1:06:52", "remaining_time": "1:10:31", "throughput": 5181.51, "total_tokens": 20789248} {"current_steps": 20650, "total_steps": 42420, "loss": 0.3776, "lr": 0.0006094546479773238, "epoch": 9.735973597359736, "percentage": 48.68, "elapsed_time": "1:06:53", "remaining_time": "1:10:31", "throughput": 5181.77, "total_tokens": 20797120} {"current_steps": 20655, "total_steps": 42420, "loss": 0.3447, "lr": 0.0006092539083246326, "epoch": 9.738330975954739, "percentage": 48.69, "elapsed_time": "1:06:54", "remaining_time": "1:10:30", "throughput": 5181.79, "total_tokens": 20801568} {"current_steps": 20660, "total_steps": 42420, "loss": 0.3587, "lr": 0.0006090531501771098, "epoch": 9.740688354549741, "percentage": 48.7, "elapsed_time": "1:06:55", "remaining_time": "1:10:29", "throughput": 5181.83, "total_tokens": 20806496} {"current_steps": 20665, "total_steps": 42420, "loss": 0.3377, "lr": 0.0006088523735687404, "epoch": 9.743045733144744, "percentage": 48.72, "elapsed_time": "1:06:56", "remaining_time": "1:10:27", "throughput": 5181.82, "total_tokens": 20810816} {"current_steps": 20670, "total_steps": 42420, "loss": 0.3511, "lr": 0.0006086515785335124, "epoch": 9.745403111739746, "percentage": 48.73, "elapsed_time": "1:06:57", "remaining_time": "1:10:26", "throughput": 5181.85, "total_tokens": 20815616} {"current_steps": 20675, "total_steps": 42420, "loss": 0.3361, "lr": 0.0006084507651054168, "epoch": 9.747760490334748, "percentage": 48.74, "elapsed_time": "1:06:57", "remaining_time": "1:10:25", "throughput": 5181.9, "total_tokens": 20820448} {"current_steps": 20680, "total_steps": 42420, "loss": 0.3426, "lr": 0.0006082499333184482, "epoch": 9.75011786892975, "percentage": 48.75, "elapsed_time": "1:06:58", "remaining_time": "1:10:24", "throughput": 5181.93, "total_tokens": 20825248} {"current_steps": 20685, "total_steps": 42420, "loss": 0.3323, "lr": 0.0006080490832066037, "epoch": 9.752475247524753, "percentage": 48.76, "elapsed_time": "1:06:59", "remaining_time": "1:10:23", "throughput": 5181.99, "total_tokens": 20830176} {"current_steps": 20690, "total_steps": 42420, "loss": 0.3099, "lr": 0.0006078482148038842, "epoch": 9.754832626119756, "percentage": 48.77, "elapsed_time": "1:07:00", "remaining_time": "1:10:22", "throughput": 5182.07, "total_tokens": 20835360} {"current_steps": 20695, "total_steps": 42420, "loss": 0.3248, "lr": 0.0006076473281442929, "epoch": 9.757190004714758, "percentage": 48.79, "elapsed_time": "1:07:01", "remaining_time": "1:10:21", "throughput": 5182.12, "total_tokens": 20840224} {"current_steps": 20700, "total_steps": 42420, "loss": 0.3322, "lr": 0.0006074464232618367, "epoch": 9.75954738330976, "percentage": 48.8, "elapsed_time": "1:07:02", "remaining_time": "1:10:20", "throughput": 5182.29, "total_tokens": 20846432} {"current_steps": 20705, "total_steps": 42420, "loss": 0.4327, "lr": 0.0006072455001905255, "epoch": 9.761904761904763, "percentage": 48.81, "elapsed_time": "1:07:03", "remaining_time": "1:10:19", "throughput": 5182.33, "total_tokens": 20851168} {"current_steps": 20710, "total_steps": 42420, "loss": 0.3018, "lr": 0.000607044558964372, "epoch": 9.764262140499763, "percentage": 48.82, "elapsed_time": "1:07:04", "remaining_time": "1:10:18", "throughput": 5182.38, "total_tokens": 20855936} {"current_steps": 20715, "total_steps": 42420, "loss": 0.2989, "lr": 0.0006068435996173923, "epoch": 9.766619519094768, "percentage": 48.83, "elapsed_time": "1:07:05", "remaining_time": "1:10:17", "throughput": 5182.38, "total_tokens": 20860224} {"current_steps": 20720, "total_steps": 42420, "loss": 0.2647, "lr": 0.0006066426221836051, "epoch": 9.768976897689768, "percentage": 48.84, "elapsed_time": "1:07:06", "remaining_time": "1:10:16", "throughput": 5182.41, "total_tokens": 20864992} {"current_steps": 20725, "total_steps": 42420, "loss": 0.3433, "lr": 0.0006064416266970327, "epoch": 9.77133427628477, "percentage": 48.86, "elapsed_time": "1:07:07", "remaining_time": "1:10:15", "throughput": 5182.52, "total_tokens": 20870784} {"current_steps": 20730, "total_steps": 42420, "loss": 0.3772, "lr": 0.0006062406131917002, "epoch": 9.773691654879773, "percentage": 48.87, "elapsed_time": "1:07:08", "remaining_time": "1:10:14", "throughput": 5182.72, "total_tokens": 20877440} {"current_steps": 20735, "total_steps": 42420, "loss": 0.3604, "lr": 0.0006060395817016358, "epoch": 9.776049033474775, "percentage": 48.88, "elapsed_time": "1:07:09", "remaining_time": "1:10:13", "throughput": 5182.74, "total_tokens": 20882240} {"current_steps": 20740, "total_steps": 42420, "loss": 0.2959, "lr": 0.0006058385322608705, "epoch": 9.778406412069778, "percentage": 48.89, "elapsed_time": "1:07:10", "remaining_time": "1:10:12", "throughput": 5182.86, "total_tokens": 20888096} {"current_steps": 20745, "total_steps": 42420, "loss": 0.2832, "lr": 0.0006056374649034389, "epoch": 9.78076379066478, "percentage": 48.9, "elapsed_time": "1:07:11", "remaining_time": "1:10:11", "throughput": 5182.85, "total_tokens": 20892192} {"current_steps": 20750, "total_steps": 42420, "loss": 0.3397, "lr": 0.000605436379663378, "epoch": 9.783121169259783, "percentage": 48.92, "elapsed_time": "1:07:11", "remaining_time": "1:10:10", "throughput": 5182.89, "total_tokens": 20896864} {"current_steps": 20755, "total_steps": 42420, "loss": 0.3539, "lr": 0.0006052352765747284, "epoch": 9.785478547854785, "percentage": 48.93, "elapsed_time": "1:07:12", "remaining_time": "1:10:09", "throughput": 5182.91, "total_tokens": 20901408} {"current_steps": 20760, "total_steps": 42420, "loss": 0.3421, "lr": 0.000605034155671533, "epoch": 9.787835926449787, "percentage": 48.94, "elapsed_time": "1:07:13", "remaining_time": "1:10:08", "throughput": 5183.0, "total_tokens": 20906496} {"current_steps": 20765, "total_steps": 42420, "loss": 0.3765, "lr": 0.0006048330169878386, "epoch": 9.79019330504479, "percentage": 48.95, "elapsed_time": "1:07:14", "remaining_time": "1:10:07", "throughput": 5183.14, "total_tokens": 20912320} {"current_steps": 20770, "total_steps": 42420, "loss": 0.3499, "lr": 0.0006046318605576942, "epoch": 9.792550683639792, "percentage": 48.96, "elapsed_time": "1:07:15", "remaining_time": "1:10:06", "throughput": 5183.21, "total_tokens": 20917536} {"current_steps": 20775, "total_steps": 42420, "loss": 0.3705, "lr": 0.0006044306864151527, "epoch": 9.794908062234795, "percentage": 48.97, "elapsed_time": "1:07:16", "remaining_time": "1:10:05", "throughput": 5183.38, "total_tokens": 20923648} {"current_steps": 20780, "total_steps": 42420, "loss": 0.3465, "lr": 0.0006042294945942688, "epoch": 9.797265440829797, "percentage": 48.99, "elapsed_time": "1:07:17", "remaining_time": "1:10:04", "throughput": 5183.43, "total_tokens": 20928576} {"current_steps": 20785, "total_steps": 42420, "loss": 0.3363, "lr": 0.0006040282851291014, "epoch": 9.7996228194248, "percentage": 49.0, "elapsed_time": "1:07:18", "remaining_time": "1:10:03", "throughput": 5183.42, "total_tokens": 20933024} {"current_steps": 20790, "total_steps": 42420, "loss": 0.3179, "lr": 0.0006038270580537114, "epoch": 9.801980198019802, "percentage": 49.01, "elapsed_time": "1:07:19", "remaining_time": "1:10:02", "throughput": 5183.5, "total_tokens": 20938336} {"current_steps": 20795, "total_steps": 42420, "loss": 0.3538, "lr": 0.0006036258134021637, "epoch": 9.804337576614804, "percentage": 49.02, "elapsed_time": "1:07:20", "remaining_time": "1:10:01", "throughput": 5183.54, "total_tokens": 20942944} {"current_steps": 20800, "total_steps": 42420, "loss": 0.3771, "lr": 0.0006034245512085249, "epoch": 9.806694955209807, "percentage": 49.03, "elapsed_time": "1:07:21", "remaining_time": "1:10:00", "throughput": 5183.55, "total_tokens": 20947488} {"current_steps": 20805, "total_steps": 42420, "loss": 0.4393, "lr": 0.0006032232715068658, "epoch": 9.809052333804809, "percentage": 49.05, "elapsed_time": "1:07:22", "remaining_time": "1:09:59", "throughput": 5183.63, "total_tokens": 20953024} {"current_steps": 20810, "total_steps": 42420, "loss": 0.3411, "lr": 0.0006030219743312594, "epoch": 9.811409712399811, "percentage": 49.06, "elapsed_time": "1:07:23", "remaining_time": "1:09:58", "throughput": 5183.65, "total_tokens": 20957504} {"current_steps": 20815, "total_steps": 42420, "loss": 0.3628, "lr": 0.000602820659715782, "epoch": 9.813767090994814, "percentage": 49.07, "elapsed_time": "1:07:24", "remaining_time": "1:09:57", "throughput": 5183.76, "total_tokens": 20963232} {"current_steps": 20820, "total_steps": 42420, "loss": 0.3432, "lr": 0.0006026193276945128, "epoch": 9.816124469589816, "percentage": 49.08, "elapsed_time": "1:07:24", "remaining_time": "1:09:56", "throughput": 5183.81, "total_tokens": 20968000} {"current_steps": 20825, "total_steps": 42420, "loss": 0.3221, "lr": 0.0006024179783015338, "epoch": 9.818481848184819, "percentage": 49.09, "elapsed_time": "1:07:25", "remaining_time": "1:09:55", "throughput": 5183.86, "total_tokens": 20972736} {"current_steps": 20830, "total_steps": 42420, "loss": 0.28, "lr": 0.0006022166115709299, "epoch": 9.820839226779821, "percentage": 49.1, "elapsed_time": "1:07:26", "remaining_time": "1:09:54", "throughput": 5183.96, "total_tokens": 20978112} {"current_steps": 20835, "total_steps": 42420, "loss": 0.3979, "lr": 0.0006020152275367891, "epoch": 9.823196605374823, "percentage": 49.12, "elapsed_time": "1:07:27", "remaining_time": "1:09:53", "throughput": 5183.97, "total_tokens": 20982752} {"current_steps": 20840, "total_steps": 42420, "loss": 0.4395, "lr": 0.0006018138262332027, "epoch": 9.825553983969826, "percentage": 49.13, "elapsed_time": "1:07:28", "remaining_time": "1:09:52", "throughput": 5184.05, "total_tokens": 20987872} {"current_steps": 20845, "total_steps": 42420, "loss": 0.3675, "lr": 0.0006016124076942639, "epoch": 9.827911362564828, "percentage": 49.14, "elapsed_time": "1:07:29", "remaining_time": "1:09:51", "throughput": 5184.05, "total_tokens": 20992384} {"current_steps": 20850, "total_steps": 42420, "loss": 0.3413, "lr": 0.00060141097195407, "epoch": 9.83026874115983, "percentage": 49.15, "elapsed_time": "1:07:30", "remaining_time": "1:09:50", "throughput": 5184.13, "total_tokens": 20997536} {"current_steps": 20855, "total_steps": 42420, "loss": 0.365, "lr": 0.0006012095190467203, "epoch": 9.832626119754833, "percentage": 49.16, "elapsed_time": "1:07:31", "remaining_time": "1:09:49", "throughput": 5184.13, "total_tokens": 21002112} {"current_steps": 20860, "total_steps": 42420, "loss": 0.3615, "lr": 0.0006010080490063175, "epoch": 9.834983498349835, "percentage": 49.17, "elapsed_time": "1:07:32", "remaining_time": "1:09:48", "throughput": 5184.35, "total_tokens": 21008832} {"current_steps": 20865, "total_steps": 42420, "loss": 0.3146, "lr": 0.0006008065618669671, "epoch": 9.837340876944838, "percentage": 49.19, "elapsed_time": "1:07:33", "remaining_time": "1:09:47", "throughput": 5184.41, "total_tokens": 21014112} {"current_steps": 20870, "total_steps": 42420, "loss": 0.3815, "lr": 0.0006006050576627772, "epoch": 9.83969825553984, "percentage": 49.2, "elapsed_time": "1:07:34", "remaining_time": "1:09:46", "throughput": 5184.42, "total_tokens": 21018592} {"current_steps": 20875, "total_steps": 42420, "loss": 0.3427, "lr": 0.0006004035364278593, "epoch": 9.842055634134843, "percentage": 49.21, "elapsed_time": "1:07:35", "remaining_time": "1:09:45", "throughput": 5184.43, "total_tokens": 21023136} {"current_steps": 20880, "total_steps": 42420, "loss": 0.3295, "lr": 0.0006002019981963274, "epoch": 9.844413012729845, "percentage": 49.22, "elapsed_time": "1:07:36", "remaining_time": "1:09:44", "throughput": 5184.5, "total_tokens": 21028448} {"current_steps": 20885, "total_steps": 42420, "loss": 0.3221, "lr": 0.0006000004430022986, "epoch": 9.846770391324847, "percentage": 49.23, "elapsed_time": "1:07:36", "remaining_time": "1:09:43", "throughput": 5184.57, "total_tokens": 21033696} {"current_steps": 20890, "total_steps": 42420, "loss": 0.3337, "lr": 0.0005997988708798927, "epoch": 9.84912776991985, "percentage": 49.25, "elapsed_time": "1:07:38", "remaining_time": "1:09:42", "throughput": 5184.74, "total_tokens": 21039904} {"current_steps": 20895, "total_steps": 42420, "loss": 0.3147, "lr": 0.0005995972818632324, "epoch": 9.851485148514852, "percentage": 49.26, "elapsed_time": "1:07:38", "remaining_time": "1:09:41", "throughput": 5184.84, "total_tokens": 21045152} {"current_steps": 20900, "total_steps": 42420, "loss": 0.3106, "lr": 0.0005993956759864434, "epoch": 9.853842527109855, "percentage": 49.27, "elapsed_time": "1:07:39", "remaining_time": "1:09:40", "throughput": 5184.87, "total_tokens": 21049696} {"current_steps": 20905, "total_steps": 42420, "loss": 0.3043, "lr": 0.000599194053283654, "epoch": 9.856199905704855, "percentage": 49.28, "elapsed_time": "1:07:40", "remaining_time": "1:09:39", "throughput": 5184.9, "total_tokens": 21054592} {"current_steps": 20910, "total_steps": 42420, "loss": 0.3295, "lr": 0.0005989924137889957, "epoch": 9.85855728429986, "percentage": 49.29, "elapsed_time": "1:07:41", "remaining_time": "1:09:38", "throughput": 5184.96, "total_tokens": 21059488} {"current_steps": 20915, "total_steps": 42420, "loss": 0.3092, "lr": 0.0005987907575366025, "epoch": 9.86091466289486, "percentage": 49.3, "elapsed_time": "1:07:42", "remaining_time": "1:09:37", "throughput": 5185.0, "total_tokens": 21064224} {"current_steps": 20920, "total_steps": 42420, "loss": 0.356, "lr": 0.0005985890845606114, "epoch": 9.863272041489862, "percentage": 49.32, "elapsed_time": "1:07:43", "remaining_time": "1:09:36", "throughput": 5185.07, "total_tokens": 21069216} {"current_steps": 20925, "total_steps": 42420, "loss": 0.3043, "lr": 0.0005983873948951624, "epoch": 9.865629420084865, "percentage": 49.33, "elapsed_time": "1:07:44", "remaining_time": "1:09:35", "throughput": 5185.16, "total_tokens": 21074688} {"current_steps": 20930, "total_steps": 42420, "loss": 0.3148, "lr": 0.0005981856885743977, "epoch": 9.867986798679867, "percentage": 49.34, "elapsed_time": "1:07:45", "remaining_time": "1:09:34", "throughput": 5185.2, "total_tokens": 21079520} {"current_steps": 20935, "total_steps": 42420, "loss": 0.3146, "lr": 0.0005979839656324633, "epoch": 9.87034417727487, "percentage": 49.35, "elapsed_time": "1:07:46", "remaining_time": "1:09:32", "throughput": 5185.21, "total_tokens": 21083904} {"current_steps": 20940, "total_steps": 42420, "loss": 0.3528, "lr": 0.000597782226103507, "epoch": 9.872701555869872, "percentage": 49.36, "elapsed_time": "1:07:47", "remaining_time": "1:09:31", "throughput": 5185.26, "total_tokens": 21088768} {"current_steps": 20945, "total_steps": 42420, "loss": 0.3613, "lr": 0.0005975804700216802, "epoch": 9.875058934464874, "percentage": 49.38, "elapsed_time": "1:07:47", "remaining_time": "1:09:30", "throughput": 5185.27, "total_tokens": 21093312} {"current_steps": 20950, "total_steps": 42420, "loss": 0.3016, "lr": 0.0005973786974211365, "epoch": 9.877416313059877, "percentage": 49.39, "elapsed_time": "1:07:48", "remaining_time": "1:09:29", "throughput": 5185.28, "total_tokens": 21097920} {"current_steps": 20955, "total_steps": 42420, "loss": 0.3413, "lr": 0.0005971769083360327, "epoch": 9.87977369165488, "percentage": 49.4, "elapsed_time": "1:07:49", "remaining_time": "1:09:28", "throughput": 5185.33, "total_tokens": 21102944} {"current_steps": 20960, "total_steps": 42420, "loss": 0.3395, "lr": 0.0005969751028005284, "epoch": 9.882131070249882, "percentage": 49.41, "elapsed_time": "1:07:50", "remaining_time": "1:09:27", "throughput": 5185.4, "total_tokens": 21108032} {"current_steps": 20965, "total_steps": 42420, "loss": 0.3335, "lr": 0.0005967732808487858, "epoch": 9.884488448844884, "percentage": 49.42, "elapsed_time": "1:07:51", "remaining_time": "1:09:26", "throughput": 5185.41, "total_tokens": 21112512} {"current_steps": 20970, "total_steps": 42420, "loss": 0.3542, "lr": 0.0005965714425149697, "epoch": 9.886845827439886, "percentage": 49.43, "elapsed_time": "1:07:52", "remaining_time": "1:09:25", "throughput": 5185.39, "total_tokens": 21116640} {"current_steps": 20975, "total_steps": 42420, "loss": 0.2726, "lr": 0.0005963695878332482, "epoch": 9.889203206034889, "percentage": 49.45, "elapsed_time": "1:07:53", "remaining_time": "1:09:24", "throughput": 5185.44, "total_tokens": 21121344} {"current_steps": 20980, "total_steps": 42420, "loss": 0.3237, "lr": 0.0005961677168377917, "epoch": 9.891560584629891, "percentage": 49.46, "elapsed_time": "1:07:54", "remaining_time": "1:09:23", "throughput": 5185.44, "total_tokens": 21125696} {"current_steps": 20985, "total_steps": 42420, "loss": 0.3273, "lr": 0.0005959658295627737, "epoch": 9.893917963224894, "percentage": 49.47, "elapsed_time": "1:07:54", "remaining_time": "1:09:22", "throughput": 5185.5, "total_tokens": 21130560} {"current_steps": 20990, "total_steps": 42420, "loss": 0.3658, "lr": 0.00059576392604237, "epoch": 9.896275341819896, "percentage": 49.48, "elapsed_time": "1:07:55", "remaining_time": "1:09:21", "throughput": 5185.56, "total_tokens": 21135840} {"current_steps": 20995, "total_steps": 42420, "loss": 0.3418, "lr": 0.0005955620063107595, "epoch": 9.898632720414899, "percentage": 49.49, "elapsed_time": "1:07:56", "remaining_time": "1:09:20", "throughput": 5185.66, "total_tokens": 21141120} {"current_steps": 21000, "total_steps": 42420, "loss": 0.3326, "lr": 0.000595360070402124, "epoch": 9.900990099009901, "percentage": 49.5, "elapsed_time": "1:07:57", "remaining_time": "1:09:19", "throughput": 5185.71, "total_tokens": 21146016} {"current_steps": 21005, "total_steps": 42420, "loss": 0.3823, "lr": 0.0005951581183506478, "epoch": 9.903347477604903, "percentage": 49.52, "elapsed_time": "1:07:59", "remaining_time": "1:09:18", "throughput": 5185.99, "total_tokens": 21154080} {"current_steps": 21010, "total_steps": 42420, "loss": 0.36, "lr": 0.0005949561501905179, "epoch": 9.905704856199906, "percentage": 49.53, "elapsed_time": "1:07:59", "remaining_time": "1:09:17", "throughput": 5186.03, "total_tokens": 21158848} {"current_steps": 21015, "total_steps": 42420, "loss": 0.3454, "lr": 0.0005947541659559238, "epoch": 9.908062234794908, "percentage": 49.54, "elapsed_time": "1:08:00", "remaining_time": "1:09:16", "throughput": 5186.16, "total_tokens": 21164704} {"current_steps": 21020, "total_steps": 42420, "loss": 0.3591, "lr": 0.0005945521656810583, "epoch": 9.91041961338991, "percentage": 49.55, "elapsed_time": "1:08:01", "remaining_time": "1:09:15", "throughput": 5186.15, "total_tokens": 21168928} {"current_steps": 21025, "total_steps": 42420, "loss": 0.3413, "lr": 0.0005943501494001166, "epoch": 9.912776991984913, "percentage": 49.56, "elapsed_time": "1:08:02", "remaining_time": "1:09:14", "throughput": 5186.2, "total_tokens": 21173760} {"current_steps": 21030, "total_steps": 42420, "loss": 0.3636, "lr": 0.0005941481171472967, "epoch": 9.915134370579915, "percentage": 49.58, "elapsed_time": "1:08:03", "remaining_time": "1:09:13", "throughput": 5186.47, "total_tokens": 21181248} {"current_steps": 21035, "total_steps": 42420, "loss": 0.3328, "lr": 0.000593946068956799, "epoch": 9.917491749174918, "percentage": 49.59, "elapsed_time": "1:08:05", "remaining_time": "1:09:13", "throughput": 5186.68, "total_tokens": 21188640} {"current_steps": 21040, "total_steps": 42420, "loss": 0.3192, "lr": 0.0005937440048628268, "epoch": 9.91984912776992, "percentage": 49.6, "elapsed_time": "1:08:06", "remaining_time": "1:09:12", "throughput": 5186.7, "total_tokens": 21193024} {"current_steps": 21045, "total_steps": 42420, "loss": 0.3229, "lr": 0.0005935419248995864, "epoch": 9.922206506364923, "percentage": 49.61, "elapsed_time": "1:08:06", "remaining_time": "1:09:11", "throughput": 5186.79, "total_tokens": 21198400} {"current_steps": 21050, "total_steps": 42420, "loss": 0.3987, "lr": 0.0005933398291012866, "epoch": 9.924563884959925, "percentage": 49.62, "elapsed_time": "1:08:08", "remaining_time": "1:09:10", "throughput": 5186.94, "total_tokens": 21204352} {"current_steps": 21055, "total_steps": 42420, "loss": 0.2888, "lr": 0.0005931377175021382, "epoch": 9.926921263554927, "percentage": 49.63, "elapsed_time": "1:08:08", "remaining_time": "1:09:09", "throughput": 5186.95, "total_tokens": 21208960} {"current_steps": 21060, "total_steps": 42420, "loss": 0.3593, "lr": 0.0005929355901363556, "epoch": 9.92927864214993, "percentage": 49.65, "elapsed_time": "1:08:09", "remaining_time": "1:09:07", "throughput": 5186.91, "total_tokens": 21213024} {"current_steps": 21065, "total_steps": 42420, "loss": 0.3529, "lr": 0.0005927334470381558, "epoch": 9.931636020744932, "percentage": 49.66, "elapsed_time": "1:08:10", "remaining_time": "1:09:06", "throughput": 5186.96, "total_tokens": 21217696} {"current_steps": 21070, "total_steps": 42420, "loss": 0.3535, "lr": 0.0005925312882417578, "epoch": 9.933993399339935, "percentage": 49.67, "elapsed_time": "1:08:11", "remaining_time": "1:09:05", "throughput": 5186.96, "total_tokens": 21222144} {"current_steps": 21075, "total_steps": 42420, "loss": 0.3497, "lr": 0.0005923291137813839, "epoch": 9.936350777934937, "percentage": 49.68, "elapsed_time": "1:08:12", "remaining_time": "1:09:04", "throughput": 5187.11, "total_tokens": 21228256} {"current_steps": 21080, "total_steps": 42420, "loss": 0.3543, "lr": 0.0005921269236912586, "epoch": 9.93870815652994, "percentage": 49.69, "elapsed_time": "1:08:13", "remaining_time": "1:09:03", "throughput": 5187.14, "total_tokens": 21232928} {"current_steps": 21085, "total_steps": 42420, "loss": 0.3102, "lr": 0.0005919247180056093, "epoch": 9.941065535124942, "percentage": 49.71, "elapsed_time": "1:08:14", "remaining_time": "1:09:03", "throughput": 5187.27, "total_tokens": 21239072} {"current_steps": 21090, "total_steps": 42420, "loss": 0.3662, "lr": 0.000591722496758666, "epoch": 9.943422913719944, "percentage": 49.72, "elapsed_time": "1:08:15", "remaining_time": "1:09:01", "throughput": 5187.33, "total_tokens": 21244160} {"current_steps": 21095, "total_steps": 42420, "loss": 0.3667, "lr": 0.0005915202599846615, "epoch": 9.945780292314947, "percentage": 49.73, "elapsed_time": "1:08:16", "remaining_time": "1:09:00", "throughput": 5187.36, "total_tokens": 21248992} {"current_steps": 21100, "total_steps": 42420, "loss": 0.346, "lr": 0.0005913180077178308, "epoch": 9.948137670909949, "percentage": 49.74, "elapsed_time": "1:08:17", "remaining_time": "1:09:00", "throughput": 5187.56, "total_tokens": 21256000} {"current_steps": 21105, "total_steps": 42420, "loss": 0.3574, "lr": 0.000591115739992412, "epoch": 9.950495049504951, "percentage": 49.75, "elapsed_time": "1:08:18", "remaining_time": "1:08:59", "throughput": 5187.6, "total_tokens": 21260640} {"current_steps": 21110, "total_steps": 42420, "loss": 0.3231, "lr": 0.0005909134568426454, "epoch": 9.952852428099952, "percentage": 49.76, "elapsed_time": "1:08:19", "remaining_time": "1:08:58", "throughput": 5187.62, "total_tokens": 21265216} {"current_steps": 21115, "total_steps": 42420, "loss": 0.3764, "lr": 0.0005907111583027743, "epoch": 9.955209806694956, "percentage": 49.78, "elapsed_time": "1:08:20", "remaining_time": "1:08:57", "throughput": 5187.65, "total_tokens": 21269920} {"current_steps": 21120, "total_steps": 42420, "loss": 0.2906, "lr": 0.0005905088444070441, "epoch": 9.957567185289957, "percentage": 49.79, "elapsed_time": "1:08:20", "remaining_time": "1:08:55", "throughput": 5187.62, "total_tokens": 21274016} {"current_steps": 21125, "total_steps": 42420, "loss": 0.3503, "lr": 0.0005903065151897034, "epoch": 9.95992456388496, "percentage": 49.8, "elapsed_time": "1:08:21", "remaining_time": "1:08:54", "throughput": 5187.63, "total_tokens": 21278400} {"current_steps": 21130, "total_steps": 42420, "loss": 0.3257, "lr": 0.000590104170685003, "epoch": 9.962281942479962, "percentage": 49.81, "elapsed_time": "1:08:22", "remaining_time": "1:08:53", "throughput": 5187.65, "total_tokens": 21283072} {"current_steps": 21135, "total_steps": 42420, "loss": 0.3461, "lr": 0.0005899018109271963, "epoch": 9.964639321074964, "percentage": 49.82, "elapsed_time": "1:08:23", "remaining_time": "1:08:52", "throughput": 5187.66, "total_tokens": 21287744} {"current_steps": 21140, "total_steps": 42420, "loss": 0.298, "lr": 0.0005896994359505394, "epoch": 9.966996699669966, "percentage": 49.83, "elapsed_time": "1:08:24", "remaining_time": "1:08:51", "throughput": 5187.7, "total_tokens": 21292512} {"current_steps": 21145, "total_steps": 42420, "loss": 0.2899, "lr": 0.0005894970457892912, "epoch": 9.969354078264969, "percentage": 49.85, "elapsed_time": "1:08:25", "remaining_time": "1:08:50", "throughput": 5187.71, "total_tokens": 21297184} {"current_steps": 21150, "total_steps": 42420, "loss": 0.3073, "lr": 0.0005892946404777126, "epoch": 9.971711456859971, "percentage": 49.86, "elapsed_time": "1:08:26", "remaining_time": "1:08:49", "throughput": 5187.75, "total_tokens": 21301920} {"current_steps": 21155, "total_steps": 42420, "loss": 0.3451, "lr": 0.0005890922200500676, "epoch": 9.974068835454974, "percentage": 49.87, "elapsed_time": "1:08:27", "remaining_time": "1:08:48", "throughput": 5187.89, "total_tokens": 21308160} {"current_steps": 21160, "total_steps": 42420, "loss": 0.3962, "lr": 0.0005888897845406225, "epoch": 9.976426214049976, "percentage": 49.88, "elapsed_time": "1:08:28", "remaining_time": "1:08:47", "throughput": 5188.0, "total_tokens": 21313760} {"current_steps": 21165, "total_steps": 42420, "loss": 0.3703, "lr": 0.0005886873339836459, "epoch": 9.978783592644978, "percentage": 49.89, "elapsed_time": "1:08:29", "remaining_time": "1:08:46", "throughput": 5188.11, "total_tokens": 21319680} {"current_steps": 21170, "total_steps": 42420, "loss": 0.314, "lr": 0.0005884848684134097, "epoch": 9.98114097123998, "percentage": 49.91, "elapsed_time": "1:08:30", "remaining_time": "1:08:45", "throughput": 5188.11, "total_tokens": 21324064} {"current_steps": 21175, "total_steps": 42420, "loss": 0.3841, "lr": 0.0005882823878641876, "epoch": 9.983498349834983, "percentage": 49.92, "elapsed_time": "1:08:31", "remaining_time": "1:08:44", "throughput": 5188.13, "total_tokens": 21328768} {"current_steps": 21180, "total_steps": 42420, "loss": 0.3599, "lr": 0.0005880798923702561, "epoch": 9.985855728429986, "percentage": 49.93, "elapsed_time": "1:08:32", "remaining_time": "1:08:43", "throughput": 5188.19, "total_tokens": 21333856} {"current_steps": 21185, "total_steps": 42420, "loss": 0.3964, "lr": 0.0005878773819658943, "epoch": 9.988213107024988, "percentage": 49.94, "elapsed_time": "1:08:32", "remaining_time": "1:08:42", "throughput": 5188.19, "total_tokens": 21338368} {"current_steps": 21190, "total_steps": 42420, "loss": 0.3697, "lr": 0.0005876748566853839, "epoch": 9.99057048561999, "percentage": 49.95, "elapsed_time": "1:08:33", "remaining_time": "1:08:41", "throughput": 5188.21, "total_tokens": 21343136} {"current_steps": 21195, "total_steps": 42420, "loss": 0.3824, "lr": 0.000587472316563009, "epoch": 9.992927864214993, "percentage": 49.96, "elapsed_time": "1:08:34", "remaining_time": "1:08:40", "throughput": 5188.28, "total_tokens": 21348224} {"current_steps": 21200, "total_steps": 42420, "loss": 0.366, "lr": 0.000587269761633056, "epoch": 9.995285242809995, "percentage": 49.98, "elapsed_time": "1:08:35", "remaining_time": "1:08:39", "throughput": 5188.42, "total_tokens": 21354112} {"current_steps": 21205, "total_steps": 42420, "loss": 0.3805, "lr": 0.0005870671919298141, "epoch": 9.997642621404998, "percentage": 49.99, "elapsed_time": "1:08:36", "remaining_time": "1:08:38", "throughput": 5188.39, "total_tokens": 21358400} {"current_steps": 21210, "total_steps": 42420, "loss": 0.3382, "lr": 0.0005868646074875749, "epoch": 10.0, "percentage": 50.0, "elapsed_time": "1:08:37", "remaining_time": "1:08:37", "throughput": 5188.3, "total_tokens": 21363392} {"current_steps": 21210, "total_steps": 42420, "eval_loss": 0.3546566963195801, "epoch": 10.0, "percentage": 50.0, "elapsed_time": "1:08:56", "remaining_time": "1:08:56", "throughput": 5164.48, "total_tokens": 21363392} {"current_steps": 21215, "total_steps": 42420, "loss": 0.3706, "lr": 0.0005866620083406325, "epoch": 10.002357378595002, "percentage": 50.01, "elapsed_time": "1:08:59", "remaining_time": "1:08:57", "throughput": 5162.06, "total_tokens": 21368800} {"current_steps": 21220, "total_steps": 42420, "loss": 0.3491, "lr": 0.0005864593945232836, "epoch": 10.004714757190005, "percentage": 50.02, "elapsed_time": "1:09:00", "remaining_time": "1:08:56", "throughput": 5162.16, "total_tokens": 21373984} {"current_steps": 21225, "total_steps": 42420, "loss": 0.2953, "lr": 0.0005862567660698271, "epoch": 10.007072135785007, "percentage": 50.04, "elapsed_time": "1:09:01", "remaining_time": "1:08:55", "throughput": 5162.14, "total_tokens": 21378016} {"current_steps": 21230, "total_steps": 42420, "loss": 0.3703, "lr": 0.0005860541230145648, "epoch": 10.00942951438001, "percentage": 50.05, "elapsed_time": "1:09:02", "remaining_time": "1:08:54", "throughput": 5162.34, "total_tokens": 21384768} {"current_steps": 21235, "total_steps": 42420, "loss": 0.3091, "lr": 0.0005858514653918006, "epoch": 10.011786892975012, "percentage": 50.06, "elapsed_time": "1:09:03", "remaining_time": "1:08:53", "throughput": 5162.35, "total_tokens": 21389152} {"current_steps": 21240, "total_steps": 42420, "loss": 0.2723, "lr": 0.000585648793235841, "epoch": 10.014144271570014, "percentage": 50.07, "elapsed_time": "1:09:04", "remaining_time": "1:08:52", "throughput": 5162.36, "total_tokens": 21393536} {"current_steps": 21245, "total_steps": 42420, "loss": 0.3263, "lr": 0.0005854461065809948, "epoch": 10.016501650165017, "percentage": 50.08, "elapsed_time": "1:09:05", "remaining_time": "1:08:51", "throughput": 5162.39, "total_tokens": 21398208} {"current_steps": 21250, "total_steps": 42420, "loss": 0.4017, "lr": 0.0005852434054615735, "epoch": 10.01885902876002, "percentage": 50.09, "elapsed_time": "1:09:05", "remaining_time": "1:08:50", "throughput": 5162.33, "total_tokens": 21401888} {"current_steps": 21255, "total_steps": 42420, "loss": 0.3233, "lr": 0.0005850406899118912, "epoch": 10.021216407355022, "percentage": 50.11, "elapsed_time": "1:09:06", "remaining_time": "1:08:49", "throughput": 5162.35, "total_tokens": 21406592} {"current_steps": 21260, "total_steps": 42420, "loss": 0.3834, "lr": 0.0005848379599662639, "epoch": 10.023573785950024, "percentage": 50.12, "elapsed_time": "1:09:07", "remaining_time": "1:08:48", "throughput": 5162.39, "total_tokens": 21411488} {"current_steps": 21265, "total_steps": 42420, "loss": 0.3507, "lr": 0.0005846352156590104, "epoch": 10.025931164545026, "percentage": 50.13, "elapsed_time": "1:09:08", "remaining_time": "1:08:47", "throughput": 5162.45, "total_tokens": 21416480} {"current_steps": 21270, "total_steps": 42420, "loss": 0.3451, "lr": 0.000584432457024452, "epoch": 10.028288543140029, "percentage": 50.14, "elapsed_time": "1:09:09", "remaining_time": "1:08:46", "throughput": 5162.54, "total_tokens": 21421888} {"current_steps": 21275, "total_steps": 42420, "loss": 0.3482, "lr": 0.0005842296840969124, "epoch": 10.030645921735031, "percentage": 50.15, "elapsed_time": "1:09:10", "remaining_time": "1:08:44", "throughput": 5162.57, "total_tokens": 21426272} {"current_steps": 21280, "total_steps": 42420, "loss": 0.3618, "lr": 0.0005840268969107172, "epoch": 10.033003300330034, "percentage": 50.17, "elapsed_time": "1:09:11", "remaining_time": "1:08:43", "throughput": 5162.62, "total_tokens": 21431072} {"current_steps": 21285, "total_steps": 42420, "loss": 0.3357, "lr": 0.000583824095500195, "epoch": 10.035360678925036, "percentage": 50.18, "elapsed_time": "1:09:12", "remaining_time": "1:08:42", "throughput": 5162.69, "total_tokens": 21436384} {"current_steps": 21290, "total_steps": 42420, "loss": 0.3688, "lr": 0.0005836212798996768, "epoch": 10.037718057520038, "percentage": 50.19, "elapsed_time": "1:09:13", "remaining_time": "1:08:41", "throughput": 5162.76, "total_tokens": 21441504} {"current_steps": 21295, "total_steps": 42420, "loss": 0.3023, "lr": 0.0005834184501434955, "epoch": 10.04007543611504, "percentage": 50.2, "elapsed_time": "1:09:14", "remaining_time": "1:08:40", "throughput": 5162.81, "total_tokens": 21446368} {"current_steps": 21300, "total_steps": 42420, "loss": 0.2735, "lr": 0.0005832156062659873, "epoch": 10.042432814710043, "percentage": 50.21, "elapsed_time": "1:09:14", "remaining_time": "1:08:39", "throughput": 5162.89, "total_tokens": 21451392} {"current_steps": 21305, "total_steps": 42420, "loss": 0.3223, "lr": 0.0005830127483014895, "epoch": 10.044790193305046, "percentage": 50.22, "elapsed_time": "1:09:15", "remaining_time": "1:08:38", "throughput": 5162.96, "total_tokens": 21456288} {"current_steps": 21310, "total_steps": 42420, "loss": 0.3145, "lr": 0.0005828098762843432, "epoch": 10.047147571900048, "percentage": 50.24, "elapsed_time": "1:09:16", "remaining_time": "1:08:37", "throughput": 5163.07, "total_tokens": 21461888} {"current_steps": 21315, "total_steps": 42420, "loss": 0.3183, "lr": 0.0005826069902488909, "epoch": 10.049504950495049, "percentage": 50.25, "elapsed_time": "1:09:17", "remaining_time": "1:08:36", "throughput": 5163.2, "total_tokens": 21468000} {"current_steps": 21320, "total_steps": 42420, "loss": 0.369, "lr": 0.0005824040902294774, "epoch": 10.051862329090051, "percentage": 50.26, "elapsed_time": "1:09:18", "remaining_time": "1:08:35", "throughput": 5163.38, "total_tokens": 21474016} {"current_steps": 21325, "total_steps": 42420, "loss": 0.3263, "lr": 0.0005822011762604508, "epoch": 10.054219707685053, "percentage": 50.27, "elapsed_time": "1:09:19", "remaining_time": "1:08:34", "throughput": 5163.44, "total_tokens": 21479072} {"current_steps": 21330, "total_steps": 42420, "loss": 0.3486, "lr": 0.0005819982483761607, "epoch": 10.056577086280056, "percentage": 50.28, "elapsed_time": "1:09:20", "remaining_time": "1:08:34", "throughput": 5163.59, "total_tokens": 21485280} {"current_steps": 21335, "total_steps": 42420, "loss": 0.3236, "lr": 0.0005817953066109595, "epoch": 10.058934464875058, "percentage": 50.29, "elapsed_time": "1:09:21", "remaining_time": "1:08:33", "throughput": 5163.68, "total_tokens": 21490592} {"current_steps": 21340, "total_steps": 42420, "loss": 0.3226, "lr": 0.0005815923509992016, "epoch": 10.06129184347006, "percentage": 50.31, "elapsed_time": "1:09:22", "remaining_time": "1:08:32", "throughput": 5163.74, "total_tokens": 21495552} {"current_steps": 21345, "total_steps": 42420, "loss": 0.3173, "lr": 0.000581389381575244, "epoch": 10.063649222065063, "percentage": 50.32, "elapsed_time": "1:09:23", "remaining_time": "1:08:31", "throughput": 5163.92, "total_tokens": 21501952} {"current_steps": 21350, "total_steps": 42420, "loss": 0.3022, "lr": 0.000581186398373446, "epoch": 10.066006600660065, "percentage": 50.33, "elapsed_time": "1:09:24", "remaining_time": "1:08:30", "throughput": 5163.95, "total_tokens": 21506752} {"current_steps": 21355, "total_steps": 42420, "loss": 0.3592, "lr": 0.0005809834014281693, "epoch": 10.068363979255068, "percentage": 50.34, "elapsed_time": "1:09:25", "remaining_time": "1:08:29", "throughput": 5164.0, "total_tokens": 21511424} {"current_steps": 21360, "total_steps": 42420, "loss": 0.293, "lr": 0.0005807803907737778, "epoch": 10.07072135785007, "percentage": 50.35, "elapsed_time": "1:09:26", "remaining_time": "1:08:28", "throughput": 5164.03, "total_tokens": 21516096} {"current_steps": 21365, "total_steps": 42420, "loss": 0.2688, "lr": 0.0005805773664446376, "epoch": 10.073078736445073, "percentage": 50.37, "elapsed_time": "1:09:27", "remaining_time": "1:08:26", "throughput": 5164.06, "total_tokens": 21520864} {"current_steps": 21370, "total_steps": 42420, "loss": 0.3183, "lr": 0.0005803743284751175, "epoch": 10.075436115040075, "percentage": 50.38, "elapsed_time": "1:09:28", "remaining_time": "1:08:25", "throughput": 5164.14, "total_tokens": 21526272} {"current_steps": 21375, "total_steps": 42420, "loss": 0.4255, "lr": 0.0005801712768995882, "epoch": 10.077793493635077, "percentage": 50.39, "elapsed_time": "1:09:29", "remaining_time": "1:08:25", "throughput": 5164.38, "total_tokens": 21533216} {"current_steps": 21380, "total_steps": 42420, "loss": 0.3535, "lr": 0.0005799682117524229, "epoch": 10.08015087223008, "percentage": 50.4, "elapsed_time": "1:09:30", "remaining_time": "1:08:24", "throughput": 5164.38, "total_tokens": 21537504} {"current_steps": 21385, "total_steps": 42420, "loss": 0.3123, "lr": 0.0005797651330679974, "epoch": 10.082508250825082, "percentage": 50.41, "elapsed_time": "1:09:31", "remaining_time": "1:08:23", "throughput": 5164.47, "total_tokens": 21542656} {"current_steps": 21390, "total_steps": 42420, "loss": 0.3582, "lr": 0.000579562040880689, "epoch": 10.084865629420085, "percentage": 50.42, "elapsed_time": "1:09:32", "remaining_time": "1:08:21", "throughput": 5164.49, "total_tokens": 21547392} {"current_steps": 21395, "total_steps": 42420, "loss": 0.3512, "lr": 0.0005793589352248779, "epoch": 10.087223008015087, "percentage": 50.44, "elapsed_time": "1:09:33", "remaining_time": "1:08:20", "throughput": 5164.53, "total_tokens": 21552256} {"current_steps": 21400, "total_steps": 42420, "loss": 0.3264, "lr": 0.0005791558161349468, "epoch": 10.08958038661009, "percentage": 50.45, "elapsed_time": "1:09:33", "remaining_time": "1:08:19", "throughput": 5164.53, "total_tokens": 21556480} {"current_steps": 21405, "total_steps": 42420, "loss": 0.3419, "lr": 0.0005789526836452797, "epoch": 10.091937765205092, "percentage": 50.46, "elapsed_time": "1:09:34", "remaining_time": "1:08:18", "throughput": 5164.59, "total_tokens": 21561376} {"current_steps": 21410, "total_steps": 42420, "loss": 0.3574, "lr": 0.0005787495377902639, "epoch": 10.094295143800094, "percentage": 50.47, "elapsed_time": "1:09:35", "remaining_time": "1:08:17", "throughput": 5164.61, "total_tokens": 21566176} {"current_steps": 21415, "total_steps": 42420, "loss": 0.3317, "lr": 0.0005785463786042883, "epoch": 10.096652522395097, "percentage": 50.48, "elapsed_time": "1:09:36", "remaining_time": "1:08:16", "throughput": 5164.73, "total_tokens": 21571776} {"current_steps": 21420, "total_steps": 42420, "loss": 0.369, "lr": 0.0005783432061217446, "epoch": 10.099009900990099, "percentage": 50.5, "elapsed_time": "1:09:37", "remaining_time": "1:08:15", "throughput": 5164.76, "total_tokens": 21576448} {"current_steps": 21425, "total_steps": 42420, "loss": 0.2942, "lr": 0.0005781400203770263, "epoch": 10.101367279585101, "percentage": 50.51, "elapsed_time": "1:09:38", "remaining_time": "1:08:14", "throughput": 5164.85, "total_tokens": 21581952} {"current_steps": 21430, "total_steps": 42420, "loss": 0.3181, "lr": 0.0005779368214045293, "epoch": 10.103724658180104, "percentage": 50.52, "elapsed_time": "1:09:39", "remaining_time": "1:08:13", "throughput": 5164.92, "total_tokens": 21587072} {"current_steps": 21435, "total_steps": 42420, "loss": 0.3541, "lr": 0.0005777336092386516, "epoch": 10.106082036775106, "percentage": 50.53, "elapsed_time": "1:09:40", "remaining_time": "1:08:12", "throughput": 5165.01, "total_tokens": 21592544} {"current_steps": 21440, "total_steps": 42420, "loss": 0.2961, "lr": 0.0005775303839137939, "epoch": 10.108439415370109, "percentage": 50.54, "elapsed_time": "1:09:41", "remaining_time": "1:08:11", "throughput": 5165.05, "total_tokens": 21597312} {"current_steps": 21445, "total_steps": 42420, "loss": 0.298, "lr": 0.0005773271454643585, "epoch": 10.110796793965111, "percentage": 50.55, "elapsed_time": "1:09:42", "remaining_time": "1:08:10", "throughput": 5165.04, "total_tokens": 21601760} {"current_steps": 21450, "total_steps": 42420, "loss": 0.3707, "lr": 0.0005771238939247503, "epoch": 10.113154172560114, "percentage": 50.57, "elapsed_time": "1:09:43", "remaining_time": "1:08:09", "throughput": 5165.09, "total_tokens": 21606528} {"current_steps": 21455, "total_steps": 42420, "loss": 0.3246, "lr": 0.0005769206293293762, "epoch": 10.115511551155116, "percentage": 50.58, "elapsed_time": "1:09:44", "remaining_time": "1:08:08", "throughput": 5165.07, "total_tokens": 21610944} {"current_steps": 21460, "total_steps": 42420, "loss": 0.3377, "lr": 0.0005767173517126458, "epoch": 10.117868929750118, "percentage": 50.59, "elapsed_time": "1:09:44", "remaining_time": "1:08:07", "throughput": 5165.08, "total_tokens": 21615392} {"current_steps": 21465, "total_steps": 42420, "loss": 0.3271, "lr": 0.0005765140611089703, "epoch": 10.12022630834512, "percentage": 50.6, "elapsed_time": "1:09:45", "remaining_time": "1:08:06", "throughput": 5165.15, "total_tokens": 21620256} {"current_steps": 21470, "total_steps": 42420, "loss": 0.3578, "lr": 0.0005763107575527634, "epoch": 10.122583686940123, "percentage": 50.61, "elapsed_time": "1:09:46", "remaining_time": "1:08:05", "throughput": 5165.24, "total_tokens": 21625536} {"current_steps": 21475, "total_steps": 42420, "loss": 0.3151, "lr": 0.0005761074410784408, "epoch": 10.124941065535126, "percentage": 50.62, "elapsed_time": "1:09:47", "remaining_time": "1:08:04", "throughput": 5165.33, "total_tokens": 21630784} {"current_steps": 21480, "total_steps": 42420, "loss": 0.3033, "lr": 0.0005759041117204209, "epoch": 10.127298444130128, "percentage": 50.64, "elapsed_time": "1:09:48", "remaining_time": "1:08:03", "throughput": 5165.45, "total_tokens": 21636320} {"current_steps": 21485, "total_steps": 42420, "loss": 0.3055, "lr": 0.0005757007695131234, "epoch": 10.12965582272513, "percentage": 50.65, "elapsed_time": "1:09:49", "remaining_time": "1:08:02", "throughput": 5165.46, "total_tokens": 21640896} {"current_steps": 21490, "total_steps": 42420, "loss": 0.3252, "lr": 0.0005754974144909712, "epoch": 10.132013201320133, "percentage": 50.66, "elapsed_time": "1:09:50", "remaining_time": "1:08:01", "throughput": 5165.55, "total_tokens": 21646208} {"current_steps": 21495, "total_steps": 42420, "loss": 0.379, "lr": 0.0005752940466883884, "epoch": 10.134370579915135, "percentage": 50.67, "elapsed_time": "1:09:51", "remaining_time": "1:08:00", "throughput": 5165.65, "total_tokens": 21652000} {"current_steps": 21500, "total_steps": 42420, "loss": 0.3723, "lr": 0.000575090666139802, "epoch": 10.136727958510138, "percentage": 50.68, "elapsed_time": "1:09:52", "remaining_time": "1:07:59", "throughput": 5165.65, "total_tokens": 21656320} {"current_steps": 21505, "total_steps": 42420, "loss": 0.3395, "lr": 0.0005748872728796408, "epoch": 10.13908533710514, "percentage": 50.7, "elapsed_time": "1:09:53", "remaining_time": "1:07:58", "throughput": 5165.85, "total_tokens": 21662752} {"current_steps": 21510, "total_steps": 42420, "loss": 0.3065, "lr": 0.0005746838669423358, "epoch": 10.14144271570014, "percentage": 50.71, "elapsed_time": "1:09:54", "remaining_time": "1:07:57", "throughput": 5165.86, "total_tokens": 21667104} {"current_steps": 21515, "total_steps": 42420, "loss": 0.3104, "lr": 0.0005744804483623202, "epoch": 10.143800094295143, "percentage": 50.72, "elapsed_time": "1:09:55", "remaining_time": "1:07:56", "throughput": 5165.88, "total_tokens": 21671488} {"current_steps": 21520, "total_steps": 42420, "loss": 0.3178, "lr": 0.0005742770171740292, "epoch": 10.146157472890145, "percentage": 50.73, "elapsed_time": "1:09:55", "remaining_time": "1:07:55", "throughput": 5165.93, "total_tokens": 21676256} {"current_steps": 21525, "total_steps": 42420, "loss": 0.3283, "lr": 0.0005740735734119005, "epoch": 10.148514851485148, "percentage": 50.74, "elapsed_time": "1:09:56", "remaining_time": "1:07:54", "throughput": 5165.99, "total_tokens": 21681440} {"current_steps": 21530, "total_steps": 42420, "loss": 0.354, "lr": 0.0005738701171103734, "epoch": 10.15087223008015, "percentage": 50.75, "elapsed_time": "1:09:57", "remaining_time": "1:07:53", "throughput": 5166.09, "total_tokens": 21686720} {"current_steps": 21535, "total_steps": 42420, "loss": 0.3293, "lr": 0.0005736666483038898, "epoch": 10.153229608675153, "percentage": 50.77, "elapsed_time": "1:09:58", "remaining_time": "1:07:52", "throughput": 5166.09, "total_tokens": 21691136} {"current_steps": 21540, "total_steps": 42420, "loss": 0.3618, "lr": 0.0005734631670268933, "epoch": 10.155586987270155, "percentage": 50.78, "elapsed_time": "1:09:59", "remaining_time": "1:07:51", "throughput": 5166.19, "total_tokens": 21696736} {"current_steps": 21545, "total_steps": 42420, "loss": 0.3151, "lr": 0.0005732596733138302, "epoch": 10.157944365865157, "percentage": 50.79, "elapsed_time": "1:10:00", "remaining_time": "1:07:50", "throughput": 5166.25, "total_tokens": 21701728} {"current_steps": 21550, "total_steps": 42420, "loss": 0.3575, "lr": 0.000573056167199148, "epoch": 10.16030174446016, "percentage": 50.8, "elapsed_time": "1:10:01", "remaining_time": "1:07:49", "throughput": 5166.38, "total_tokens": 21707264} {"current_steps": 21555, "total_steps": 42420, "loss": 0.3431, "lr": 0.0005728526487172972, "epoch": 10.162659123055162, "percentage": 50.81, "elapsed_time": "1:10:02", "remaining_time": "1:07:48", "throughput": 5166.51, "total_tokens": 21712896} {"current_steps": 21560, "total_steps": 42420, "loss": 0.3679, "lr": 0.00057264911790273, "epoch": 10.165016501650165, "percentage": 50.83, "elapsed_time": "1:10:03", "remaining_time": "1:07:47", "throughput": 5166.63, "total_tokens": 21718272} {"current_steps": 21565, "total_steps": 42420, "loss": 0.3303, "lr": 0.0005724455747899007, "epoch": 10.167373880245167, "percentage": 50.84, "elapsed_time": "1:10:04", "remaining_time": "1:07:46", "throughput": 5166.82, "total_tokens": 21724832} {"current_steps": 21570, "total_steps": 42420, "loss": 0.3543, "lr": 0.0005722420194132656, "epoch": 10.16973125884017, "percentage": 50.85, "elapsed_time": "1:10:05", "remaining_time": "1:07:45", "throughput": 5166.85, "total_tokens": 21729504} {"current_steps": 21575, "total_steps": 42420, "loss": 0.3211, "lr": 0.0005720384518072831, "epoch": 10.172088637435172, "percentage": 50.86, "elapsed_time": "1:10:06", "remaining_time": "1:07:44", "throughput": 5166.91, "total_tokens": 21734496} {"current_steps": 21580, "total_steps": 42420, "loss": 0.3586, "lr": 0.0005718348720064139, "epoch": 10.174446016030174, "percentage": 50.87, "elapsed_time": "1:10:07", "remaining_time": "1:07:43", "throughput": 5166.98, "total_tokens": 21739520} {"current_steps": 21585, "total_steps": 42420, "loss": 0.2905, "lr": 0.0005716312800451206, "epoch": 10.176803394625177, "percentage": 50.88, "elapsed_time": "1:10:08", "remaining_time": "1:07:42", "throughput": 5167.05, "total_tokens": 21744416} {"current_steps": 21590, "total_steps": 42420, "loss": 0.3812, "lr": 0.0005714276759578679, "epoch": 10.179160773220179, "percentage": 50.9, "elapsed_time": "1:10:09", "remaining_time": "1:07:41", "throughput": 5167.16, "total_tokens": 21750048} {"current_steps": 21595, "total_steps": 42420, "loss": 0.3208, "lr": 0.0005712240597791222, "epoch": 10.181518151815181, "percentage": 50.91, "elapsed_time": "1:10:10", "remaining_time": "1:07:40", "throughput": 5167.11, "total_tokens": 21754176} {"current_steps": 21600, "total_steps": 42420, "loss": 0.3031, "lr": 0.0005710204315433527, "epoch": 10.183875530410184, "percentage": 50.92, "elapsed_time": "1:10:10", "remaining_time": "1:07:38", "throughput": 5167.04, "total_tokens": 21757536} {"current_steps": 21605, "total_steps": 42420, "loss": 0.34, "lr": 0.00057081679128503, "epoch": 10.186232909005186, "percentage": 50.93, "elapsed_time": "1:10:11", "remaining_time": "1:07:37", "throughput": 5167.04, "total_tokens": 21762176} {"current_steps": 21610, "total_steps": 42420, "loss": 0.2862, "lr": 0.0005706131390386269, "epoch": 10.188590287600189, "percentage": 50.94, "elapsed_time": "1:10:12", "remaining_time": "1:07:36", "throughput": 5167.09, "total_tokens": 21766944} {"current_steps": 21615, "total_steps": 42420, "loss": 0.323, "lr": 0.0005704094748386184, "epoch": 10.190947666195191, "percentage": 50.95, "elapsed_time": "1:10:13", "remaining_time": "1:07:35", "throughput": 5167.16, "total_tokens": 21772064} {"current_steps": 21620, "total_steps": 42420, "loss": 0.3756, "lr": 0.0005702057987194811, "epoch": 10.193305044790193, "percentage": 50.97, "elapsed_time": "1:10:14", "remaining_time": "1:07:34", "throughput": 5167.2, "total_tokens": 21776736} {"current_steps": 21625, "total_steps": 42420, "loss": 0.2956, "lr": 0.0005700021107156942, "epoch": 10.195662423385196, "percentage": 50.98, "elapsed_time": "1:10:15", "remaining_time": "1:07:33", "throughput": 5167.27, "total_tokens": 21781888} {"current_steps": 21630, "total_steps": 42420, "loss": 0.3135, "lr": 0.0005697984108617386, "epoch": 10.198019801980198, "percentage": 50.99, "elapsed_time": "1:10:16", "remaining_time": "1:07:32", "throughput": 5167.32, "total_tokens": 21786656} {"current_steps": 21635, "total_steps": 42420, "loss": 0.2845, "lr": 0.000569594699192097, "epoch": 10.2003771805752, "percentage": 51.0, "elapsed_time": "1:10:17", "remaining_time": "1:07:31", "throughput": 5167.41, "total_tokens": 21791872} {"current_steps": 21640, "total_steps": 42420, "loss": 0.3895, "lr": 0.0005693909757412545, "epoch": 10.202734559170203, "percentage": 51.01, "elapsed_time": "1:10:18", "remaining_time": "1:07:30", "throughput": 5167.48, "total_tokens": 21796864} {"current_steps": 21645, "total_steps": 42420, "loss": 0.3736, "lr": 0.000569187240543698, "epoch": 10.205091937765205, "percentage": 51.03, "elapsed_time": "1:10:18", "remaining_time": "1:07:29", "throughput": 5167.57, "total_tokens": 21801888} {"current_steps": 21650, "total_steps": 42420, "loss": 0.3192, "lr": 0.0005689834936339165, "epoch": 10.207449316360208, "percentage": 51.04, "elapsed_time": "1:10:19", "remaining_time": "1:07:28", "throughput": 5167.63, "total_tokens": 21807072} {"current_steps": 21655, "total_steps": 42420, "loss": 0.3099, "lr": 0.0005687797350464006, "epoch": 10.20980669495521, "percentage": 51.05, "elapsed_time": "1:10:20", "remaining_time": "1:07:27", "throughput": 5167.65, "total_tokens": 21811488} {"current_steps": 21660, "total_steps": 42420, "loss": 0.289, "lr": 0.0005685759648156437, "epoch": 10.212164073550213, "percentage": 51.06, "elapsed_time": "1:10:21", "remaining_time": "1:07:26", "throughput": 5167.66, "total_tokens": 21816192} {"current_steps": 21665, "total_steps": 42420, "loss": 0.3657, "lr": 0.00056837218297614, "epoch": 10.214521452145215, "percentage": 51.07, "elapsed_time": "1:10:22", "remaining_time": "1:07:25", "throughput": 5167.64, "total_tokens": 21820384} {"current_steps": 21670, "total_steps": 42420, "loss": 0.3759, "lr": 0.0005681683895623868, "epoch": 10.216878830740217, "percentage": 51.08, "elapsed_time": "1:10:23", "remaining_time": "1:07:24", "throughput": 5167.7, "total_tokens": 21825120} {"current_steps": 21675, "total_steps": 42420, "loss": 0.3264, "lr": 0.0005679645846088826, "epoch": 10.21923620933522, "percentage": 51.1, "elapsed_time": "1:10:24", "remaining_time": "1:07:23", "throughput": 5167.74, "total_tokens": 21829856} {"current_steps": 21680, "total_steps": 42420, "loss": 0.3346, "lr": 0.0005677607681501282, "epoch": 10.221593587930222, "percentage": 51.11, "elapsed_time": "1:10:25", "remaining_time": "1:07:21", "throughput": 5167.81, "total_tokens": 21834784} {"current_steps": 21685, "total_steps": 42420, "loss": 0.3267, "lr": 0.0005675569402206262, "epoch": 10.223950966525225, "percentage": 51.12, "elapsed_time": "1:10:26", "remaining_time": "1:07:20", "throughput": 5167.87, "total_tokens": 21840096} {"current_steps": 21690, "total_steps": 42420, "loss": 0.3193, "lr": 0.0005673531008548815, "epoch": 10.226308345120227, "percentage": 51.13, "elapsed_time": "1:10:27", "remaining_time": "1:07:19", "throughput": 5168.0, "total_tokens": 21845568} {"current_steps": 21695, "total_steps": 42420, "loss": 0.3652, "lr": 0.0005671492500874003, "epoch": 10.22866572371523, "percentage": 51.14, "elapsed_time": "1:10:27", "remaining_time": "1:07:18", "throughput": 5168.02, "total_tokens": 21850016} {"current_steps": 21700, "total_steps": 42420, "loss": 0.3855, "lr": 0.000566945387952691, "epoch": 10.231023102310232, "percentage": 51.16, "elapsed_time": "1:10:28", "remaining_time": "1:07:17", "throughput": 5168.12, "total_tokens": 21855232} {"current_steps": 21705, "total_steps": 42420, "loss": 0.3508, "lr": 0.0005667415144852643, "epoch": 10.233380480905234, "percentage": 51.17, "elapsed_time": "1:10:29", "remaining_time": "1:07:16", "throughput": 5168.13, "total_tokens": 21859776} {"current_steps": 21710, "total_steps": 42420, "loss": 0.3374, "lr": 0.0005665376297196327, "epoch": 10.235737859500237, "percentage": 51.18, "elapsed_time": "1:10:30", "remaining_time": "1:07:15", "throughput": 5168.21, "total_tokens": 21865088} {"current_steps": 21715, "total_steps": 42420, "loss": 0.3566, "lr": 0.0005663337336903097, "epoch": 10.238095238095237, "percentage": 51.19, "elapsed_time": "1:10:31", "remaining_time": "1:07:14", "throughput": 5168.23, "total_tokens": 21869760} {"current_steps": 21720, "total_steps": 42420, "loss": 0.3629, "lr": 0.000566129826431812, "epoch": 10.24045261669024, "percentage": 51.2, "elapsed_time": "1:10:32", "remaining_time": "1:07:13", "throughput": 5168.28, "total_tokens": 21874560} {"current_steps": 21725, "total_steps": 42420, "loss": 0.3063, "lr": 0.0005659259079786578, "epoch": 10.242809995285242, "percentage": 51.21, "elapsed_time": "1:10:33", "remaining_time": "1:07:12", "throughput": 5168.32, "total_tokens": 21879296} {"current_steps": 21730, "total_steps": 42420, "loss": 0.333, "lr": 0.0005657219783653664, "epoch": 10.245167373880244, "percentage": 51.23, "elapsed_time": "1:10:34", "remaining_time": "1:07:11", "throughput": 5168.36, "total_tokens": 21884448} {"current_steps": 21735, "total_steps": 42420, "loss": 0.3002, "lr": 0.0005655180376264601, "epoch": 10.247524752475247, "percentage": 51.24, "elapsed_time": "1:10:35", "remaining_time": "1:07:10", "throughput": 5168.44, "total_tokens": 21889792} {"current_steps": 21740, "total_steps": 42420, "loss": 0.393, "lr": 0.0005653140857964624, "epoch": 10.24988213107025, "percentage": 51.25, "elapsed_time": "1:10:36", "remaining_time": "1:07:09", "throughput": 5168.63, "total_tokens": 21896320} {"current_steps": 21745, "total_steps": 42420, "loss": 0.437, "lr": 0.0005651101229098989, "epoch": 10.252239509665252, "percentage": 51.26, "elapsed_time": "1:10:37", "remaining_time": "1:07:08", "throughput": 5168.64, "total_tokens": 21900768} {"current_steps": 21750, "total_steps": 42420, "loss": 0.3922, "lr": 0.0005649061490012972, "epoch": 10.254596888260254, "percentage": 51.27, "elapsed_time": "1:10:38", "remaining_time": "1:07:07", "throughput": 5168.63, "total_tokens": 21904896} {"current_steps": 21755, "total_steps": 42420, "loss": 0.4249, "lr": 0.0005647021641051865, "epoch": 10.256954266855256, "percentage": 51.28, "elapsed_time": "1:10:39", "remaining_time": "1:07:06", "throughput": 5168.78, "total_tokens": 21910944} {"current_steps": 21760, "total_steps": 42420, "loss": 0.4215, "lr": 0.0005644981682560979, "epoch": 10.259311645450259, "percentage": 51.3, "elapsed_time": "1:10:40", "remaining_time": "1:07:05", "throughput": 5168.83, "total_tokens": 21915872} {"current_steps": 21765, "total_steps": 42420, "loss": 0.3813, "lr": 0.0005642941614885646, "epoch": 10.261669024045261, "percentage": 51.31, "elapsed_time": "1:10:41", "remaining_time": "1:07:04", "throughput": 5169.0, "total_tokens": 21922240} {"current_steps": 21770, "total_steps": 42420, "loss": 0.3306, "lr": 0.0005640901438371214, "epoch": 10.264026402640264, "percentage": 51.32, "elapsed_time": "1:10:41", "remaining_time": "1:07:03", "throughput": 5169.04, "total_tokens": 21926880} {"current_steps": 21775, "total_steps": 42420, "loss": 0.4185, "lr": 0.0005638861153363052, "epoch": 10.266383781235266, "percentage": 51.33, "elapsed_time": "1:10:42", "remaining_time": "1:07:02", "throughput": 5169.15, "total_tokens": 21932448} {"current_steps": 21780, "total_steps": 42420, "loss": 0.3679, "lr": 0.0005636820760206542, "epoch": 10.268741159830268, "percentage": 51.34, "elapsed_time": "1:10:44", "remaining_time": "1:07:01", "throughput": 5169.3, "total_tokens": 21938784} {"current_steps": 21785, "total_steps": 42420, "loss": 0.3667, "lr": 0.000563478025924709, "epoch": 10.27109853842527, "percentage": 51.36, "elapsed_time": "1:10:44", "remaining_time": "1:07:00", "throughput": 5169.37, "total_tokens": 21943680} {"current_steps": 21790, "total_steps": 42420, "loss": 0.3566, "lr": 0.0005632739650830119, "epoch": 10.273455917020273, "percentage": 51.37, "elapsed_time": "1:10:45", "remaining_time": "1:06:59", "throughput": 5169.46, "total_tokens": 21949024} {"current_steps": 21795, "total_steps": 42420, "loss": 0.3448, "lr": 0.0005630698935301068, "epoch": 10.275813295615276, "percentage": 51.38, "elapsed_time": "1:10:46", "remaining_time": "1:06:58", "throughput": 5169.46, "total_tokens": 21953536} {"current_steps": 21800, "total_steps": 42420, "loss": 0.3184, "lr": 0.0005628658113005396, "epoch": 10.278170674210278, "percentage": 51.39, "elapsed_time": "1:10:47", "remaining_time": "1:06:57", "throughput": 5169.49, "total_tokens": 21958304} {"current_steps": 21805, "total_steps": 42420, "loss": 0.3054, "lr": 0.0005626617184288579, "epoch": 10.28052805280528, "percentage": 51.4, "elapsed_time": "1:10:48", "remaining_time": "1:06:56", "throughput": 5169.59, "total_tokens": 21963968} {"current_steps": 21810, "total_steps": 42420, "loss": 0.3326, "lr": 0.0005624576149496111, "epoch": 10.282885431400283, "percentage": 51.41, "elapsed_time": "1:10:49", "remaining_time": "1:06:55", "throughput": 5169.67, "total_tokens": 21969024} {"current_steps": 21815, "total_steps": 42420, "loss": 0.3583, "lr": 0.0005622535008973508, "epoch": 10.285242809995285, "percentage": 51.43, "elapsed_time": "1:10:50", "remaining_time": "1:06:54", "throughput": 5169.7, "total_tokens": 21973440} {"current_steps": 21820, "total_steps": 42420, "loss": 0.3917, "lr": 0.0005620493763066296, "epoch": 10.287600188590288, "percentage": 51.44, "elapsed_time": "1:10:51", "remaining_time": "1:06:53", "throughput": 5169.87, "total_tokens": 21979744} {"current_steps": 21825, "total_steps": 42420, "loss": 0.3062, "lr": 0.0005618452412120027, "epoch": 10.28995756718529, "percentage": 51.45, "elapsed_time": "1:10:52", "remaining_time": "1:06:52", "throughput": 5169.9, "total_tokens": 21984320} {"current_steps": 21830, "total_steps": 42420, "loss": 0.3482, "lr": 0.0005616410956480265, "epoch": 10.292314945780292, "percentage": 51.46, "elapsed_time": "1:10:53", "remaining_time": "1:06:51", "throughput": 5170.0, "total_tokens": 21989888} {"current_steps": 21835, "total_steps": 42420, "loss": 0.3595, "lr": 0.0005614369396492593, "epoch": 10.294672324375295, "percentage": 51.47, "elapsed_time": "1:10:54", "remaining_time": "1:06:50", "throughput": 5170.07, "total_tokens": 21995168} {"current_steps": 21840, "total_steps": 42420, "loss": 0.3147, "lr": 0.0005612327732502616, "epoch": 10.297029702970297, "percentage": 51.49, "elapsed_time": "1:10:55", "remaining_time": "1:06:49", "throughput": 5170.1, "total_tokens": 21999968} {"current_steps": 21845, "total_steps": 42420, "loss": 0.2851, "lr": 0.000561028596485595, "epoch": 10.2993870815653, "percentage": 51.5, "elapsed_time": "1:10:56", "remaining_time": "1:06:48", "throughput": 5170.18, "total_tokens": 22005344} {"current_steps": 21850, "total_steps": 42420, "loss": 0.3368, "lr": 0.000560824409389823, "epoch": 10.301744460160302, "percentage": 51.51, "elapsed_time": "1:10:57", "remaining_time": "1:06:47", "throughput": 5170.29, "total_tokens": 22010880} {"current_steps": 21855, "total_steps": 42420, "loss": 0.2613, "lr": 0.0005606202119975112, "epoch": 10.304101838755304, "percentage": 51.52, "elapsed_time": "1:10:58", "remaining_time": "1:06:46", "throughput": 5170.35, "total_tokens": 22015808} {"current_steps": 21860, "total_steps": 42420, "loss": 0.3892, "lr": 0.000560416004343227, "epoch": 10.306459217350307, "percentage": 51.53, "elapsed_time": "1:10:58", "remaining_time": "1:06:45", "throughput": 5170.37, "total_tokens": 22020288} {"current_steps": 21865, "total_steps": 42420, "loss": 0.3248, "lr": 0.000560211786461539, "epoch": 10.30881659594531, "percentage": 51.54, "elapsed_time": "1:10:59", "remaining_time": "1:06:44", "throughput": 5170.45, "total_tokens": 22025536} {"current_steps": 21870, "total_steps": 42420, "loss": 0.3124, "lr": 0.0005600075583870179, "epoch": 10.311173974540312, "percentage": 51.56, "elapsed_time": "1:11:00", "remaining_time": "1:06:43", "throughput": 5170.49, "total_tokens": 22030272} {"current_steps": 21875, "total_steps": 42420, "loss": 0.3475, "lr": 0.000559803320154236, "epoch": 10.313531353135314, "percentage": 51.57, "elapsed_time": "1:11:01", "remaining_time": "1:06:42", "throughput": 5170.59, "total_tokens": 22035872} {"current_steps": 21880, "total_steps": 42420, "loss": 0.3571, "lr": 0.0005595990717977675, "epoch": 10.315888731730317, "percentage": 51.58, "elapsed_time": "1:11:02", "remaining_time": "1:06:41", "throughput": 5170.63, "total_tokens": 22040544} {"current_steps": 21885, "total_steps": 42420, "loss": 0.3071, "lr": 0.000559394813352188, "epoch": 10.318246110325319, "percentage": 51.59, "elapsed_time": "1:11:03", "remaining_time": "1:06:40", "throughput": 5170.57, "total_tokens": 22044576} {"current_steps": 21890, "total_steps": 42420, "loss": 0.3382, "lr": 0.000559190544852075, "epoch": 10.320603488920321, "percentage": 51.6, "elapsed_time": "1:11:04", "remaining_time": "1:06:39", "throughput": 5170.71, "total_tokens": 22050848} {"current_steps": 21895, "total_steps": 42420, "loss": 0.3797, "lr": 0.0005589862663320078, "epoch": 10.322960867515324, "percentage": 51.61, "elapsed_time": "1:11:05", "remaining_time": "1:06:38", "throughput": 5170.73, "total_tokens": 22055488} {"current_steps": 21900, "total_steps": 42420, "loss": 0.4181, "lr": 0.0005587819778265674, "epoch": 10.325318246110326, "percentage": 51.63, "elapsed_time": "1:11:06", "remaining_time": "1:06:37", "throughput": 5170.85, "total_tokens": 22061408} {"current_steps": 21905, "total_steps": 42420, "loss": 0.3655, "lr": 0.000558577679370336, "epoch": 10.327675624705329, "percentage": 51.64, "elapsed_time": "1:11:07", "remaining_time": "1:06:36", "throughput": 5170.99, "total_tokens": 22067520} {"current_steps": 21910, "total_steps": 42420, "loss": 0.3544, "lr": 0.0005583733709978981, "epoch": 10.33003300330033, "percentage": 51.65, "elapsed_time": "1:11:08", "remaining_time": "1:06:35", "throughput": 5171.04, "total_tokens": 22072416} {"current_steps": 21915, "total_steps": 42420, "loss": 0.3431, "lr": 0.0005581690527438396, "epoch": 10.332390381895332, "percentage": 51.66, "elapsed_time": "1:11:09", "remaining_time": "1:06:34", "throughput": 5171.1, "total_tokens": 22077440} {"current_steps": 21920, "total_steps": 42420, "loss": 0.3432, "lr": 0.0005579647246427482, "epoch": 10.334747760490334, "percentage": 51.67, "elapsed_time": "1:11:10", "remaining_time": "1:06:33", "throughput": 5171.15, "total_tokens": 22082176} {"current_steps": 21925, "total_steps": 42420, "loss": 0.3453, "lr": 0.0005577603867292131, "epoch": 10.337105139085336, "percentage": 51.69, "elapsed_time": "1:11:11", "remaining_time": "1:06:32", "throughput": 5171.13, "total_tokens": 22086400} {"current_steps": 21930, "total_steps": 42420, "loss": 0.3551, "lr": 0.0005575560390378252, "epoch": 10.339462517680339, "percentage": 51.7, "elapsed_time": "1:11:12", "remaining_time": "1:06:31", "throughput": 5171.17, "total_tokens": 22091360} {"current_steps": 21935, "total_steps": 42420, "loss": 0.3236, "lr": 0.0005573516816031773, "epoch": 10.341819896275341, "percentage": 51.71, "elapsed_time": "1:11:13", "remaining_time": "1:06:30", "throughput": 5171.29, "total_tokens": 22097056} {"current_steps": 21940, "total_steps": 42420, "loss": 0.273, "lr": 0.0005571473144598633, "epoch": 10.344177274870344, "percentage": 51.72, "elapsed_time": "1:11:13", "remaining_time": "1:06:29", "throughput": 5171.34, "total_tokens": 22101888} {"current_steps": 21945, "total_steps": 42420, "loss": 0.3206, "lr": 0.0005569429376424793, "epoch": 10.346534653465346, "percentage": 51.73, "elapsed_time": "1:11:14", "remaining_time": "1:06:28", "throughput": 5171.31, "total_tokens": 22106048} {"current_steps": 21950, "total_steps": 42420, "loss": 0.3243, "lr": 0.000556738551185623, "epoch": 10.348892032060348, "percentage": 51.74, "elapsed_time": "1:11:15", "remaining_time": "1:06:27", "throughput": 5171.38, "total_tokens": 22111264} {"current_steps": 21955, "total_steps": 42420, "loss": 0.3792, "lr": 0.0005565341551238931, "epoch": 10.35124941065535, "percentage": 51.76, "elapsed_time": "1:11:16", "remaining_time": "1:06:26", "throughput": 5171.5, "total_tokens": 22116928} {"current_steps": 21960, "total_steps": 42420, "loss": 0.31, "lr": 0.0005563297494918907, "epoch": 10.353606789250353, "percentage": 51.77, "elapsed_time": "1:11:17", "remaining_time": "1:06:25", "throughput": 5171.61, "total_tokens": 22122848} {"current_steps": 21965, "total_steps": 42420, "loss": 0.3591, "lr": 0.0005561253343242181, "epoch": 10.355964167845356, "percentage": 51.78, "elapsed_time": "1:11:18", "remaining_time": "1:06:24", "throughput": 5171.64, "total_tokens": 22127488} {"current_steps": 21970, "total_steps": 42420, "loss": 0.3308, "lr": 0.0005559209096554795, "epoch": 10.358321546440358, "percentage": 51.79, "elapsed_time": "1:11:19", "remaining_time": "1:06:23", "throughput": 5171.71, "total_tokens": 22132832} {"current_steps": 21975, "total_steps": 42420, "loss": 0.3497, "lr": 0.0005557164755202801, "epoch": 10.36067892503536, "percentage": 51.8, "elapsed_time": "1:11:20", "remaining_time": "1:06:22", "throughput": 5171.83, "total_tokens": 22138464} {"current_steps": 21980, "total_steps": 42420, "loss": 0.3523, "lr": 0.0005555120319532276, "epoch": 10.363036303630363, "percentage": 51.82, "elapsed_time": "1:11:21", "remaining_time": "1:06:21", "throughput": 5171.91, "total_tokens": 22143648} {"current_steps": 21985, "total_steps": 42420, "loss": 0.3519, "lr": 0.0005553075789889307, "epoch": 10.365393682225365, "percentage": 51.83, "elapsed_time": "1:11:22", "remaining_time": "1:06:20", "throughput": 5171.93, "total_tokens": 22148192} {"current_steps": 21990, "total_steps": 42420, "loss": 0.3363, "lr": 0.0005551031166619997, "epoch": 10.367751060820368, "percentage": 51.84, "elapsed_time": "1:11:23", "remaining_time": "1:06:19", "throughput": 5171.95, "total_tokens": 22152672} {"current_steps": 21995, "total_steps": 42420, "loss": 0.3192, "lr": 0.0005548986450070464, "epoch": 10.37010843941537, "percentage": 51.85, "elapsed_time": "1:11:24", "remaining_time": "1:06:18", "throughput": 5172.04, "total_tokens": 22158336} {"current_steps": 22000, "total_steps": 42420, "loss": 0.3342, "lr": 0.0005546941640586849, "epoch": 10.372465818010372, "percentage": 51.86, "elapsed_time": "1:11:25", "remaining_time": "1:06:17", "throughput": 5172.06, "total_tokens": 22162912} {"current_steps": 22005, "total_steps": 42420, "loss": 0.3346, "lr": 0.00055448967385153, "epoch": 10.374823196605375, "percentage": 51.87, "elapsed_time": "1:11:26", "remaining_time": "1:06:16", "throughput": 5172.21, "total_tokens": 22168864} {"current_steps": 22010, "total_steps": 42420, "loss": 0.4286, "lr": 0.0005542851744201985, "epoch": 10.377180575200377, "percentage": 51.89, "elapsed_time": "1:11:27", "remaining_time": "1:06:15", "throughput": 5172.31, "total_tokens": 22174368} {"current_steps": 22015, "total_steps": 42420, "loss": 0.3214, "lr": 0.0005540806657993087, "epoch": 10.37953795379538, "percentage": 51.9, "elapsed_time": "1:11:28", "remaining_time": "1:06:14", "throughput": 5172.33, "total_tokens": 22179040} {"current_steps": 22020, "total_steps": 42420, "loss": 0.3395, "lr": 0.0005538761480234804, "epoch": 10.381895332390382, "percentage": 51.91, "elapsed_time": "1:11:28", "remaining_time": "1:06:13", "throughput": 5172.36, "total_tokens": 22183712} {"current_steps": 22025, "total_steps": 42420, "loss": 0.3288, "lr": 0.000553671621127335, "epoch": 10.384252710985384, "percentage": 51.92, "elapsed_time": "1:11:29", "remaining_time": "1:06:12", "throughput": 5172.48, "total_tokens": 22189408} {"current_steps": 22030, "total_steps": 42420, "loss": 0.3307, "lr": 0.0005534670851454955, "epoch": 10.386610089580387, "percentage": 51.93, "elapsed_time": "1:11:30", "remaining_time": "1:06:11", "throughput": 5172.63, "total_tokens": 22195424} {"current_steps": 22035, "total_steps": 42420, "loss": 0.3149, "lr": 0.0005532625401125862, "epoch": 10.38896746817539, "percentage": 51.94, "elapsed_time": "1:11:31", "remaining_time": "1:06:10", "throughput": 5172.66, "total_tokens": 22200224} {"current_steps": 22040, "total_steps": 42420, "loss": 0.3181, "lr": 0.0005530579860632332, "epoch": 10.391324846770392, "percentage": 51.96, "elapsed_time": "1:11:32", "remaining_time": "1:06:09", "throughput": 5172.76, "total_tokens": 22205632} {"current_steps": 22045, "total_steps": 42420, "loss": 0.3447, "lr": 0.0005528534230320641, "epoch": 10.393682225365394, "percentage": 51.97, "elapsed_time": "1:11:33", "remaining_time": "1:06:08", "throughput": 5172.82, "total_tokens": 22210496} {"current_steps": 22050, "total_steps": 42420, "loss": 0.3873, "lr": 0.0005526488510537079, "epoch": 10.396039603960396, "percentage": 51.98, "elapsed_time": "1:11:34", "remaining_time": "1:06:07", "throughput": 5173.03, "total_tokens": 22217664} {"current_steps": 22055, "total_steps": 42420, "loss": 0.3312, "lr": 0.000552444270162795, "epoch": 10.398396982555399, "percentage": 51.99, "elapsed_time": "1:11:35", "remaining_time": "1:06:06", "throughput": 5173.0, "total_tokens": 22221440} {"current_steps": 22060, "total_steps": 42420, "loss": 0.3357, "lr": 0.0005522396803939578, "epoch": 10.400754361150401, "percentage": 52.0, "elapsed_time": "1:11:36", "remaining_time": "1:06:05", "throughput": 5173.09, "total_tokens": 22226816} {"current_steps": 22065, "total_steps": 42420, "loss": 0.3301, "lr": 0.0005520350817818296, "epoch": 10.403111739745404, "percentage": 52.02, "elapsed_time": "1:11:37", "remaining_time": "1:06:04", "throughput": 5173.2, "total_tokens": 22232480} {"current_steps": 22070, "total_steps": 42420, "loss": 0.3547, "lr": 0.0005518304743610457, "epoch": 10.405469118340406, "percentage": 52.03, "elapsed_time": "1:11:38", "remaining_time": "1:06:03", "throughput": 5173.25, "total_tokens": 22237312} {"current_steps": 22075, "total_steps": 42420, "loss": 0.3503, "lr": 0.0005516258581662423, "epoch": 10.407826496935408, "percentage": 52.04, "elapsed_time": "1:11:39", "remaining_time": "1:06:02", "throughput": 5173.34, "total_tokens": 22242784} {"current_steps": 22080, "total_steps": 42420, "loss": 0.3296, "lr": 0.0005514212332320578, "epoch": 10.41018387553041, "percentage": 52.05, "elapsed_time": "1:11:40", "remaining_time": "1:06:01", "throughput": 5173.43, "total_tokens": 22248480} {"current_steps": 22085, "total_steps": 42420, "loss": 0.3542, "lr": 0.0005512165995931315, "epoch": 10.412541254125413, "percentage": 52.06, "elapsed_time": "1:11:41", "remaining_time": "1:06:00", "throughput": 5173.5, "total_tokens": 22253632} {"current_steps": 22090, "total_steps": 42420, "loss": 0.3367, "lr": 0.0005510119572841045, "epoch": 10.414898632720416, "percentage": 52.07, "elapsed_time": "1:11:42", "remaining_time": "1:05:59", "throughput": 5173.5, "total_tokens": 22258016} {"current_steps": 22095, "total_steps": 42420, "loss": 0.3716, "lr": 0.0005508073063396194, "epoch": 10.417256011315418, "percentage": 52.09, "elapsed_time": "1:11:43", "remaining_time": "1:05:58", "throughput": 5173.46, "total_tokens": 22262080} {"current_steps": 22100, "total_steps": 42420, "loss": 0.337, "lr": 0.0005506026467943199, "epoch": 10.41961338991042, "percentage": 52.1, "elapsed_time": "1:11:44", "remaining_time": "1:05:57", "throughput": 5173.58, "total_tokens": 22268096} {"current_steps": 22105, "total_steps": 42420, "loss": 0.3405, "lr": 0.0005503979786828514, "epoch": 10.421970768505423, "percentage": 52.11, "elapsed_time": "1:11:45", "remaining_time": "1:05:56", "throughput": 5173.61, "total_tokens": 22272768} {"current_steps": 22110, "total_steps": 42420, "loss": 0.3643, "lr": 0.000550193302039861, "epoch": 10.424328147100425, "percentage": 52.12, "elapsed_time": "1:11:45", "remaining_time": "1:05:55", "throughput": 5173.63, "total_tokens": 22277312} {"current_steps": 22115, "total_steps": 42420, "loss": 0.3451, "lr": 0.0005499886168999968, "epoch": 10.426685525695426, "percentage": 52.13, "elapsed_time": "1:11:46", "remaining_time": "1:05:54", "throughput": 5173.66, "total_tokens": 22282016} {"current_steps": 22120, "total_steps": 42420, "loss": 0.3214, "lr": 0.0005497839232979084, "epoch": 10.429042904290428, "percentage": 52.15, "elapsed_time": "1:11:47", "remaining_time": "1:05:53", "throughput": 5173.82, "total_tokens": 22288160} {"current_steps": 22125, "total_steps": 42420, "loss": 0.3391, "lr": 0.000549579221268247, "epoch": 10.43140028288543, "percentage": 52.16, "elapsed_time": "1:11:48", "remaining_time": "1:05:52", "throughput": 5173.85, "total_tokens": 22293120} {"current_steps": 22130, "total_steps": 42420, "loss": 0.3362, "lr": 0.0005493745108456655, "epoch": 10.433757661480433, "percentage": 52.17, "elapsed_time": "1:11:49", "remaining_time": "1:05:51", "throughput": 5173.98, "total_tokens": 22298944} {"current_steps": 22135, "total_steps": 42420, "loss": 0.2901, "lr": 0.0005491697920648174, "epoch": 10.436115040075435, "percentage": 52.18, "elapsed_time": "1:11:50", "remaining_time": "1:05:50", "throughput": 5174.05, "total_tokens": 22304000} {"current_steps": 22140, "total_steps": 42420, "loss": 0.3593, "lr": 0.0005489650649603584, "epoch": 10.438472418670438, "percentage": 52.19, "elapsed_time": "1:11:51", "remaining_time": "1:05:49", "throughput": 5174.12, "total_tokens": 22308992} {"current_steps": 22145, "total_steps": 42420, "loss": 0.3322, "lr": 0.0005487603295669454, "epoch": 10.44082979726544, "percentage": 52.2, "elapsed_time": "1:11:52", "remaining_time": "1:05:48", "throughput": 5174.18, "total_tokens": 22314016} {"current_steps": 22150, "total_steps": 42420, "loss": 0.2968, "lr": 0.0005485555859192365, "epoch": 10.443187175860443, "percentage": 52.22, "elapsed_time": "1:11:53", "remaining_time": "1:05:47", "throughput": 5174.24, "total_tokens": 22319072} {"current_steps": 22155, "total_steps": 42420, "loss": 0.2624, "lr": 0.0005483508340518912, "epoch": 10.445544554455445, "percentage": 52.23, "elapsed_time": "1:11:54", "remaining_time": "1:05:46", "throughput": 5174.34, "total_tokens": 22324448} {"current_steps": 22160, "total_steps": 42420, "loss": 0.3032, "lr": 0.0005481460739995706, "epoch": 10.447901933050447, "percentage": 52.24, "elapsed_time": "1:11:55", "remaining_time": "1:05:45", "throughput": 5174.4, "total_tokens": 22329696} {"current_steps": 22165, "total_steps": 42420, "loss": 0.2497, "lr": 0.0005479413057969374, "epoch": 10.45025931164545, "percentage": 52.25, "elapsed_time": "1:11:56", "remaining_time": "1:05:44", "throughput": 5174.48, "total_tokens": 22334912} {"current_steps": 22170, "total_steps": 42420, "loss": 0.347, "lr": 0.0005477365294786549, "epoch": 10.452616690240452, "percentage": 52.26, "elapsed_time": "1:11:57", "remaining_time": "1:05:43", "throughput": 5174.55, "total_tokens": 22339936} {"current_steps": 22175, "total_steps": 42420, "loss": 0.3598, "lr": 0.0005475317450793888, "epoch": 10.454974068835455, "percentage": 52.27, "elapsed_time": "1:11:58", "remaining_time": "1:05:42", "throughput": 5174.73, "total_tokens": 22346688} {"current_steps": 22180, "total_steps": 42420, "loss": 0.3252, "lr": 0.000547326952633805, "epoch": 10.457331447430457, "percentage": 52.29, "elapsed_time": "1:11:59", "remaining_time": "1:05:41", "throughput": 5174.71, "total_tokens": 22350624} {"current_steps": 22185, "total_steps": 42420, "loss": 0.2824, "lr": 0.0005471221521765717, "epoch": 10.45968882602546, "percentage": 52.3, "elapsed_time": "1:12:00", "remaining_time": "1:05:40", "throughput": 5174.72, "total_tokens": 22355520} {"current_steps": 22190, "total_steps": 42420, "loss": 0.4321, "lr": 0.0005469173437423581, "epoch": 10.462046204620462, "percentage": 52.31, "elapsed_time": "1:12:01", "remaining_time": "1:05:39", "throughput": 5174.77, "total_tokens": 22360448} {"current_steps": 22195, "total_steps": 42420, "loss": 0.3608, "lr": 0.000546712527365835, "epoch": 10.464403583215464, "percentage": 52.32, "elapsed_time": "1:12:01", "remaining_time": "1:05:38", "throughput": 5174.74, "total_tokens": 22364448} {"current_steps": 22200, "total_steps": 42420, "loss": 0.339, "lr": 0.0005465077030816739, "epoch": 10.466760961810467, "percentage": 52.33, "elapsed_time": "1:12:02", "remaining_time": "1:05:37", "throughput": 5174.72, "total_tokens": 22368544} {"current_steps": 22205, "total_steps": 42420, "loss": 0.3639, "lr": 0.0005463028709245485, "epoch": 10.469118340405469, "percentage": 52.35, "elapsed_time": "1:12:03", "remaining_time": "1:05:36", "throughput": 5174.74, "total_tokens": 22373344} {"current_steps": 22210, "total_steps": 42420, "loss": 0.3651, "lr": 0.0005460980309291332, "epoch": 10.471475719000471, "percentage": 52.36, "elapsed_time": "1:12:04", "remaining_time": "1:05:35", "throughput": 5174.82, "total_tokens": 22378400} {"current_steps": 22215, "total_steps": 42420, "loss": 0.3306, "lr": 0.0005458931831301041, "epoch": 10.473833097595474, "percentage": 52.37, "elapsed_time": "1:12:05", "remaining_time": "1:05:34", "throughput": 5174.88, "total_tokens": 22383360} {"current_steps": 22220, "total_steps": 42420, "loss": 0.3585, "lr": 0.000545688327562138, "epoch": 10.476190476190476, "percentage": 52.38, "elapsed_time": "1:12:06", "remaining_time": "1:05:32", "throughput": 5174.88, "total_tokens": 22387808} {"current_steps": 22225, "total_steps": 42420, "loss": 0.3508, "lr": 0.000545483464259914, "epoch": 10.478547854785479, "percentage": 52.39, "elapsed_time": "1:12:07", "remaining_time": "1:05:31", "throughput": 5174.88, "total_tokens": 22392288} {"current_steps": 22230, "total_steps": 42420, "loss": 0.2871, "lr": 0.0005452785932581116, "epoch": 10.480905233380481, "percentage": 52.4, "elapsed_time": "1:12:07", "remaining_time": "1:05:30", "throughput": 5174.89, "total_tokens": 22396832} {"current_steps": 22235, "total_steps": 42420, "loss": 0.3524, "lr": 0.0005450737145914125, "epoch": 10.483262611975483, "percentage": 52.42, "elapsed_time": "1:12:09", "remaining_time": "1:05:29", "throughput": 5175.01, "total_tokens": 22402656} {"current_steps": 22240, "total_steps": 42420, "loss": 0.3551, "lr": 0.0005448688282944985, "epoch": 10.485619990570486, "percentage": 52.43, "elapsed_time": "1:12:09", "remaining_time": "1:05:28", "throughput": 5175.11, "total_tokens": 22408032} {"current_steps": 22245, "total_steps": 42420, "loss": 0.2946, "lr": 0.0005446639344020539, "epoch": 10.487977369165488, "percentage": 52.44, "elapsed_time": "1:12:10", "remaining_time": "1:05:27", "throughput": 5175.1, "total_tokens": 22412192} {"current_steps": 22250, "total_steps": 42420, "loss": 0.3444, "lr": 0.0005444590329487635, "epoch": 10.49033474776049, "percentage": 52.45, "elapsed_time": "1:12:11", "remaining_time": "1:05:26", "throughput": 5175.17, "total_tokens": 22417440} {"current_steps": 22255, "total_steps": 42420, "loss": 0.2641, "lr": 0.0005442541239693136, "epoch": 10.492692126355493, "percentage": 52.46, "elapsed_time": "1:12:12", "remaining_time": "1:05:25", "throughput": 5175.18, "total_tokens": 22422112} {"current_steps": 22260, "total_steps": 42420, "loss": 0.2679, "lr": 0.000544049207498392, "epoch": 10.495049504950495, "percentage": 52.48, "elapsed_time": "1:12:13", "remaining_time": "1:05:24", "throughput": 5175.23, "total_tokens": 22426784} {"current_steps": 22265, "total_steps": 42420, "loss": 0.3361, "lr": 0.0005438442835706874, "epoch": 10.497406883545498, "percentage": 52.49, "elapsed_time": "1:12:14", "remaining_time": "1:05:23", "throughput": 5175.23, "total_tokens": 22431200} {"current_steps": 22270, "total_steps": 42420, "loss": 0.4189, "lr": 0.00054363935222089, "epoch": 10.4997642621405, "percentage": 52.5, "elapsed_time": "1:12:15", "remaining_time": "1:05:22", "throughput": 5175.25, "total_tokens": 22435840} {"current_steps": 22275, "total_steps": 42420, "loss": 0.3028, "lr": 0.0005434344134836914, "epoch": 10.502121640735503, "percentage": 52.51, "elapsed_time": "1:12:16", "remaining_time": "1:05:21", "throughput": 5175.31, "total_tokens": 22440960} {"current_steps": 22280, "total_steps": 42420, "loss": 0.3288, "lr": 0.0005432294673937838, "epoch": 10.504479019330505, "percentage": 52.52, "elapsed_time": "1:12:17", "remaining_time": "1:05:20", "throughput": 5175.36, "total_tokens": 22445792} {"current_steps": 22285, "total_steps": 42420, "loss": 0.3442, "lr": 0.0005430245139858616, "epoch": 10.506836397925507, "percentage": 52.53, "elapsed_time": "1:12:18", "remaining_time": "1:05:19", "throughput": 5175.49, "total_tokens": 22451744} {"current_steps": 22290, "total_steps": 42420, "loss": 0.3605, "lr": 0.0005428195532946197, "epoch": 10.50919377652051, "percentage": 52.55, "elapsed_time": "1:12:18", "remaining_time": "1:05:18", "throughput": 5175.57, "total_tokens": 22456800} {"current_steps": 22295, "total_steps": 42420, "loss": 0.3702, "lr": 0.0005426145853547545, "epoch": 10.511551155115512, "percentage": 52.56, "elapsed_time": "1:12:19", "remaining_time": "1:05:17", "throughput": 5175.61, "total_tokens": 22461536} {"current_steps": 22300, "total_steps": 42420, "loss": 0.3397, "lr": 0.0005424096102009633, "epoch": 10.513908533710515, "percentage": 52.57, "elapsed_time": "1:12:20", "remaining_time": "1:05:16", "throughput": 5175.71, "total_tokens": 22467232} {"current_steps": 22305, "total_steps": 42420, "loss": 0.3616, "lr": 0.0005422046278679454, "epoch": 10.516265912305517, "percentage": 52.58, "elapsed_time": "1:12:21", "remaining_time": "1:05:15", "throughput": 5175.81, "total_tokens": 22472736} {"current_steps": 22310, "total_steps": 42420, "loss": 0.2995, "lr": 0.0005419996383904005, "epoch": 10.518623290900518, "percentage": 52.59, "elapsed_time": "1:12:22", "remaining_time": "1:05:14", "throughput": 5175.81, "total_tokens": 22476992} {"current_steps": 22315, "total_steps": 42420, "loss": 0.3497, "lr": 0.00054179464180303, "epoch": 10.520980669495522, "percentage": 52.6, "elapsed_time": "1:12:23", "remaining_time": "1:05:13", "throughput": 5175.86, "total_tokens": 22481952} {"current_steps": 22320, "total_steps": 42420, "loss": 0.3855, "lr": 0.0005415896381405362, "epoch": 10.523338048090523, "percentage": 52.62, "elapsed_time": "1:12:24", "remaining_time": "1:05:12", "throughput": 5176.0, "total_tokens": 22487968} {"current_steps": 22325, "total_steps": 42420, "loss": 0.3364, "lr": 0.0005413846274376227, "epoch": 10.525695426685525, "percentage": 52.63, "elapsed_time": "1:12:25", "remaining_time": "1:05:11", "throughput": 5176.02, "total_tokens": 22492416} {"current_steps": 22330, "total_steps": 42420, "loss": 0.3404, "lr": 0.0005411796097289946, "epoch": 10.528052805280527, "percentage": 52.64, "elapsed_time": "1:12:26", "remaining_time": "1:05:10", "throughput": 5176.1, "total_tokens": 22497952} {"current_steps": 22335, "total_steps": 42420, "loss": 0.3293, "lr": 0.0005409745850493579, "epoch": 10.53041018387553, "percentage": 52.65, "elapsed_time": "1:12:27", "remaining_time": "1:05:09", "throughput": 5176.11, "total_tokens": 22502464} {"current_steps": 22340, "total_steps": 42420, "loss": 0.3303, "lr": 0.0005407695534334194, "epoch": 10.532767562470532, "percentage": 52.66, "elapsed_time": "1:12:28", "remaining_time": "1:05:08", "throughput": 5176.14, "total_tokens": 22507104} {"current_steps": 22345, "total_steps": 42420, "loss": 0.3622, "lr": 0.0005405645149158878, "epoch": 10.535124941065535, "percentage": 52.68, "elapsed_time": "1:12:29", "remaining_time": "1:05:07", "throughput": 5176.11, "total_tokens": 22511104} {"current_steps": 22350, "total_steps": 42420, "loss": 0.3189, "lr": 0.0005403594695314725, "epoch": 10.537482319660537, "percentage": 52.69, "elapsed_time": "1:12:29", "remaining_time": "1:05:06", "throughput": 5176.14, "total_tokens": 22515744} {"current_steps": 22355, "total_steps": 42420, "loss": 0.3872, "lr": 0.0005401544173148844, "epoch": 10.53983969825554, "percentage": 52.7, "elapsed_time": "1:12:30", "remaining_time": "1:05:05", "throughput": 5176.16, "total_tokens": 22520288} {"current_steps": 22360, "total_steps": 42420, "loss": 0.3814, "lr": 0.0005399493583008353, "epoch": 10.542197076850542, "percentage": 52.71, "elapsed_time": "1:12:31", "remaining_time": "1:05:04", "throughput": 5176.25, "total_tokens": 22525760} {"current_steps": 22365, "total_steps": 42420, "loss": 0.341, "lr": 0.000539744292524038, "epoch": 10.544554455445544, "percentage": 52.72, "elapsed_time": "1:12:32", "remaining_time": "1:05:03", "throughput": 5176.26, "total_tokens": 22530272} {"current_steps": 22370, "total_steps": 42420, "loss": 0.3488, "lr": 0.0005395392200192069, "epoch": 10.546911834040547, "percentage": 52.73, "elapsed_time": "1:12:33", "remaining_time": "1:05:02", "throughput": 5176.28, "total_tokens": 22535072} {"current_steps": 22375, "total_steps": 42420, "loss": 0.3337, "lr": 0.0005393341408210571, "epoch": 10.549269212635549, "percentage": 52.75, "elapsed_time": "1:12:34", "remaining_time": "1:05:00", "throughput": 5176.27, "total_tokens": 22539200} {"current_steps": 22380, "total_steps": 42420, "loss": 0.3577, "lr": 0.0005391290549643051, "epoch": 10.551626591230551, "percentage": 52.76, "elapsed_time": "1:12:35", "remaining_time": "1:04:59", "throughput": 5176.38, "total_tokens": 22545056} {"current_steps": 22385, "total_steps": 42420, "loss": 0.3415, "lr": 0.0005389239624836685, "epoch": 10.553983969825554, "percentage": 52.77, "elapsed_time": "1:12:36", "remaining_time": "1:04:59", "throughput": 5176.45, "total_tokens": 22550432} {"current_steps": 22390, "total_steps": 42420, "loss": 0.2621, "lr": 0.0005387188634138659, "epoch": 10.556341348420556, "percentage": 52.78, "elapsed_time": "1:12:37", "remaining_time": "1:04:57", "throughput": 5176.43, "total_tokens": 22554656} {"current_steps": 22395, "total_steps": 42420, "loss": 0.3007, "lr": 0.0005385137577896173, "epoch": 10.558698727015559, "percentage": 52.79, "elapsed_time": "1:12:38", "remaining_time": "1:04:56", "throughput": 5176.57, "total_tokens": 22560480} {"current_steps": 22400, "total_steps": 42420, "loss": 0.3082, "lr": 0.0005383086456456434, "epoch": 10.561056105610561, "percentage": 52.81, "elapsed_time": "1:12:38", "remaining_time": "1:04:55", "throughput": 5176.54, "total_tokens": 22564544} {"current_steps": 22405, "total_steps": 42420, "loss": 0.4131, "lr": 0.000538103527016666, "epoch": 10.563413484205563, "percentage": 52.82, "elapsed_time": "1:12:39", "remaining_time": "1:04:54", "throughput": 5176.51, "total_tokens": 22568544} {"current_steps": 22410, "total_steps": 42420, "loss": 0.3857, "lr": 0.0005378984019374087, "epoch": 10.565770862800566, "percentage": 52.83, "elapsed_time": "1:12:40", "remaining_time": "1:04:53", "throughput": 5176.52, "total_tokens": 22572960} {"current_steps": 22415, "total_steps": 42420, "loss": 0.3286, "lr": 0.0005376932704425953, "epoch": 10.568128241395568, "percentage": 52.84, "elapsed_time": "1:12:41", "remaining_time": "1:04:52", "throughput": 5176.62, "total_tokens": 22578496} {"current_steps": 22420, "total_steps": 42420, "loss": 0.3324, "lr": 0.0005374881325669513, "epoch": 10.57048561999057, "percentage": 52.85, "elapsed_time": "1:12:42", "remaining_time": "1:04:51", "throughput": 5176.76, "total_tokens": 22584576} {"current_steps": 22425, "total_steps": 42420, "loss": 0.3006, "lr": 0.0005372829883452028, "epoch": 10.572842998585573, "percentage": 52.86, "elapsed_time": "1:12:43", "remaining_time": "1:04:50", "throughput": 5176.84, "total_tokens": 22589952} {"current_steps": 22430, "total_steps": 42420, "loss": 0.3155, "lr": 0.0005370778378120775, "epoch": 10.575200377180575, "percentage": 52.88, "elapsed_time": "1:12:44", "remaining_time": "1:04:49", "throughput": 5176.84, "total_tokens": 22594400} {"current_steps": 22435, "total_steps": 42420, "loss": 0.38, "lr": 0.0005368726810023038, "epoch": 10.577557755775578, "percentage": 52.89, "elapsed_time": "1:12:45", "remaining_time": "1:04:48", "throughput": 5176.93, "total_tokens": 22599808} {"current_steps": 22440, "total_steps": 42420, "loss": 0.388, "lr": 0.0005366675179506114, "epoch": 10.57991513437058, "percentage": 52.9, "elapsed_time": "1:12:46", "remaining_time": "1:04:47", "throughput": 5176.96, "total_tokens": 22604416} {"current_steps": 22445, "total_steps": 42420, "loss": 0.3367, "lr": 0.0005364623486917306, "epoch": 10.582272512965583, "percentage": 52.91, "elapsed_time": "1:12:47", "remaining_time": "1:04:46", "throughput": 5177.1, "total_tokens": 22610176} {"current_steps": 22450, "total_steps": 42420, "loss": 0.333, "lr": 0.0005362571732603934, "epoch": 10.584629891560585, "percentage": 52.92, "elapsed_time": "1:12:48", "remaining_time": "1:04:45", "throughput": 5177.1, "total_tokens": 22614560} {"current_steps": 22455, "total_steps": 42420, "loss": 0.3433, "lr": 0.0005360519916913322, "epoch": 10.586987270155587, "percentage": 52.93, "elapsed_time": "1:12:49", "remaining_time": "1:04:44", "throughput": 5177.11, "total_tokens": 22618976} {"current_steps": 22460, "total_steps": 42420, "loss": 0.3738, "lr": 0.0005358468040192811, "epoch": 10.58934464875059, "percentage": 52.95, "elapsed_time": "1:12:49", "remaining_time": "1:04:43", "throughput": 5177.21, "total_tokens": 22624256} {"current_steps": 22465, "total_steps": 42420, "loss": 0.3507, "lr": 0.0005356416102789746, "epoch": 10.591702027345592, "percentage": 52.96, "elapsed_time": "1:12:50", "remaining_time": "1:04:42", "throughput": 5177.28, "total_tokens": 22629248} {"current_steps": 22470, "total_steps": 42420, "loss": 0.3338, "lr": 0.0005354364105051487, "epoch": 10.594059405940595, "percentage": 52.97, "elapsed_time": "1:12:51", "remaining_time": "1:04:41", "throughput": 5177.31, "total_tokens": 22633760} {"current_steps": 22475, "total_steps": 42420, "loss": 0.3549, "lr": 0.0005352312047325402, "epoch": 10.596416784535597, "percentage": 52.98, "elapsed_time": "1:12:52", "remaining_time": "1:04:40", "throughput": 5177.36, "total_tokens": 22638976} {"current_steps": 22480, "total_steps": 42420, "loss": 0.3456, "lr": 0.000535025992995887, "epoch": 10.5987741631306, "percentage": 52.99, "elapsed_time": "1:12:53", "remaining_time": "1:04:39", "throughput": 5177.37, "total_tokens": 22643392} {"current_steps": 22485, "total_steps": 42420, "loss": 0.3291, "lr": 0.0005348207753299278, "epoch": 10.601131541725602, "percentage": 53.01, "elapsed_time": "1:12:54", "remaining_time": "1:04:38", "throughput": 5177.39, "total_tokens": 22648032} {"current_steps": 22490, "total_steps": 42420, "loss": 0.2986, "lr": 0.0005346155517694024, "epoch": 10.603488920320604, "percentage": 53.02, "elapsed_time": "1:12:55", "remaining_time": "1:04:37", "throughput": 5177.37, "total_tokens": 22652352} {"current_steps": 22495, "total_steps": 42420, "loss": 0.3713, "lr": 0.0005344103223490521, "epoch": 10.605846298915607, "percentage": 53.03, "elapsed_time": "1:12:56", "remaining_time": "1:04:36", "throughput": 5177.35, "total_tokens": 22656608} {"current_steps": 22500, "total_steps": 42420, "loss": 0.4051, "lr": 0.0005342050871036183, "epoch": 10.608203677510609, "percentage": 53.04, "elapsed_time": "1:12:57", "remaining_time": "1:04:35", "throughput": 5177.5, "total_tokens": 22663200} {"current_steps": 22505, "total_steps": 42420, "loss": 0.3233, "lr": 0.0005339998460678441, "epoch": 10.61056105610561, "percentage": 53.05, "elapsed_time": "1:12:58", "remaining_time": "1:04:34", "throughput": 5177.5, "total_tokens": 22667776} {"current_steps": 22510, "total_steps": 42420, "loss": 0.3076, "lr": 0.0005337945992764733, "epoch": 10.612918434700614, "percentage": 53.06, "elapsed_time": "1:12:59", "remaining_time": "1:04:33", "throughput": 5177.63, "total_tokens": 22673952} {"current_steps": 22515, "total_steps": 42420, "loss": 0.3439, "lr": 0.0005335893467642507, "epoch": 10.615275813295614, "percentage": 53.08, "elapsed_time": "1:13:00", "remaining_time": "1:04:32", "throughput": 5177.72, "total_tokens": 22679232} {"current_steps": 22520, "total_steps": 42420, "loss": 0.3358, "lr": 0.0005333840885659218, "epoch": 10.617633191890617, "percentage": 53.09, "elapsed_time": "1:13:01", "remaining_time": "1:04:31", "throughput": 5177.81, "total_tokens": 22684448} {"current_steps": 22525, "total_steps": 42420, "loss": 0.2933, "lr": 0.0005331788247162337, "epoch": 10.61999057048562, "percentage": 53.1, "elapsed_time": "1:13:01", "remaining_time": "1:04:30", "throughput": 5177.81, "total_tokens": 22688672} {"current_steps": 22530, "total_steps": 42420, "loss": 0.2845, "lr": 0.0005329735552499339, "epoch": 10.622347949080622, "percentage": 53.11, "elapsed_time": "1:13:02", "remaining_time": "1:04:29", "throughput": 5177.89, "total_tokens": 22693824} {"current_steps": 22535, "total_steps": 42420, "loss": 0.3402, "lr": 0.0005327682802017709, "epoch": 10.624705327675624, "percentage": 53.12, "elapsed_time": "1:13:03", "remaining_time": "1:04:28", "throughput": 5178.04, "total_tokens": 22700288} {"current_steps": 22540, "total_steps": 42420, "loss": 0.3662, "lr": 0.0005325629996064945, "epoch": 10.627062706270626, "percentage": 53.14, "elapsed_time": "1:13:04", "remaining_time": "1:04:27", "throughput": 5178.13, "total_tokens": 22705728} {"current_steps": 22545, "total_steps": 42420, "loss": 0.3617, "lr": 0.0005323577134988553, "epoch": 10.629420084865629, "percentage": 53.15, "elapsed_time": "1:13:05", "remaining_time": "1:04:26", "throughput": 5178.13, "total_tokens": 22710144} {"current_steps": 22550, "total_steps": 42420, "loss": 0.2782, "lr": 0.0005321524219136045, "epoch": 10.631777463460631, "percentage": 53.16, "elapsed_time": "1:13:06", "remaining_time": "1:04:25", "throughput": 5178.26, "total_tokens": 22715872} {"current_steps": 22555, "total_steps": 42420, "loss": 0.2568, "lr": 0.0005319471248854945, "epoch": 10.634134842055634, "percentage": 53.17, "elapsed_time": "1:13:07", "remaining_time": "1:04:24", "throughput": 5178.24, "total_tokens": 22720160} {"current_steps": 22560, "total_steps": 42420, "loss": 0.3549, "lr": 0.0005317418224492787, "epoch": 10.636492220650636, "percentage": 53.18, "elapsed_time": "1:13:08", "remaining_time": "1:04:23", "throughput": 5178.23, "total_tokens": 22724416} {"current_steps": 22565, "total_steps": 42420, "loss": 0.3544, "lr": 0.0005315365146397116, "epoch": 10.638849599245638, "percentage": 53.19, "elapsed_time": "1:13:09", "remaining_time": "1:04:22", "throughput": 5178.3, "total_tokens": 22729504} {"current_steps": 22570, "total_steps": 42420, "loss": 0.355, "lr": 0.0005313312014915476, "epoch": 10.64120697784064, "percentage": 53.21, "elapsed_time": "1:13:10", "remaining_time": "1:04:21", "throughput": 5178.31, "total_tokens": 22734240} {"current_steps": 22575, "total_steps": 42420, "loss": 0.3395, "lr": 0.0005311258830395433, "epoch": 10.643564356435643, "percentage": 53.22, "elapsed_time": "1:13:11", "remaining_time": "1:04:20", "throughput": 5178.26, "total_tokens": 22738016} {"current_steps": 22580, "total_steps": 42420, "loss": 0.3488, "lr": 0.0005309205593184556, "epoch": 10.645921735030646, "percentage": 53.23, "elapsed_time": "1:13:12", "remaining_time": "1:04:19", "throughput": 5178.34, "total_tokens": 22743264} {"current_steps": 22585, "total_steps": 42420, "loss": 0.3089, "lr": 0.0005307152303630423, "epoch": 10.648279113625648, "percentage": 53.24, "elapsed_time": "1:13:12", "remaining_time": "1:04:17", "throughput": 5178.36, "total_tokens": 22747872} {"current_steps": 22590, "total_steps": 42420, "loss": 0.3459, "lr": 0.0005305098962080618, "epoch": 10.65063649222065, "percentage": 53.25, "elapsed_time": "1:13:13", "remaining_time": "1:04:17", "throughput": 5178.5, "total_tokens": 22753952} {"current_steps": 22595, "total_steps": 42420, "loss": 0.3737, "lr": 0.000530304556888274, "epoch": 10.652993870815653, "percentage": 53.26, "elapsed_time": "1:13:14", "remaining_time": "1:04:16", "throughput": 5178.6, "total_tokens": 22759520} {"current_steps": 22600, "total_steps": 42420, "loss": 0.3432, "lr": 0.0005300992124384391, "epoch": 10.655351249410655, "percentage": 53.28, "elapsed_time": "1:13:15", "remaining_time": "1:04:15", "throughput": 5178.58, "total_tokens": 22763712} {"current_steps": 22605, "total_steps": 42420, "loss": 0.3644, "lr": 0.0005298938628933189, "epoch": 10.657708628005658, "percentage": 53.29, "elapsed_time": "1:13:16", "remaining_time": "1:04:13", "throughput": 5178.59, "total_tokens": 22767968} {"current_steps": 22610, "total_steps": 42420, "loss": 0.3524, "lr": 0.0005296885082876751, "epoch": 10.66006600660066, "percentage": 53.3, "elapsed_time": "1:13:17", "remaining_time": "1:04:12", "throughput": 5178.61, "total_tokens": 22772544} {"current_steps": 22615, "total_steps": 42420, "loss": 0.3613, "lr": 0.0005294831486562709, "epoch": 10.662423385195662, "percentage": 53.31, "elapsed_time": "1:13:18", "remaining_time": "1:04:11", "throughput": 5178.56, "total_tokens": 22776256} {"current_steps": 22620, "total_steps": 42420, "loss": 0.3567, "lr": 0.0005292777840338703, "epoch": 10.664780763790665, "percentage": 53.32, "elapsed_time": "1:13:19", "remaining_time": "1:04:10", "throughput": 5178.74, "total_tokens": 22783040} {"current_steps": 22625, "total_steps": 42420, "loss": 0.3371, "lr": 0.0005290724144552379, "epoch": 10.667138142385667, "percentage": 53.34, "elapsed_time": "1:13:20", "remaining_time": "1:04:09", "throughput": 5178.82, "total_tokens": 22788416} {"current_steps": 22630, "total_steps": 42420, "loss": 0.3625, "lr": 0.0005288670399551395, "epoch": 10.66949552098067, "percentage": 53.35, "elapsed_time": "1:13:21", "remaining_time": "1:04:08", "throughput": 5178.87, "total_tokens": 22793248} {"current_steps": 22635, "total_steps": 42420, "loss": 0.3607, "lr": 0.0005286616605683413, "epoch": 10.671852899575672, "percentage": 53.36, "elapsed_time": "1:13:22", "remaining_time": "1:04:07", "throughput": 5178.94, "total_tokens": 22798432} {"current_steps": 22640, "total_steps": 42420, "loss": 0.3371, "lr": 0.0005284562763296106, "epoch": 10.674210278170674, "percentage": 53.37, "elapsed_time": "1:13:23", "remaining_time": "1:04:06", "throughput": 5179.04, "total_tokens": 22804096} {"current_steps": 22645, "total_steps": 42420, "loss": 0.3536, "lr": 0.0005282508872737154, "epoch": 10.676567656765677, "percentage": 53.38, "elapsed_time": "1:13:24", "remaining_time": "1:04:05", "throughput": 5179.1, "total_tokens": 22809568} {"current_steps": 22650, "total_steps": 42420, "loss": 0.3141, "lr": 0.0005280454934354248, "epoch": 10.67892503536068, "percentage": 53.39, "elapsed_time": "1:13:25", "remaining_time": "1:04:04", "throughput": 5179.16, "total_tokens": 22814816} {"current_steps": 22655, "total_steps": 42420, "loss": 0.3306, "lr": 0.0005278400948495081, "epoch": 10.681282413955682, "percentage": 53.41, "elapsed_time": "1:13:25", "remaining_time": "1:04:03", "throughput": 5179.18, "total_tokens": 22819360} {"current_steps": 22660, "total_steps": 42420, "loss": 0.3472, "lr": 0.0005276346915507363, "epoch": 10.683639792550684, "percentage": 53.42, "elapsed_time": "1:13:26", "remaining_time": "1:04:02", "throughput": 5179.25, "total_tokens": 22824320} {"current_steps": 22665, "total_steps": 42420, "loss": 0.3644, "lr": 0.0005274292835738802, "epoch": 10.685997171145686, "percentage": 53.43, "elapsed_time": "1:13:27", "remaining_time": "1:04:01", "throughput": 5179.24, "total_tokens": 22828768} {"current_steps": 22670, "total_steps": 42420, "loss": 0.4191, "lr": 0.0005272238709537124, "epoch": 10.688354549740689, "percentage": 53.44, "elapsed_time": "1:13:28", "remaining_time": "1:04:00", "throughput": 5179.27, "total_tokens": 22833536} {"current_steps": 22675, "total_steps": 42420, "loss": 0.3612, "lr": 0.0005270184537250052, "epoch": 10.690711928335691, "percentage": 53.45, "elapsed_time": "1:13:29", "remaining_time": "1:03:59", "throughput": 5179.4, "total_tokens": 22839552} {"current_steps": 22680, "total_steps": 42420, "loss": 0.3462, "lr": 0.0005268130319225328, "epoch": 10.693069306930694, "percentage": 53.47, "elapsed_time": "1:13:30", "remaining_time": "1:03:58", "throughput": 5179.44, "total_tokens": 22844448} {"current_steps": 22685, "total_steps": 42420, "loss": 0.386, "lr": 0.0005266076055810691, "epoch": 10.695426685525696, "percentage": 53.48, "elapsed_time": "1:13:31", "remaining_time": "1:03:57", "throughput": 5179.42, "total_tokens": 22848544} {"current_steps": 22690, "total_steps": 42420, "loss": 0.3491, "lr": 0.0005264021747353897, "epoch": 10.697784064120698, "percentage": 53.49, "elapsed_time": "1:13:32", "remaining_time": "1:03:56", "throughput": 5179.52, "total_tokens": 22854112} {"current_steps": 22695, "total_steps": 42420, "loss": 0.3695, "lr": 0.0005261967394202704, "epoch": 10.700141442715701, "percentage": 53.5, "elapsed_time": "1:13:33", "remaining_time": "1:03:55", "throughput": 5179.59, "total_tokens": 22859232} {"current_steps": 22700, "total_steps": 42420, "loss": 0.3358, "lr": 0.0005259912996704877, "epoch": 10.702498821310703, "percentage": 53.51, "elapsed_time": "1:13:34", "remaining_time": "1:03:55", "throughput": 5179.82, "total_tokens": 22866592} {"current_steps": 22705, "total_steps": 42420, "loss": 0.3693, "lr": 0.0005257858555208195, "epoch": 10.704856199905706, "percentage": 53.52, "elapsed_time": "1:13:35", "remaining_time": "1:03:53", "throughput": 5179.8, "total_tokens": 22870784} {"current_steps": 22710, "total_steps": 42420, "loss": 0.3289, "lr": 0.0005255804070060438, "epoch": 10.707213578500706, "percentage": 53.54, "elapsed_time": "1:13:36", "remaining_time": "1:03:52", "throughput": 5179.89, "total_tokens": 22876192} {"current_steps": 22715, "total_steps": 42420, "loss": 0.3226, "lr": 0.0005253749541609394, "epoch": 10.70957095709571, "percentage": 53.55, "elapsed_time": "1:13:37", "remaining_time": "1:03:51", "throughput": 5179.88, "total_tokens": 22880288} {"current_steps": 22720, "total_steps": 42420, "loss": 0.3827, "lr": 0.0005251694970202862, "epoch": 10.711928335690711, "percentage": 53.56, "elapsed_time": "1:13:38", "remaining_time": "1:03:51", "throughput": 5180.1, "total_tokens": 22887968} {"current_steps": 22725, "total_steps": 42420, "loss": 0.3522, "lr": 0.0005249640356188645, "epoch": 10.714285714285714, "percentage": 53.57, "elapsed_time": "1:13:39", "remaining_time": "1:03:50", "throughput": 5180.23, "total_tokens": 22893984} {"current_steps": 22730, "total_steps": 42420, "loss": 0.3451, "lr": 0.0005247585699914554, "epoch": 10.716643092880716, "percentage": 53.58, "elapsed_time": "1:13:40", "remaining_time": "1:03:49", "throughput": 5180.28, "total_tokens": 22899040} {"current_steps": 22735, "total_steps": 42420, "loss": 0.3059, "lr": 0.0005245531001728407, "epoch": 10.719000471475718, "percentage": 53.6, "elapsed_time": "1:13:41", "remaining_time": "1:03:48", "throughput": 5180.27, "total_tokens": 22903200} {"current_steps": 22740, "total_steps": 42420, "loss": 0.3228, "lr": 0.0005243476261978032, "epoch": 10.72135785007072, "percentage": 53.61, "elapsed_time": "1:13:42", "remaining_time": "1:03:47", "throughput": 5180.29, "total_tokens": 22907808} {"current_steps": 22745, "total_steps": 42420, "loss": 0.3606, "lr": 0.0005241421481011258, "epoch": 10.723715228665723, "percentage": 53.62, "elapsed_time": "1:13:43", "remaining_time": "1:03:46", "throughput": 5180.47, "total_tokens": 22914464} {"current_steps": 22750, "total_steps": 42420, "loss": 0.3679, "lr": 0.0005239366659175927, "epoch": 10.726072607260726, "percentage": 53.63, "elapsed_time": "1:13:44", "remaining_time": "1:03:45", "throughput": 5180.52, "total_tokens": 22919392} {"current_steps": 22755, "total_steps": 42420, "loss": 0.3482, "lr": 0.0005237311796819886, "epoch": 10.728429985855728, "percentage": 53.64, "elapsed_time": "1:13:44", "remaining_time": "1:03:44", "throughput": 5180.52, "total_tokens": 22923776} {"current_steps": 22760, "total_steps": 42420, "loss": 0.3252, "lr": 0.0005235256894290986, "epoch": 10.73078736445073, "percentage": 53.65, "elapsed_time": "1:13:45", "remaining_time": "1:03:43", "throughput": 5180.53, "total_tokens": 22928384} {"current_steps": 22765, "total_steps": 42420, "loss": 0.3377, "lr": 0.0005233201951937088, "epoch": 10.733144743045733, "percentage": 53.67, "elapsed_time": "1:13:46", "remaining_time": "1:03:42", "throughput": 5180.59, "total_tokens": 22933632} {"current_steps": 22770, "total_steps": 42420, "loss": 0.3482, "lr": 0.0005231146970106059, "epoch": 10.735502121640735, "percentage": 53.68, "elapsed_time": "1:13:47", "remaining_time": "1:03:41", "throughput": 5180.57, "total_tokens": 22938016} {"current_steps": 22775, "total_steps": 42420, "loss": 0.3759, "lr": 0.0005229091949145774, "epoch": 10.737859500235738, "percentage": 53.69, "elapsed_time": "1:13:48", "remaining_time": "1:03:39", "throughput": 5180.66, "total_tokens": 22943104} {"current_steps": 22780, "total_steps": 42420, "loss": 0.3298, "lr": 0.0005227036889404111, "epoch": 10.74021687883074, "percentage": 53.7, "elapsed_time": "1:13:49", "remaining_time": "1:03:38", "throughput": 5180.69, "total_tokens": 22947744} {"current_steps": 22785, "total_steps": 42420, "loss": 0.3488, "lr": 0.0005224981791228957, "epoch": 10.742574257425742, "percentage": 53.71, "elapsed_time": "1:13:50", "remaining_time": "1:03:37", "throughput": 5180.75, "total_tokens": 22952736} {"current_steps": 22790, "total_steps": 42420, "loss": 0.3363, "lr": 0.0005222926654968206, "epoch": 10.744931636020745, "percentage": 53.72, "elapsed_time": "1:13:51", "remaining_time": "1:03:36", "throughput": 5180.81, "total_tokens": 22957664} {"current_steps": 22795, "total_steps": 42420, "loss": 0.3253, "lr": 0.0005220871480969757, "epoch": 10.747289014615747, "percentage": 53.74, "elapsed_time": "1:13:52", "remaining_time": "1:03:35", "throughput": 5180.97, "total_tokens": 22963936} {"current_steps": 22800, "total_steps": 42420, "loss": 0.2858, "lr": 0.0005218816269581518, "epoch": 10.74964639321075, "percentage": 53.75, "elapsed_time": "1:13:53", "remaining_time": "1:03:34", "throughput": 5180.97, "total_tokens": 22968544} {"current_steps": 22805, "total_steps": 42420, "loss": 0.4026, "lr": 0.0005216761021151398, "epoch": 10.752003771805752, "percentage": 53.76, "elapsed_time": "1:13:54", "remaining_time": "1:03:33", "throughput": 5181.04, "total_tokens": 22973664} {"current_steps": 22810, "total_steps": 42420, "loss": 0.3129, "lr": 0.0005214705736027317, "epoch": 10.754361150400754, "percentage": 53.77, "elapsed_time": "1:13:55", "remaining_time": "1:03:32", "throughput": 5181.07, "total_tokens": 22978496} {"current_steps": 22815, "total_steps": 42420, "loss": 0.3224, "lr": 0.0005212650414557203, "epoch": 10.756718528995757, "percentage": 53.78, "elapsed_time": "1:13:55", "remaining_time": "1:03:31", "throughput": 5181.08, "total_tokens": 22982880} {"current_steps": 22820, "total_steps": 42420, "loss": 0.3637, "lr": 0.0005210595057088983, "epoch": 10.75907590759076, "percentage": 53.8, "elapsed_time": "1:13:56", "remaining_time": "1:03:30", "throughput": 5181.12, "total_tokens": 22987712} {"current_steps": 22825, "total_steps": 42420, "loss": 0.3246, "lr": 0.0005208539663970595, "epoch": 10.761433286185762, "percentage": 53.81, "elapsed_time": "1:13:57", "remaining_time": "1:03:29", "throughput": 5181.15, "total_tokens": 22992288} {"current_steps": 22830, "total_steps": 42420, "loss": 0.3137, "lr": 0.0005206484235549983, "epoch": 10.763790664780764, "percentage": 53.82, "elapsed_time": "1:13:58", "remaining_time": "1:03:28", "throughput": 5181.28, "total_tokens": 22998208} {"current_steps": 22835, "total_steps": 42420, "loss": 0.2664, "lr": 0.0005204428772175098, "epoch": 10.766148043375766, "percentage": 53.83, "elapsed_time": "1:13:59", "remaining_time": "1:03:27", "throughput": 5181.36, "total_tokens": 23003744} {"current_steps": 22840, "total_steps": 42420, "loss": 0.3292, "lr": 0.000520237327419389, "epoch": 10.768505421970769, "percentage": 53.84, "elapsed_time": "1:14:00", "remaining_time": "1:03:26", "throughput": 5181.5, "total_tokens": 23010112} {"current_steps": 22845, "total_steps": 42420, "loss": 0.2868, "lr": 0.0005200317741954324, "epoch": 10.770862800565771, "percentage": 53.85, "elapsed_time": "1:14:01", "remaining_time": "1:03:26", "throughput": 5181.62, "total_tokens": 23015744} {"current_steps": 22850, "total_steps": 42420, "loss": 0.4191, "lr": 0.0005198262175804366, "epoch": 10.773220179160774, "percentage": 53.87, "elapsed_time": "1:14:02", "remaining_time": "1:03:24", "throughput": 5181.64, "total_tokens": 23020416} {"current_steps": 22855, "total_steps": 42420, "loss": 0.3004, "lr": 0.0005196206576091986, "epoch": 10.775577557755776, "percentage": 53.88, "elapsed_time": "1:14:03", "remaining_time": "1:03:23", "throughput": 5181.71, "total_tokens": 23025536} {"current_steps": 22860, "total_steps": 42420, "loss": 0.3189, "lr": 0.0005194150943165167, "epoch": 10.777934936350778, "percentage": 53.89, "elapsed_time": "1:14:04", "remaining_time": "1:03:22", "throughput": 5181.72, "total_tokens": 23030048} {"current_steps": 22865, "total_steps": 42420, "loss": 0.316, "lr": 0.0005192095277371888, "epoch": 10.78029231494578, "percentage": 53.9, "elapsed_time": "1:14:05", "remaining_time": "1:03:21", "throughput": 5181.8, "total_tokens": 23035392} {"current_steps": 22870, "total_steps": 42420, "loss": 0.3328, "lr": 0.0005190039579060139, "epoch": 10.782649693540783, "percentage": 53.91, "elapsed_time": "1:14:06", "remaining_time": "1:03:20", "throughput": 5181.84, "total_tokens": 23040416} {"current_steps": 22875, "total_steps": 42420, "loss": 0.3318, "lr": 0.0005187983848577917, "epoch": 10.785007072135786, "percentage": 53.93, "elapsed_time": "1:14:07", "remaining_time": "1:03:19", "throughput": 5181.92, "total_tokens": 23045440} {"current_steps": 22880, "total_steps": 42420, "loss": 0.2764, "lr": 0.0005185928086273222, "epoch": 10.787364450730788, "percentage": 53.94, "elapsed_time": "1:14:08", "remaining_time": "1:03:18", "throughput": 5181.96, "total_tokens": 23050208} {"current_steps": 22885, "total_steps": 42420, "loss": 0.3605, "lr": 0.0005183872292494058, "epoch": 10.78972182932579, "percentage": 53.95, "elapsed_time": "1:14:09", "remaining_time": "1:03:17", "throughput": 5182.11, "total_tokens": 23056448} {"current_steps": 22890, "total_steps": 42420, "loss": 0.3143, "lr": 0.0005181816467588437, "epoch": 10.792079207920793, "percentage": 53.96, "elapsed_time": "1:14:10", "remaining_time": "1:03:16", "throughput": 5182.18, "total_tokens": 23061568} {"current_steps": 22895, "total_steps": 42420, "loss": 0.3654, "lr": 0.0005179760611904374, "epoch": 10.794436586515795, "percentage": 53.97, "elapsed_time": "1:14:11", "remaining_time": "1:03:15", "throughput": 5182.29, "total_tokens": 23067232} {"current_steps": 22900, "total_steps": 42420, "loss": 0.3827, "lr": 0.0005177704725789895, "epoch": 10.796793965110798, "percentage": 53.98, "elapsed_time": "1:14:12", "remaining_time": "1:03:14", "throughput": 5182.35, "total_tokens": 23072416} {"current_steps": 22905, "total_steps": 42420, "loss": 0.3373, "lr": 0.000517564880959302, "epoch": 10.799151343705798, "percentage": 54.0, "elapsed_time": "1:14:12", "remaining_time": "1:03:13", "throughput": 5182.37, "total_tokens": 23076736} {"current_steps": 22910, "total_steps": 42420, "loss": 0.3435, "lr": 0.0005173592863661785, "epoch": 10.801508722300802, "percentage": 54.01, "elapsed_time": "1:14:13", "remaining_time": "1:03:12", "throughput": 5182.5, "total_tokens": 23082528} {"current_steps": 22915, "total_steps": 42420, "loss": 0.3469, "lr": 0.0005171536888344225, "epoch": 10.803866100895803, "percentage": 54.02, "elapsed_time": "1:14:15", "remaining_time": "1:03:12", "throughput": 5182.67, "total_tokens": 23088960} {"current_steps": 22920, "total_steps": 42420, "loss": 0.3323, "lr": 0.0005169480883988383, "epoch": 10.806223479490805, "percentage": 54.03, "elapsed_time": "1:14:15", "remaining_time": "1:03:10", "throughput": 5182.65, "total_tokens": 23093248} {"current_steps": 22925, "total_steps": 42420, "loss": 0.3168, "lr": 0.0005167424850942305, "epoch": 10.808580858085808, "percentage": 54.04, "elapsed_time": "1:14:16", "remaining_time": "1:03:10", "throughput": 5182.77, "total_tokens": 23099232} {"current_steps": 22930, "total_steps": 42420, "loss": 0.3097, "lr": 0.0005165368789554043, "epoch": 10.81093823668081, "percentage": 54.05, "elapsed_time": "1:14:17", "remaining_time": "1:03:09", "throughput": 5182.86, "total_tokens": 23104896} {"current_steps": 22935, "total_steps": 42420, "loss": 0.3917, "lr": 0.0005163312700171651, "epoch": 10.813295615275813, "percentage": 54.07, "elapsed_time": "1:14:18", "remaining_time": "1:03:08", "throughput": 5182.92, "total_tokens": 23109984} {"current_steps": 22940, "total_steps": 42420, "loss": 0.3178, "lr": 0.0005161256583143194, "epoch": 10.815652993870815, "percentage": 54.08, "elapsed_time": "1:14:19", "remaining_time": "1:03:07", "throughput": 5182.95, "total_tokens": 23114912} {"current_steps": 22945, "total_steps": 42420, "loss": 0.3329, "lr": 0.0005159200438816734, "epoch": 10.818010372465817, "percentage": 54.09, "elapsed_time": "1:14:20", "remaining_time": "1:03:06", "throughput": 5183.05, "total_tokens": 23120448} {"current_steps": 22950, "total_steps": 42420, "loss": 0.3586, "lr": 0.0005157144267540341, "epoch": 10.82036775106082, "percentage": 54.1, "elapsed_time": "1:14:21", "remaining_time": "1:03:05", "throughput": 5183.1, "total_tokens": 23125600} {"current_steps": 22955, "total_steps": 42420, "loss": 0.4025, "lr": 0.0005155088069662094, "epoch": 10.822725129655822, "percentage": 54.11, "elapsed_time": "1:14:22", "remaining_time": "1:03:04", "throughput": 5183.16, "total_tokens": 23130752} {"current_steps": 22960, "total_steps": 42420, "loss": 0.3257, "lr": 0.000515303184553007, "epoch": 10.825082508250825, "percentage": 54.13, "elapsed_time": "1:14:23", "remaining_time": "1:03:03", "throughput": 5183.23, "total_tokens": 23135840} {"current_steps": 22965, "total_steps": 42420, "loss": 0.3452, "lr": 0.0005150975595492353, "epoch": 10.827439886845827, "percentage": 54.14, "elapsed_time": "1:14:24", "remaining_time": "1:03:02", "throughput": 5183.26, "total_tokens": 23140544} {"current_steps": 22970, "total_steps": 42420, "loss": 0.3465, "lr": 0.000514891931989703, "epoch": 10.82979726544083, "percentage": 54.15, "elapsed_time": "1:14:25", "remaining_time": "1:03:01", "throughput": 5183.29, "total_tokens": 23145184} {"current_steps": 22975, "total_steps": 42420, "loss": 0.3686, "lr": 0.0005146863019092196, "epoch": 10.832154644035832, "percentage": 54.16, "elapsed_time": "1:14:26", "remaining_time": "1:03:00", "throughput": 5183.35, "total_tokens": 23150592} {"current_steps": 22980, "total_steps": 42420, "loss": 0.3162, "lr": 0.0005144806693425945, "epoch": 10.834512022630834, "percentage": 54.17, "elapsed_time": "1:14:27", "remaining_time": "1:02:59", "throughput": 5183.47, "total_tokens": 23156384} {"current_steps": 22985, "total_steps": 42420, "loss": 0.3283, "lr": 0.0005142750343246382, "epoch": 10.836869401225837, "percentage": 54.18, "elapsed_time": "1:14:28", "remaining_time": "1:02:58", "throughput": 5183.63, "total_tokens": 23162720} {"current_steps": 22990, "total_steps": 42420, "loss": 0.3082, "lr": 0.0005140693968901607, "epoch": 10.839226779820839, "percentage": 54.2, "elapsed_time": "1:14:29", "remaining_time": "1:02:57", "throughput": 5183.68, "total_tokens": 23167648} {"current_steps": 22995, "total_steps": 42420, "loss": 0.3542, "lr": 0.0005138637570739732, "epoch": 10.841584158415841, "percentage": 54.21, "elapsed_time": "1:14:30", "remaining_time": "1:02:56", "throughput": 5183.76, "total_tokens": 23173088} {"current_steps": 23000, "total_steps": 42420, "loss": 0.3999, "lr": 0.0005136581149108871, "epoch": 10.843941537010844, "percentage": 54.22, "elapsed_time": "1:14:31", "remaining_time": "1:02:55", "throughput": 5183.76, "total_tokens": 23177760} {"current_steps": 23005, "total_steps": 42420, "loss": 0.2963, "lr": 0.0005134524704357142, "epoch": 10.846298915605846, "percentage": 54.23, "elapsed_time": "1:14:32", "remaining_time": "1:02:54", "throughput": 5183.77, "total_tokens": 23182240} {"current_steps": 23010, "total_steps": 42420, "loss": 0.3387, "lr": 0.0005132468236832662, "epoch": 10.848656294200849, "percentage": 54.24, "elapsed_time": "1:14:33", "remaining_time": "1:02:53", "throughput": 5183.81, "total_tokens": 23187328} {"current_steps": 23015, "total_steps": 42420, "loss": 0.3813, "lr": 0.000513041174688356, "epoch": 10.851013672795851, "percentage": 54.26, "elapsed_time": "1:14:33", "remaining_time": "1:02:52", "throughput": 5183.82, "total_tokens": 23191616} {"current_steps": 23020, "total_steps": 42420, "loss": 0.3517, "lr": 0.0005128355234857963, "epoch": 10.853371051390853, "percentage": 54.27, "elapsed_time": "1:14:34", "remaining_time": "1:02:51", "throughput": 5183.85, "total_tokens": 23196416} {"current_steps": 23025, "total_steps": 42420, "loss": 0.3678, "lr": 0.0005126298701104005, "epoch": 10.855728429985856, "percentage": 54.28, "elapsed_time": "1:14:35", "remaining_time": "1:02:50", "throughput": 5183.96, "total_tokens": 23201952} {"current_steps": 23030, "total_steps": 42420, "loss": 0.3702, "lr": 0.000512424214596982, "epoch": 10.858085808580858, "percentage": 54.29, "elapsed_time": "1:14:36", "remaining_time": "1:02:49", "throughput": 5184.0, "total_tokens": 23206816} {"current_steps": 23035, "total_steps": 42420, "loss": 0.3576, "lr": 0.000512218556980355, "epoch": 10.86044318717586, "percentage": 54.3, "elapsed_time": "1:14:37", "remaining_time": "1:02:48", "throughput": 5184.04, "total_tokens": 23211808} {"current_steps": 23040, "total_steps": 42420, "loss": 0.35, "lr": 0.0005120128972953337, "epoch": 10.862800565770863, "percentage": 54.31, "elapsed_time": "1:14:38", "remaining_time": "1:02:47", "throughput": 5184.08, "total_tokens": 23216896} {"current_steps": 23045, "total_steps": 42420, "loss": 0.3474, "lr": 0.000511807235576733, "epoch": 10.865157944365865, "percentage": 54.33, "elapsed_time": "1:14:39", "remaining_time": "1:02:46", "throughput": 5184.14, "total_tokens": 23221952} {"current_steps": 23050, "total_steps": 42420, "loss": 0.2864, "lr": 0.0005116015718593678, "epoch": 10.867515322960868, "percentage": 54.34, "elapsed_time": "1:14:40", "remaining_time": "1:02:45", "throughput": 5184.23, "total_tokens": 23227328} {"current_steps": 23055, "total_steps": 42420, "loss": 0.2643, "lr": 0.0005113959061780532, "epoch": 10.86987270155587, "percentage": 54.35, "elapsed_time": "1:14:41", "remaining_time": "1:02:43", "throughput": 5184.23, "total_tokens": 23231520} {"current_steps": 23060, "total_steps": 42420, "loss": 0.2923, "lr": 0.0005111902385676054, "epoch": 10.872230080150873, "percentage": 54.36, "elapsed_time": "1:14:41", "remaining_time": "1:02:42", "throughput": 5184.19, "total_tokens": 23235296} {"current_steps": 23065, "total_steps": 42420, "loss": 0.2804, "lr": 0.0005109845690628402, "epoch": 10.874587458745875, "percentage": 54.37, "elapsed_time": "1:14:43", "remaining_time": "1:02:41", "throughput": 5184.33, "total_tokens": 23241568} {"current_steps": 23070, "total_steps": 42420, "loss": 0.4136, "lr": 0.0005107788976985739, "epoch": 10.876944837340877, "percentage": 54.38, "elapsed_time": "1:14:43", "remaining_time": "1:02:40", "throughput": 5184.39, "total_tokens": 23246624} {"current_steps": 23075, "total_steps": 42420, "loss": 0.3405, "lr": 0.0005105732245096233, "epoch": 10.87930221593588, "percentage": 54.4, "elapsed_time": "1:14:44", "remaining_time": "1:02:39", "throughput": 5184.35, "total_tokens": 23250592} {"current_steps": 23080, "total_steps": 42420, "loss": 0.3558, "lr": 0.0005103675495308054, "epoch": 10.881659594530882, "percentage": 54.41, "elapsed_time": "1:14:45", "remaining_time": "1:02:38", "throughput": 5184.33, "total_tokens": 23254944} {"current_steps": 23085, "total_steps": 42420, "loss": 0.3599, "lr": 0.0005101618727969371, "epoch": 10.884016973125885, "percentage": 54.42, "elapsed_time": "1:14:46", "remaining_time": "1:02:37", "throughput": 5184.29, "total_tokens": 23259104} {"current_steps": 23090, "total_steps": 42420, "loss": 0.3621, "lr": 0.0005099561943428365, "epoch": 10.886374351720887, "percentage": 54.43, "elapsed_time": "1:14:47", "remaining_time": "1:02:36", "throughput": 5184.3, "total_tokens": 23263616} {"current_steps": 23095, "total_steps": 42420, "loss": 0.3528, "lr": 0.0005097505142033212, "epoch": 10.88873173031589, "percentage": 54.44, "elapsed_time": "1:14:48", "remaining_time": "1:02:35", "throughput": 5184.31, "total_tokens": 23268032} {"current_steps": 23100, "total_steps": 42420, "loss": 0.3521, "lr": 0.0005095448324132093, "epoch": 10.891089108910892, "percentage": 54.46, "elapsed_time": "1:14:49", "remaining_time": "1:02:34", "throughput": 5184.34, "total_tokens": 23272736} {"current_steps": 23105, "total_steps": 42420, "loss": 0.3801, "lr": 0.0005093391490073194, "epoch": 10.893446487505894, "percentage": 54.47, "elapsed_time": "1:14:49", "remaining_time": "1:02:33", "throughput": 5184.32, "total_tokens": 23277152} {"current_steps": 23110, "total_steps": 42420, "loss": 0.3254, "lr": 0.0005091334640204702, "epoch": 10.895803866100895, "percentage": 54.48, "elapsed_time": "1:14:50", "remaining_time": "1:02:32", "throughput": 5184.45, "total_tokens": 23283264} {"current_steps": 23115, "total_steps": 42420, "loss": 0.391, "lr": 0.0005089277774874806, "epoch": 10.898161244695899, "percentage": 54.49, "elapsed_time": "1:14:51", "remaining_time": "1:02:31", "throughput": 5184.53, "total_tokens": 23288512} {"current_steps": 23120, "total_steps": 42420, "loss": 0.3442, "lr": 0.0005087220894431697, "epoch": 10.9005186232909, "percentage": 54.5, "elapsed_time": "1:14:52", "remaining_time": "1:02:30", "throughput": 5184.54, "total_tokens": 23293184} {"current_steps": 23125, "total_steps": 42420, "loss": 0.3153, "lr": 0.0005085163999223575, "epoch": 10.902876001885902, "percentage": 54.51, "elapsed_time": "1:14:53", "remaining_time": "1:02:29", "throughput": 5184.56, "total_tokens": 23297760} {"current_steps": 23130, "total_steps": 42420, "loss": 0.2959, "lr": 0.0005083107089598632, "epoch": 10.905233380480905, "percentage": 54.53, "elapsed_time": "1:14:54", "remaining_time": "1:02:28", "throughput": 5184.64, "total_tokens": 23303136} {"current_steps": 23135, "total_steps": 42420, "loss": 0.3384, "lr": 0.0005081050165905069, "epoch": 10.907590759075907, "percentage": 54.54, "elapsed_time": "1:14:55", "remaining_time": "1:02:27", "throughput": 5184.73, "total_tokens": 23308384} {"current_steps": 23140, "total_steps": 42420, "loss": 0.378, "lr": 0.0005078993228491089, "epoch": 10.90994813767091, "percentage": 54.55, "elapsed_time": "1:14:56", "remaining_time": "1:02:26", "throughput": 5184.72, "total_tokens": 23312800} {"current_steps": 23145, "total_steps": 42420, "loss": 0.3819, "lr": 0.0005076936277704898, "epoch": 10.912305516265912, "percentage": 54.56, "elapsed_time": "1:14:57", "remaining_time": "1:02:25", "throughput": 5184.69, "total_tokens": 23316640} {"current_steps": 23150, "total_steps": 42420, "loss": 0.3383, "lr": 0.0005074879313894701, "epoch": 10.914662894860914, "percentage": 54.57, "elapsed_time": "1:14:58", "remaining_time": "1:02:24", "throughput": 5184.74, "total_tokens": 23321504} {"current_steps": 23155, "total_steps": 42420, "loss": 0.366, "lr": 0.0005072822337408707, "epoch": 10.917020273455917, "percentage": 54.59, "elapsed_time": "1:14:59", "remaining_time": "1:02:23", "throughput": 5184.84, "total_tokens": 23327040} {"current_steps": 23160, "total_steps": 42420, "loss": 0.3287, "lr": 0.0005070765348595128, "epoch": 10.919377652050919, "percentage": 54.6, "elapsed_time": "1:15:00", "remaining_time": "1:02:22", "throughput": 5184.93, "total_tokens": 23332800} {"current_steps": 23165, "total_steps": 42420, "loss": 0.3056, "lr": 0.0005068708347802177, "epoch": 10.921735030645921, "percentage": 54.61, "elapsed_time": "1:15:01", "remaining_time": "1:02:21", "throughput": 5184.97, "total_tokens": 23337824} {"current_steps": 23170, "total_steps": 42420, "loss": 0.3316, "lr": 0.000506665133537807, "epoch": 10.924092409240924, "percentage": 54.62, "elapsed_time": "1:15:02", "remaining_time": "1:02:20", "throughput": 5185.17, "total_tokens": 23344992} {"current_steps": 23175, "total_steps": 42420, "loss": 0.3646, "lr": 0.0005064594311671022, "epoch": 10.926449787835926, "percentage": 54.63, "elapsed_time": "1:15:03", "remaining_time": "1:02:19", "throughput": 5185.26, "total_tokens": 23350560} {"current_steps": 23180, "total_steps": 42420, "loss": 0.4214, "lr": 0.0005062537277029254, "epoch": 10.928807166430929, "percentage": 54.64, "elapsed_time": "1:15:04", "remaining_time": "1:02:18", "throughput": 5185.29, "total_tokens": 23355200} {"current_steps": 23185, "total_steps": 42420, "loss": 0.3746, "lr": 0.0005060480231800988, "epoch": 10.931164545025931, "percentage": 54.66, "elapsed_time": "1:15:05", "remaining_time": "1:02:17", "throughput": 5185.31, "total_tokens": 23359840} {"current_steps": 23190, "total_steps": 42420, "loss": 0.3092, "lr": 0.0005058423176334446, "epoch": 10.933521923620933, "percentage": 54.67, "elapsed_time": "1:15:05", "remaining_time": "1:02:16", "throughput": 5185.33, "total_tokens": 23364320} {"current_steps": 23195, "total_steps": 42420, "loss": 0.3017, "lr": 0.0005056366110977853, "epoch": 10.935879302215936, "percentage": 54.68, "elapsed_time": "1:15:06", "remaining_time": "1:02:15", "throughput": 5185.47, "total_tokens": 23370336} {"current_steps": 23200, "total_steps": 42420, "loss": 0.3589, "lr": 0.0005054309036079432, "epoch": 10.938236680810938, "percentage": 54.69, "elapsed_time": "1:15:07", "remaining_time": "1:02:14", "throughput": 5185.48, "total_tokens": 23374848} {"current_steps": 23205, "total_steps": 42420, "loss": 0.2843, "lr": 0.0005052251951987415, "epoch": 10.94059405940594, "percentage": 54.7, "elapsed_time": "1:15:08", "remaining_time": "1:02:13", "throughput": 5185.49, "total_tokens": 23379328} {"current_steps": 23210, "total_steps": 42420, "loss": 0.367, "lr": 0.0005050194859050028, "epoch": 10.942951438000943, "percentage": 54.71, "elapsed_time": "1:15:09", "remaining_time": "1:02:12", "throughput": 5185.57, "total_tokens": 23384640} {"current_steps": 23215, "total_steps": 42420, "loss": 0.3681, "lr": 0.0005048137757615505, "epoch": 10.945308816595945, "percentage": 54.73, "elapsed_time": "1:15:10", "remaining_time": "1:02:11", "throughput": 5185.64, "total_tokens": 23389920} {"current_steps": 23220, "total_steps": 42420, "loss": 0.2851, "lr": 0.0005046080648032076, "epoch": 10.947666195190948, "percentage": 54.74, "elapsed_time": "1:15:11", "remaining_time": "1:02:10", "throughput": 5185.65, "total_tokens": 23394432} {"current_steps": 23225, "total_steps": 42420, "loss": 0.3153, "lr": 0.0005044023530647976, "epoch": 10.95002357378595, "percentage": 54.75, "elapsed_time": "1:15:12", "remaining_time": "1:02:09", "throughput": 5185.67, "total_tokens": 23399104} {"current_steps": 23230, "total_steps": 42420, "loss": 0.2225, "lr": 0.0005041966405811441, "epoch": 10.952380952380953, "percentage": 54.76, "elapsed_time": "1:15:13", "remaining_time": "1:02:08", "throughput": 5185.83, "total_tokens": 23405632} {"current_steps": 23235, "total_steps": 42420, "loss": 0.406, "lr": 0.0005039909273870704, "epoch": 10.954738330975955, "percentage": 54.77, "elapsed_time": "1:15:14", "remaining_time": "1:02:07", "throughput": 5185.82, "total_tokens": 23410208} {"current_steps": 23240, "total_steps": 42420, "loss": 0.3174, "lr": 0.0005037852135174006, "epoch": 10.957095709570957, "percentage": 54.79, "elapsed_time": "1:15:15", "remaining_time": "1:02:06", "throughput": 5185.89, "total_tokens": 23415360} {"current_steps": 23245, "total_steps": 42420, "loss": 0.3098, "lr": 0.0005035794990069584, "epoch": 10.95945308816596, "percentage": 54.8, "elapsed_time": "1:15:16", "remaining_time": "1:02:05", "throughput": 5185.94, "total_tokens": 23420512} {"current_steps": 23250, "total_steps": 42420, "loss": 0.3403, "lr": 0.0005033737838905677, "epoch": 10.961810466760962, "percentage": 54.81, "elapsed_time": "1:15:17", "remaining_time": "1:02:04", "throughput": 5186.03, "total_tokens": 23426112} {"current_steps": 23255, "total_steps": 42420, "loss": 0.3625, "lr": 0.0005031680682030528, "epoch": 10.964167845355965, "percentage": 54.82, "elapsed_time": "1:15:18", "remaining_time": "1:02:03", "throughput": 5186.12, "total_tokens": 23431648} {"current_steps": 23260, "total_steps": 42420, "loss": 0.3756, "lr": 0.0005029623519792377, "epoch": 10.966525223950967, "percentage": 54.83, "elapsed_time": "1:15:19", "remaining_time": "1:02:02", "throughput": 5186.23, "total_tokens": 23437312} {"current_steps": 23265, "total_steps": 42420, "loss": 0.3445, "lr": 0.0005027566352539467, "epoch": 10.96888260254597, "percentage": 54.84, "elapsed_time": "1:15:19", "remaining_time": "1:02:01", "throughput": 5186.21, "total_tokens": 23441600} {"current_steps": 23270, "total_steps": 42420, "loss": 0.3482, "lr": 0.0005025509180620041, "epoch": 10.971239981140972, "percentage": 54.86, "elapsed_time": "1:15:20", "remaining_time": "1:02:00", "throughput": 5186.32, "total_tokens": 23447296} {"current_steps": 23275, "total_steps": 42420, "loss": 0.3602, "lr": 0.0005023452004382345, "epoch": 10.973597359735974, "percentage": 54.87, "elapsed_time": "1:15:21", "remaining_time": "1:01:59", "throughput": 5186.35, "total_tokens": 23451648} {"current_steps": 23280, "total_steps": 42420, "loss": 0.3497, "lr": 0.0005021394824174622, "epoch": 10.975954738330977, "percentage": 54.88, "elapsed_time": "1:15:22", "remaining_time": "1:01:58", "throughput": 5186.36, "total_tokens": 23456160} {"current_steps": 23285, "total_steps": 42420, "loss": 0.3211, "lr": 0.0005019337640345119, "epoch": 10.978312116925979, "percentage": 54.89, "elapsed_time": "1:15:23", "remaining_time": "1:01:57", "throughput": 5186.32, "total_tokens": 23459936} {"current_steps": 23290, "total_steps": 42420, "loss": 0.39, "lr": 0.000501728045324208, "epoch": 10.980669495520981, "percentage": 54.9, "elapsed_time": "1:15:24", "remaining_time": "1:01:56", "throughput": 5186.51, "total_tokens": 23466720} {"current_steps": 23295, "total_steps": 42420, "loss": 0.3569, "lr": 0.0005015223263213756, "epoch": 10.983026874115984, "percentage": 54.92, "elapsed_time": "1:15:25", "remaining_time": "1:01:55", "throughput": 5186.57, "total_tokens": 23471840} {"current_steps": 23300, "total_steps": 42420, "loss": 0.2708, "lr": 0.0005013166070608392, "epoch": 10.985384252710986, "percentage": 54.93, "elapsed_time": "1:15:26", "remaining_time": "1:01:54", "throughput": 5186.65, "total_tokens": 23477408} {"current_steps": 23305, "total_steps": 42420, "loss": 0.3267, "lr": 0.0005011108875774235, "epoch": 10.987741631305987, "percentage": 54.94, "elapsed_time": "1:15:27", "remaining_time": "1:01:53", "throughput": 5186.74, "total_tokens": 23482720} {"current_steps": 23310, "total_steps": 42420, "loss": 0.3259, "lr": 0.0005009051679059534, "epoch": 10.990099009900991, "percentage": 54.95, "elapsed_time": "1:15:28", "remaining_time": "1:01:52", "throughput": 5186.84, "total_tokens": 23488448} {"current_steps": 23315, "total_steps": 42420, "loss": 0.3648, "lr": 0.0005006994480812538, "epoch": 10.992456388495992, "percentage": 54.96, "elapsed_time": "1:15:29", "remaining_time": "1:01:51", "throughput": 5186.94, "total_tokens": 23494080} {"current_steps": 23320, "total_steps": 42420, "loss": 0.2927, "lr": 0.0005004937281381495, "epoch": 10.994813767090994, "percentage": 54.97, "elapsed_time": "1:15:30", "remaining_time": "1:01:50", "throughput": 5186.97, "total_tokens": 23498912} {"current_steps": 23325, "total_steps": 42420, "loss": 0.359, "lr": 0.0005002880081114653, "epoch": 10.997171145685996, "percentage": 54.99, "elapsed_time": "1:15:31", "remaining_time": "1:01:49", "throughput": 5187.04, "total_tokens": 23504128} {"current_steps": 23330, "total_steps": 42420, "loss": 0.3377, "lr": 0.0005000822880360261, "epoch": 10.999528524280999, "percentage": 55.0, "elapsed_time": "1:15:32", "remaining_time": "1:01:48", "throughput": 5187.12, "total_tokens": 23509408} {"current_steps": 23331, "total_steps": 42420, "eval_loss": 0.3380082845687866, "epoch": 11.0, "percentage": 55.0, "elapsed_time": "1:15:51", "remaining_time": "1:02:04", "throughput": 5165.33, "total_tokens": 23510432} {"current_steps": 23335, "total_steps": 42420, "loss": 0.3405, "lr": 0.0004998765679466572, "epoch": 11.001885902876001, "percentage": 55.01, "elapsed_time": "1:15:54", "remaining_time": "1:02:04", "throughput": 5163.19, "total_tokens": 23514688} {"current_steps": 23340, "total_steps": 42420, "loss": 0.354, "lr": 0.0004996708478781832, "epoch": 11.004243281471004, "percentage": 55.02, "elapsed_time": "1:15:55", "remaining_time": "1:02:03", "throughput": 5163.3, "total_tokens": 23520640} {"current_steps": 23345, "total_steps": 42420, "loss": 0.3217, "lr": 0.000499465127865429, "epoch": 11.006600660066006, "percentage": 55.03, "elapsed_time": "1:15:56", "remaining_time": "1:02:02", "throughput": 5163.37, "total_tokens": 23525696} {"current_steps": 23350, "total_steps": 42420, "loss": 0.3488, "lr": 0.0004992594079432195, "epoch": 11.008958038661008, "percentage": 55.04, "elapsed_time": "1:15:57", "remaining_time": "1:02:01", "throughput": 5163.46, "total_tokens": 23531200} {"current_steps": 23355, "total_steps": 42420, "loss": 0.3147, "lr": 0.00049905368814638, "epoch": 11.01131541725601, "percentage": 55.06, "elapsed_time": "1:15:58", "remaining_time": "1:02:00", "throughput": 5163.5, "total_tokens": 23535872} {"current_steps": 23360, "total_steps": 42420, "loss": 0.2947, "lr": 0.0004988479685097348, "epoch": 11.013672795851013, "percentage": 55.07, "elapsed_time": "1:15:59", "remaining_time": "1:01:59", "throughput": 5163.6, "total_tokens": 23541504} {"current_steps": 23365, "total_steps": 42420, "loss": 0.385, "lr": 0.000498642249068109, "epoch": 11.016030174446016, "percentage": 55.08, "elapsed_time": "1:16:00", "remaining_time": "1:01:58", "throughput": 5163.66, "total_tokens": 23546464} {"current_steps": 23370, "total_steps": 42420, "loss": 0.3093, "lr": 0.0004984365298563273, "epoch": 11.018387553041018, "percentage": 55.09, "elapsed_time": "1:16:00", "remaining_time": "1:01:57", "throughput": 5163.76, "total_tokens": 23551872} {"current_steps": 23375, "total_steps": 42420, "loss": 0.376, "lr": 0.0004982308109092146, "epoch": 11.02074493163602, "percentage": 55.1, "elapsed_time": "1:16:01", "remaining_time": "1:01:56", "throughput": 5163.79, "total_tokens": 23556512} {"current_steps": 23380, "total_steps": 42420, "loss": 0.3009, "lr": 0.0004980250922615956, "epoch": 11.023102310231023, "percentage": 55.12, "elapsed_time": "1:16:02", "remaining_time": "1:01:55", "throughput": 5163.93, "total_tokens": 23562400} {"current_steps": 23385, "total_steps": 42420, "loss": 0.3603, "lr": 0.0004978193739482949, "epoch": 11.025459688826025, "percentage": 55.13, "elapsed_time": "1:16:03", "remaining_time": "1:01:54", "throughput": 5163.95, "total_tokens": 23567104} {"current_steps": 23390, "total_steps": 42420, "loss": 0.3145, "lr": 0.0004976136560041371, "epoch": 11.027817067421028, "percentage": 55.14, "elapsed_time": "1:16:04", "remaining_time": "1:01:53", "throughput": 5163.99, "total_tokens": 23571776} {"current_steps": 23395, "total_steps": 42420, "loss": 0.3626, "lr": 0.0004974079384639469, "epoch": 11.03017444601603, "percentage": 55.15, "elapsed_time": "1:16:05", "remaining_time": "1:01:52", "throughput": 5164.03, "total_tokens": 23576480} {"current_steps": 23400, "total_steps": 42420, "loss": 0.3472, "lr": 0.0004972022213625488, "epoch": 11.032531824611032, "percentage": 55.16, "elapsed_time": "1:16:06", "remaining_time": "1:01:51", "throughput": 5164.09, "total_tokens": 23581504} {"current_steps": 23405, "total_steps": 42420, "loss": 0.3617, "lr": 0.000496996504734767, "epoch": 11.034889203206035, "percentage": 55.17, "elapsed_time": "1:16:07", "remaining_time": "1:01:50", "throughput": 5164.13, "total_tokens": 23586144} {"current_steps": 23410, "total_steps": 42420, "loss": 0.3409, "lr": 0.0004967907886154259, "epoch": 11.037246581801037, "percentage": 55.19, "elapsed_time": "1:16:08", "remaining_time": "1:01:49", "throughput": 5164.22, "total_tokens": 23591424} {"current_steps": 23415, "total_steps": 42420, "loss": 0.3308, "lr": 0.0004965850730393499, "epoch": 11.03960396039604, "percentage": 55.2, "elapsed_time": "1:16:09", "remaining_time": "1:01:48", "throughput": 5164.32, "total_tokens": 23596896} {"current_steps": 23420, "total_steps": 42420, "loss": 0.3477, "lr": 0.0004963793580413632, "epoch": 11.041961338991042, "percentage": 55.21, "elapsed_time": "1:16:10", "remaining_time": "1:01:47", "throughput": 5164.36, "total_tokens": 23601856} {"current_steps": 23425, "total_steps": 42420, "loss": 0.3374, "lr": 0.0004961736436562894, "epoch": 11.044318717586044, "percentage": 55.22, "elapsed_time": "1:16:10", "remaining_time": "1:01:46", "throughput": 5164.38, "total_tokens": 23606368} {"current_steps": 23430, "total_steps": 42420, "loss": 0.3555, "lr": 0.0004959679299189529, "epoch": 11.046676096181047, "percentage": 55.23, "elapsed_time": "1:16:12", "remaining_time": "1:01:45", "throughput": 5164.57, "total_tokens": 23613312} {"current_steps": 23435, "total_steps": 42420, "loss": 0.33, "lr": 0.0004957622168641774, "epoch": 11.04903347477605, "percentage": 55.25, "elapsed_time": "1:16:13", "remaining_time": "1:01:44", "throughput": 5164.64, "total_tokens": 23618528} {"current_steps": 23440, "total_steps": 42420, "loss": 0.362, "lr": 0.0004955565045267866, "epoch": 11.051390853371052, "percentage": 55.26, "elapsed_time": "1:16:14", "remaining_time": "1:01:43", "throughput": 5164.72, "total_tokens": 23623872} {"current_steps": 23445, "total_steps": 42420, "loss": 0.3229, "lr": 0.0004953507929416043, "epoch": 11.053748231966054, "percentage": 55.27, "elapsed_time": "1:16:15", "remaining_time": "1:01:42", "throughput": 5164.85, "total_tokens": 23629536} {"current_steps": 23450, "total_steps": 42420, "loss": 0.3923, "lr": 0.0004951450821434536, "epoch": 11.056105610561056, "percentage": 55.28, "elapsed_time": "1:16:16", "remaining_time": "1:01:41", "throughput": 5164.98, "total_tokens": 23635232} {"current_steps": 23455, "total_steps": 42420, "loss": 0.3576, "lr": 0.0004949393721671581, "epoch": 11.058462989156059, "percentage": 55.29, "elapsed_time": "1:16:16", "remaining_time": "1:01:40", "throughput": 5165.0, "total_tokens": 23639872} {"current_steps": 23460, "total_steps": 42420, "loss": 0.3577, "lr": 0.000494733663047541, "epoch": 11.060820367751061, "percentage": 55.3, "elapsed_time": "1:16:17", "remaining_time": "1:01:39", "throughput": 5165.0, "total_tokens": 23644032} {"current_steps": 23465, "total_steps": 42420, "loss": 0.335, "lr": 0.0004945279548194254, "epoch": 11.063177746346064, "percentage": 55.32, "elapsed_time": "1:16:18", "remaining_time": "1:01:38", "throughput": 5165.04, "total_tokens": 23648704} {"current_steps": 23470, "total_steps": 42420, "loss": 0.3342, "lr": 0.000494322247517634, "epoch": 11.065535124941066, "percentage": 55.33, "elapsed_time": "1:16:19", "remaining_time": "1:01:37", "throughput": 5165.17, "total_tokens": 23654624} {"current_steps": 23475, "total_steps": 42420, "loss": 0.2988, "lr": 0.0004941165411769899, "epoch": 11.067892503536068, "percentage": 55.34, "elapsed_time": "1:16:20", "remaining_time": "1:01:36", "throughput": 5165.31, "total_tokens": 23660672} {"current_steps": 23480, "total_steps": 42420, "loss": 0.4095, "lr": 0.0004939108358323154, "epoch": 11.07024988213107, "percentage": 55.35, "elapsed_time": "1:16:21", "remaining_time": "1:01:35", "throughput": 5165.39, "total_tokens": 23665824} {"current_steps": 23485, "total_steps": 42420, "loss": 0.3353, "lr": 0.0004937051315184329, "epoch": 11.072607260726073, "percentage": 55.36, "elapsed_time": "1:16:22", "remaining_time": "1:01:34", "throughput": 5165.46, "total_tokens": 23670976} {"current_steps": 23490, "total_steps": 42420, "loss": 0.381, "lr": 0.0004934994282701648, "epoch": 11.074964639321076, "percentage": 55.37, "elapsed_time": "1:16:23", "remaining_time": "1:01:33", "throughput": 5165.52, "total_tokens": 23676096} {"current_steps": 23495, "total_steps": 42420, "loss": 0.355, "lr": 0.0004932937261223331, "epoch": 11.077322017916078, "percentage": 55.39, "elapsed_time": "1:16:24", "remaining_time": "1:01:32", "throughput": 5165.52, "total_tokens": 23680608} {"current_steps": 23500, "total_steps": 42420, "loss": 0.3389, "lr": 0.0004930880251097598, "epoch": 11.07967939651108, "percentage": 55.4, "elapsed_time": "1:16:25", "remaining_time": "1:01:31", "throughput": 5165.54, "total_tokens": 23685408} {"current_steps": 23505, "total_steps": 42420, "loss": 0.3032, "lr": 0.0004928823252672665, "epoch": 11.082036775106083, "percentage": 55.41, "elapsed_time": "1:16:26", "remaining_time": "1:01:30", "throughput": 5165.6, "total_tokens": 23690336} {"current_steps": 23510, "total_steps": 42420, "loss": 0.3732, "lr": 0.0004926766266296745, "epoch": 11.084394153701085, "percentage": 55.42, "elapsed_time": "1:16:27", "remaining_time": "1:01:29", "throughput": 5165.69, "total_tokens": 23695776} {"current_steps": 23515, "total_steps": 42420, "loss": 0.3335, "lr": 0.0004924709292318054, "epoch": 11.086751532296086, "percentage": 55.43, "elapsed_time": "1:16:28", "remaining_time": "1:01:28", "throughput": 5165.73, "total_tokens": 23700544} {"current_steps": 23520, "total_steps": 42420, "loss": 0.3667, "lr": 0.0004922652331084802, "epoch": 11.089108910891088, "percentage": 55.45, "elapsed_time": "1:16:28", "remaining_time": "1:01:27", "throughput": 5165.78, "total_tokens": 23705344} {"current_steps": 23525, "total_steps": 42420, "loss": 0.3286, "lr": 0.0004920595382945195, "epoch": 11.09146628948609, "percentage": 55.46, "elapsed_time": "1:16:29", "remaining_time": "1:01:26", "throughput": 5165.8, "total_tokens": 23709824} {"current_steps": 23530, "total_steps": 42420, "loss": 0.3549, "lr": 0.0004918538448247443, "epoch": 11.093823668081093, "percentage": 55.47, "elapsed_time": "1:16:30", "remaining_time": "1:01:25", "throughput": 5165.83, "total_tokens": 23714688} {"current_steps": 23535, "total_steps": 42420, "loss": 0.3149, "lr": 0.0004916481527339748, "epoch": 11.096181046676096, "percentage": 55.48, "elapsed_time": "1:16:31", "remaining_time": "1:01:24", "throughput": 5165.9, "total_tokens": 23719744} {"current_steps": 23540, "total_steps": 42420, "loss": 0.3182, "lr": 0.0004914424620570313, "epoch": 11.098538425271098, "percentage": 55.49, "elapsed_time": "1:16:32", "remaining_time": "1:01:23", "throughput": 5165.91, "total_tokens": 23724032} {"current_steps": 23545, "total_steps": 42420, "loss": 0.2939, "lr": 0.0004912367728287337, "epoch": 11.1008958038661, "percentage": 55.5, "elapsed_time": "1:16:33", "remaining_time": "1:01:22", "throughput": 5165.97, "total_tokens": 23729280} {"current_steps": 23550, "total_steps": 42420, "loss": 0.3304, "lr": 0.0004910310850839016, "epoch": 11.103253182461103, "percentage": 55.52, "elapsed_time": "1:16:34", "remaining_time": "1:01:21", "throughput": 5165.93, "total_tokens": 23733184} {"current_steps": 23555, "total_steps": 42420, "loss": 0.3615, "lr": 0.0004908253988573544, "epoch": 11.105610561056105, "percentage": 55.53, "elapsed_time": "1:16:35", "remaining_time": "1:01:20", "throughput": 5165.97, "total_tokens": 23738016} {"current_steps": 23560, "total_steps": 42420, "loss": 0.3541, "lr": 0.0004906197141839116, "epoch": 11.107967939651108, "percentage": 55.54, "elapsed_time": "1:16:35", "remaining_time": "1:01:19", "throughput": 5166.07, "total_tokens": 23743232} {"current_steps": 23565, "total_steps": 42420, "loss": 0.3013, "lr": 0.0004904140310983919, "epoch": 11.11032531824611, "percentage": 55.55, "elapsed_time": "1:16:37", "remaining_time": "1:01:18", "throughput": 5166.21, "total_tokens": 23749440} {"current_steps": 23570, "total_steps": 42420, "loss": 0.3149, "lr": 0.0004902083496356139, "epoch": 11.112682696841112, "percentage": 55.56, "elapsed_time": "1:16:38", "remaining_time": "1:01:17", "throughput": 5166.32, "total_tokens": 23755072} {"current_steps": 23575, "total_steps": 42420, "loss": 0.3057, "lr": 0.000490002669830396, "epoch": 11.115040075436115, "percentage": 55.58, "elapsed_time": "1:16:39", "remaining_time": "1:01:16", "throughput": 5166.42, "total_tokens": 23760704} {"current_steps": 23580, "total_steps": 42420, "loss": 0.3653, "lr": 0.0004897969917175566, "epoch": 11.117397454031117, "percentage": 55.59, "elapsed_time": "1:16:39", "remaining_time": "1:01:15", "throughput": 5166.5, "total_tokens": 23765856} {"current_steps": 23585, "total_steps": 42420, "loss": 0.2766, "lr": 0.0004895913153319131, "epoch": 11.11975483262612, "percentage": 55.6, "elapsed_time": "1:16:40", "remaining_time": "1:01:14", "throughput": 5166.53, "total_tokens": 23770560} {"current_steps": 23590, "total_steps": 42420, "loss": 0.2584, "lr": 0.0004893856407082833, "epoch": 11.122112211221122, "percentage": 55.61, "elapsed_time": "1:16:41", "remaining_time": "1:01:13", "throughput": 5166.61, "total_tokens": 23775840} {"current_steps": 23595, "total_steps": 42420, "loss": 0.3928, "lr": 0.0004891799678814841, "epoch": 11.124469589816124, "percentage": 55.62, "elapsed_time": "1:16:42", "remaining_time": "1:01:12", "throughput": 5166.7, "total_tokens": 23781120} {"current_steps": 23600, "total_steps": 42420, "loss": 0.4017, "lr": 0.0004889742968863327, "epoch": 11.126826968411127, "percentage": 55.63, "elapsed_time": "1:16:43", "remaining_time": "1:01:11", "throughput": 5166.69, "total_tokens": 23785216} {"current_steps": 23605, "total_steps": 42420, "loss": 0.2521, "lr": 0.0004887686277576456, "epoch": 11.12918434700613, "percentage": 55.65, "elapsed_time": "1:16:44", "remaining_time": "1:01:10", "throughput": 5166.83, "total_tokens": 23791104} {"current_steps": 23610, "total_steps": 42420, "loss": 0.2883, "lr": 0.0004885629605302392, "epoch": 11.131541725601132, "percentage": 55.66, "elapsed_time": "1:16:45", "remaining_time": "1:01:09", "throughput": 5166.97, "total_tokens": 23797184} {"current_steps": 23615, "total_steps": 42420, "loss": 0.392, "lr": 0.0004883572952389294, "epoch": 11.133899104196134, "percentage": 55.67, "elapsed_time": "1:16:46", "remaining_time": "1:01:08", "throughput": 5167.09, "total_tokens": 23803520} {"current_steps": 23620, "total_steps": 42420, "loss": 0.3075, "lr": 0.0004881516319185317, "epoch": 11.136256482791136, "percentage": 55.68, "elapsed_time": "1:16:47", "remaining_time": "1:01:07", "throughput": 5167.09, "total_tokens": 23807968} {"current_steps": 23625, "total_steps": 42420, "loss": 0.3493, "lr": 0.0004879459706038616, "epoch": 11.138613861386139, "percentage": 55.69, "elapsed_time": "1:16:48", "remaining_time": "1:01:06", "throughput": 5167.12, "total_tokens": 23812576} {"current_steps": 23630, "total_steps": 42420, "loss": 0.3552, "lr": 0.00048774031132973414, "epoch": 11.140971239981141, "percentage": 55.7, "elapsed_time": "1:16:49", "remaining_time": "1:01:05", "throughput": 5167.17, "total_tokens": 23817504} {"current_steps": 23635, "total_steps": 42420, "loss": 0.3227, "lr": 0.00048753465413096367, "epoch": 11.143328618576144, "percentage": 55.72, "elapsed_time": "1:16:50", "remaining_time": "1:01:04", "throughput": 5167.21, "total_tokens": 23822176} {"current_steps": 23640, "total_steps": 42420, "loss": 0.3105, "lr": 0.0004873289990423646, "epoch": 11.145685997171146, "percentage": 55.73, "elapsed_time": "1:16:51", "remaining_time": "1:01:03", "throughput": 5167.19, "total_tokens": 23826496} {"current_steps": 23645, "total_steps": 42420, "loss": 0.3457, "lr": 0.00048712334609875085, "epoch": 11.148043375766148, "percentage": 55.74, "elapsed_time": "1:16:52", "remaining_time": "1:01:02", "throughput": 5167.26, "total_tokens": 23831680} {"current_steps": 23650, "total_steps": 42420, "loss": 0.3324, "lr": 0.00048691769533493613, "epoch": 11.15040075436115, "percentage": 55.75, "elapsed_time": "1:16:52", "remaining_time": "1:01:01", "throughput": 5167.26, "total_tokens": 23836064} {"current_steps": 23655, "total_steps": 42420, "loss": 0.3497, "lr": 0.0004867120467857333, "epoch": 11.152758132956153, "percentage": 55.76, "elapsed_time": "1:16:53", "remaining_time": "1:00:59", "throughput": 5167.29, "total_tokens": 23840672} {"current_steps": 23660, "total_steps": 42420, "loss": 0.3426, "lr": 0.00048650640048595543, "epoch": 11.155115511551156, "percentage": 55.78, "elapsed_time": "1:16:54", "remaining_time": "1:00:58", "throughput": 5167.36, "total_tokens": 23845472} {"current_steps": 23665, "total_steps": 42420, "loss": 0.3134, "lr": 0.0004863007564704148, "epoch": 11.157472890146158, "percentage": 55.79, "elapsed_time": "1:16:55", "remaining_time": "1:00:57", "throughput": 5167.43, "total_tokens": 23850432} {"current_steps": 23670, "total_steps": 42420, "loss": 0.3459, "lr": 0.0004860951147739235, "epoch": 11.15983026874116, "percentage": 55.8, "elapsed_time": "1:16:56", "remaining_time": "1:00:56", "throughput": 5167.53, "total_tokens": 23856256} {"current_steps": 23675, "total_steps": 42420, "loss": 0.3512, "lr": 0.00048588947543129317, "epoch": 11.162187647336163, "percentage": 55.81, "elapsed_time": "1:16:57", "remaining_time": "1:00:55", "throughput": 5167.54, "total_tokens": 23860800} {"current_steps": 23680, "total_steps": 42420, "loss": 0.315, "lr": 0.0004856838384773351, "epoch": 11.164545025931165, "percentage": 55.82, "elapsed_time": "1:16:58", "remaining_time": "1:00:54", "throughput": 5167.55, "total_tokens": 23865408} {"current_steps": 23685, "total_steps": 42420, "loss": 0.3314, "lr": 0.00048547820394686007, "epoch": 11.166902404526168, "percentage": 55.83, "elapsed_time": "1:16:59", "remaining_time": "1:00:53", "throughput": 5167.66, "total_tokens": 23871040} {"current_steps": 23690, "total_steps": 42420, "loss": 0.3336, "lr": 0.0004852725718746786, "epoch": 11.16925978312117, "percentage": 55.85, "elapsed_time": "1:17:00", "remaining_time": "1:00:52", "throughput": 5167.66, "total_tokens": 23875232} {"current_steps": 23695, "total_steps": 42420, "loss": 0.3341, "lr": 0.00048506694229560063, "epoch": 11.171617161716172, "percentage": 55.86, "elapsed_time": "1:17:01", "remaining_time": "1:00:51", "throughput": 5167.85, "total_tokens": 23882048} {"current_steps": 23700, "total_steps": 42420, "loss": 0.3475, "lr": 0.0004848613152444357, "epoch": 11.173974540311175, "percentage": 55.87, "elapsed_time": "1:17:02", "remaining_time": "1:00:50", "throughput": 5167.85, "total_tokens": 23886336} {"current_steps": 23705, "total_steps": 42420, "loss": 0.2182, "lr": 0.0004846556907559931, "epoch": 11.176331918906177, "percentage": 55.88, "elapsed_time": "1:17:03", "remaining_time": "1:00:49", "throughput": 5167.91, "total_tokens": 23891264} {"current_steps": 23710, "total_steps": 42420, "loss": 0.2654, "lr": 0.00048445006886508164, "epoch": 11.17868929750118, "percentage": 55.89, "elapsed_time": "1:17:03", "remaining_time": "1:00:48", "throughput": 5167.94, "total_tokens": 23895936} {"current_steps": 23715, "total_steps": 42420, "loss": 0.3714, "lr": 0.00048424444960650935, "epoch": 11.18104667609618, "percentage": 55.91, "elapsed_time": "1:17:04", "remaining_time": "1:00:47", "throughput": 5167.97, "total_tokens": 23900576} {"current_steps": 23720, "total_steps": 42420, "loss": 0.3135, "lr": 0.00048403883301508427, "epoch": 11.183404054691183, "percentage": 55.92, "elapsed_time": "1:17:05", "remaining_time": "1:00:46", "throughput": 5168.0, "total_tokens": 23905472} {"current_steps": 23725, "total_steps": 42420, "loss": 0.3816, "lr": 0.00048383321912561373, "epoch": 11.185761433286185, "percentage": 55.93, "elapsed_time": "1:17:06", "remaining_time": "1:00:45", "throughput": 5168.05, "total_tokens": 23910400} {"current_steps": 23730, "total_steps": 42420, "loss": 0.3788, "lr": 0.0004836276079729048, "epoch": 11.188118811881187, "percentage": 55.94, "elapsed_time": "1:17:07", "remaining_time": "1:00:44", "throughput": 5168.04, "total_tokens": 23914496} {"current_steps": 23735, "total_steps": 42420, "loss": 0.3515, "lr": 0.00048342199959176374, "epoch": 11.19047619047619, "percentage": 55.95, "elapsed_time": "1:17:08", "remaining_time": "1:00:43", "throughput": 5168.07, "total_tokens": 23919456} {"current_steps": 23740, "total_steps": 42420, "loss": 0.3413, "lr": 0.00048321639401699667, "epoch": 11.192833569071192, "percentage": 55.96, "elapsed_time": "1:17:09", "remaining_time": "1:00:42", "throughput": 5168.22, "total_tokens": 23925632} {"current_steps": 23745, "total_steps": 42420, "loss": 0.3416, "lr": 0.0004830107912834092, "epoch": 11.195190947666195, "percentage": 55.98, "elapsed_time": "1:17:10", "remaining_time": "1:00:41", "throughput": 5168.29, "total_tokens": 23930880} {"current_steps": 23750, "total_steps": 42420, "loss": 0.3568, "lr": 0.00048280519142580635, "epoch": 11.197548326261197, "percentage": 55.99, "elapsed_time": "1:17:11", "remaining_time": "1:00:40", "throughput": 5168.38, "total_tokens": 23936672} {"current_steps": 23755, "total_steps": 42420, "loss": 0.3683, "lr": 0.00048259959447899257, "epoch": 11.1999057048562, "percentage": 56.0, "elapsed_time": "1:17:12", "remaining_time": "1:00:39", "throughput": 5168.4, "total_tokens": 23941184} {"current_steps": 23760, "total_steps": 42420, "loss": 0.3264, "lr": 0.00048239400047777193, "epoch": 11.202263083451202, "percentage": 56.01, "elapsed_time": "1:17:13", "remaining_time": "1:00:38", "throughput": 5168.47, "total_tokens": 23946624} {"current_steps": 23765, "total_steps": 42420, "loss": 0.3664, "lr": 0.0004821884094569481, "epoch": 11.204620462046204, "percentage": 56.02, "elapsed_time": "1:17:14", "remaining_time": "1:00:37", "throughput": 5168.48, "total_tokens": 23951296} {"current_steps": 23770, "total_steps": 42420, "loss": 0.3503, "lr": 0.0004819828214513243, "epoch": 11.206977840641207, "percentage": 56.03, "elapsed_time": "1:17:14", "remaining_time": "1:00:36", "throughput": 5168.52, "total_tokens": 23956064} {"current_steps": 23775, "total_steps": 42420, "loss": 0.3721, "lr": 0.0004817772364957027, "epoch": 11.209335219236209, "percentage": 56.05, "elapsed_time": "1:17:15", "remaining_time": "1:00:35", "throughput": 5168.58, "total_tokens": 23961024} {"current_steps": 23780, "total_steps": 42420, "loss": 0.3435, "lr": 0.00048157165462488563, "epoch": 11.211692597831211, "percentage": 56.06, "elapsed_time": "1:17:16", "remaining_time": "1:00:34", "throughput": 5168.66, "total_tokens": 23966624} {"current_steps": 23785, "total_steps": 42420, "loss": 0.3535, "lr": 0.00048136607587367453, "epoch": 11.214049976426214, "percentage": 56.07, "elapsed_time": "1:17:17", "remaining_time": "1:00:33", "throughput": 5168.65, "total_tokens": 23970848} {"current_steps": 23790, "total_steps": 42420, "loss": 0.3371, "lr": 0.00048116050027687045, "epoch": 11.216407355021216, "percentage": 56.08, "elapsed_time": "1:17:18", "remaining_time": "1:00:32", "throughput": 5168.73, "total_tokens": 23976064} {"current_steps": 23795, "total_steps": 42420, "loss": 0.3321, "lr": 0.00048095492786927373, "epoch": 11.218764733616219, "percentage": 56.09, "elapsed_time": "1:17:19", "remaining_time": "1:00:31", "throughput": 5168.86, "total_tokens": 23982176} {"current_steps": 23800, "total_steps": 42420, "loss": 0.2883, "lr": 0.0004807493586856844, "epoch": 11.221122112211221, "percentage": 56.11, "elapsed_time": "1:17:20", "remaining_time": "1:00:30", "throughput": 5168.93, "total_tokens": 23987456} {"current_steps": 23805, "total_steps": 42420, "loss": 0.3923, "lr": 0.000480543792760902, "epoch": 11.223479490806223, "percentage": 56.12, "elapsed_time": "1:17:21", "remaining_time": "1:00:29", "throughput": 5168.92, "total_tokens": 23991840} {"current_steps": 23810, "total_steps": 42420, "loss": 0.2728, "lr": 0.0004803382301297248, "epoch": 11.225836869401226, "percentage": 56.13, "elapsed_time": "1:17:22", "remaining_time": "1:00:28", "throughput": 5168.95, "total_tokens": 23996480} {"current_steps": 23815, "total_steps": 42420, "loss": 0.3429, "lr": 0.0004801326708269516, "epoch": 11.228194247996228, "percentage": 56.14, "elapsed_time": "1:17:23", "remaining_time": "1:00:27", "throughput": 5168.95, "total_tokens": 24000864} {"current_steps": 23820, "total_steps": 42420, "loss": 0.3311, "lr": 0.00047992711488737983, "epoch": 11.23055162659123, "percentage": 56.15, "elapsed_time": "1:17:24", "remaining_time": "1:00:26", "throughput": 5169.07, "total_tokens": 24006560} {"current_steps": 23825, "total_steps": 42420, "loss": 0.3102, "lr": 0.00047972156234580665, "epoch": 11.232909005186233, "percentage": 56.16, "elapsed_time": "1:17:25", "remaining_time": "1:00:25", "throughput": 5169.14, "total_tokens": 24012000} {"current_steps": 23830, "total_steps": 42420, "loss": 0.3493, "lr": 0.00047951601323702864, "epoch": 11.235266383781235, "percentage": 56.18, "elapsed_time": "1:17:26", "remaining_time": "1:00:24", "throughput": 5169.15, "total_tokens": 24016896} {"current_steps": 23835, "total_steps": 42420, "loss": 0.2906, "lr": 0.0004793104675958418, "epoch": 11.237623762376238, "percentage": 56.19, "elapsed_time": "1:17:27", "remaining_time": "1:00:23", "throughput": 5169.23, "total_tokens": 24022368} {"current_steps": 23840, "total_steps": 42420, "loss": 0.3322, "lr": 0.0004791049254570417, "epoch": 11.23998114097124, "percentage": 56.2, "elapsed_time": "1:17:28", "remaining_time": "1:00:22", "throughput": 5169.31, "total_tokens": 24027616} {"current_steps": 23845, "total_steps": 42420, "loss": 0.3472, "lr": 0.00047889938685542267, "epoch": 11.242338519566243, "percentage": 56.21, "elapsed_time": "1:17:29", "remaining_time": "1:00:21", "throughput": 5169.32, "total_tokens": 24032256} {"current_steps": 23850, "total_steps": 42420, "loss": 0.3581, "lr": 0.00047869385182577927, "epoch": 11.244695898161245, "percentage": 56.22, "elapsed_time": "1:17:29", "remaining_time": "1:00:20", "throughput": 5169.36, "total_tokens": 24037120} {"current_steps": 23855, "total_steps": 42420, "loss": 0.2468, "lr": 0.000478488320402905, "epoch": 11.247053276756247, "percentage": 56.24, "elapsed_time": "1:17:30", "remaining_time": "1:00:19", "throughput": 5169.37, "total_tokens": 24041472} {"current_steps": 23860, "total_steps": 42420, "loss": 0.4102, "lr": 0.0004782827926215929, "epoch": 11.24941065535125, "percentage": 56.25, "elapsed_time": "1:17:31", "remaining_time": "1:00:18", "throughput": 5169.38, "total_tokens": 24046048} {"current_steps": 23865, "total_steps": 42420, "loss": 0.3852, "lr": 0.00047807726851663523, "epoch": 11.251768033946252, "percentage": 56.26, "elapsed_time": "1:17:32", "remaining_time": "1:00:17", "throughput": 5169.46, "total_tokens": 24051072} {"current_steps": 23870, "total_steps": 42420, "loss": 0.2992, "lr": 0.0004778717481228237, "epoch": 11.254125412541255, "percentage": 56.27, "elapsed_time": "1:17:33", "remaining_time": "1:00:16", "throughput": 5169.55, "total_tokens": 24056640} {"current_steps": 23875, "total_steps": 42420, "loss": 0.3264, "lr": 0.0004776662314749496, "epoch": 11.256482791136257, "percentage": 56.28, "elapsed_time": "1:17:34", "remaining_time": "1:00:15", "throughput": 5169.55, "total_tokens": 24061184} {"current_steps": 23880, "total_steps": 42420, "loss": 0.3193, "lr": 0.00047746071860780333, "epoch": 11.25884016973126, "percentage": 56.29, "elapsed_time": "1:17:35", "remaining_time": "1:00:14", "throughput": 5169.58, "total_tokens": 24065856} {"current_steps": 23885, "total_steps": 42420, "loss": 0.2931, "lr": 0.00047725520955617466, "epoch": 11.261197548326262, "percentage": 56.31, "elapsed_time": "1:17:36", "remaining_time": "1:00:13", "throughput": 5169.54, "total_tokens": 24069984} {"current_steps": 23890, "total_steps": 42420, "loss": 0.307, "lr": 0.0004770497043548527, "epoch": 11.263554926921264, "percentage": 56.32, "elapsed_time": "1:17:37", "remaining_time": "1:00:12", "throughput": 5169.59, "total_tokens": 24075104} {"current_steps": 23895, "total_steps": 42420, "loss": 0.297, "lr": 0.00047684420303862625, "epoch": 11.265912305516267, "percentage": 56.33, "elapsed_time": "1:17:37", "remaining_time": "1:00:11", "throughput": 5169.6, "total_tokens": 24079648} {"current_steps": 23900, "total_steps": 42420, "loss": 0.4177, "lr": 0.00047663870564228314, "epoch": 11.268269684111269, "percentage": 56.34, "elapsed_time": "1:17:38", "remaining_time": "1:00:10", "throughput": 5169.65, "total_tokens": 24084672} {"current_steps": 23905, "total_steps": 42420, "loss": 0.381, "lr": 0.00047643321220061045, "epoch": 11.270627062706271, "percentage": 56.35, "elapsed_time": "1:17:39", "remaining_time": "1:00:09", "throughput": 5169.71, "total_tokens": 24089632} {"current_steps": 23910, "total_steps": 42420, "loss": 0.3676, "lr": 0.0004762277227483947, "epoch": 11.272984441301272, "percentage": 56.36, "elapsed_time": "1:17:40", "remaining_time": "1:00:08", "throughput": 5169.76, "total_tokens": 24094496} {"current_steps": 23915, "total_steps": 42420, "loss": 0.3574, "lr": 0.0004760222373204219, "epoch": 11.275341819896274, "percentage": 56.38, "elapsed_time": "1:17:41", "remaining_time": "1:00:07", "throughput": 5169.77, "total_tokens": 24099104} {"current_steps": 23920, "total_steps": 42420, "loss": 0.3432, "lr": 0.00047581675595147725, "epoch": 11.277699198491277, "percentage": 56.39, "elapsed_time": "1:17:42", "remaining_time": "1:00:06", "throughput": 5169.88, "total_tokens": 24104864} {"current_steps": 23925, "total_steps": 42420, "loss": 0.3272, "lr": 0.0004756112786763452, "epoch": 11.28005657708628, "percentage": 56.4, "elapsed_time": "1:17:43", "remaining_time": "1:00:05", "throughput": 5169.93, "total_tokens": 24109696} {"current_steps": 23930, "total_steps": 42420, "loss": 0.3034, "lr": 0.0004754058055298095, "epoch": 11.282413955681282, "percentage": 56.41, "elapsed_time": "1:17:44", "remaining_time": "1:00:04", "throughput": 5169.98, "total_tokens": 24114816} {"current_steps": 23935, "total_steps": 42420, "loss": 0.3505, "lr": 0.0004752003365466534, "epoch": 11.284771334276284, "percentage": 56.42, "elapsed_time": "1:17:45", "remaining_time": "1:00:02", "throughput": 5169.98, "total_tokens": 24119296} {"current_steps": 23940, "total_steps": 42420, "loss": 0.3791, "lr": 0.0004749948717616593, "epoch": 11.287128712871286, "percentage": 56.44, "elapsed_time": "1:17:46", "remaining_time": "1:00:02", "throughput": 5170.08, "total_tokens": 24124896} {"current_steps": 23945, "total_steps": 42420, "loss": 0.3813, "lr": 0.0004747894112096088, "epoch": 11.289486091466289, "percentage": 56.45, "elapsed_time": "1:17:47", "remaining_time": "1:00:01", "throughput": 5170.14, "total_tokens": 24130432} {"current_steps": 23950, "total_steps": 42420, "loss": 0.3418, "lr": 0.00047458395492528283, "epoch": 11.291843470061291, "percentage": 56.46, "elapsed_time": "1:17:48", "remaining_time": "1:00:00", "throughput": 5170.21, "total_tokens": 24135936} {"current_steps": 23955, "total_steps": 42420, "loss": 0.3373, "lr": 0.00047437850294346185, "epoch": 11.294200848656294, "percentage": 56.47, "elapsed_time": "1:17:49", "remaining_time": "0:59:59", "throughput": 5170.2, "total_tokens": 24140288} {"current_steps": 23960, "total_steps": 42420, "loss": 0.3324, "lr": 0.00047417305529892537, "epoch": 11.296558227251296, "percentage": 56.48, "elapsed_time": "1:17:50", "remaining_time": "0:59:58", "throughput": 5170.35, "total_tokens": 24146560} {"current_steps": 23965, "total_steps": 42420, "loss": 0.3158, "lr": 0.000473967612026452, "epoch": 11.298915605846299, "percentage": 56.49, "elapsed_time": "1:17:51", "remaining_time": "0:59:57", "throughput": 5170.39, "total_tokens": 24151200} {"current_steps": 23970, "total_steps": 42420, "loss": 0.3205, "lr": 0.00047376217316081985, "epoch": 11.301272984441301, "percentage": 56.51, "elapsed_time": "1:17:51", "remaining_time": "0:59:56", "throughput": 5170.41, "total_tokens": 24155648} {"current_steps": 23975, "total_steps": 42420, "loss": 0.3762, "lr": 0.0004735567387368064, "epoch": 11.303630363036303, "percentage": 56.52, "elapsed_time": "1:17:52", "remaining_time": "0:59:55", "throughput": 5170.56, "total_tokens": 24161984} {"current_steps": 23980, "total_steps": 42420, "loss": 0.3373, "lr": 0.0004733513087891881, "epoch": 11.305987741631306, "percentage": 56.53, "elapsed_time": "1:17:53", "remaining_time": "0:59:54", "throughput": 5170.64, "total_tokens": 24167264} {"current_steps": 23985, "total_steps": 42420, "loss": 0.3122, "lr": 0.00047314588335274074, "epoch": 11.308345120226308, "percentage": 56.54, "elapsed_time": "1:17:54", "remaining_time": "0:59:53", "throughput": 5170.65, "total_tokens": 24171840} {"current_steps": 23990, "total_steps": 42420, "loss": 0.3436, "lr": 0.00047294046246223933, "epoch": 11.31070249882131, "percentage": 56.55, "elapsed_time": "1:17:55", "remaining_time": "0:59:52", "throughput": 5170.78, "total_tokens": 24177728} {"current_steps": 23995, "total_steps": 42420, "loss": 0.3218, "lr": 0.0004727350461524582, "epoch": 11.313059877416313, "percentage": 56.57, "elapsed_time": "1:17:56", "remaining_time": "0:59:51", "throughput": 5170.84, "total_tokens": 24182848} {"current_steps": 24000, "total_steps": 42420, "loss": 0.3063, "lr": 0.0004725296344581709, "epoch": 11.315417256011315, "percentage": 56.58, "elapsed_time": "1:17:57", "remaining_time": "0:59:50", "throughput": 5170.8, "total_tokens": 24187008} {"current_steps": 24005, "total_steps": 42420, "loss": 0.3814, "lr": 0.00047232422741415, "epoch": 11.317774634606318, "percentage": 56.59, "elapsed_time": "1:17:58", "remaining_time": "0:59:49", "throughput": 5170.85, "total_tokens": 24191936} {"current_steps": 24010, "total_steps": 42420, "loss": 0.3357, "lr": 0.00047211882505516746, "epoch": 11.32013201320132, "percentage": 56.6, "elapsed_time": "1:17:59", "remaining_time": "0:59:47", "throughput": 5170.83, "total_tokens": 24195968} {"current_steps": 24015, "total_steps": 42420, "loss": 0.3333, "lr": 0.0004719134274159945, "epoch": 11.322489391796323, "percentage": 56.61, "elapsed_time": "1:18:00", "remaining_time": "0:59:46", "throughput": 5170.82, "total_tokens": 24200224} {"current_steps": 24020, "total_steps": 42420, "loss": 0.3161, "lr": 0.0004717080345314014, "epoch": 11.324846770391325, "percentage": 56.62, "elapsed_time": "1:18:01", "remaining_time": "0:59:45", "throughput": 5170.96, "total_tokens": 24206176} {"current_steps": 24025, "total_steps": 42420, "loss": 0.3343, "lr": 0.00047150264643615764, "epoch": 11.327204148986327, "percentage": 56.64, "elapsed_time": "1:18:02", "remaining_time": "0:59:45", "throughput": 5171.12, "total_tokens": 24212608} {"current_steps": 24030, "total_steps": 42420, "loss": 0.3171, "lr": 0.00047129726316503216, "epoch": 11.32956152758133, "percentage": 56.65, "elapsed_time": "1:18:03", "remaining_time": "0:59:44", "throughput": 5171.18, "total_tokens": 24217664} {"current_steps": 24035, "total_steps": 42420, "loss": 0.33, "lr": 0.00047109188475279255, "epoch": 11.331918906176332, "percentage": 56.66, "elapsed_time": "1:18:04", "remaining_time": "0:59:42", "throughput": 5171.2, "total_tokens": 24222272} {"current_steps": 24040, "total_steps": 42420, "loss": 0.3512, "lr": 0.0004708865112342059, "epoch": 11.334276284771335, "percentage": 56.67, "elapsed_time": "1:18:04", "remaining_time": "0:59:41", "throughput": 5171.25, "total_tokens": 24227168} {"current_steps": 24045, "total_steps": 42420, "loss": 0.3329, "lr": 0.00047068114264403865, "epoch": 11.336633663366337, "percentage": 56.68, "elapsed_time": "1:18:05", "remaining_time": "0:59:40", "throughput": 5171.27, "total_tokens": 24231744} {"current_steps": 24050, "total_steps": 42420, "loss": 0.3333, "lr": 0.00047047577901705625, "epoch": 11.33899104196134, "percentage": 56.69, "elapsed_time": "1:18:06", "remaining_time": "0:59:39", "throughput": 5171.38, "total_tokens": 24237632} {"current_steps": 24055, "total_steps": 42420, "loss": 0.3502, "lr": 0.000470270420388023, "epoch": 11.341348420556342, "percentage": 56.71, "elapsed_time": "1:18:07", "remaining_time": "0:59:39", "throughput": 5171.48, "total_tokens": 24243328} {"current_steps": 24060, "total_steps": 42420, "loss": 0.3111, "lr": 0.0004700650667917029, "epoch": 11.343705799151344, "percentage": 56.72, "elapsed_time": "1:18:08", "remaining_time": "0:59:37", "throughput": 5171.51, "total_tokens": 24248224} {"current_steps": 24065, "total_steps": 42420, "loss": 0.2755, "lr": 0.0004698597182628587, "epoch": 11.346063177746347, "percentage": 56.73, "elapsed_time": "1:18:09", "remaining_time": "0:59:37", "throughput": 5171.61, "total_tokens": 24254176} {"current_steps": 24070, "total_steps": 42420, "loss": 0.3754, "lr": 0.0004696543748362525, "epoch": 11.348420556341349, "percentage": 56.74, "elapsed_time": "1:18:10", "remaining_time": "0:59:36", "throughput": 5171.64, "total_tokens": 24258976} {"current_steps": 24075, "total_steps": 42420, "loss": 0.2866, "lr": 0.00046944903654664536, "epoch": 11.350777934936351, "percentage": 56.75, "elapsed_time": "1:18:11", "remaining_time": "0:59:35", "throughput": 5171.71, "total_tokens": 24264192} {"current_steps": 24080, "total_steps": 42420, "loss": 0.3891, "lr": 0.0004692437034287975, "epoch": 11.353135313531354, "percentage": 56.77, "elapsed_time": "1:18:12", "remaining_time": "0:59:34", "throughput": 5171.8, "total_tokens": 24269504} {"current_steps": 24085, "total_steps": 42420, "loss": 0.4193, "lr": 0.00046903837551746854, "epoch": 11.355492692126356, "percentage": 56.78, "elapsed_time": "1:18:13", "remaining_time": "0:59:33", "throughput": 5171.83, "total_tokens": 24274240} {"current_steps": 24090, "total_steps": 42420, "loss": 0.3583, "lr": 0.000468833052847417, "epoch": 11.357850070721359, "percentage": 56.79, "elapsed_time": "1:18:14", "remaining_time": "0:59:31", "throughput": 5171.82, "total_tokens": 24278752} {"current_steps": 24095, "total_steps": 42420, "loss": 0.3476, "lr": 0.00046862773545340027, "epoch": 11.360207449316361, "percentage": 56.8, "elapsed_time": "1:18:15", "remaining_time": "0:59:30", "throughput": 5171.83, "total_tokens": 24283072} {"current_steps": 24100, "total_steps": 42420, "loss": 0.3597, "lr": 0.0004684224233701753, "epoch": 11.362564827911363, "percentage": 56.81, "elapsed_time": "1:18:16", "remaining_time": "0:59:29", "throughput": 5171.87, "total_tokens": 24288160} {"current_steps": 24105, "total_steps": 42420, "loss": 0.3432, "lr": 0.00046821711663249787, "epoch": 11.364922206506366, "percentage": 56.82, "elapsed_time": "1:18:17", "remaining_time": "0:59:28", "throughput": 5171.89, "total_tokens": 24292768} {"current_steps": 24110, "total_steps": 42420, "loss": 0.355, "lr": 0.0004680118152751231, "epoch": 11.367279585101368, "percentage": 56.84, "elapsed_time": "1:18:17", "remaining_time": "0:59:27", "throughput": 5171.93, "total_tokens": 24297472} {"current_steps": 24115, "total_steps": 42420, "loss": 0.3303, "lr": 0.0004678065193328047, "epoch": 11.369636963696369, "percentage": 56.85, "elapsed_time": "1:18:19", "remaining_time": "0:59:27", "throughput": 5172.14, "total_tokens": 24305152} {"current_steps": 24120, "total_steps": 42420, "loss": 0.3447, "lr": 0.00046760122884029605, "epoch": 11.371994342291371, "percentage": 56.86, "elapsed_time": "1:18:20", "remaining_time": "0:59:26", "throughput": 5172.16, "total_tokens": 24309792} {"current_steps": 24125, "total_steps": 42420, "loss": 0.3014, "lr": 0.0004673959438323492, "epoch": 11.374351720886374, "percentage": 56.87, "elapsed_time": "1:18:20", "remaining_time": "0:59:24", "throughput": 5172.17, "total_tokens": 24314240} {"current_steps": 24130, "total_steps": 42420, "loss": 0.3134, "lr": 0.0004671906643437156, "epoch": 11.376709099481376, "percentage": 56.88, "elapsed_time": "1:18:22", "remaining_time": "0:59:24", "throughput": 5172.31, "total_tokens": 24320512} {"current_steps": 24135, "total_steps": 42420, "loss": 0.3902, "lr": 0.00046698539040914534, "epoch": 11.379066478076378, "percentage": 56.9, "elapsed_time": "1:18:23", "remaining_time": "0:59:23", "throughput": 5172.39, "total_tokens": 24325760} {"current_steps": 24140, "total_steps": 42420, "loss": 0.3775, "lr": 0.00046678012206338796, "epoch": 11.38142385667138, "percentage": 56.91, "elapsed_time": "1:18:23", "remaining_time": "0:59:22", "throughput": 5172.49, "total_tokens": 24331200} {"current_steps": 24145, "total_steps": 42420, "loss": 0.3201, "lr": 0.00046657485934119183, "epoch": 11.383781235266383, "percentage": 56.92, "elapsed_time": "1:18:24", "remaining_time": "0:59:21", "throughput": 5172.55, "total_tokens": 24336512} {"current_steps": 24150, "total_steps": 42420, "loss": 0.3272, "lr": 0.00046636960227730464, "epoch": 11.386138613861386, "percentage": 56.93, "elapsed_time": "1:18:25", "remaining_time": "0:59:20", "throughput": 5172.53, "total_tokens": 24340736} {"current_steps": 24155, "total_steps": 42420, "loss": 0.3673, "lr": 0.0004661643509064726, "epoch": 11.388495992456388, "percentage": 56.94, "elapsed_time": "1:18:26", "remaining_time": "0:59:18", "throughput": 5172.53, "total_tokens": 24344992} {"current_steps": 24160, "total_steps": 42420, "loss": 0.3234, "lr": 0.00046595910526344147, "epoch": 11.39085337105139, "percentage": 56.95, "elapsed_time": "1:18:27", "remaining_time": "0:59:17", "throughput": 5172.62, "total_tokens": 24350560} {"current_steps": 24165, "total_steps": 42420, "loss": 0.3589, "lr": 0.00046575386538295584, "epoch": 11.393210749646393, "percentage": 56.97, "elapsed_time": "1:18:28", "remaining_time": "0:59:16", "throughput": 5172.66, "total_tokens": 24355392} {"current_steps": 24170, "total_steps": 42420, "loss": 0.3573, "lr": 0.0004655486312997595, "epoch": 11.395568128241395, "percentage": 56.98, "elapsed_time": "1:18:29", "remaining_time": "0:59:15", "throughput": 5172.73, "total_tokens": 24360480} {"current_steps": 24175, "total_steps": 42420, "loss": 0.3577, "lr": 0.0004653434030485947, "epoch": 11.397925506836398, "percentage": 56.99, "elapsed_time": "1:18:30", "remaining_time": "0:59:14", "throughput": 5172.83, "total_tokens": 24366048} {"current_steps": 24180, "total_steps": 42420, "loss": 0.3382, "lr": 0.0004651381806642035, "epoch": 11.4002828854314, "percentage": 57.0, "elapsed_time": "1:18:31", "remaining_time": "0:59:13", "throughput": 5172.82, "total_tokens": 24370272} {"current_steps": 24185, "total_steps": 42420, "loss": 0.3538, "lr": 0.00046493296418132624, "epoch": 11.402640264026402, "percentage": 57.01, "elapsed_time": "1:18:32", "remaining_time": "0:59:12", "throughput": 5172.9, "total_tokens": 24375648} {"current_steps": 24190, "total_steps": 42420, "loss": 0.2979, "lr": 0.00046472775363470294, "epoch": 11.404997642621405, "percentage": 57.02, "elapsed_time": "1:18:33", "remaining_time": "0:59:11", "throughput": 5172.92, "total_tokens": 24380288} {"current_steps": 24195, "total_steps": 42420, "loss": 0.3136, "lr": 0.00046452254905907195, "epoch": 11.407355021216407, "percentage": 57.04, "elapsed_time": "1:18:34", "remaining_time": "0:59:10", "throughput": 5173.01, "total_tokens": 24385920} {"current_steps": 24200, "total_steps": 42420, "loss": 0.3583, "lr": 0.000464317350489171, "epoch": 11.40971239981141, "percentage": 57.05, "elapsed_time": "1:18:35", "remaining_time": "0:59:10", "throughput": 5173.14, "total_tokens": 24392320} {"current_steps": 24205, "total_steps": 42420, "loss": 0.3834, "lr": 0.00046411215795973674, "epoch": 11.412069778406412, "percentage": 57.06, "elapsed_time": "1:18:36", "remaining_time": "0:59:08", "throughput": 5173.13, "total_tokens": 24396800} {"current_steps": 24210, "total_steps": 42420, "loss": 0.287, "lr": 0.00046390697150550494, "epoch": 11.414427157001414, "percentage": 57.07, "elapsed_time": "1:18:37", "remaining_time": "0:59:08", "throughput": 5173.2, "total_tokens": 24402336} {"current_steps": 24215, "total_steps": 42420, "loss": 0.3246, "lr": 0.0004637017911612099, "epoch": 11.416784535596417, "percentage": 57.08, "elapsed_time": "1:18:38", "remaining_time": "0:59:07", "throughput": 5173.27, "total_tokens": 24407776} {"current_steps": 24220, "total_steps": 42420, "loss": 0.3551, "lr": 0.0004634966169615853, "epoch": 11.41914191419142, "percentage": 57.1, "elapsed_time": "1:18:38", "remaining_time": "0:59:06", "throughput": 5173.3, "total_tokens": 24412384} {"current_steps": 24225, "total_steps": 42420, "loss": 0.3449, "lr": 0.0004632914489413637, "epoch": 11.421499292786422, "percentage": 57.11, "elapsed_time": "1:18:39", "remaining_time": "0:59:05", "throughput": 5173.38, "total_tokens": 24417728} {"current_steps": 24230, "total_steps": 42420, "loss": 0.317, "lr": 0.00046308628713527663, "epoch": 11.423856671381424, "percentage": 57.12, "elapsed_time": "1:18:40", "remaining_time": "0:59:03", "throughput": 5173.39, "total_tokens": 24422176} {"current_steps": 24235, "total_steps": 42420, "loss": 0.3352, "lr": 0.0004628811315780542, "epoch": 11.426214049976426, "percentage": 57.13, "elapsed_time": "1:18:41", "remaining_time": "0:59:02", "throughput": 5173.4, "total_tokens": 24426912} {"current_steps": 24240, "total_steps": 42420, "loss": 0.361, "lr": 0.000462675982304426, "epoch": 11.428571428571429, "percentage": 57.14, "elapsed_time": "1:18:42", "remaining_time": "0:59:01", "throughput": 5173.42, "total_tokens": 24431840} {"current_steps": 24245, "total_steps": 42420, "loss": 0.3652, "lr": 0.0004624708393491203, "epoch": 11.430928807166431, "percentage": 57.15, "elapsed_time": "1:18:43", "remaining_time": "0:59:00", "throughput": 5173.54, "total_tokens": 24437664} {"current_steps": 24250, "total_steps": 42420, "loss": 0.3312, "lr": 0.0004622657027468644, "epoch": 11.433286185761434, "percentage": 57.17, "elapsed_time": "1:18:44", "remaining_time": "0:58:59", "throughput": 5173.53, "total_tokens": 24441952} {"current_steps": 24255, "total_steps": 42420, "loss": 0.2862, "lr": 0.00046206057253238443, "epoch": 11.435643564356436, "percentage": 57.18, "elapsed_time": "1:18:45", "remaining_time": "0:58:58", "throughput": 5173.6, "total_tokens": 24447040} {"current_steps": 24260, "total_steps": 42420, "loss": 0.285, "lr": 0.0004618554487404051, "epoch": 11.438000942951438, "percentage": 57.19, "elapsed_time": "1:18:46", "remaining_time": "0:58:57", "throughput": 5173.59, "total_tokens": 24451392} {"current_steps": 24265, "total_steps": 42420, "loss": 0.3335, "lr": 0.00046165033140565067, "epoch": 11.44035832154644, "percentage": 57.2, "elapsed_time": "1:18:47", "remaining_time": "0:58:56", "throughput": 5173.58, "total_tokens": 24455680} {"current_steps": 24270, "total_steps": 42420, "loss": 0.3221, "lr": 0.000461445220562844, "epoch": 11.442715700141443, "percentage": 57.21, "elapsed_time": "1:18:47", "remaining_time": "0:58:55", "throughput": 5173.55, "total_tokens": 24459808} {"current_steps": 24275, "total_steps": 42420, "loss": 0.2937, "lr": 0.000461240116246707, "epoch": 11.445073078736446, "percentage": 57.23, "elapsed_time": "1:18:48", "remaining_time": "0:58:54", "throughput": 5173.59, "total_tokens": 24464416} {"current_steps": 24280, "total_steps": 42420, "loss": 0.3091, "lr": 0.00046103501849196007, "epoch": 11.447430457331448, "percentage": 57.24, "elapsed_time": "1:18:49", "remaining_time": "0:58:53", "throughput": 5173.64, "total_tokens": 24469728} {"current_steps": 24285, "total_steps": 42420, "loss": 0.3386, "lr": 0.00046082992733332287, "epoch": 11.44978783592645, "percentage": 57.25, "elapsed_time": "1:18:50", "remaining_time": "0:58:52", "throughput": 5173.7, "total_tokens": 24474976} {"current_steps": 24290, "total_steps": 42420, "loss": 0.3555, "lr": 0.000460624842805514, "epoch": 11.452145214521453, "percentage": 57.26, "elapsed_time": "1:18:51", "remaining_time": "0:58:51", "throughput": 5173.74, "total_tokens": 24479968} {"current_steps": 24295, "total_steps": 42420, "loss": 0.3497, "lr": 0.0004604197649432507, "epoch": 11.454502593116455, "percentage": 57.27, "elapsed_time": "1:18:52", "remaining_time": "0:58:50", "throughput": 5173.82, "total_tokens": 24485280} {"current_steps": 24300, "total_steps": 42420, "loss": 0.3154, "lr": 0.00046021469378124916, "epoch": 11.456859971711458, "percentage": 57.28, "elapsed_time": "1:18:53", "remaining_time": "0:58:49", "throughput": 5173.76, "total_tokens": 24489120} {"current_steps": 24305, "total_steps": 42420, "loss": 0.2911, "lr": 0.0004600096293542244, "epoch": 11.45921735030646, "percentage": 57.3, "elapsed_time": "1:18:54", "remaining_time": "0:58:48", "throughput": 5173.79, "total_tokens": 24493728} {"current_steps": 24310, "total_steps": 42420, "loss": 0.324, "lr": 0.0004598045716968903, "epoch": 11.46157472890146, "percentage": 57.31, "elapsed_time": "1:18:55", "remaining_time": "0:58:47", "throughput": 5173.86, "total_tokens": 24498752} {"current_steps": 24315, "total_steps": 42420, "loss": 0.3054, "lr": 0.00045959952084395996, "epoch": 11.463932107496463, "percentage": 57.32, "elapsed_time": "1:18:55", "remaining_time": "0:58:46", "throughput": 5173.82, "total_tokens": 24502752} {"current_steps": 24320, "total_steps": 42420, "loss": 0.3342, "lr": 0.0004593944768301446, "epoch": 11.466289486091465, "percentage": 57.33, "elapsed_time": "1:18:56", "remaining_time": "0:58:45", "throughput": 5173.9, "total_tokens": 24508128} {"current_steps": 24325, "total_steps": 42420, "loss": 0.3762, "lr": 0.00045918943969015493, "epoch": 11.468646864686468, "percentage": 57.34, "elapsed_time": "1:18:57", "remaining_time": "0:58:44", "throughput": 5173.92, "total_tokens": 24513088} {"current_steps": 24330, "total_steps": 42420, "loss": 0.4127, "lr": 0.0004589844094587002, "epoch": 11.47100424328147, "percentage": 57.36, "elapsed_time": "1:18:58", "remaining_time": "0:58:43", "throughput": 5173.99, "total_tokens": 24518240} {"current_steps": 24335, "total_steps": 42420, "loss": 0.3178, "lr": 0.00045877938617048863, "epoch": 11.473361621876473, "percentage": 57.37, "elapsed_time": "1:18:59", "remaining_time": "0:58:42", "throughput": 5173.94, "total_tokens": 24522144} {"current_steps": 24340, "total_steps": 42420, "loss": 0.3114, "lr": 0.0004585743698602269, "epoch": 11.475719000471475, "percentage": 57.38, "elapsed_time": "1:19:00", "remaining_time": "0:58:41", "throughput": 5174.04, "total_tokens": 24527904} {"current_steps": 24345, "total_steps": 42420, "loss": 0.2836, "lr": 0.00045836936056262113, "epoch": 11.478076379066477, "percentage": 57.39, "elapsed_time": "1:19:01", "remaining_time": "0:58:40", "throughput": 5174.03, "total_tokens": 24532256} {"current_steps": 24350, "total_steps": 42420, "loss": 0.3476, "lr": 0.0004581643583123757, "epoch": 11.48043375766148, "percentage": 57.4, "elapsed_time": "1:19:02", "remaining_time": "0:58:39", "throughput": 5174.0, "total_tokens": 24536384} {"current_steps": 24355, "total_steps": 42420, "loss": 0.4187, "lr": 0.00045795936314419416, "epoch": 11.482791136256482, "percentage": 57.41, "elapsed_time": "1:19:03", "remaining_time": "0:58:38", "throughput": 5174.04, "total_tokens": 24541120} {"current_steps": 24360, "total_steps": 42420, "loss": 0.4251, "lr": 0.0004577543750927786, "epoch": 11.485148514851485, "percentage": 57.43, "elapsed_time": "1:19:03", "remaining_time": "0:58:37", "throughput": 5174.01, "total_tokens": 24545120} {"current_steps": 24365, "total_steps": 42420, "loss": 0.3366, "lr": 0.00045754939419283, "epoch": 11.487505893446487, "percentage": 57.44, "elapsed_time": "1:19:04", "remaining_time": "0:58:36", "throughput": 5174.05, "total_tokens": 24549984} {"current_steps": 24370, "total_steps": 42420, "loss": 0.3476, "lr": 0.00045734442047904816, "epoch": 11.48986327204149, "percentage": 57.45, "elapsed_time": "1:19:05", "remaining_time": "0:58:34", "throughput": 5174.02, "total_tokens": 24554080} {"current_steps": 24375, "total_steps": 42420, "loss": 0.3616, "lr": 0.0004571394539861319, "epoch": 11.492220650636492, "percentage": 57.46, "elapsed_time": "1:19:06", "remaining_time": "0:58:33", "throughput": 5174.1, "total_tokens": 24559520} {"current_steps": 24380, "total_steps": 42420, "loss": 0.3323, "lr": 0.0004569344947487781, "epoch": 11.494578029231494, "percentage": 57.47, "elapsed_time": "1:19:07", "remaining_time": "0:58:32", "throughput": 5174.12, "total_tokens": 24564064} {"current_steps": 24385, "total_steps": 42420, "loss": 0.3396, "lr": 0.0004567295428016832, "epoch": 11.496935407826497, "percentage": 57.48, "elapsed_time": "1:19:08", "remaining_time": "0:58:31", "throughput": 5174.19, "total_tokens": 24569600} {"current_steps": 24390, "total_steps": 42420, "loss": 0.3251, "lr": 0.00045652459817954207, "epoch": 11.499292786421499, "percentage": 57.5, "elapsed_time": "1:19:09", "remaining_time": "0:58:30", "throughput": 5174.18, "total_tokens": 24573824} {"current_steps": 24395, "total_steps": 42420, "loss": 0.3115, "lr": 0.0004563196609170483, "epoch": 11.501650165016502, "percentage": 57.51, "elapsed_time": "1:19:10", "remaining_time": "0:58:30", "throughput": 5174.34, "total_tokens": 24580352} {"current_steps": 24400, "total_steps": 42420, "loss": 0.341, "lr": 0.00045611473104889425, "epoch": 11.504007543611504, "percentage": 57.52, "elapsed_time": "1:19:11", "remaining_time": "0:58:28", "throughput": 5174.35, "total_tokens": 24584864} {"current_steps": 24405, "total_steps": 42420, "loss": 0.3374, "lr": 0.00045590980860977105, "epoch": 11.506364922206506, "percentage": 57.53, "elapsed_time": "1:19:12", "remaining_time": "0:58:27", "throughput": 5174.35, "total_tokens": 24589280} {"current_steps": 24410, "total_steps": 42420, "loss": 0.357, "lr": 0.0004557048936343687, "epoch": 11.508722300801509, "percentage": 57.54, "elapsed_time": "1:19:13", "remaining_time": "0:58:26", "throughput": 5174.42, "total_tokens": 24594656} {"current_steps": 24415, "total_steps": 42420, "loss": 0.3176, "lr": 0.0004554999861573758, "epoch": 11.511079679396511, "percentage": 57.56, "elapsed_time": "1:19:14", "remaining_time": "0:58:25", "throughput": 5174.42, "total_tokens": 24599264} {"current_steps": 24420, "total_steps": 42420, "loss": 0.3877, "lr": 0.00045529508621347964, "epoch": 11.513437057991514, "percentage": 57.57, "elapsed_time": "1:19:14", "remaining_time": "0:58:24", "throughput": 5174.47, "total_tokens": 24604224} {"current_steps": 24425, "total_steps": 42420, "loss": 0.3512, "lr": 0.0004550901938373663, "epoch": 11.515794436586516, "percentage": 57.58, "elapsed_time": "1:19:15", "remaining_time": "0:58:23", "throughput": 5174.48, "total_tokens": 24608640} {"current_steps": 24430, "total_steps": 42420, "loss": 0.343, "lr": 0.00045488530906372067, "epoch": 11.518151815181518, "percentage": 57.59, "elapsed_time": "1:19:16", "remaining_time": "0:58:22", "throughput": 5174.49, "total_tokens": 24613504} {"current_steps": 24435, "total_steps": 42420, "loss": 0.3699, "lr": 0.00045468043192722625, "epoch": 11.52050919377652, "percentage": 57.6, "elapsed_time": "1:19:17", "remaining_time": "0:58:21", "throughput": 5174.59, "total_tokens": 24619072} {"current_steps": 24440, "total_steps": 42420, "loss": 0.3489, "lr": 0.00045447556246256516, "epoch": 11.522866572371523, "percentage": 57.61, "elapsed_time": "1:19:18", "remaining_time": "0:58:20", "throughput": 5174.65, "total_tokens": 24624256} {"current_steps": 24445, "total_steps": 42420, "loss": 0.309, "lr": 0.00045427070070441837, "epoch": 11.525223950966526, "percentage": 57.63, "elapsed_time": "1:19:19", "remaining_time": "0:58:19", "throughput": 5174.71, "total_tokens": 24629440} {"current_steps": 24450, "total_steps": 42420, "loss": 0.4083, "lr": 0.00045406584668746555, "epoch": 11.527581329561528, "percentage": 57.64, "elapsed_time": "1:19:20", "remaining_time": "0:58:18", "throughput": 5174.73, "total_tokens": 24633792} {"current_steps": 24455, "total_steps": 42420, "loss": 0.3263, "lr": 0.000453861000446385, "epoch": 11.52993870815653, "percentage": 57.65, "elapsed_time": "1:19:21", "remaining_time": "0:58:17", "throughput": 5174.82, "total_tokens": 24639456} {"current_steps": 24460, "total_steps": 42420, "loss": 0.3542, "lr": 0.0004536561620158536, "epoch": 11.532296086751533, "percentage": 57.66, "elapsed_time": "1:19:22", "remaining_time": "0:58:16", "throughput": 5174.81, "total_tokens": 24643680} {"current_steps": 24465, "total_steps": 42420, "loss": 0.314, "lr": 0.00045345133143054713, "epoch": 11.534653465346535, "percentage": 57.67, "elapsed_time": "1:19:23", "remaining_time": "0:58:15", "throughput": 5174.93, "total_tokens": 24649408} {"current_steps": 24470, "total_steps": 42420, "loss": 0.3373, "lr": 0.00045324650872513984, "epoch": 11.537010843941538, "percentage": 57.69, "elapsed_time": "1:19:24", "remaining_time": "0:58:14", "throughput": 5175.04, "total_tokens": 24655424} {"current_steps": 24475, "total_steps": 42420, "loss": 0.3649, "lr": 0.00045304169393430493, "epoch": 11.53936822253654, "percentage": 57.7, "elapsed_time": "1:19:25", "remaining_time": "0:58:13", "throughput": 5175.05, "total_tokens": 24660032} {"current_steps": 24480, "total_steps": 42420, "loss": 0.2967, "lr": 0.0004528368870927139, "epoch": 11.541725601131542, "percentage": 57.71, "elapsed_time": "1:19:26", "remaining_time": "0:58:12", "throughput": 5175.09, "total_tokens": 24664832} {"current_steps": 24485, "total_steps": 42420, "loss": 0.3364, "lr": 0.00045263208823503724, "epoch": 11.544082979726545, "percentage": 57.72, "elapsed_time": "1:19:26", "remaining_time": "0:58:11", "throughput": 5175.12, "total_tokens": 24669792} {"current_steps": 24490, "total_steps": 42420, "loss": 0.3153, "lr": 0.0004524272973959436, "epoch": 11.546440358321547, "percentage": 57.73, "elapsed_time": "1:19:27", "remaining_time": "0:58:10", "throughput": 5175.12, "total_tokens": 24674208} {"current_steps": 24495, "total_steps": 42420, "loss": 0.3316, "lr": 0.0004522225146101007, "epoch": 11.54879773691655, "percentage": 57.74, "elapsed_time": "1:19:28", "remaining_time": "0:58:09", "throughput": 5175.11, "total_tokens": 24678496} {"current_steps": 24500, "total_steps": 42420, "loss": 0.322, "lr": 0.00045201773991217486, "epoch": 11.551155115511552, "percentage": 57.76, "elapsed_time": "1:19:29", "remaining_time": "0:58:08", "throughput": 5175.07, "total_tokens": 24682592} {"current_steps": 24505, "total_steps": 42420, "loss": 0.3356, "lr": 0.0004518129733368311, "epoch": 11.553512494106554, "percentage": 57.77, "elapsed_time": "1:19:30", "remaining_time": "0:58:07", "throughput": 5175.11, "total_tokens": 24687584} {"current_steps": 24510, "total_steps": 42420, "loss": 0.385, "lr": 0.0004516082149187327, "epoch": 11.555869872701557, "percentage": 57.78, "elapsed_time": "1:19:31", "remaining_time": "0:58:06", "throughput": 5175.12, "total_tokens": 24692288} {"current_steps": 24515, "total_steps": 42420, "loss": 0.3403, "lr": 0.0004514034646925419, "epoch": 11.558227251296557, "percentage": 57.79, "elapsed_time": "1:19:32", "remaining_time": "0:58:05", "throughput": 5175.14, "total_tokens": 24697024} {"current_steps": 24520, "total_steps": 42420, "loss": 0.3365, "lr": 0.00045119872269291933, "epoch": 11.56058462989156, "percentage": 57.8, "elapsed_time": "1:19:33", "remaining_time": "0:58:04", "throughput": 5175.16, "total_tokens": 24701728} {"current_steps": 24525, "total_steps": 42420, "loss": 0.3393, "lr": 0.0004509939889545246, "epoch": 11.562942008486562, "percentage": 57.81, "elapsed_time": "1:19:33", "remaining_time": "0:58:03", "throughput": 5175.15, "total_tokens": 24706144} {"current_steps": 24530, "total_steps": 42420, "loss": 0.3259, "lr": 0.00045078926351201526, "epoch": 11.565299387081565, "percentage": 57.83, "elapsed_time": "1:19:34", "remaining_time": "0:58:02", "throughput": 5175.15, "total_tokens": 24710400} {"current_steps": 24535, "total_steps": 42420, "loss": 0.324, "lr": 0.0004505845464000481, "epoch": 11.567656765676567, "percentage": 57.84, "elapsed_time": "1:19:35", "remaining_time": "0:58:01", "throughput": 5175.22, "total_tokens": 24715712} {"current_steps": 24540, "total_steps": 42420, "loss": 0.2545, "lr": 0.0004503798376532782, "epoch": 11.57001414427157, "percentage": 57.85, "elapsed_time": "1:19:36", "remaining_time": "0:58:00", "throughput": 5175.22, "total_tokens": 24719968} {"current_steps": 24545, "total_steps": 42420, "loss": 0.3377, "lr": 0.00045017513730635946, "epoch": 11.572371522866572, "percentage": 57.86, "elapsed_time": "1:19:37", "remaining_time": "0:57:59", "throughput": 5175.41, "total_tokens": 24726880} {"current_steps": 24550, "total_steps": 42420, "loss": 0.2417, "lr": 0.00044997044539394384, "epoch": 11.574728901461574, "percentage": 57.87, "elapsed_time": "1:19:38", "remaining_time": "0:57:58", "throughput": 5175.42, "total_tokens": 24731200} {"current_steps": 24555, "total_steps": 42420, "loss": 0.3293, "lr": 0.0004497657619506823, "epoch": 11.577086280056577, "percentage": 57.89, "elapsed_time": "1:19:39", "remaining_time": "0:57:57", "throughput": 5175.42, "total_tokens": 24735488} {"current_steps": 24560, "total_steps": 42420, "loss": 0.3194, "lr": 0.0004495610870112244, "epoch": 11.579443658651579, "percentage": 57.9, "elapsed_time": "1:19:40", "remaining_time": "0:57:56", "throughput": 5175.49, "total_tokens": 24740640} {"current_steps": 24565, "total_steps": 42420, "loss": 0.3271, "lr": 0.00044935642061021814, "epoch": 11.581801037246581, "percentage": 57.91, "elapsed_time": "1:19:41", "remaining_time": "0:57:55", "throughput": 5175.66, "total_tokens": 24747552} {"current_steps": 24570, "total_steps": 42420, "loss": 0.2916, "lr": 0.00044915176278230993, "epoch": 11.584158415841584, "percentage": 57.92, "elapsed_time": "1:19:42", "remaining_time": "0:57:54", "throughput": 5175.72, "total_tokens": 24752608} {"current_steps": 24575, "total_steps": 42420, "loss": 0.2973, "lr": 0.0004489471135621449, "epoch": 11.586515794436586, "percentage": 57.93, "elapsed_time": "1:19:43", "remaining_time": "0:57:53", "throughput": 5175.89, "total_tokens": 24759264} {"current_steps": 24580, "total_steps": 42420, "loss": 0.3184, "lr": 0.0004487424729843667, "epoch": 11.588873173031589, "percentage": 57.94, "elapsed_time": "1:19:44", "remaining_time": "0:57:52", "throughput": 5175.94, "total_tokens": 24764384} {"current_steps": 24585, "total_steps": 42420, "loss": 0.2908, "lr": 0.0004485378410836176, "epoch": 11.591230551626591, "percentage": 57.96, "elapsed_time": "1:19:45", "remaining_time": "0:57:51", "throughput": 5176.06, "total_tokens": 24770464} {"current_steps": 24590, "total_steps": 42420, "loss": 0.328, "lr": 0.00044833321789453817, "epoch": 11.593587930221593, "percentage": 57.97, "elapsed_time": "1:19:46", "remaining_time": "0:57:50", "throughput": 5176.01, "total_tokens": 24774400} {"current_steps": 24595, "total_steps": 42420, "loss": 0.1953, "lr": 0.0004481286034517677, "epoch": 11.595945308816596, "percentage": 57.98, "elapsed_time": "1:19:47", "remaining_time": "0:57:49", "throughput": 5176.01, "total_tokens": 24778720} {"current_steps": 24600, "total_steps": 42420, "loss": 0.2893, "lr": 0.00044792399778994396, "epoch": 11.598302687411598, "percentage": 57.99, "elapsed_time": "1:19:48", "remaining_time": "0:57:48", "throughput": 5176.12, "total_tokens": 24784832} {"current_steps": 24605, "total_steps": 42420, "loss": 0.4253, "lr": 0.0004477194009437033, "epoch": 11.6006600660066, "percentage": 58.0, "elapsed_time": "1:19:49", "remaining_time": "0:57:47", "throughput": 5176.15, "total_tokens": 24789696} {"current_steps": 24610, "total_steps": 42420, "loss": 0.3782, "lr": 0.0004475148129476802, "epoch": 11.603017444601603, "percentage": 58.02, "elapsed_time": "1:19:50", "remaining_time": "0:57:46", "throughput": 5176.24, "total_tokens": 24795040} {"current_steps": 24615, "total_steps": 42420, "loss": 0.3339, "lr": 0.0004473102338365081, "epoch": 11.605374823196605, "percentage": 58.03, "elapsed_time": "1:19:51", "remaining_time": "0:57:45", "throughput": 5176.27, "total_tokens": 24799776} {"current_steps": 24620, "total_steps": 42420, "loss": 0.3434, "lr": 0.00044710566364481895, "epoch": 11.607732201791608, "percentage": 58.04, "elapsed_time": "1:19:51", "remaining_time": "0:57:44", "throughput": 5176.3, "total_tokens": 24804416} {"current_steps": 24625, "total_steps": 42420, "loss": 0.359, "lr": 0.0004469011024072429, "epoch": 11.61008958038661, "percentage": 58.05, "elapsed_time": "1:19:52", "remaining_time": "0:57:43", "throughput": 5176.37, "total_tokens": 24809984} {"current_steps": 24630, "total_steps": 42420, "loss": 0.3504, "lr": 0.0004466965501584085, "epoch": 11.612446958981613, "percentage": 58.06, "elapsed_time": "1:19:53", "remaining_time": "0:57:42", "throughput": 5176.35, "total_tokens": 24814208} {"current_steps": 24635, "total_steps": 42420, "loss": 0.3266, "lr": 0.00044649200693294304, "epoch": 11.614804337576615, "percentage": 58.07, "elapsed_time": "1:19:54", "remaining_time": "0:57:41", "throughput": 5176.51, "total_tokens": 24820576} {"current_steps": 24640, "total_steps": 42420, "loss": 0.3314, "lr": 0.00044628747276547246, "epoch": 11.617161716171617, "percentage": 58.09, "elapsed_time": "1:19:55", "remaining_time": "0:57:40", "throughput": 5176.53, "total_tokens": 24825408} {"current_steps": 24645, "total_steps": 42420, "loss": 0.3286, "lr": 0.00044608294769062074, "epoch": 11.61951909476662, "percentage": 58.1, "elapsed_time": "1:19:56", "remaining_time": "0:57:39", "throughput": 5176.53, "total_tokens": 24829792} {"current_steps": 24650, "total_steps": 42420, "loss": 0.3574, "lr": 0.00044587843174301046, "epoch": 11.621876473361622, "percentage": 58.11, "elapsed_time": "1:19:57", "remaining_time": "0:57:38", "throughput": 5176.67, "total_tokens": 24836032} {"current_steps": 24655, "total_steps": 42420, "loss": 0.3316, "lr": 0.00044567392495726275, "epoch": 11.624233851956625, "percentage": 58.12, "elapsed_time": "1:19:58", "remaining_time": "0:57:37", "throughput": 5176.73, "total_tokens": 24841376} {"current_steps": 24660, "total_steps": 42420, "loss": 0.2807, "lr": 0.0004454694273679971, "epoch": 11.626591230551627, "percentage": 58.13, "elapsed_time": "1:19:59", "remaining_time": "0:57:36", "throughput": 5176.78, "total_tokens": 24846048} {"current_steps": 24665, "total_steps": 42420, "loss": 0.3647, "lr": 0.00044526493900983165, "epoch": 11.62894860914663, "percentage": 58.14, "elapsed_time": "1:20:00", "remaining_time": "0:57:35", "throughput": 5176.87, "total_tokens": 24851808} {"current_steps": 24670, "total_steps": 42420, "loss": 0.3274, "lr": 0.0004450604599173826, "epoch": 11.631305987741632, "percentage": 58.16, "elapsed_time": "1:20:01", "remaining_time": "0:57:34", "throughput": 5176.96, "total_tokens": 24857376} {"current_steps": 24675, "total_steps": 42420, "loss": 0.35, "lr": 0.00044485599012526484, "epoch": 11.633663366336634, "percentage": 58.17, "elapsed_time": "1:20:02", "remaining_time": "0:57:33", "throughput": 5176.99, "total_tokens": 24862176} {"current_steps": 24680, "total_steps": 42420, "loss": 0.3114, "lr": 0.0004446515296680916, "epoch": 11.636020744931637, "percentage": 58.18, "elapsed_time": "1:20:03", "remaining_time": "0:57:32", "throughput": 5177.06, "total_tokens": 24867424} {"current_steps": 24685, "total_steps": 42420, "loss": 0.292, "lr": 0.00044444707858047486, "epoch": 11.638378123526639, "percentage": 58.19, "elapsed_time": "1:20:04", "remaining_time": "0:57:31", "throughput": 5177.11, "total_tokens": 24872224} {"current_steps": 24690, "total_steps": 42420, "loss": 0.34, "lr": 0.0004442426368970243, "epoch": 11.640735502121641, "percentage": 58.2, "elapsed_time": "1:20:05", "remaining_time": "0:57:30", "throughput": 5177.14, "total_tokens": 24877152} {"current_steps": 24695, "total_steps": 42420, "loss": 0.3853, "lr": 0.00044403820465234857, "epoch": 11.643092880716644, "percentage": 58.22, "elapsed_time": "1:20:06", "remaining_time": "0:57:29", "throughput": 5177.23, "total_tokens": 24882592} {"current_steps": 24700, "total_steps": 42420, "loss": 0.3943, "lr": 0.00044383378188105467, "epoch": 11.645450259311646, "percentage": 58.23, "elapsed_time": "1:20:07", "remaining_time": "0:57:28", "throughput": 5177.39, "total_tokens": 24889184} {"current_steps": 24705, "total_steps": 42420, "loss": 0.357, "lr": 0.00044362936861774794, "epoch": 11.647807637906649, "percentage": 58.24, "elapsed_time": "1:20:08", "remaining_time": "0:57:27", "throughput": 5177.48, "total_tokens": 24894848} {"current_steps": 24710, "total_steps": 42420, "loss": 0.3453, "lr": 0.000443424964897032, "epoch": 11.65016501650165, "percentage": 58.25, "elapsed_time": "1:20:09", "remaining_time": "0:57:26", "throughput": 5177.52, "total_tokens": 24899744} {"current_steps": 24715, "total_steps": 42420, "loss": 0.368, "lr": 0.00044322057075350894, "epoch": 11.652522395096653, "percentage": 58.26, "elapsed_time": "1:20:10", "remaining_time": "0:57:25", "throughput": 5177.57, "total_tokens": 24904512} {"current_steps": 24720, "total_steps": 42420, "loss": 0.3089, "lr": 0.000443016186221779, "epoch": 11.654879773691654, "percentage": 58.27, "elapsed_time": "1:20:10", "remaining_time": "0:57:24", "throughput": 5177.54, "total_tokens": 24908480} {"current_steps": 24725, "total_steps": 42420, "loss": 0.3671, "lr": 0.00044281181133644127, "epoch": 11.657237152286656, "percentage": 58.29, "elapsed_time": "1:20:11", "remaining_time": "0:57:23", "throughput": 5177.54, "total_tokens": 24912768} {"current_steps": 24730, "total_steps": 42420, "loss": 0.3557, "lr": 0.0004426074461320929, "epoch": 11.659594530881659, "percentage": 58.3, "elapsed_time": "1:20:12", "remaining_time": "0:57:22", "throughput": 5177.56, "total_tokens": 24917344} {"current_steps": 24735, "total_steps": 42420, "loss": 0.353, "lr": 0.00044240309064332944, "epoch": 11.661951909476661, "percentage": 58.31, "elapsed_time": "1:20:13", "remaining_time": "0:57:21", "throughput": 5177.64, "total_tokens": 24922592} {"current_steps": 24740, "total_steps": 42420, "loss": 0.365, "lr": 0.00044219874490474475, "epoch": 11.664309288071664, "percentage": 58.32, "elapsed_time": "1:20:14", "remaining_time": "0:57:20", "throughput": 5177.67, "total_tokens": 24927392} {"current_steps": 24745, "total_steps": 42420, "loss": 0.3311, "lr": 0.00044199440895093114, "epoch": 11.666666666666666, "percentage": 58.33, "elapsed_time": "1:20:15", "remaining_time": "0:57:19", "throughput": 5177.75, "total_tokens": 24932864} {"current_steps": 24750, "total_steps": 42420, "loss": 0.3518, "lr": 0.0004417900828164792, "epoch": 11.669024045261668, "percentage": 58.35, "elapsed_time": "1:20:16", "remaining_time": "0:57:18", "throughput": 5177.76, "total_tokens": 24937376} {"current_steps": 24755, "total_steps": 42420, "loss": 0.3627, "lr": 0.00044158576653597806, "epoch": 11.67138142385667, "percentage": 58.36, "elapsed_time": "1:20:17", "remaining_time": "0:57:17", "throughput": 5177.77, "total_tokens": 24941632} {"current_steps": 24760, "total_steps": 42420, "loss": 0.3233, "lr": 0.00044138146014401466, "epoch": 11.673738802451673, "percentage": 58.37, "elapsed_time": "1:20:17", "remaining_time": "0:57:16", "throughput": 5177.78, "total_tokens": 24946112} {"current_steps": 24765, "total_steps": 42420, "loss": 0.3679, "lr": 0.00044117716367517477, "epoch": 11.676096181046676, "percentage": 58.38, "elapsed_time": "1:20:19", "remaining_time": "0:57:15", "throughput": 5177.91, "total_tokens": 24952384} {"current_steps": 24770, "total_steps": 42420, "loss": 0.3378, "lr": 0.00044097287716404233, "epoch": 11.678453559641678, "percentage": 58.39, "elapsed_time": "1:20:19", "remaining_time": "0:57:14", "throughput": 5177.89, "total_tokens": 24956544} {"current_steps": 24775, "total_steps": 42420, "loss": 0.3715, "lr": 0.0004407686006451996, "epoch": 11.68081093823668, "percentage": 58.4, "elapsed_time": "1:20:20", "remaining_time": "0:57:13", "throughput": 5178.0, "total_tokens": 24962336} {"current_steps": 24780, "total_steps": 42420, "loss": 0.3571, "lr": 0.0004405643341532271, "epoch": 11.683168316831683, "percentage": 58.42, "elapsed_time": "1:20:21", "remaining_time": "0:57:12", "throughput": 5178.12, "total_tokens": 24968416} {"current_steps": 24785, "total_steps": 42420, "loss": 0.3535, "lr": 0.0004403600777227036, "epoch": 11.685525695426685, "percentage": 58.43, "elapsed_time": "1:20:22", "remaining_time": "0:57:11", "throughput": 5178.18, "total_tokens": 24973568} {"current_steps": 24790, "total_steps": 42420, "loss": 0.3416, "lr": 0.00044015583138820634, "epoch": 11.687883074021688, "percentage": 58.44, "elapsed_time": "1:20:23", "remaining_time": "0:57:10", "throughput": 5178.2, "total_tokens": 24978176} {"current_steps": 24795, "total_steps": 42420, "loss": 0.3175, "lr": 0.0004399515951843108, "epoch": 11.69024045261669, "percentage": 58.45, "elapsed_time": "1:20:24", "remaining_time": "0:57:09", "throughput": 5178.23, "total_tokens": 24983008} {"current_steps": 24800, "total_steps": 42420, "loss": 0.2935, "lr": 0.00043974736914559055, "epoch": 11.692597831211692, "percentage": 58.46, "elapsed_time": "1:20:25", "remaining_time": "0:57:08", "throughput": 5178.29, "total_tokens": 24988096} {"current_steps": 24805, "total_steps": 42420, "loss": 0.3928, "lr": 0.0004395431533066177, "epoch": 11.694955209806695, "percentage": 58.47, "elapsed_time": "1:20:26", "remaining_time": "0:57:07", "throughput": 5178.35, "total_tokens": 24993152} {"current_steps": 24810, "total_steps": 42420, "loss": 0.3323, "lr": 0.0004393389477019625, "epoch": 11.697312588401697, "percentage": 58.49, "elapsed_time": "1:20:27", "remaining_time": "0:57:06", "throughput": 5178.39, "total_tokens": 24997984} {"current_steps": 24815, "total_steps": 42420, "loss": 0.3229, "lr": 0.0004391347523661935, "epoch": 11.6996699669967, "percentage": 58.5, "elapsed_time": "1:20:28", "remaining_time": "0:57:05", "throughput": 5178.51, "total_tokens": 25003936} {"current_steps": 24820, "total_steps": 42420, "loss": 0.3496, "lr": 0.0004389305673338775, "epoch": 11.702027345591702, "percentage": 58.51, "elapsed_time": "1:20:29", "remaining_time": "0:57:04", "throughput": 5178.54, "total_tokens": 25008640} {"current_steps": 24825, "total_steps": 42420, "loss": 0.3053, "lr": 0.0004387263926395795, "epoch": 11.704384724186705, "percentage": 58.52, "elapsed_time": "1:20:30", "remaining_time": "0:57:03", "throughput": 5178.61, "total_tokens": 25013856} {"current_steps": 24830, "total_steps": 42420, "loss": 0.3542, "lr": 0.0004385222283178628, "epoch": 11.706742102781707, "percentage": 58.53, "elapsed_time": "1:20:31", "remaining_time": "0:57:02", "throughput": 5178.65, "total_tokens": 25018592} {"current_steps": 24835, "total_steps": 42420, "loss": 0.3435, "lr": 0.0004383180744032892, "epoch": 11.70909948137671, "percentage": 58.55, "elapsed_time": "1:20:32", "remaining_time": "0:57:01", "throughput": 5178.7, "total_tokens": 25023552} {"current_steps": 24840, "total_steps": 42420, "loss": 0.3119, "lr": 0.00043811393093041816, "epoch": 11.711456859971712, "percentage": 58.56, "elapsed_time": "1:20:32", "remaining_time": "0:57:00", "throughput": 5178.72, "total_tokens": 25028000} {"current_steps": 24845, "total_steps": 42420, "loss": 0.3859, "lr": 0.00043790979793380784, "epoch": 11.713814238566714, "percentage": 58.57, "elapsed_time": "1:20:33", "remaining_time": "0:56:59", "throughput": 5178.76, "total_tokens": 25033088} {"current_steps": 24850, "total_steps": 42420, "loss": 0.3389, "lr": 0.00043770567544801456, "epoch": 11.716171617161717, "percentage": 58.58, "elapsed_time": "1:20:34", "remaining_time": "0:56:58", "throughput": 5178.79, "total_tokens": 25037600} {"current_steps": 24855, "total_steps": 42420, "loss": 0.3604, "lr": 0.00043750156350759277, "epoch": 11.718528995756719, "percentage": 58.59, "elapsed_time": "1:20:35", "remaining_time": "0:56:57", "throughput": 5178.93, "total_tokens": 25043904} {"current_steps": 24860, "total_steps": 42420, "loss": 0.3523, "lr": 0.000437297462147095, "epoch": 11.720886374351721, "percentage": 58.6, "elapsed_time": "1:20:36", "remaining_time": "0:56:56", "throughput": 5178.99, "total_tokens": 25049248} {"current_steps": 24865, "total_steps": 42420, "loss": 0.3451, "lr": 0.00043709337140107233, "epoch": 11.723243752946724, "percentage": 58.62, "elapsed_time": "1:20:37", "remaining_time": "0:56:55", "throughput": 5179.14, "total_tokens": 25055744} {"current_steps": 24870, "total_steps": 42420, "loss": 0.3529, "lr": 0.00043688929130407373, "epoch": 11.725601131541726, "percentage": 58.63, "elapsed_time": "1:20:38", "remaining_time": "0:56:54", "throughput": 5179.17, "total_tokens": 25060544} {"current_steps": 24875, "total_steps": 42420, "loss": 0.2932, "lr": 0.0004366852218906467, "epoch": 11.727958510136729, "percentage": 58.64, "elapsed_time": "1:20:39", "remaining_time": "0:56:53", "throughput": 5179.21, "total_tokens": 25065248} {"current_steps": 24880, "total_steps": 42420, "loss": 0.2872, "lr": 0.0004364811631953365, "epoch": 11.730315888731731, "percentage": 58.65, "elapsed_time": "1:20:40", "remaining_time": "0:56:52", "throughput": 5179.23, "total_tokens": 25069792} {"current_steps": 24885, "total_steps": 42420, "loss": 0.3691, "lr": 0.00043627711525268684, "epoch": 11.732673267326733, "percentage": 58.66, "elapsed_time": "1:20:41", "remaining_time": "0:56:51", "throughput": 5179.3, "total_tokens": 25075232} {"current_steps": 24890, "total_steps": 42420, "loss": 0.2989, "lr": 0.00043607307809723954, "epoch": 11.735030645921736, "percentage": 58.68, "elapsed_time": "1:20:42", "remaining_time": "0:56:50", "throughput": 5179.38, "total_tokens": 25080576} {"current_steps": 24895, "total_steps": 42420, "loss": 0.3594, "lr": 0.00043586905176353494, "epoch": 11.737388024516738, "percentage": 58.69, "elapsed_time": "1:20:43", "remaining_time": "0:56:49", "throughput": 5179.39, "total_tokens": 25085472} {"current_steps": 24900, "total_steps": 42420, "loss": 0.3816, "lr": 0.0004356650362861108, "epoch": 11.73974540311174, "percentage": 58.7, "elapsed_time": "1:20:44", "remaining_time": "0:56:48", "throughput": 5179.39, "total_tokens": 25089728} {"current_steps": 24905, "total_steps": 42420, "loss": 0.369, "lr": 0.0004354610316995037, "epoch": 11.742102781706743, "percentage": 58.71, "elapsed_time": "1:20:44", "remaining_time": "0:56:47", "throughput": 5179.39, "total_tokens": 25093984} {"current_steps": 24910, "total_steps": 42420, "loss": 0.3408, "lr": 0.0004352570380382481, "epoch": 11.744460160301745, "percentage": 58.72, "elapsed_time": "1:20:45", "remaining_time": "0:56:46", "throughput": 5179.42, "total_tokens": 25098848} {"current_steps": 24915, "total_steps": 42420, "loss": 0.345, "lr": 0.0004350530553368768, "epoch": 11.746817538896746, "percentage": 58.73, "elapsed_time": "1:20:46", "remaining_time": "0:56:45", "throughput": 5179.53, "total_tokens": 25104544} {"current_steps": 24920, "total_steps": 42420, "loss": 0.3225, "lr": 0.00043484908362992044, "epoch": 11.749174917491748, "percentage": 58.75, "elapsed_time": "1:20:47", "remaining_time": "0:56:44", "throughput": 5179.58, "total_tokens": 25109376} {"current_steps": 24925, "total_steps": 42420, "loss": 0.3519, "lr": 0.000434645122951908, "epoch": 11.75153229608675, "percentage": 58.76, "elapsed_time": "1:20:48", "remaining_time": "0:56:43", "throughput": 5179.59, "total_tokens": 25113824} {"current_steps": 24930, "total_steps": 42420, "loss": 0.3491, "lr": 0.0004344411733373665, "epoch": 11.753889674681753, "percentage": 58.77, "elapsed_time": "1:20:49", "remaining_time": "0:56:42", "throughput": 5179.66, "total_tokens": 25119072} {"current_steps": 24935, "total_steps": 42420, "loss": 0.3501, "lr": 0.00043423723482082156, "epoch": 11.756247053276756, "percentage": 58.78, "elapsed_time": "1:20:50", "remaining_time": "0:56:41", "throughput": 5179.63, "total_tokens": 25123168} {"current_steps": 24940, "total_steps": 42420, "loss": 0.351, "lr": 0.0004340333074367959, "epoch": 11.758604431871758, "percentage": 58.79, "elapsed_time": "1:20:51", "remaining_time": "0:56:40", "throughput": 5179.67, "total_tokens": 25128128} {"current_steps": 24945, "total_steps": 42420, "loss": 0.3398, "lr": 0.0004338293912198114, "epoch": 11.76096181046676, "percentage": 58.8, "elapsed_time": "1:20:52", "remaining_time": "0:56:39", "throughput": 5179.71, "total_tokens": 25132864} {"current_steps": 24950, "total_steps": 42420, "loss": 0.315, "lr": 0.0004336254862043873, "epoch": 11.763319189061763, "percentage": 58.82, "elapsed_time": "1:20:53", "remaining_time": "0:56:38", "throughput": 5179.69, "total_tokens": 25137216} {"current_steps": 24955, "total_steps": 42420, "loss": 0.3393, "lr": 0.00043342159242504145, "epoch": 11.765676567656765, "percentage": 58.83, "elapsed_time": "1:20:54", "remaining_time": "0:56:37", "throughput": 5179.8, "total_tokens": 25142976} {"current_steps": 24960, "total_steps": 42420, "loss": 0.2992, "lr": 0.0004332177099162895, "epoch": 11.768033946251768, "percentage": 58.84, "elapsed_time": "1:20:54", "remaining_time": "0:56:36", "throughput": 5179.84, "total_tokens": 25147776} {"current_steps": 24965, "total_steps": 42420, "loss": 0.3564, "lr": 0.0004330138387126455, "epoch": 11.77039132484677, "percentage": 58.85, "elapsed_time": "1:20:55", "remaining_time": "0:56:35", "throughput": 5179.82, "total_tokens": 25151872} {"current_steps": 24970, "total_steps": 42420, "loss": 0.3136, "lr": 0.00043280997884862117, "epoch": 11.772748703441772, "percentage": 58.86, "elapsed_time": "1:20:56", "remaining_time": "0:56:33", "throughput": 5179.81, "total_tokens": 25156064} {"current_steps": 24975, "total_steps": 42420, "loss": 0.352, "lr": 0.0004326061303587265, "epoch": 11.775106082036775, "percentage": 58.88, "elapsed_time": "1:20:57", "remaining_time": "0:56:32", "throughput": 5179.83, "total_tokens": 25160864} {"current_steps": 24980, "total_steps": 42420, "loss": 0.3505, "lr": 0.0004324022932774698, "epoch": 11.777463460631777, "percentage": 58.89, "elapsed_time": "1:20:58", "remaining_time": "0:56:31", "throughput": 5179.87, "total_tokens": 25165440} {"current_steps": 24985, "total_steps": 42420, "loss": 0.357, "lr": 0.00043219846763935715, "epoch": 11.77982083922678, "percentage": 58.9, "elapsed_time": "1:20:59", "remaining_time": "0:56:30", "throughput": 5179.98, "total_tokens": 25171488} {"current_steps": 24990, "total_steps": 42420, "loss": 0.3786, "lr": 0.0004319946534788927, "epoch": 11.782178217821782, "percentage": 58.91, "elapsed_time": "1:21:00", "remaining_time": "0:56:30", "throughput": 5180.07, "total_tokens": 25177056} {"current_steps": 24995, "total_steps": 42420, "loss": 0.3399, "lr": 0.0004317908508305786, "epoch": 11.784535596416784, "percentage": 58.92, "elapsed_time": "1:21:01", "remaining_time": "0:56:29", "throughput": 5180.18, "total_tokens": 25182880} {"current_steps": 25000, "total_steps": 42420, "loss": 0.3422, "lr": 0.00043158705972891543, "epoch": 11.786892975011787, "percentage": 58.93, "elapsed_time": "1:21:02", "remaining_time": "0:56:28", "throughput": 5180.25, "total_tokens": 25188128} {"current_steps": 25005, "total_steps": 42420, "loss": 0.3271, "lr": 0.00043138328020840154, "epoch": 11.78925035360679, "percentage": 58.95, "elapsed_time": "1:21:03", "remaining_time": "0:56:27", "throughput": 5180.26, "total_tokens": 25192608} {"current_steps": 25010, "total_steps": 42420, "loss": 0.373, "lr": 0.00043117951230353316, "epoch": 11.791607732201792, "percentage": 58.96, "elapsed_time": "1:21:04", "remaining_time": "0:56:26", "throughput": 5180.35, "total_tokens": 25198144} {"current_steps": 25015, "total_steps": 42420, "loss": 0.3244, "lr": 0.00043097575604880483, "epoch": 11.793965110796794, "percentage": 58.97, "elapsed_time": "1:21:05", "remaining_time": "0:56:25", "throughput": 5180.4, "total_tokens": 25203232} {"current_steps": 25020, "total_steps": 42420, "loss": 0.3257, "lr": 0.00043077201147870904, "epoch": 11.796322489391796, "percentage": 58.98, "elapsed_time": "1:21:05", "remaining_time": "0:56:24", "throughput": 5180.41, "total_tokens": 25207840} {"current_steps": 25025, "total_steps": 42420, "loss": 0.3091, "lr": 0.0004305682786277364, "epoch": 11.798679867986799, "percentage": 58.99, "elapsed_time": "1:21:07", "remaining_time": "0:56:23", "throughput": 5180.59, "total_tokens": 25214912} {"current_steps": 25030, "total_steps": 42420, "loss": 0.3303, "lr": 0.00043036455753037525, "epoch": 11.801037246581801, "percentage": 59.01, "elapsed_time": "1:21:08", "remaining_time": "0:56:22", "throughput": 5180.74, "total_tokens": 25221632} {"current_steps": 25035, "total_steps": 42420, "loss": 0.401, "lr": 0.0004301608482211121, "epoch": 11.803394625176804, "percentage": 59.02, "elapsed_time": "1:21:09", "remaining_time": "0:56:21", "throughput": 5180.79, "total_tokens": 25226528} {"current_steps": 25040, "total_steps": 42420, "loss": 0.3464, "lr": 0.0004299571507344315, "epoch": 11.805752003771806, "percentage": 59.03, "elapsed_time": "1:21:10", "remaining_time": "0:56:20", "throughput": 5180.8, "total_tokens": 25231168} {"current_steps": 25045, "total_steps": 42420, "loss": 0.3342, "lr": 0.00042975346510481626, "epoch": 11.808109382366808, "percentage": 59.04, "elapsed_time": "1:21:10", "remaining_time": "0:56:19", "throughput": 5180.79, "total_tokens": 25235488} {"current_steps": 25050, "total_steps": 42420, "loss": 0.3269, "lr": 0.0004295497913667465, "epoch": 11.81046676096181, "percentage": 59.05, "elapsed_time": "1:21:11", "remaining_time": "0:56:18", "throughput": 5180.87, "total_tokens": 25240960} {"current_steps": 25055, "total_steps": 42420, "loss": 0.344, "lr": 0.0004293461295547009, "epoch": 11.812824139556813, "percentage": 59.06, "elapsed_time": "1:21:12", "remaining_time": "0:56:17", "throughput": 5180.88, "total_tokens": 25245728} {"current_steps": 25060, "total_steps": 42420, "loss": 0.3292, "lr": 0.00042914247970315587, "epoch": 11.815181518151816, "percentage": 59.08, "elapsed_time": "1:21:13", "remaining_time": "0:56:16", "throughput": 5180.94, "total_tokens": 25250752} {"current_steps": 25065, "total_steps": 42420, "loss": 0.3648, "lr": 0.00042893884184658606, "epoch": 11.817538896746818, "percentage": 59.09, "elapsed_time": "1:21:14", "remaining_time": "0:56:15", "throughput": 5180.94, "total_tokens": 25254976} {"current_steps": 25070, "total_steps": 42420, "loss": 0.3043, "lr": 0.0004287352160194636, "epoch": 11.81989627534182, "percentage": 59.1, "elapsed_time": "1:21:15", "remaining_time": "0:56:14", "throughput": 5180.94, "total_tokens": 25259616} {"current_steps": 25075, "total_steps": 42420, "loss": 0.3511, "lr": 0.0004285316022562592, "epoch": 11.822253653936823, "percentage": 59.11, "elapsed_time": "1:21:16", "remaining_time": "0:56:13", "throughput": 5181.0, "total_tokens": 25264864} {"current_steps": 25080, "total_steps": 42420, "loss": 0.3511, "lr": 0.000428328000591441, "epoch": 11.824611032531825, "percentage": 59.12, "elapsed_time": "1:21:17", "remaining_time": "0:56:12", "throughput": 5181.03, "total_tokens": 25269600} {"current_steps": 25085, "total_steps": 42420, "loss": 0.3711, "lr": 0.0004281244110594755, "epoch": 11.826968411126828, "percentage": 59.13, "elapsed_time": "1:21:18", "remaining_time": "0:56:11", "throughput": 5181.05, "total_tokens": 25274240} {"current_steps": 25090, "total_steps": 42420, "loss": 0.3588, "lr": 0.0004279208336948267, "epoch": 11.82932578972183, "percentage": 59.15, "elapsed_time": "1:21:19", "remaining_time": "0:56:10", "throughput": 5181.04, "total_tokens": 25278528} {"current_steps": 25095, "total_steps": 42420, "loss": 0.3706, "lr": 0.00042771726853195696, "epoch": 11.831683168316832, "percentage": 59.16, "elapsed_time": "1:21:19", "remaining_time": "0:56:08", "throughput": 5181.06, "total_tokens": 25283136} {"current_steps": 25100, "total_steps": 42420, "loss": 0.3492, "lr": 0.0004275137156053264, "epoch": 11.834040546911835, "percentage": 59.17, "elapsed_time": "1:21:20", "remaining_time": "0:56:07", "throughput": 5181.11, "total_tokens": 25288192} {"current_steps": 25105, "total_steps": 42420, "loss": 0.3542, "lr": 0.00042731017494939305, "epoch": 11.836397925506837, "percentage": 59.18, "elapsed_time": "1:21:21", "remaining_time": "0:56:06", "throughput": 5181.16, "total_tokens": 25293120} {"current_steps": 25110, "total_steps": 42420, "loss": 0.3258, "lr": 0.00042710664659861283, "epoch": 11.838755304101838, "percentage": 59.19, "elapsed_time": "1:21:22", "remaining_time": "0:56:06", "throughput": 5181.25, "total_tokens": 25298880} {"current_steps": 25115, "total_steps": 42420, "loss": 0.3433, "lr": 0.0004269031305874397, "epoch": 11.841112682696842, "percentage": 59.21, "elapsed_time": "1:21:23", "remaining_time": "0:56:05", "throughput": 5181.36, "total_tokens": 25304512} {"current_steps": 25120, "total_steps": 42420, "loss": 0.3239, "lr": 0.00042669962695032537, "epoch": 11.843470061291843, "percentage": 59.22, "elapsed_time": "1:21:24", "remaining_time": "0:56:04", "throughput": 5181.39, "total_tokens": 25309312} {"current_steps": 25125, "total_steps": 42420, "loss": 0.3392, "lr": 0.0004264961357217198, "epoch": 11.845827439886845, "percentage": 59.23, "elapsed_time": "1:21:25", "remaining_time": "0:56:03", "throughput": 5181.43, "total_tokens": 25314208} {"current_steps": 25130, "total_steps": 42420, "loss": 0.2735, "lr": 0.00042629265693607026, "epoch": 11.848184818481847, "percentage": 59.24, "elapsed_time": "1:21:26", "remaining_time": "0:56:01", "throughput": 5181.38, "total_tokens": 25318048} {"current_steps": 25135, "total_steps": 42420, "loss": 0.3725, "lr": 0.0004260891906278225, "epoch": 11.85054219707685, "percentage": 59.25, "elapsed_time": "1:21:27", "remaining_time": "0:56:00", "throughput": 5181.36, "total_tokens": 25322112} {"current_steps": 25140, "total_steps": 42420, "loss": 0.2696, "lr": 0.0004258857368314198, "epoch": 11.852899575671852, "percentage": 59.26, "elapsed_time": "1:21:28", "remaining_time": "0:55:59", "throughput": 5181.43, "total_tokens": 25327392} {"current_steps": 25145, "total_steps": 42420, "loss": 0.3467, "lr": 0.00042568229558130355, "epoch": 11.855256954266855, "percentage": 59.28, "elapsed_time": "1:21:29", "remaining_time": "0:55:58", "throughput": 5181.56, "total_tokens": 25333440} {"current_steps": 25150, "total_steps": 42420, "loss": 0.4072, "lr": 0.0004254788669119127, "epoch": 11.857614332861857, "percentage": 59.29, "elapsed_time": "1:21:30", "remaining_time": "0:55:57", "throughput": 5181.59, "total_tokens": 25338272} {"current_steps": 25155, "total_steps": 42420, "loss": 0.3184, "lr": 0.0004252754508576845, "epoch": 11.85997171145686, "percentage": 59.3, "elapsed_time": "1:21:30", "remaining_time": "0:55:56", "throughput": 5181.62, "total_tokens": 25343008} {"current_steps": 25160, "total_steps": 42420, "loss": 0.3273, "lr": 0.00042507204745305383, "epoch": 11.862329090051862, "percentage": 59.31, "elapsed_time": "1:21:31", "remaining_time": "0:55:55", "throughput": 5181.7, "total_tokens": 25348576} {"current_steps": 25165, "total_steps": 42420, "loss": 0.3563, "lr": 0.0004248686567324531, "epoch": 11.864686468646864, "percentage": 59.32, "elapsed_time": "1:21:32", "remaining_time": "0:55:54", "throughput": 5181.72, "total_tokens": 25353216} {"current_steps": 25170, "total_steps": 42420, "loss": 0.3278, "lr": 0.00042466527873031315, "epoch": 11.867043847241867, "percentage": 59.34, "elapsed_time": "1:21:33", "remaining_time": "0:55:53", "throughput": 5181.74, "total_tokens": 25357696} {"current_steps": 25175, "total_steps": 42420, "loss": 0.3165, "lr": 0.0004244619134810625, "epoch": 11.869401225836869, "percentage": 59.35, "elapsed_time": "1:21:34", "remaining_time": "0:55:52", "throughput": 5181.88, "total_tokens": 25364128} {"current_steps": 25180, "total_steps": 42420, "loss": 0.3896, "lr": 0.00042425856101912726, "epoch": 11.871758604431871, "percentage": 59.36, "elapsed_time": "1:21:35", "remaining_time": "0:55:51", "throughput": 5181.89, "total_tokens": 25368672} {"current_steps": 25185, "total_steps": 42420, "loss": 0.3291, "lr": 0.0004240552213789316, "epoch": 11.874115983026874, "percentage": 59.37, "elapsed_time": "1:21:36", "remaining_time": "0:55:50", "throughput": 5181.88, "total_tokens": 25372800} {"current_steps": 25190, "total_steps": 42420, "loss": 0.326, "lr": 0.0004238518945948975, "epoch": 11.876473361621876, "percentage": 59.38, "elapsed_time": "1:21:37", "remaining_time": "0:55:49", "throughput": 5181.9, "total_tokens": 25377344} {"current_steps": 25195, "total_steps": 42420, "loss": 0.3499, "lr": 0.0004236485807014448, "epoch": 11.878830740216879, "percentage": 59.39, "elapsed_time": "1:21:38", "remaining_time": "0:55:48", "throughput": 5181.93, "total_tokens": 25382272} {"current_steps": 25200, "total_steps": 42420, "loss": 0.2941, "lr": 0.00042344527973299094, "epoch": 11.881188118811881, "percentage": 59.41, "elapsed_time": "1:21:39", "remaining_time": "0:55:47", "throughput": 5182.01, "total_tokens": 25387744} {"current_steps": 25205, "total_steps": 42420, "loss": 0.2863, "lr": 0.00042324199172395137, "epoch": 11.883545497406883, "percentage": 59.42, "elapsed_time": "1:21:40", "remaining_time": "0:55:46", "throughput": 5182.07, "total_tokens": 25392736} {"current_steps": 25210, "total_steps": 42420, "loss": 0.3572, "lr": 0.0004230387167087393, "epoch": 11.885902876001886, "percentage": 59.43, "elapsed_time": "1:21:41", "remaining_time": "0:55:45", "throughput": 5182.15, "total_tokens": 25398016} {"current_steps": 25215, "total_steps": 42420, "loss": 0.3329, "lr": 0.0004228354547217659, "epoch": 11.888260254596888, "percentage": 59.44, "elapsed_time": "1:21:41", "remaining_time": "0:55:44", "throughput": 5182.14, "total_tokens": 25402400} {"current_steps": 25220, "total_steps": 42420, "loss": 0.3552, "lr": 0.0004226322057974398, "epoch": 11.89061763319189, "percentage": 59.45, "elapsed_time": "1:21:43", "remaining_time": "0:55:43", "throughput": 5182.28, "total_tokens": 25408864} {"current_steps": 25225, "total_steps": 42420, "loss": 0.3164, "lr": 0.00042242896997016753, "epoch": 11.892975011786893, "percentage": 59.46, "elapsed_time": "1:21:44", "remaining_time": "0:55:42", "throughput": 5182.42, "total_tokens": 25415168} {"current_steps": 25230, "total_steps": 42420, "loss": 0.371, "lr": 0.0004222257472743536, "epoch": 11.895332390381895, "percentage": 59.48, "elapsed_time": "1:21:45", "remaining_time": "0:55:41", "throughput": 5182.48, "total_tokens": 25420320} {"current_steps": 25235, "total_steps": 42420, "loss": 0.36, "lr": 0.00042202253774440026, "epoch": 11.897689768976898, "percentage": 59.49, "elapsed_time": "1:21:45", "remaining_time": "0:55:40", "throughput": 5182.53, "total_tokens": 25425344} {"current_steps": 25240, "total_steps": 42420, "loss": 0.328, "lr": 0.00042181934141470714, "epoch": 11.9000471475719, "percentage": 59.5, "elapsed_time": "1:21:46", "remaining_time": "0:55:39", "throughput": 5182.57, "total_tokens": 25430144} {"current_steps": 25245, "total_steps": 42420, "loss": 0.3319, "lr": 0.00042161615831967213, "epoch": 11.902404526166903, "percentage": 59.51, "elapsed_time": "1:21:47", "remaining_time": "0:55:38", "throughput": 5182.67, "total_tokens": 25435648} {"current_steps": 25250, "total_steps": 42420, "loss": 0.2844, "lr": 0.00042141298849369057, "epoch": 11.904761904761905, "percentage": 59.52, "elapsed_time": "1:21:48", "remaining_time": "0:55:37", "throughput": 5182.7, "total_tokens": 25440800} {"current_steps": 25255, "total_steps": 42420, "loss": 0.3251, "lr": 0.00042120983197115595, "epoch": 11.907119283356908, "percentage": 59.54, "elapsed_time": "1:21:49", "remaining_time": "0:55:36", "throughput": 5182.73, "total_tokens": 25445696} {"current_steps": 25260, "total_steps": 42420, "loss": 0.3062, "lr": 0.0004210066887864588, "epoch": 11.90947666195191, "percentage": 59.55, "elapsed_time": "1:21:50", "remaining_time": "0:55:35", "throughput": 5182.69, "total_tokens": 25449408} {"current_steps": 25265, "total_steps": 42420, "loss": 0.3967, "lr": 0.00042080355897398806, "epoch": 11.911834040546912, "percentage": 59.56, "elapsed_time": "1:21:51", "remaining_time": "0:55:34", "throughput": 5182.74, "total_tokens": 25454560} {"current_steps": 25270, "total_steps": 42420, "loss": 0.29, "lr": 0.00042060044256813, "epoch": 11.914191419141915, "percentage": 59.57, "elapsed_time": "1:21:52", "remaining_time": "0:55:33", "throughput": 5182.91, "total_tokens": 25460960} {"current_steps": 25275, "total_steps": 42420, "loss": 0.2779, "lr": 0.0004203973396032691, "epoch": 11.916548797736917, "percentage": 59.58, "elapsed_time": "1:21:53", "remaining_time": "0:55:32", "throughput": 5182.95, "total_tokens": 25465856} {"current_steps": 25280, "total_steps": 42420, "loss": 0.3304, "lr": 0.00042019425011378687, "epoch": 11.91890617633192, "percentage": 59.59, "elapsed_time": "1:21:54", "remaining_time": "0:55:31", "throughput": 5182.93, "total_tokens": 25470432} {"current_steps": 25285, "total_steps": 42420, "loss": 0.2991, "lr": 0.00041999117413406306, "epoch": 11.921263554926922, "percentage": 59.61, "elapsed_time": "1:21:55", "remaining_time": "0:55:30", "throughput": 5183.02, "total_tokens": 25476192} {"current_steps": 25290, "total_steps": 42420, "loss": 0.3568, "lr": 0.00041978811169847505, "epoch": 11.923620933521924, "percentage": 59.62, "elapsed_time": "1:21:56", "remaining_time": "0:55:29", "throughput": 5183.05, "total_tokens": 25480736} {"current_steps": 25295, "total_steps": 42420, "loss": 0.2984, "lr": 0.00041958506284139787, "epoch": 11.925978312116927, "percentage": 59.63, "elapsed_time": "1:21:57", "remaining_time": "0:55:28", "throughput": 5183.07, "total_tokens": 25485440} {"current_steps": 25300, "total_steps": 42420, "loss": 0.4207, "lr": 0.00041938202759720413, "epoch": 11.92833569071193, "percentage": 59.64, "elapsed_time": "1:21:57", "remaining_time": "0:55:27", "throughput": 5183.11, "total_tokens": 25490464} {"current_steps": 25305, "total_steps": 42420, "loss": 0.3678, "lr": 0.00041917900600026424, "epoch": 11.930693069306932, "percentage": 59.65, "elapsed_time": "1:21:58", "remaining_time": "0:55:26", "throughput": 5183.14, "total_tokens": 25495040} {"current_steps": 25310, "total_steps": 42420, "loss": 0.327, "lr": 0.00041897599808494643, "epoch": 11.933050447901934, "percentage": 59.67, "elapsed_time": "1:21:59", "remaining_time": "0:55:25", "throughput": 5183.16, "total_tokens": 25500032} {"current_steps": 25315, "total_steps": 42420, "loss": 0.3588, "lr": 0.00041877300388561643, "epoch": 11.935407826496935, "percentage": 59.68, "elapsed_time": "1:22:00", "remaining_time": "0:55:24", "throughput": 5183.28, "total_tokens": 25506048} {"current_steps": 25320, "total_steps": 42420, "loss": 0.3498, "lr": 0.0004185700234366378, "epoch": 11.937765205091937, "percentage": 59.69, "elapsed_time": "1:22:01", "remaining_time": "0:55:24", "throughput": 5183.42, "total_tokens": 25512800} {"current_steps": 25325, "total_steps": 42420, "loss": 0.346, "lr": 0.0004183670567723714, "epoch": 11.94012258368694, "percentage": 59.7, "elapsed_time": "1:22:02", "remaining_time": "0:55:23", "throughput": 5183.4, "total_tokens": 25517120} {"current_steps": 25330, "total_steps": 42420, "loss": 0.3521, "lr": 0.00041816410392717633, "epoch": 11.942479962281942, "percentage": 59.71, "elapsed_time": "1:22:04", "remaining_time": "0:55:22", "throughput": 5183.58, "total_tokens": 25524160} {"current_steps": 25335, "total_steps": 42420, "loss": 0.3509, "lr": 0.00041796116493540914, "epoch": 11.944837340876944, "percentage": 59.72, "elapsed_time": "1:22:04", "remaining_time": "0:55:21", "throughput": 5183.61, "total_tokens": 25528928} {"current_steps": 25340, "total_steps": 42420, "loss": 0.3465, "lr": 0.00041775823983142355, "epoch": 11.947194719471947, "percentage": 59.74, "elapsed_time": "1:22:05", "remaining_time": "0:55:20", "throughput": 5183.62, "total_tokens": 25533312} {"current_steps": 25345, "total_steps": 42420, "loss": 0.3124, "lr": 0.0004175553286495717, "epoch": 11.949552098066949, "percentage": 59.75, "elapsed_time": "1:22:06", "remaining_time": "0:55:19", "throughput": 5183.67, "total_tokens": 25538336} {"current_steps": 25350, "total_steps": 42420, "loss": 0.3267, "lr": 0.0004173524314242028, "epoch": 11.951909476661951, "percentage": 59.76, "elapsed_time": "1:22:07", "remaining_time": "0:55:18", "throughput": 5183.7, "total_tokens": 25543072} {"current_steps": 25355, "total_steps": 42420, "loss": 0.3433, "lr": 0.0004171495481896641, "epoch": 11.954266855256954, "percentage": 59.77, "elapsed_time": "1:22:08", "remaining_time": "0:55:17", "throughput": 5183.71, "total_tokens": 25547680} {"current_steps": 25360, "total_steps": 42420, "loss": 0.3484, "lr": 0.0004169466789803001, "epoch": 11.956624233851956, "percentage": 59.78, "elapsed_time": "1:22:09", "remaining_time": "0:55:16", "throughput": 5183.73, "total_tokens": 25552256} {"current_steps": 25365, "total_steps": 42420, "loss": 0.3727, "lr": 0.0004167438238304531, "epoch": 11.958981612446959, "percentage": 59.79, "elapsed_time": "1:22:10", "remaining_time": "0:55:15", "throughput": 5183.8, "total_tokens": 25557536} {"current_steps": 25370, "total_steps": 42420, "loss": 0.3191, "lr": 0.0004165409827744632, "epoch": 11.961338991041961, "percentage": 59.81, "elapsed_time": "1:22:11", "remaining_time": "0:55:14", "throughput": 5183.88, "total_tokens": 25562848} {"current_steps": 25375, "total_steps": 42420, "loss": 0.342, "lr": 0.00041633815584666797, "epoch": 11.963696369636963, "percentage": 59.82, "elapsed_time": "1:22:12", "remaining_time": "0:55:12", "throughput": 5183.85, "total_tokens": 25567072} {"current_steps": 25380, "total_steps": 42420, "loss": 0.3366, "lr": 0.0004161353430814023, "epoch": 11.966053748231966, "percentage": 59.83, "elapsed_time": "1:22:12", "remaining_time": "0:55:11", "throughput": 5183.89, "total_tokens": 25571968} {"current_steps": 25385, "total_steps": 42420, "loss": 0.2973, "lr": 0.00041593254451299935, "epoch": 11.968411126826968, "percentage": 59.84, "elapsed_time": "1:22:13", "remaining_time": "0:55:11", "throughput": 5183.97, "total_tokens": 25577472} {"current_steps": 25390, "total_steps": 42420, "loss": 0.3279, "lr": 0.000415729760175789, "epoch": 11.97076850542197, "percentage": 59.85, "elapsed_time": "1:22:14", "remaining_time": "0:55:10", "throughput": 5184.07, "total_tokens": 25583168} {"current_steps": 25395, "total_steps": 42420, "loss": 0.3628, "lr": 0.00041552699010409945, "epoch": 11.973125884016973, "percentage": 59.87, "elapsed_time": "1:22:16", "remaining_time": "0:55:09", "throughput": 5184.21, "total_tokens": 25589408} {"current_steps": 25400, "total_steps": 42420, "loss": 0.3824, "lr": 0.0004153242343322563, "epoch": 11.975483262611975, "percentage": 59.88, "elapsed_time": "1:22:16", "remaining_time": "0:55:08", "throughput": 5184.24, "total_tokens": 25594176} {"current_steps": 25405, "total_steps": 42420, "loss": 0.3109, "lr": 0.0004151214928945825, "epoch": 11.977840641206978, "percentage": 59.89, "elapsed_time": "1:22:18", "remaining_time": "0:55:07", "throughput": 5184.4, "total_tokens": 25600640} {"current_steps": 25410, "total_steps": 42420, "loss": 0.3498, "lr": 0.00041491876582539866, "epoch": 11.98019801980198, "percentage": 59.9, "elapsed_time": "1:22:18", "remaining_time": "0:55:06", "throughput": 5184.41, "total_tokens": 25605568} {"current_steps": 25415, "total_steps": 42420, "loss": 0.3141, "lr": 0.0004147160531590233, "epoch": 11.982555398396983, "percentage": 59.91, "elapsed_time": "1:22:19", "remaining_time": "0:55:05", "throughput": 5184.46, "total_tokens": 25610528} {"current_steps": 25420, "total_steps": 42420, "loss": 0.3309, "lr": 0.00041451335492977213, "epoch": 11.984912776991985, "percentage": 59.92, "elapsed_time": "1:22:20", "remaining_time": "0:55:04", "throughput": 5184.54, "total_tokens": 25616000} {"current_steps": 25425, "total_steps": 42420, "loss": 0.3711, "lr": 0.00041431067117195843, "epoch": 11.987270155586987, "percentage": 59.94, "elapsed_time": "1:22:21", "remaining_time": "0:55:03", "throughput": 5184.66, "total_tokens": 25622112} {"current_steps": 25430, "total_steps": 42420, "loss": 0.3153, "lr": 0.00041410800191989317, "epoch": 11.98962753418199, "percentage": 59.95, "elapsed_time": "1:22:22", "remaining_time": "0:55:02", "throughput": 5184.67, "total_tokens": 25626688} {"current_steps": 25435, "total_steps": 42420, "loss": 0.3545, "lr": 0.00041390534720788484, "epoch": 11.991984912776992, "percentage": 59.96, "elapsed_time": "1:22:23", "remaining_time": "0:55:01", "throughput": 5184.71, "total_tokens": 25631840} {"current_steps": 25440, "total_steps": 42420, "loss": 0.3625, "lr": 0.0004137027070702395, "epoch": 11.994342291371995, "percentage": 59.97, "elapsed_time": "1:22:24", "remaining_time": "0:55:00", "throughput": 5184.72, "total_tokens": 25636320} {"current_steps": 25445, "total_steps": 42420, "loss": 0.3469, "lr": 0.00041350008154126053, "epoch": 11.996699669966997, "percentage": 59.98, "elapsed_time": "1:22:25", "remaining_time": "0:54:59", "throughput": 5184.74, "total_tokens": 25640928} {"current_steps": 25450, "total_steps": 42420, "loss": 0.3508, "lr": 0.000413297470655249, "epoch": 11.999057048562, "percentage": 60.0, "elapsed_time": "1:22:26", "remaining_time": "0:54:58", "throughput": 5184.76, "total_tokens": 25645568} {"current_steps": 25452, "total_steps": 42420, "eval_loss": 0.35587719082832336, "epoch": 12.0, "percentage": 60.0, "elapsed_time": "1:22:45", "remaining_time": "0:55:10", "throughput": 5164.89, "total_tokens": 25647872} {"current_steps": 25455, "total_steps": 42420, "loss": 0.3484, "lr": 0.0004130948744465037, "epoch": 12.001414427157002, "percentage": 60.01, "elapsed_time": "1:22:48", "remaining_time": "0:55:11", "throughput": 5162.82, "total_tokens": 25650464} {"current_steps": 25460, "total_steps": 42420, "loss": 0.3827, "lr": 0.00041289229294932055, "epoch": 12.003771805752004, "percentage": 60.02, "elapsed_time": "1:22:49", "remaining_time": "0:55:10", "throughput": 5162.79, "total_tokens": 25654432} {"current_steps": 25465, "total_steps": 42420, "loss": 0.3916, "lr": 0.00041268972619799317, "epoch": 12.006129184347007, "percentage": 60.03, "elapsed_time": "1:22:49", "remaining_time": "0:55:09", "throughput": 5162.79, "total_tokens": 25658816} {"current_steps": 25470, "total_steps": 42420, "loss": 0.3425, "lr": 0.0004124871742268127, "epoch": 12.008486562942009, "percentage": 60.04, "elapsed_time": "1:22:50", "remaining_time": "0:55:08", "throughput": 5162.82, "total_tokens": 25664192} {"current_steps": 25475, "total_steps": 42420, "loss": 0.3484, "lr": 0.00041228463707006777, "epoch": 12.010843941537011, "percentage": 60.05, "elapsed_time": "1:22:51", "remaining_time": "0:55:07", "throughput": 5162.83, "total_tokens": 25668608} {"current_steps": 25480, "total_steps": 42420, "loss": 0.3965, "lr": 0.0004120821147620445, "epoch": 12.013201320132014, "percentage": 60.07, "elapsed_time": "1:22:52", "remaining_time": "0:55:05", "throughput": 5162.84, "total_tokens": 25673024} {"current_steps": 25485, "total_steps": 42420, "loss": 0.4013, "lr": 0.00041187960733702635, "epoch": 12.015558698727016, "percentage": 60.08, "elapsed_time": "1:22:53", "remaining_time": "0:55:04", "throughput": 5162.93, "total_tokens": 25678208} {"current_steps": 25490, "total_steps": 42420, "loss": 0.3365, "lr": 0.0004116771148292945, "epoch": 12.017916077322019, "percentage": 60.09, "elapsed_time": "1:22:54", "remaining_time": "0:55:03", "throughput": 5162.95, "total_tokens": 25683040} {"current_steps": 25495, "total_steps": 42420, "loss": 0.3356, "lr": 0.0004114746372731275, "epoch": 12.020273455917021, "percentage": 60.1, "elapsed_time": "1:22:55", "remaining_time": "0:55:02", "throughput": 5163.01, "total_tokens": 25688160} {"current_steps": 25500, "total_steps": 42420, "loss": 0.3462, "lr": 0.00041127217470280145, "epoch": 12.022630834512023, "percentage": 60.11, "elapsed_time": "1:22:56", "remaining_time": "0:55:01", "throughput": 5163.11, "total_tokens": 25693664} {"current_steps": 25505, "total_steps": 42420, "loss": 0.3422, "lr": 0.0004110697271525895, "epoch": 12.024988213107026, "percentage": 60.12, "elapsed_time": "1:22:57", "remaining_time": "0:55:00", "throughput": 5163.09, "total_tokens": 25697792} {"current_steps": 25510, "total_steps": 42420, "loss": 0.3584, "lr": 0.0004108672946567629, "epoch": 12.027345591702028, "percentage": 60.14, "elapsed_time": "1:22:58", "remaining_time": "0:54:59", "throughput": 5163.13, "total_tokens": 25702560} {"current_steps": 25515, "total_steps": 42420, "loss": 0.325, "lr": 0.00041066487724958985, "epoch": 12.029702970297029, "percentage": 60.15, "elapsed_time": "1:22:59", "remaining_time": "0:54:58", "throughput": 5163.23, "total_tokens": 25708416} {"current_steps": 25520, "total_steps": 42420, "loss": 0.3441, "lr": 0.0004104624749653364, "epoch": 12.032060348892031, "percentage": 60.16, "elapsed_time": "1:23:00", "remaining_time": "0:54:57", "throughput": 5163.28, "total_tokens": 25713376} {"current_steps": 25525, "total_steps": 42420, "loss": 0.3151, "lr": 0.0004102600878382656, "epoch": 12.034417727487034, "percentage": 60.17, "elapsed_time": "1:23:01", "remaining_time": "0:54:56", "throughput": 5163.39, "total_tokens": 25718976} {"current_steps": 25530, "total_steps": 42420, "loss": 0.3335, "lr": 0.0004100577159026382, "epoch": 12.036775106082036, "percentage": 60.18, "elapsed_time": "1:23:02", "remaining_time": "0:54:56", "throughput": 5163.53, "total_tokens": 25725056} {"current_steps": 25535, "total_steps": 42420, "loss": 0.4084, "lr": 0.00040985535919271234, "epoch": 12.039132484677038, "percentage": 60.2, "elapsed_time": "1:23:02", "remaining_time": "0:54:54", "throughput": 5163.55, "total_tokens": 25729600} {"current_steps": 25540, "total_steps": 42420, "loss": 0.3666, "lr": 0.00040965301774274375, "epoch": 12.04148986327204, "percentage": 60.21, "elapsed_time": "1:23:03", "remaining_time": "0:54:53", "throughput": 5163.54, "total_tokens": 25734016} {"current_steps": 25545, "total_steps": 42420, "loss": 0.3509, "lr": 0.0004094506915869851, "epoch": 12.043847241867043, "percentage": 60.22, "elapsed_time": "1:23:04", "remaining_time": "0:54:52", "throughput": 5163.6, "total_tokens": 25738912} {"current_steps": 25550, "total_steps": 42420, "loss": 0.3468, "lr": 0.00040924838075968693, "epoch": 12.046204620462046, "percentage": 60.23, "elapsed_time": "1:23:05", "remaining_time": "0:54:52", "throughput": 5163.81, "total_tokens": 25746176} {"current_steps": 25555, "total_steps": 42420, "loss": 0.3407, "lr": 0.0004090460852950971, "epoch": 12.048561999057048, "percentage": 60.24, "elapsed_time": "1:23:06", "remaining_time": "0:54:51", "throughput": 5163.88, "total_tokens": 25751296} {"current_steps": 25560, "total_steps": 42420, "loss": 0.3478, "lr": 0.00040884380522746076, "epoch": 12.05091937765205, "percentage": 60.25, "elapsed_time": "1:23:07", "remaining_time": "0:54:50", "throughput": 5163.92, "total_tokens": 25756224} {"current_steps": 25565, "total_steps": 42420, "loss": 0.332, "lr": 0.00040864154059102036, "epoch": 12.053276756247053, "percentage": 60.27, "elapsed_time": "1:23:08", "remaining_time": "0:54:48", "throughput": 5163.95, "total_tokens": 25760928} {"current_steps": 25570, "total_steps": 42420, "loss": 0.3612, "lr": 0.00040843929142001596, "epoch": 12.055634134842055, "percentage": 60.28, "elapsed_time": "1:23:09", "remaining_time": "0:54:47", "throughput": 5163.96, "total_tokens": 25765248} {"current_steps": 25575, "total_steps": 42420, "loss": 0.3536, "lr": 0.0004082370577486849, "epoch": 12.057991513437058, "percentage": 60.29, "elapsed_time": "1:23:10", "remaining_time": "0:54:46", "throughput": 5163.96, "total_tokens": 25769664} {"current_steps": 25580, "total_steps": 42420, "loss": 0.3327, "lr": 0.000408034839611262, "epoch": 12.06034889203206, "percentage": 60.3, "elapsed_time": "1:23:11", "remaining_time": "0:54:45", "throughput": 5163.94, "total_tokens": 25773824} {"current_steps": 25585, "total_steps": 42420, "loss": 0.3411, "lr": 0.00040783263704197923, "epoch": 12.062706270627062, "percentage": 60.31, "elapsed_time": "1:23:12", "remaining_time": "0:54:44", "throughput": 5164.01, "total_tokens": 25779232} {"current_steps": 25590, "total_steps": 42420, "loss": 0.354, "lr": 0.00040763045007506606, "epoch": 12.065063649222065, "percentage": 60.33, "elapsed_time": "1:23:13", "remaining_time": "0:54:43", "throughput": 5164.09, "total_tokens": 25784896} {"current_steps": 25595, "total_steps": 42420, "loss": 0.3652, "lr": 0.0004074282787447494, "epoch": 12.067421027817067, "percentage": 60.34, "elapsed_time": "1:23:13", "remaining_time": "0:54:42", "throughput": 5164.09, "total_tokens": 25789152} {"current_steps": 25600, "total_steps": 42420, "loss": 0.354, "lr": 0.00040722612308525337, "epoch": 12.06977840641207, "percentage": 60.35, "elapsed_time": "1:23:14", "remaining_time": "0:54:41", "throughput": 5164.11, "total_tokens": 25793888} {"current_steps": 25605, "total_steps": 42420, "loss": 0.3835, "lr": 0.0004070239831307995, "epoch": 12.072135785007072, "percentage": 60.36, "elapsed_time": "1:23:15", "remaining_time": "0:54:40", "throughput": 5164.14, "total_tokens": 25798688} {"current_steps": 25610, "total_steps": 42420, "loss": 0.3688, "lr": 0.00040682185891560674, "epoch": 12.074493163602074, "percentage": 60.37, "elapsed_time": "1:23:16", "remaining_time": "0:54:39", "throughput": 5164.18, "total_tokens": 25803520} {"current_steps": 25615, "total_steps": 42420, "loss": 0.3476, "lr": 0.00040661975047389103, "epoch": 12.076850542197077, "percentage": 60.38, "elapsed_time": "1:23:17", "remaining_time": "0:54:38", "throughput": 5164.23, "total_tokens": 25808448} {"current_steps": 25620, "total_steps": 42420, "loss": 0.3564, "lr": 0.0004064176578398661, "epoch": 12.07920792079208, "percentage": 60.4, "elapsed_time": "1:23:18", "remaining_time": "0:54:37", "throughput": 5164.37, "total_tokens": 25814752} {"current_steps": 25625, "total_steps": 42420, "loss": 0.3381, "lr": 0.0004062155810477427, "epoch": 12.081565299387082, "percentage": 60.41, "elapsed_time": "1:23:19", "remaining_time": "0:54:36", "throughput": 5164.38, "total_tokens": 25819136} {"current_steps": 25630, "total_steps": 42420, "loss": 0.3344, "lr": 0.0004060135201317292, "epoch": 12.083922677982084, "percentage": 60.42, "elapsed_time": "1:23:20", "remaining_time": "0:54:35", "throughput": 5164.4, "total_tokens": 25823872} {"current_steps": 25635, "total_steps": 42420, "loss": 0.354, "lr": 0.0004058114751260308, "epoch": 12.086280056577086, "percentage": 60.43, "elapsed_time": "1:23:21", "remaining_time": "0:54:34", "throughput": 5164.4, "total_tokens": 25828256} {"current_steps": 25640, "total_steps": 42420, "loss": 0.3644, "lr": 0.0004056094460648505, "epoch": 12.088637435172089, "percentage": 60.44, "elapsed_time": "1:23:22", "remaining_time": "0:54:33", "throughput": 5164.44, "total_tokens": 25833248} {"current_steps": 25645, "total_steps": 42420, "loss": 0.3232, "lr": 0.00040540743298238827, "epoch": 12.090994813767091, "percentage": 60.45, "elapsed_time": "1:23:23", "remaining_time": "0:54:32", "throughput": 5164.5, "total_tokens": 25838144} {"current_steps": 25650, "total_steps": 42420, "loss": 0.3432, "lr": 0.00040520543591284164, "epoch": 12.093352192362094, "percentage": 60.47, "elapsed_time": "1:23:24", "remaining_time": "0:54:31", "throughput": 5164.56, "total_tokens": 25843488} {"current_steps": 25655, "total_steps": 42420, "loss": 0.3185, "lr": 0.00040500345489040515, "epoch": 12.095709570957096, "percentage": 60.48, "elapsed_time": "1:23:25", "remaining_time": "0:54:30", "throughput": 5164.65, "total_tokens": 25849344} {"current_steps": 25660, "total_steps": 42420, "loss": 0.3392, "lr": 0.0004048014899492707, "epoch": 12.098066949552098, "percentage": 60.49, "elapsed_time": "1:23:25", "remaining_time": "0:54:29", "throughput": 5164.72, "total_tokens": 25854400} {"current_steps": 25665, "total_steps": 42420, "loss": 0.3342, "lr": 0.0004045995411236277, "epoch": 12.100424328147101, "percentage": 60.5, "elapsed_time": "1:23:26", "remaining_time": "0:54:28", "throughput": 5164.78, "total_tokens": 25859808} {"current_steps": 25670, "total_steps": 42420, "loss": 0.336, "lr": 0.00040439760844766267, "epoch": 12.102781706742103, "percentage": 60.51, "elapsed_time": "1:23:27", "remaining_time": "0:54:27", "throughput": 5164.83, "total_tokens": 25864928} {"current_steps": 25675, "total_steps": 42420, "loss": 0.3108, "lr": 0.00040419569195555914, "epoch": 12.105139085337106, "percentage": 60.53, "elapsed_time": "1:23:28", "remaining_time": "0:54:26", "throughput": 5164.9, "total_tokens": 25870144} {"current_steps": 25680, "total_steps": 42420, "loss": 0.3248, "lr": 0.00040399379168149836, "epoch": 12.107496463932108, "percentage": 60.54, "elapsed_time": "1:23:29", "remaining_time": "0:54:25", "throughput": 5164.98, "total_tokens": 25875328} {"current_steps": 25685, "total_steps": 42420, "loss": 0.3444, "lr": 0.0004037919076596585, "epoch": 12.10985384252711, "percentage": 60.55, "elapsed_time": "1:23:30", "remaining_time": "0:54:24", "throughput": 5164.97, "total_tokens": 25879488} {"current_steps": 25690, "total_steps": 42420, "loss": 0.3381, "lr": 0.0004035900399242153, "epoch": 12.112211221122113, "percentage": 60.56, "elapsed_time": "1:23:31", "remaining_time": "0:54:23", "throughput": 5165.05, "total_tokens": 25884864} {"current_steps": 25695, "total_steps": 42420, "loss": 0.344, "lr": 0.0004033881885093413, "epoch": 12.114568599717115, "percentage": 60.57, "elapsed_time": "1:23:32", "remaining_time": "0:54:22", "throughput": 5165.16, "total_tokens": 25890944} {"current_steps": 25700, "total_steps": 42420, "loss": 0.3286, "lr": 0.0004031863534492066, "epoch": 12.116925978312118, "percentage": 60.58, "elapsed_time": "1:23:33", "remaining_time": "0:54:21", "throughput": 5165.28, "total_tokens": 25897024} {"current_steps": 25705, "total_steps": 42420, "loss": 0.367, "lr": 0.0004029845347779784, "epoch": 12.11928335690712, "percentage": 60.6, "elapsed_time": "1:23:34", "remaining_time": "0:54:20", "throughput": 5165.36, "total_tokens": 25902272} {"current_steps": 25710, "total_steps": 42420, "loss": 0.3177, "lr": 0.0004027827325298214, "epoch": 12.121640735502123, "percentage": 60.61, "elapsed_time": "1:23:35", "remaining_time": "0:54:19", "throughput": 5165.4, "total_tokens": 25906816} {"current_steps": 25715, "total_steps": 42420, "loss": 0.3225, "lr": 0.000402580946738897, "epoch": 12.123998114097123, "percentage": 60.62, "elapsed_time": "1:23:36", "remaining_time": "0:54:18", "throughput": 5165.46, "total_tokens": 25911968} {"current_steps": 25720, "total_steps": 42420, "loss": 0.3261, "lr": 0.00040237917743936427, "epoch": 12.126355492692126, "percentage": 60.63, "elapsed_time": "1:23:37", "remaining_time": "0:54:17", "throughput": 5165.45, "total_tokens": 25916256} {"current_steps": 25725, "total_steps": 42420, "loss": 0.3734, "lr": 0.0004021774246653793, "epoch": 12.128712871287128, "percentage": 60.64, "elapsed_time": "1:23:38", "remaining_time": "0:54:16", "throughput": 5165.48, "total_tokens": 25921056} {"current_steps": 25730, "total_steps": 42420, "loss": 0.3327, "lr": 0.00040197568845109545, "epoch": 12.13107024988213, "percentage": 60.66, "elapsed_time": "1:23:38", "remaining_time": "0:54:15", "throughput": 5165.47, "total_tokens": 25925184} {"current_steps": 25735, "total_steps": 42420, "loss": 0.3401, "lr": 0.00040177396883066306, "epoch": 12.133427628477133, "percentage": 60.67, "elapsed_time": "1:23:39", "remaining_time": "0:54:14", "throughput": 5165.49, "total_tokens": 25929824} {"current_steps": 25740, "total_steps": 42420, "loss": 0.3155, "lr": 0.00040157226583823, "epoch": 12.135785007072135, "percentage": 60.68, "elapsed_time": "1:23:40", "remaining_time": "0:54:13", "throughput": 5165.51, "total_tokens": 25934336} {"current_steps": 25745, "total_steps": 42420, "loss": 0.3545, "lr": 0.00040137057950794116, "epoch": 12.138142385667138, "percentage": 60.69, "elapsed_time": "1:23:41", "remaining_time": "0:54:12", "throughput": 5165.56, "total_tokens": 25939456} {"current_steps": 25750, "total_steps": 42420, "loss": 0.3024, "lr": 0.00040116890987393865, "epoch": 12.14049976426214, "percentage": 60.7, "elapsed_time": "1:23:42", "remaining_time": "0:54:11", "throughput": 5165.63, "total_tokens": 25944640} {"current_steps": 25755, "total_steps": 42420, "loss": 0.3684, "lr": 0.0004009672569703615, "epoch": 12.142857142857142, "percentage": 60.71, "elapsed_time": "1:23:43", "remaining_time": "0:54:10", "throughput": 5165.69, "total_tokens": 25950112} {"current_steps": 25760, "total_steps": 42420, "loss": 0.3285, "lr": 0.0004007656208313464, "epoch": 12.145214521452145, "percentage": 60.73, "elapsed_time": "1:23:44", "remaining_time": "0:54:09", "throughput": 5165.73, "total_tokens": 25955168} {"current_steps": 25765, "total_steps": 42420, "loss": 0.342, "lr": 0.00040056400149102666, "epoch": 12.147571900047147, "percentage": 60.74, "elapsed_time": "1:23:45", "remaining_time": "0:54:08", "throughput": 5165.77, "total_tokens": 25959872} {"current_steps": 25770, "total_steps": 42420, "loss": 0.3106, "lr": 0.00040036239898353325, "epoch": 12.14992927864215, "percentage": 60.75, "elapsed_time": "1:23:46", "remaining_time": "0:54:07", "throughput": 5165.74, "total_tokens": 25963968} {"current_steps": 25775, "total_steps": 42420, "loss": 0.3238, "lr": 0.00040016081334299386, "epoch": 12.152286657237152, "percentage": 60.76, "elapsed_time": "1:23:47", "remaining_time": "0:54:06", "throughput": 5165.8, "total_tokens": 25969184} {"current_steps": 25780, "total_steps": 42420, "loss": 0.3516, "lr": 0.00039995924460353363, "epoch": 12.154644035832154, "percentage": 60.77, "elapsed_time": "1:23:48", "remaining_time": "0:54:05", "throughput": 5165.87, "total_tokens": 25974304} {"current_steps": 25785, "total_steps": 42420, "loss": 0.3723, "lr": 0.00039975769279927464, "epoch": 12.157001414427157, "percentage": 60.79, "elapsed_time": "1:23:48", "remaining_time": "0:54:04", "throughput": 5165.94, "total_tokens": 25979488} {"current_steps": 25790, "total_steps": 42420, "loss": 0.3575, "lr": 0.0003995561579643364, "epoch": 12.15935879302216, "percentage": 60.8, "elapsed_time": "1:23:49", "remaining_time": "0:54:03", "throughput": 5166.01, "total_tokens": 25984928} {"current_steps": 25795, "total_steps": 42420, "loss": 0.3385, "lr": 0.00039935464013283507, "epoch": 12.161716171617162, "percentage": 60.81, "elapsed_time": "1:23:50", "remaining_time": "0:54:02", "throughput": 5166.1, "total_tokens": 25990464} {"current_steps": 25800, "total_steps": 42420, "loss": 0.3395, "lr": 0.00039915313933888425, "epoch": 12.164073550212164, "percentage": 60.82, "elapsed_time": "1:23:52", "remaining_time": "0:54:01", "throughput": 5166.25, "total_tokens": 25997088} {"current_steps": 25805, "total_steps": 42420, "loss": 0.3464, "lr": 0.0003989516556165947, "epoch": 12.166430928807166, "percentage": 60.83, "elapsed_time": "1:23:53", "remaining_time": "0:54:00", "throughput": 5166.29, "total_tokens": 26001984} {"current_steps": 25810, "total_steps": 42420, "loss": 0.3118, "lr": 0.00039875018900007423, "epoch": 12.168788307402169, "percentage": 60.84, "elapsed_time": "1:23:53", "remaining_time": "0:53:59", "throughput": 5166.3, "total_tokens": 26006464} {"current_steps": 25815, "total_steps": 42420, "loss": 0.3116, "lr": 0.00039854873952342756, "epoch": 12.171145685997171, "percentage": 60.86, "elapsed_time": "1:23:54", "remaining_time": "0:53:58", "throughput": 5166.35, "total_tokens": 26011552} {"current_steps": 25820, "total_steps": 42420, "loss": 0.3868, "lr": 0.0003983473072207567, "epoch": 12.173503064592174, "percentage": 60.87, "elapsed_time": "1:23:55", "remaining_time": "0:53:57", "throughput": 5166.43, "total_tokens": 26017184} {"current_steps": 25825, "total_steps": 42420, "loss": 0.2841, "lr": 0.0003981458921261608, "epoch": 12.175860443187176, "percentage": 60.88, "elapsed_time": "1:23:56", "remaining_time": "0:53:56", "throughput": 5166.44, "total_tokens": 26021664} {"current_steps": 25830, "total_steps": 42420, "loss": 0.3992, "lr": 0.00039794449427373615, "epoch": 12.178217821782178, "percentage": 60.89, "elapsed_time": "1:23:57", "remaining_time": "0:53:55", "throughput": 5166.51, "total_tokens": 26026880} {"current_steps": 25835, "total_steps": 42420, "loss": 0.3105, "lr": 0.0003977431136975757, "epoch": 12.18057520037718, "percentage": 60.9, "elapsed_time": "1:23:58", "remaining_time": "0:53:54", "throughput": 5166.54, "total_tokens": 26031776} {"current_steps": 25840, "total_steps": 42420, "loss": 0.3759, "lr": 0.00039754175043177, "epoch": 12.182932578972183, "percentage": 60.91, "elapsed_time": "1:23:59", "remaining_time": "0:53:53", "throughput": 5166.51, "total_tokens": 26035968} {"current_steps": 25845, "total_steps": 42420, "loss": 0.3199, "lr": 0.0003973404045104061, "epoch": 12.185289957567186, "percentage": 60.93, "elapsed_time": "1:24:00", "remaining_time": "0:53:52", "throughput": 5166.62, "total_tokens": 26041792} {"current_steps": 25850, "total_steps": 42420, "loss": 0.3517, "lr": 0.00039713907596756863, "epoch": 12.187647336162188, "percentage": 60.94, "elapsed_time": "1:24:01", "remaining_time": "0:53:51", "throughput": 5166.62, "total_tokens": 26046016} {"current_steps": 25855, "total_steps": 42420, "loss": 0.3663, "lr": 0.0003969377648373392, "epoch": 12.19000471475719, "percentage": 60.95, "elapsed_time": "1:24:02", "remaining_time": "0:53:50", "throughput": 5166.73, "total_tokens": 26052224} {"current_steps": 25860, "total_steps": 42420, "loss": 0.3591, "lr": 0.0003967364711537964, "epoch": 12.192362093352193, "percentage": 60.96, "elapsed_time": "1:24:03", "remaining_time": "0:53:49", "throughput": 5166.72, "total_tokens": 26056640} {"current_steps": 25865, "total_steps": 42420, "loss": 0.3476, "lr": 0.0003965351949510155, "epoch": 12.194719471947195, "percentage": 60.97, "elapsed_time": "1:24:04", "remaining_time": "0:53:48", "throughput": 5166.82, "total_tokens": 26062400} {"current_steps": 25870, "total_steps": 42420, "loss": 0.3581, "lr": 0.00039633393626306953, "epoch": 12.197076850542198, "percentage": 60.99, "elapsed_time": "1:24:05", "remaining_time": "0:53:47", "throughput": 5166.82, "total_tokens": 26066944} {"current_steps": 25875, "total_steps": 42420, "loss": 0.3509, "lr": 0.0003961326951240279, "epoch": 12.1994342291372, "percentage": 61.0, "elapsed_time": "1:24:05", "remaining_time": "0:53:46", "throughput": 5166.83, "total_tokens": 26071552} {"current_steps": 25880, "total_steps": 42420, "loss": 0.3334, "lr": 0.0003959314715679575, "epoch": 12.201791607732202, "percentage": 61.01, "elapsed_time": "1:24:06", "remaining_time": "0:53:45", "throughput": 5166.82, "total_tokens": 26075840} {"current_steps": 25885, "total_steps": 42420, "loss": 0.3371, "lr": 0.00039573026562892197, "epoch": 12.204148986327205, "percentage": 61.02, "elapsed_time": "1:24:07", "remaining_time": "0:53:44", "throughput": 5166.92, "total_tokens": 26081568} {"current_steps": 25890, "total_steps": 42420, "loss": 0.3658, "lr": 0.00039552907734098197, "epoch": 12.206506364922207, "percentage": 61.03, "elapsed_time": "1:24:08", "remaining_time": "0:53:43", "throughput": 5166.99, "total_tokens": 26086496} {"current_steps": 25895, "total_steps": 42420, "loss": 0.3421, "lr": 0.0003953279067381954, "epoch": 12.20886374351721, "percentage": 61.04, "elapsed_time": "1:24:09", "remaining_time": "0:53:42", "throughput": 5167.02, "total_tokens": 26091552} {"current_steps": 25900, "total_steps": 42420, "loss": 0.3885, "lr": 0.0003951267538546171, "epoch": 12.211221122112212, "percentage": 61.06, "elapsed_time": "1:24:10", "remaining_time": "0:53:41", "throughput": 5167.06, "total_tokens": 26096448} {"current_steps": 25905, "total_steps": 42420, "loss": 0.3341, "lr": 0.0003949256187242986, "epoch": 12.213578500707214, "percentage": 61.07, "elapsed_time": "1:24:11", "remaining_time": "0:53:40", "throughput": 5167.19, "total_tokens": 26102656} {"current_steps": 25910, "total_steps": 42420, "loss": 0.3278, "lr": 0.0003947245013812888, "epoch": 12.215935879302217, "percentage": 61.08, "elapsed_time": "1:24:12", "remaining_time": "0:53:39", "throughput": 5167.31, "total_tokens": 26108768} {"current_steps": 25915, "total_steps": 42420, "loss": 0.3187, "lr": 0.0003945234018596335, "epoch": 12.218293257897217, "percentage": 61.09, "elapsed_time": "1:24:13", "remaining_time": "0:53:38", "throughput": 5167.3, "total_tokens": 26113216} {"current_steps": 25920, "total_steps": 42420, "loss": 0.3781, "lr": 0.0003943223201933754, "epoch": 12.22065063649222, "percentage": 61.1, "elapsed_time": "1:24:14", "remaining_time": "0:53:37", "throughput": 5167.45, "total_tokens": 26119360} {"current_steps": 25925, "total_steps": 42420, "loss": 0.3279, "lr": 0.0003941212564165541, "epoch": 12.223008015087222, "percentage": 61.12, "elapsed_time": "1:24:15", "remaining_time": "0:53:36", "throughput": 5167.49, "total_tokens": 26124288} {"current_steps": 25930, "total_steps": 42420, "loss": 0.3453, "lr": 0.0003939202105632064, "epoch": 12.225365393682225, "percentage": 61.13, "elapsed_time": "1:24:16", "remaining_time": "0:53:35", "throughput": 5167.52, "total_tokens": 26128928} {"current_steps": 25935, "total_steps": 42420, "loss": 0.3663, "lr": 0.0003937191826673659, "epoch": 12.227722772277227, "percentage": 61.14, "elapsed_time": "1:24:17", "remaining_time": "0:53:34", "throughput": 5167.54, "total_tokens": 26133568} {"current_steps": 25940, "total_steps": 42420, "loss": 0.3191, "lr": 0.00039351817276306333, "epoch": 12.23008015087223, "percentage": 61.15, "elapsed_time": "1:24:18", "remaining_time": "0:53:33", "throughput": 5167.57, "total_tokens": 26138720} {"current_steps": 25945, "total_steps": 42420, "loss": 0.3538, "lr": 0.000393317180884326, "epoch": 12.232437529467232, "percentage": 61.16, "elapsed_time": "1:24:19", "remaining_time": "0:53:32", "throughput": 5167.62, "total_tokens": 26143712} {"current_steps": 25950, "total_steps": 42420, "loss": 0.3699, "lr": 0.00039311620706517864, "epoch": 12.234794908062234, "percentage": 61.17, "elapsed_time": "1:24:20", "remaining_time": "0:53:31", "throughput": 5167.65, "total_tokens": 26148480} {"current_steps": 25955, "total_steps": 42420, "loss": 0.3441, "lr": 0.0003929152513396426, "epoch": 12.237152286657237, "percentage": 61.19, "elapsed_time": "1:24:20", "remaining_time": "0:53:30", "throughput": 5167.68, "total_tokens": 26153184} {"current_steps": 25960, "total_steps": 42420, "loss": 0.3345, "lr": 0.0003927143137417365, "epoch": 12.239509665252239, "percentage": 61.2, "elapsed_time": "1:24:21", "remaining_time": "0:53:29", "throughput": 5167.68, "total_tokens": 26157376} {"current_steps": 25965, "total_steps": 42420, "loss": 0.3215, "lr": 0.0003925133943054753, "epoch": 12.241867043847241, "percentage": 61.21, "elapsed_time": "1:24:22", "remaining_time": "0:53:28", "throughput": 5167.84, "total_tokens": 26163968} {"current_steps": 25970, "total_steps": 42420, "loss": 0.3899, "lr": 0.00039231249306487145, "epoch": 12.244224422442244, "percentage": 61.22, "elapsed_time": "1:24:23", "remaining_time": "0:53:27", "throughput": 5167.86, "total_tokens": 26168672} {"current_steps": 25975, "total_steps": 42420, "loss": 0.3622, "lr": 0.000392111610053934, "epoch": 12.246581801037246, "percentage": 61.23, "elapsed_time": "1:24:24", "remaining_time": "0:53:26", "throughput": 5167.9, "total_tokens": 26173760} {"current_steps": 25980, "total_steps": 42420, "loss": 0.3247, "lr": 0.0003919107453066694, "epoch": 12.248939179632249, "percentage": 61.24, "elapsed_time": "1:24:25", "remaining_time": "0:53:25", "throughput": 5167.88, "total_tokens": 26177824} {"current_steps": 25985, "total_steps": 42420, "loss": 0.3548, "lr": 0.00039170989885708014, "epoch": 12.251296558227251, "percentage": 61.26, "elapsed_time": "1:24:26", "remaining_time": "0:53:24", "throughput": 5167.89, "total_tokens": 26182336} {"current_steps": 25990, "total_steps": 42420, "loss": 0.3351, "lr": 0.00039150907073916643, "epoch": 12.253653936822253, "percentage": 61.27, "elapsed_time": "1:24:27", "remaining_time": "0:53:23", "throughput": 5167.99, "total_tokens": 26188000} {"current_steps": 25995, "total_steps": 42420, "loss": 0.3265, "lr": 0.0003913082609869248, "epoch": 12.256011315417256, "percentage": 61.28, "elapsed_time": "1:24:28", "remaining_time": "0:53:22", "throughput": 5168.12, "total_tokens": 26194592} {"current_steps": 26000, "total_steps": 42420, "loss": 0.3306, "lr": 0.0003911074696343494, "epoch": 12.258368694012258, "percentage": 61.29, "elapsed_time": "1:24:29", "remaining_time": "0:53:21", "throughput": 5168.15, "total_tokens": 26199168} {"current_steps": 26005, "total_steps": 42420, "loss": 0.3108, "lr": 0.00039090669671543036, "epoch": 12.26072607260726, "percentage": 61.3, "elapsed_time": "1:24:30", "remaining_time": "0:53:20", "throughput": 5168.28, "total_tokens": 26205504} {"current_steps": 26010, "total_steps": 42420, "loss": 0.3142, "lr": 0.00039070594226415525, "epoch": 12.263083451202263, "percentage": 61.32, "elapsed_time": "1:24:31", "remaining_time": "0:53:19", "throughput": 5168.32, "total_tokens": 26210656} {"current_steps": 26015, "total_steps": 42420, "loss": 0.335, "lr": 0.00039050520631450837, "epoch": 12.265440829797265, "percentage": 61.33, "elapsed_time": "1:24:32", "remaining_time": "0:53:18", "throughput": 5168.49, "total_tokens": 26217408} {"current_steps": 26020, "total_steps": 42420, "loss": 0.3127, "lr": 0.00039030448890047106, "epoch": 12.267798208392268, "percentage": 61.34, "elapsed_time": "1:24:33", "remaining_time": "0:53:17", "throughput": 5168.54, "total_tokens": 26222368} {"current_steps": 26025, "total_steps": 42420, "loss": 0.2837, "lr": 0.00039010379005602114, "epoch": 12.27015558698727, "percentage": 61.35, "elapsed_time": "1:24:34", "remaining_time": "0:53:16", "throughput": 5168.56, "total_tokens": 26227232} {"current_steps": 26030, "total_steps": 42420, "loss": 0.4203, "lr": 0.00038990310981513364, "epoch": 12.272512965582273, "percentage": 61.36, "elapsed_time": "1:24:35", "remaining_time": "0:53:15", "throughput": 5168.55, "total_tokens": 26231328} {"current_steps": 26035, "total_steps": 42420, "loss": 0.3379, "lr": 0.00038970244821178037, "epoch": 12.274870344177275, "percentage": 61.37, "elapsed_time": "1:24:36", "remaining_time": "0:53:14", "throughput": 5168.61, "total_tokens": 26236352} {"current_steps": 26040, "total_steps": 42420, "loss": 0.3473, "lr": 0.0003895018052799299, "epoch": 12.277227722772277, "percentage": 61.39, "elapsed_time": "1:24:36", "remaining_time": "0:53:13", "throughput": 5168.6, "total_tokens": 26240608} {"current_steps": 26045, "total_steps": 42420, "loss": 0.3604, "lr": 0.00038930118105354763, "epoch": 12.27958510136728, "percentage": 61.4, "elapsed_time": "1:24:37", "remaining_time": "0:53:12", "throughput": 5168.6, "total_tokens": 26244896} {"current_steps": 26050, "total_steps": 42420, "loss": 0.311, "lr": 0.0003891005755665957, "epoch": 12.281942479962282, "percentage": 61.41, "elapsed_time": "1:24:38", "remaining_time": "0:53:11", "throughput": 5168.63, "total_tokens": 26249504} {"current_steps": 26055, "total_steps": 42420, "loss": 0.3222, "lr": 0.00038889998885303353, "epoch": 12.284299858557285, "percentage": 61.42, "elapsed_time": "1:24:39", "remaining_time": "0:53:10", "throughput": 5168.63, "total_tokens": 26253760} {"current_steps": 26060, "total_steps": 42420, "loss": 0.3196, "lr": 0.000388699420946817, "epoch": 12.286657237152287, "percentage": 61.43, "elapsed_time": "1:24:40", "remaining_time": "0:53:09", "throughput": 5168.62, "total_tokens": 26258208} {"current_steps": 26065, "total_steps": 42420, "loss": 0.399, "lr": 0.00038849887188189847, "epoch": 12.28901461574729, "percentage": 61.45, "elapsed_time": "1:24:41", "remaining_time": "0:53:08", "throughput": 5168.73, "total_tokens": 26264032} {"current_steps": 26070, "total_steps": 42420, "loss": 0.3308, "lr": 0.0003882983416922279, "epoch": 12.291371994342292, "percentage": 61.46, "elapsed_time": "1:24:42", "remaining_time": "0:53:07", "throughput": 5168.75, "total_tokens": 26268608} {"current_steps": 26075, "total_steps": 42420, "loss": 0.3771, "lr": 0.00038809783041175147, "epoch": 12.293729372937294, "percentage": 61.47, "elapsed_time": "1:24:43", "remaining_time": "0:53:06", "throughput": 5168.75, "total_tokens": 26273280} {"current_steps": 26080, "total_steps": 42420, "loss": 0.3443, "lr": 0.0003878973380744122, "epoch": 12.296086751532297, "percentage": 61.48, "elapsed_time": "1:24:43", "remaining_time": "0:53:05", "throughput": 5168.76, "total_tokens": 26277600} {"current_steps": 26085, "total_steps": 42420, "loss": 0.3567, "lr": 0.0003876968647141503, "epoch": 12.298444130127299, "percentage": 61.49, "elapsed_time": "1:24:44", "remaining_time": "0:53:04", "throughput": 5168.82, "total_tokens": 26282880} {"current_steps": 26090, "total_steps": 42420, "loss": 0.3258, "lr": 0.00038749641036490257, "epoch": 12.300801508722301, "percentage": 61.5, "elapsed_time": "1:24:45", "remaining_time": "0:53:03", "throughput": 5168.88, "total_tokens": 26288192} {"current_steps": 26095, "total_steps": 42420, "loss": 0.3056, "lr": 0.00038729597506060215, "epoch": 12.303158887317304, "percentage": 61.52, "elapsed_time": "1:24:46", "remaining_time": "0:53:02", "throughput": 5168.88, "total_tokens": 26292704} {"current_steps": 26100, "total_steps": 42420, "loss": 0.3706, "lr": 0.0003870955588351796, "epoch": 12.305516265912306, "percentage": 61.53, "elapsed_time": "1:24:47", "remaining_time": "0:53:01", "throughput": 5168.95, "total_tokens": 26298176} {"current_steps": 26105, "total_steps": 42420, "loss": 0.3076, "lr": 0.0003868951617225619, "epoch": 12.307873644507309, "percentage": 61.54, "elapsed_time": "1:24:48", "remaining_time": "0:53:00", "throughput": 5169.0, "total_tokens": 26303296} {"current_steps": 26110, "total_steps": 42420, "loss": 0.2557, "lr": 0.00038669478375667304, "epoch": 12.310231023102311, "percentage": 61.55, "elapsed_time": "1:24:49", "remaining_time": "0:52:59", "throughput": 5169.01, "total_tokens": 26308000} {"current_steps": 26115, "total_steps": 42420, "loss": 0.3523, "lr": 0.00038649442497143335, "epoch": 12.312588401697312, "percentage": 61.56, "elapsed_time": "1:24:50", "remaining_time": "0:52:58", "throughput": 5169.05, "total_tokens": 26312704} {"current_steps": 26120, "total_steps": 42420, "loss": 0.3773, "lr": 0.0003862940854007603, "epoch": 12.314945780292314, "percentage": 61.57, "elapsed_time": "1:24:51", "remaining_time": "0:52:57", "throughput": 5169.1, "total_tokens": 26317856} {"current_steps": 26125, "total_steps": 42420, "loss": 0.2821, "lr": 0.000386093765078568, "epoch": 12.317303158887317, "percentage": 61.59, "elapsed_time": "1:24:52", "remaining_time": "0:52:56", "throughput": 5169.14, "total_tokens": 26322720} {"current_steps": 26130, "total_steps": 42420, "loss": 0.3667, "lr": 0.00038589346403876737, "epoch": 12.319660537482319, "percentage": 61.6, "elapsed_time": "1:24:53", "remaining_time": "0:52:55", "throughput": 5169.19, "total_tokens": 26327680} {"current_steps": 26135, "total_steps": 42420, "loss": 0.3356, "lr": 0.0003856931823152657, "epoch": 12.322017916077321, "percentage": 61.61, "elapsed_time": "1:24:54", "remaining_time": "0:52:54", "throughput": 5169.19, "total_tokens": 26332288} {"current_steps": 26140, "total_steps": 42420, "loss": 0.3331, "lr": 0.00038549291994196747, "epoch": 12.324375294672324, "percentage": 61.62, "elapsed_time": "1:24:55", "remaining_time": "0:52:53", "throughput": 5169.27, "total_tokens": 26337664} {"current_steps": 26145, "total_steps": 42420, "loss": 0.3233, "lr": 0.0003852926769527737, "epoch": 12.326732673267326, "percentage": 61.63, "elapsed_time": "1:24:56", "remaining_time": "0:52:52", "throughput": 5169.37, "total_tokens": 26343264} {"current_steps": 26150, "total_steps": 42420, "loss": 0.3252, "lr": 0.0003850924533815822, "epoch": 12.329090051862329, "percentage": 61.65, "elapsed_time": "1:24:56", "remaining_time": "0:52:51", "throughput": 5169.42, "total_tokens": 26348448} {"current_steps": 26155, "total_steps": 42420, "loss": 0.357, "lr": 0.00038489224926228715, "epoch": 12.331447430457331, "percentage": 61.66, "elapsed_time": "1:24:58", "remaining_time": "0:52:50", "throughput": 5169.53, "total_tokens": 26354496} {"current_steps": 26160, "total_steps": 42420, "loss": 0.3258, "lr": 0.00038469206462877993, "epoch": 12.333804809052333, "percentage": 61.67, "elapsed_time": "1:24:59", "remaining_time": "0:52:49", "throughput": 5169.68, "total_tokens": 26360928} {"current_steps": 26165, "total_steps": 42420, "loss": 0.3328, "lr": 0.00038449189951494836, "epoch": 12.336162187647336, "percentage": 61.68, "elapsed_time": "1:25:00", "remaining_time": "0:52:48", "throughput": 5169.69, "total_tokens": 26365536} {"current_steps": 26170, "total_steps": 42420, "loss": 0.3079, "lr": 0.0003842917539546771, "epoch": 12.338519566242338, "percentage": 61.69, "elapsed_time": "1:25:00", "remaining_time": "0:52:47", "throughput": 5169.68, "total_tokens": 26369632} {"current_steps": 26175, "total_steps": 42420, "loss": 0.3157, "lr": 0.00038409162798184717, "epoch": 12.34087694483734, "percentage": 61.7, "elapsed_time": "1:25:01", "remaining_time": "0:52:46", "throughput": 5169.72, "total_tokens": 26374656} {"current_steps": 26180, "total_steps": 42420, "loss": 0.3109, "lr": 0.00038389152163033656, "epoch": 12.343234323432343, "percentage": 61.72, "elapsed_time": "1:25:02", "remaining_time": "0:52:45", "throughput": 5169.8, "total_tokens": 26380352} {"current_steps": 26185, "total_steps": 42420, "loss": 0.3001, "lr": 0.00038369143493402, "epoch": 12.345591702027345, "percentage": 61.73, "elapsed_time": "1:25:03", "remaining_time": "0:52:44", "throughput": 5169.84, "total_tokens": 26385120} {"current_steps": 26190, "total_steps": 42420, "loss": 0.4019, "lr": 0.00038349136792676877, "epoch": 12.347949080622348, "percentage": 61.74, "elapsed_time": "1:25:04", "remaining_time": "0:52:43", "throughput": 5169.97, "total_tokens": 26391264} {"current_steps": 26195, "total_steps": 42420, "loss": 0.3466, "lr": 0.00038329132064245065, "epoch": 12.35030645921735, "percentage": 61.75, "elapsed_time": "1:25:05", "remaining_time": "0:52:42", "throughput": 5169.93, "total_tokens": 26395008} {"current_steps": 26200, "total_steps": 42420, "loss": 0.333, "lr": 0.00038309129311493034, "epoch": 12.352663837812353, "percentage": 61.76, "elapsed_time": "1:25:06", "remaining_time": "0:52:41", "throughput": 5170.0, "total_tokens": 26400672} {"current_steps": 26205, "total_steps": 42420, "loss": 0.3528, "lr": 0.0003828912853780691, "epoch": 12.355021216407355, "percentage": 61.78, "elapsed_time": "1:25:07", "remaining_time": "0:52:40", "throughput": 5170.08, "total_tokens": 26406176} {"current_steps": 26210, "total_steps": 42420, "loss": 0.3238, "lr": 0.00038269129746572496, "epoch": 12.357378595002357, "percentage": 61.79, "elapsed_time": "1:25:08", "remaining_time": "0:52:39", "throughput": 5170.17, "total_tokens": 26411712} {"current_steps": 26215, "total_steps": 42420, "loss": 0.3408, "lr": 0.0003824913294117523, "epoch": 12.35973597359736, "percentage": 61.8, "elapsed_time": "1:25:09", "remaining_time": "0:52:38", "throughput": 5170.13, "total_tokens": 26415456} {"current_steps": 26220, "total_steps": 42420, "loss": 0.3165, "lr": 0.00038229138125000237, "epoch": 12.362093352192362, "percentage": 61.81, "elapsed_time": "1:25:10", "remaining_time": "0:52:37", "throughput": 5170.18, "total_tokens": 26420416} {"current_steps": 26225, "total_steps": 42420, "loss": 0.3018, "lr": 0.000382091453014323, "epoch": 12.364450730787365, "percentage": 61.82, "elapsed_time": "1:25:11", "remaining_time": "0:52:36", "throughput": 5170.29, "total_tokens": 26426368} {"current_steps": 26230, "total_steps": 42420, "loss": 0.3297, "lr": 0.00038189154473855877, "epoch": 12.366808109382367, "percentage": 61.83, "elapsed_time": "1:25:12", "remaining_time": "0:52:35", "throughput": 5170.31, "total_tokens": 26431040} {"current_steps": 26235, "total_steps": 42420, "loss": 0.2847, "lr": 0.0003816916564565505, "epoch": 12.36916548797737, "percentage": 61.85, "elapsed_time": "1:25:13", "remaining_time": "0:52:34", "throughput": 5170.42, "total_tokens": 26437088} {"current_steps": 26240, "total_steps": 42420, "loss": 0.3693, "lr": 0.00038149178820213607, "epoch": 12.371522866572372, "percentage": 61.86, "elapsed_time": "1:25:14", "remaining_time": "0:52:33", "throughput": 5170.47, "total_tokens": 26442144} {"current_steps": 26245, "total_steps": 42420, "loss": 0.3397, "lr": 0.0003812919400091498, "epoch": 12.373880245167374, "percentage": 61.87, "elapsed_time": "1:25:15", "remaining_time": "0:52:32", "throughput": 5170.54, "total_tokens": 26447584} {"current_steps": 26250, "total_steps": 42420, "loss": 0.3788, "lr": 0.00038109211191142254, "epoch": 12.376237623762377, "percentage": 61.88, "elapsed_time": "1:25:15", "remaining_time": "0:52:31", "throughput": 5170.55, "total_tokens": 26452384} {"current_steps": 26255, "total_steps": 42420, "loss": 0.322, "lr": 0.0003808923039427817, "epoch": 12.378595002357379, "percentage": 61.89, "elapsed_time": "1:25:16", "remaining_time": "0:52:30", "throughput": 5170.63, "total_tokens": 26457792} {"current_steps": 26260, "total_steps": 42420, "loss": 0.3551, "lr": 0.0003806925161370516, "epoch": 12.380952380952381, "percentage": 61.9, "elapsed_time": "1:25:17", "remaining_time": "0:52:29", "throughput": 5170.67, "total_tokens": 26462624} {"current_steps": 26265, "total_steps": 42420, "loss": 0.3385, "lr": 0.00038049274852805265, "epoch": 12.383309759547384, "percentage": 61.92, "elapsed_time": "1:25:18", "remaining_time": "0:52:28", "throughput": 5170.72, "total_tokens": 26467680} {"current_steps": 26270, "total_steps": 42420, "loss": 0.3201, "lr": 0.0003802930011496024, "epoch": 12.385667138142386, "percentage": 61.93, "elapsed_time": "1:25:19", "remaining_time": "0:52:27", "throughput": 5170.76, "total_tokens": 26472768} {"current_steps": 26275, "total_steps": 42420, "loss": 0.3569, "lr": 0.00038009327403551446, "epoch": 12.388024516737389, "percentage": 61.94, "elapsed_time": "1:25:20", "remaining_time": "0:52:26", "throughput": 5170.75, "total_tokens": 26477152} {"current_steps": 26280, "total_steps": 42420, "loss": 0.3062, "lr": 0.0003798935672195993, "epoch": 12.390381895332391, "percentage": 61.95, "elapsed_time": "1:25:21", "remaining_time": "0:52:25", "throughput": 5170.73, "total_tokens": 26481152} {"current_steps": 26285, "total_steps": 42420, "loss": 0.3515, "lr": 0.0003796938807356642, "epoch": 12.392739273927393, "percentage": 61.96, "elapsed_time": "1:25:22", "remaining_time": "0:52:24", "throughput": 5170.68, "total_tokens": 26484928} {"current_steps": 26290, "total_steps": 42420, "loss": 0.3136, "lr": 0.0003794942146175121, "epoch": 12.395096652522396, "percentage": 61.98, "elapsed_time": "1:25:23", "remaining_time": "0:52:23", "throughput": 5170.78, "total_tokens": 26490560} {"current_steps": 26295, "total_steps": 42420, "loss": 0.3444, "lr": 0.0003792945688989436, "epoch": 12.397454031117398, "percentage": 61.99, "elapsed_time": "1:25:24", "remaining_time": "0:52:22", "throughput": 5170.81, "total_tokens": 26495360} {"current_steps": 26300, "total_steps": 42420, "loss": 0.3093, "lr": 0.00037909494361375496, "epoch": 12.3998114097124, "percentage": 62.0, "elapsed_time": "1:25:24", "remaining_time": "0:52:21", "throughput": 5170.77, "total_tokens": 26499392} {"current_steps": 26305, "total_steps": 42420, "loss": 0.3613, "lr": 0.00037889533879573953, "epoch": 12.402168788307403, "percentage": 62.01, "elapsed_time": "1:25:25", "remaining_time": "0:52:20", "throughput": 5170.92, "total_tokens": 26505536} {"current_steps": 26310, "total_steps": 42420, "loss": 0.3684, "lr": 0.00037869575447868696, "epoch": 12.404526166902404, "percentage": 62.02, "elapsed_time": "1:25:26", "remaining_time": "0:52:19", "throughput": 5170.97, "total_tokens": 26510592} {"current_steps": 26315, "total_steps": 42420, "loss": 0.3699, "lr": 0.00037849619069638363, "epoch": 12.406883545497408, "percentage": 62.03, "elapsed_time": "1:25:27", "remaining_time": "0:52:18", "throughput": 5171.02, "total_tokens": 26515648} {"current_steps": 26320, "total_steps": 42420, "loss": 0.3287, "lr": 0.00037829664748261215, "epoch": 12.409240924092408, "percentage": 62.05, "elapsed_time": "1:25:28", "remaining_time": "0:52:17", "throughput": 5171.04, "total_tokens": 26520256} {"current_steps": 26325, "total_steps": 42420, "loss": 0.4509, "lr": 0.0003780971248711518, "epoch": 12.41159830268741, "percentage": 62.06, "elapsed_time": "1:25:29", "remaining_time": "0:52:16", "throughput": 5171.04, "total_tokens": 26524736} {"current_steps": 26330, "total_steps": 42420, "loss": 0.3927, "lr": 0.00037789762289577845, "epoch": 12.413955681282413, "percentage": 62.07, "elapsed_time": "1:25:30", "remaining_time": "0:52:15", "throughput": 5171.05, "total_tokens": 26529184} {"current_steps": 26335, "total_steps": 42420, "loss": 0.3353, "lr": 0.0003776981415902645, "epoch": 12.416313059877416, "percentage": 62.08, "elapsed_time": "1:25:31", "remaining_time": "0:52:14", "throughput": 5171.05, "total_tokens": 26533728} {"current_steps": 26340, "total_steps": 42420, "loss": 0.3333, "lr": 0.00037749868098837846, "epoch": 12.418670438472418, "percentage": 62.09, "elapsed_time": "1:25:32", "remaining_time": "0:52:13", "throughput": 5171.1, "total_tokens": 26538720} {"current_steps": 26345, "total_steps": 42420, "loss": 0.3216, "lr": 0.00037729924112388577, "epoch": 12.42102781706742, "percentage": 62.11, "elapsed_time": "1:25:33", "remaining_time": "0:52:12", "throughput": 5171.18, "total_tokens": 26544416} {"current_steps": 26350, "total_steps": 42420, "loss": 0.328, "lr": 0.0003770998220305482, "epoch": 12.423385195662423, "percentage": 62.12, "elapsed_time": "1:25:34", "remaining_time": "0:52:11", "throughput": 5171.21, "total_tokens": 26549376} {"current_steps": 26355, "total_steps": 42420, "loss": 0.3437, "lr": 0.0003769004237421242, "epoch": 12.425742574257425, "percentage": 62.13, "elapsed_time": "1:25:35", "remaining_time": "0:52:10", "throughput": 5171.27, "total_tokens": 26554688} {"current_steps": 26360, "total_steps": 42420, "loss": 0.3137, "lr": 0.00037670104629236817, "epoch": 12.428099952852428, "percentage": 62.14, "elapsed_time": "1:25:35", "remaining_time": "0:52:09", "throughput": 5171.26, "total_tokens": 26558848} {"current_steps": 26365, "total_steps": 42420, "loss": 0.3435, "lr": 0.00037650168971503157, "epoch": 12.43045733144743, "percentage": 62.15, "elapsed_time": "1:25:36", "remaining_time": "0:52:08", "throughput": 5171.34, "total_tokens": 26564416} {"current_steps": 26370, "total_steps": 42420, "loss": 0.3317, "lr": 0.000376302354043862, "epoch": 12.432814710042432, "percentage": 62.16, "elapsed_time": "1:25:37", "remaining_time": "0:52:07", "throughput": 5171.37, "total_tokens": 26569472} {"current_steps": 26375, "total_steps": 42420, "loss": 0.3111, "lr": 0.0003761030393126037, "epoch": 12.435172088637435, "percentage": 62.18, "elapsed_time": "1:25:38", "remaining_time": "0:52:06", "throughput": 5171.47, "total_tokens": 26575040} {"current_steps": 26380, "total_steps": 42420, "loss": 0.3231, "lr": 0.0003759037455549971, "epoch": 12.437529467232437, "percentage": 62.19, "elapsed_time": "1:25:39", "remaining_time": "0:52:05", "throughput": 5171.48, "total_tokens": 26579392} {"current_steps": 26385, "total_steps": 42420, "loss": 0.3333, "lr": 0.00037570447280477936, "epoch": 12.43988684582744, "percentage": 62.2, "elapsed_time": "1:25:40", "remaining_time": "0:52:04", "throughput": 5171.68, "total_tokens": 26586912} {"current_steps": 26390, "total_steps": 42420, "loss": 0.3476, "lr": 0.00037550522109568396, "epoch": 12.442244224422442, "percentage": 62.21, "elapsed_time": "1:25:41", "remaining_time": "0:52:03", "throughput": 5171.71, "total_tokens": 26591744} {"current_steps": 26395, "total_steps": 42420, "loss": 0.3072, "lr": 0.00037530599046144093, "epoch": 12.444601603017444, "percentage": 62.22, "elapsed_time": "1:25:42", "remaining_time": "0:52:02", "throughput": 5171.81, "total_tokens": 26597728} {"current_steps": 26400, "total_steps": 42420, "loss": 0.2522, "lr": 0.00037510678093577644, "epoch": 12.446958981612447, "percentage": 62.23, "elapsed_time": "1:25:43", "remaining_time": "0:52:01", "throughput": 5171.88, "total_tokens": 26603200} {"current_steps": 26405, "total_steps": 42420, "loss": 0.3821, "lr": 0.00037490759255241334, "epoch": 12.44931636020745, "percentage": 62.25, "elapsed_time": "1:25:44", "remaining_time": "0:52:00", "throughput": 5171.88, "total_tokens": 26607616} {"current_steps": 26410, "total_steps": 42420, "loss": 0.3392, "lr": 0.00037470842534507087, "epoch": 12.451673738802452, "percentage": 62.26, "elapsed_time": "1:25:45", "remaining_time": "0:51:59", "throughput": 5171.92, "total_tokens": 26612480} {"current_steps": 26415, "total_steps": 42420, "loss": 0.3283, "lr": 0.0003745092793474648, "epoch": 12.454031117397454, "percentage": 62.27, "elapsed_time": "1:25:46", "remaining_time": "0:51:58", "throughput": 5171.99, "total_tokens": 26617696} {"current_steps": 26420, "total_steps": 42420, "loss": 0.3863, "lr": 0.00037431015459330684, "epoch": 12.456388495992456, "percentage": 62.28, "elapsed_time": "1:25:47", "remaining_time": "0:51:57", "throughput": 5172.02, "total_tokens": 26622528} {"current_steps": 26425, "total_steps": 42420, "loss": 0.3703, "lr": 0.00037411105111630574, "epoch": 12.458745874587459, "percentage": 62.29, "elapsed_time": "1:25:48", "remaining_time": "0:51:56", "throughput": 5172.05, "total_tokens": 26627232} {"current_steps": 26430, "total_steps": 42420, "loss": 0.3217, "lr": 0.0003739119689501661, "epoch": 12.461103253182461, "percentage": 62.31, "elapsed_time": "1:25:49", "remaining_time": "0:51:55", "throughput": 5172.11, "total_tokens": 26632512} {"current_steps": 26435, "total_steps": 42420, "loss": 0.3255, "lr": 0.00037371290812858937, "epoch": 12.463460631777464, "percentage": 62.32, "elapsed_time": "1:25:50", "remaining_time": "0:51:54", "throughput": 5172.13, "total_tokens": 26637408} {"current_steps": 26440, "total_steps": 42420, "loss": 0.3175, "lr": 0.00037351386868527303, "epoch": 12.465818010372466, "percentage": 62.33, "elapsed_time": "1:25:51", "remaining_time": "0:51:53", "throughput": 5172.17, "total_tokens": 26642464} {"current_steps": 26445, "total_steps": 42420, "loss": 0.3179, "lr": 0.0003733148506539111, "epoch": 12.468175388967468, "percentage": 62.34, "elapsed_time": "1:25:52", "remaining_time": "0:51:52", "throughput": 5172.22, "total_tokens": 26647424} {"current_steps": 26450, "total_steps": 42420, "loss": 0.3089, "lr": 0.0003731158540681939, "epoch": 12.47053276756247, "percentage": 62.35, "elapsed_time": "1:25:52", "remaining_time": "0:51:51", "throughput": 5172.24, "total_tokens": 26651904} {"current_steps": 26455, "total_steps": 42420, "loss": 0.396, "lr": 0.00037291687896180844, "epoch": 12.472890146157473, "percentage": 62.36, "elapsed_time": "1:25:53", "remaining_time": "0:51:50", "throughput": 5172.26, "total_tokens": 26656384} {"current_steps": 26460, "total_steps": 42420, "loss": 0.3239, "lr": 0.0003727179253684374, "epoch": 12.475247524752476, "percentage": 62.38, "elapsed_time": "1:25:54", "remaining_time": "0:51:49", "throughput": 5172.3, "total_tokens": 26661344} {"current_steps": 26465, "total_steps": 42420, "loss": 0.3336, "lr": 0.0003725189933217605, "epoch": 12.477604903347478, "percentage": 62.39, "elapsed_time": "1:25:55", "remaining_time": "0:51:48", "throughput": 5172.39, "total_tokens": 26666944} {"current_steps": 26470, "total_steps": 42420, "loss": 0.3042, "lr": 0.0003723200828554536, "epoch": 12.47996228194248, "percentage": 62.4, "elapsed_time": "1:25:56", "remaining_time": "0:51:47", "throughput": 5172.62, "total_tokens": 26674720} {"current_steps": 26475, "total_steps": 42420, "loss": 0.3521, "lr": 0.0003721211940031889, "epoch": 12.482319660537483, "percentage": 62.41, "elapsed_time": "1:25:57", "remaining_time": "0:51:46", "throughput": 5172.65, "total_tokens": 26679648} {"current_steps": 26480, "total_steps": 42420, "loss": 0.2839, "lr": 0.00037192232679863466, "epoch": 12.484677039132485, "percentage": 62.42, "elapsed_time": "1:25:58", "remaining_time": "0:51:45", "throughput": 5172.66, "total_tokens": 26684288} {"current_steps": 26485, "total_steps": 42420, "loss": 0.3009, "lr": 0.0003717234812754558, "epoch": 12.487034417727488, "percentage": 62.44, "elapsed_time": "1:25:59", "remaining_time": "0:51:44", "throughput": 5172.72, "total_tokens": 26689408} {"current_steps": 26490, "total_steps": 42420, "loss": 0.3109, "lr": 0.0003715246574673136, "epoch": 12.48939179632249, "percentage": 62.45, "elapsed_time": "1:26:00", "remaining_time": "0:51:43", "throughput": 5172.79, "total_tokens": 26694848} {"current_steps": 26495, "total_steps": 42420, "loss": 0.303, "lr": 0.00037132585540786557, "epoch": 12.491749174917492, "percentage": 62.46, "elapsed_time": "1:26:01", "remaining_time": "0:51:42", "throughput": 5172.77, "total_tokens": 26698880} {"current_steps": 26500, "total_steps": 42420, "loss": 0.2837, "lr": 0.0003711270751307654, "epoch": 12.494106553512495, "percentage": 62.47, "elapsed_time": "1:26:02", "remaining_time": "0:51:41", "throughput": 5172.79, "total_tokens": 26703520} {"current_steps": 26505, "total_steps": 42420, "loss": 0.3117, "lr": 0.0003709283166696633, "epoch": 12.496463932107497, "percentage": 62.48, "elapsed_time": "1:26:03", "remaining_time": "0:51:40", "throughput": 5172.89, "total_tokens": 26709024} {"current_steps": 26510, "total_steps": 42420, "loss": 0.3016, "lr": 0.00037072958005820577, "epoch": 12.4988213107025, "percentage": 62.49, "elapsed_time": "1:26:04", "remaining_time": "0:51:39", "throughput": 5172.92, "total_tokens": 26713856} {"current_steps": 26515, "total_steps": 42420, "loss": 0.2698, "lr": 0.0003705308653300353, "epoch": 12.5011786892975, "percentage": 62.51, "elapsed_time": "1:26:05", "remaining_time": "0:51:38", "throughput": 5173.04, "total_tokens": 26720256} {"current_steps": 26520, "total_steps": 42420, "loss": 0.339, "lr": 0.00037033217251879115, "epoch": 12.503536067892503, "percentage": 62.52, "elapsed_time": "1:26:06", "remaining_time": "0:51:37", "throughput": 5173.04, "total_tokens": 26724576} {"current_steps": 26525, "total_steps": 42420, "loss": 0.3074, "lr": 0.00037013350165810853, "epoch": 12.505893446487505, "percentage": 62.53, "elapsed_time": "1:26:07", "remaining_time": "0:51:36", "throughput": 5173.14, "total_tokens": 26730144} {"current_steps": 26530, "total_steps": 42420, "loss": 0.3551, "lr": 0.00036993485278161907, "epoch": 12.508250825082508, "percentage": 62.54, "elapsed_time": "1:26:07", "remaining_time": "0:51:35", "throughput": 5173.15, "total_tokens": 26734528} {"current_steps": 26535, "total_steps": 42420, "loss": 0.3132, "lr": 0.00036973622592295066, "epoch": 12.51060820367751, "percentage": 62.55, "elapsed_time": "1:26:09", "remaining_time": "0:51:34", "throughput": 5173.33, "total_tokens": 26741536} {"current_steps": 26540, "total_steps": 42420, "loss": 0.3097, "lr": 0.00036953762111572734, "epoch": 12.512965582272512, "percentage": 62.56, "elapsed_time": "1:26:10", "remaining_time": "0:51:33", "throughput": 5173.37, "total_tokens": 26746592} {"current_steps": 26545, "total_steps": 42420, "loss": 0.3502, "lr": 0.0003693390383935698, "epoch": 12.515322960867515, "percentage": 62.58, "elapsed_time": "1:26:10", "remaining_time": "0:51:32", "throughput": 5173.43, "total_tokens": 26751744} {"current_steps": 26550, "total_steps": 42420, "loss": 0.3619, "lr": 0.00036914047779009444, "epoch": 12.517680339462517, "percentage": 62.59, "elapsed_time": "1:26:11", "remaining_time": "0:51:31", "throughput": 5173.44, "total_tokens": 26756544} {"current_steps": 26555, "total_steps": 42420, "loss": 0.3317, "lr": 0.0003689419393389144, "epoch": 12.52003771805752, "percentage": 62.6, "elapsed_time": "1:26:12", "remaining_time": "0:51:30", "throughput": 5173.52, "total_tokens": 26762016} {"current_steps": 26560, "total_steps": 42420, "loss": 0.3427, "lr": 0.0003687434230736387, "epoch": 12.522395096652522, "percentage": 62.61, "elapsed_time": "1:26:13", "remaining_time": "0:51:29", "throughput": 5173.61, "total_tokens": 26767712} {"current_steps": 26565, "total_steps": 42420, "loss": 0.3103, "lr": 0.000368544929027873, "epoch": 12.524752475247524, "percentage": 62.62, "elapsed_time": "1:26:14", "remaining_time": "0:51:28", "throughput": 5173.68, "total_tokens": 26772896} {"current_steps": 26570, "total_steps": 42420, "loss": 0.3453, "lr": 0.00036834645723521874, "epoch": 12.527109853842527, "percentage": 62.64, "elapsed_time": "1:26:15", "remaining_time": "0:51:27", "throughput": 5173.74, "total_tokens": 26778112} {"current_steps": 26575, "total_steps": 42420, "loss": 0.3698, "lr": 0.0003681480077292739, "epoch": 12.52946723243753, "percentage": 62.65, "elapsed_time": "1:26:16", "remaining_time": "0:51:26", "throughput": 5173.79, "total_tokens": 26782944} {"current_steps": 26580, "total_steps": 42420, "loss": 0.3859, "lr": 0.0003679495805436327, "epoch": 12.531824611032532, "percentage": 62.66, "elapsed_time": "1:26:17", "remaining_time": "0:51:25", "throughput": 5173.86, "total_tokens": 26788192} {"current_steps": 26585, "total_steps": 42420, "loss": 0.3026, "lr": 0.00036775117571188545, "epoch": 12.534181989627534, "percentage": 62.67, "elapsed_time": "1:26:18", "remaining_time": "0:51:24", "throughput": 5173.89, "total_tokens": 26792992} {"current_steps": 26590, "total_steps": 42420, "loss": 0.3454, "lr": 0.0003675527932676186, "epoch": 12.536539368222536, "percentage": 62.68, "elapsed_time": "1:26:19", "remaining_time": "0:51:23", "throughput": 5173.94, "total_tokens": 26797952} {"current_steps": 26595, "total_steps": 42420, "loss": 0.3238, "lr": 0.0003673544332444151, "epoch": 12.538896746817539, "percentage": 62.69, "elapsed_time": "1:26:20", "remaining_time": "0:51:22", "throughput": 5173.94, "total_tokens": 26802592} {"current_steps": 26600, "total_steps": 42420, "loss": 0.37, "lr": 0.00036715609567585373, "epoch": 12.541254125412541, "percentage": 62.71, "elapsed_time": "1:26:21", "remaining_time": "0:51:21", "throughput": 5174.0, "total_tokens": 26807776} {"current_steps": 26605, "total_steps": 42420, "loss": 0.3324, "lr": 0.00036695778059551, "epoch": 12.543611504007544, "percentage": 62.72, "elapsed_time": "1:26:22", "remaining_time": "0:51:20", "throughput": 5174.04, "total_tokens": 26812608} {"current_steps": 26610, "total_steps": 42420, "loss": 0.305, "lr": 0.00036675948803695495, "epoch": 12.545968882602546, "percentage": 62.73, "elapsed_time": "1:26:22", "remaining_time": "0:51:19", "throughput": 5174.01, "total_tokens": 26816608} {"current_steps": 26615, "total_steps": 42420, "loss": 0.3301, "lr": 0.0003665612180337563, "epoch": 12.548326261197548, "percentage": 62.74, "elapsed_time": "1:26:23", "remaining_time": "0:51:18", "throughput": 5174.01, "total_tokens": 26821152} {"current_steps": 26620, "total_steps": 42420, "loss": 0.3788, "lr": 0.00036636297061947776, "epoch": 12.55068363979255, "percentage": 62.75, "elapsed_time": "1:26:24", "remaining_time": "0:51:17", "throughput": 5173.98, "total_tokens": 26825440} {"current_steps": 26625, "total_steps": 42420, "loss": 0.3686, "lr": 0.0003661647458276794, "epoch": 12.553041018387553, "percentage": 62.77, "elapsed_time": "1:26:25", "remaining_time": "0:51:16", "throughput": 5173.97, "total_tokens": 26829632} {"current_steps": 26630, "total_steps": 42420, "loss": 0.3436, "lr": 0.0003659665436919171, "epoch": 12.555398396982556, "percentage": 62.78, "elapsed_time": "1:26:26", "remaining_time": "0:51:15", "throughput": 5174.04, "total_tokens": 26835200} {"current_steps": 26635, "total_steps": 42420, "loss": 0.3242, "lr": 0.0003657683642457431, "epoch": 12.557755775577558, "percentage": 62.79, "elapsed_time": "1:26:27", "remaining_time": "0:51:14", "throughput": 5174.12, "total_tokens": 26840704} {"current_steps": 26640, "total_steps": 42420, "loss": 0.3452, "lr": 0.0003655702075227059, "epoch": 12.56011315417256, "percentage": 62.8, "elapsed_time": "1:26:28", "remaining_time": "0:51:13", "throughput": 5174.24, "total_tokens": 26847072} {"current_steps": 26645, "total_steps": 42420, "loss": 0.3087, "lr": 0.0003653720735563503, "epoch": 12.562470532767563, "percentage": 62.81, "elapsed_time": "1:26:29", "remaining_time": "0:51:12", "throughput": 5174.26, "total_tokens": 26851936} {"current_steps": 26650, "total_steps": 42420, "loss": 0.3272, "lr": 0.0003651739623802166, "epoch": 12.564827911362565, "percentage": 62.82, "elapsed_time": "1:26:30", "remaining_time": "0:51:11", "throughput": 5174.36, "total_tokens": 26857568} {"current_steps": 26655, "total_steps": 42420, "loss": 0.3349, "lr": 0.0003649758740278418, "epoch": 12.567185289957568, "percentage": 62.84, "elapsed_time": "1:26:31", "remaining_time": "0:51:10", "throughput": 5174.43, "total_tokens": 26862720} {"current_steps": 26660, "total_steps": 42420, "loss": 0.2562, "lr": 0.00036477780853275913, "epoch": 12.56954266855257, "percentage": 62.85, "elapsed_time": "1:26:32", "remaining_time": "0:51:09", "throughput": 5174.4, "total_tokens": 26866656} {"current_steps": 26665, "total_steps": 42420, "loss": 0.3616, "lr": 0.0003645797659284975, "epoch": 12.571900047147572, "percentage": 62.86, "elapsed_time": "1:26:33", "remaining_time": "0:51:08", "throughput": 5174.41, "total_tokens": 26871296} {"current_steps": 26670, "total_steps": 42420, "loss": 0.3088, "lr": 0.00036438174624858213, "epoch": 12.574257425742575, "percentage": 62.87, "elapsed_time": "1:26:34", "remaining_time": "0:51:07", "throughput": 5174.46, "total_tokens": 26876416} {"current_steps": 26675, "total_steps": 42420, "loss": 0.285, "lr": 0.00036418374952653445, "epoch": 12.576614804337577, "percentage": 62.88, "elapsed_time": "1:26:35", "remaining_time": "0:51:06", "throughput": 5174.56, "total_tokens": 26882112} {"current_steps": 26680, "total_steps": 42420, "loss": 0.3378, "lr": 0.00036398577579587187, "epoch": 12.57897218293258, "percentage": 62.89, "elapsed_time": "1:26:35", "remaining_time": "0:51:05", "throughput": 5174.6, "total_tokens": 26886816} {"current_steps": 26685, "total_steps": 42420, "loss": 0.3283, "lr": 0.00036378782509010834, "epoch": 12.581329561527582, "percentage": 62.91, "elapsed_time": "1:26:36", "remaining_time": "0:51:04", "throughput": 5174.63, "total_tokens": 26891520} {"current_steps": 26690, "total_steps": 42420, "loss": 0.3946, "lr": 0.000363589897442753, "epoch": 12.583686940122584, "percentage": 62.92, "elapsed_time": "1:26:37", "remaining_time": "0:51:03", "throughput": 5174.68, "total_tokens": 26896608} {"current_steps": 26695, "total_steps": 42420, "loss": 0.364, "lr": 0.00036339199288731194, "epoch": 12.586044318717587, "percentage": 62.93, "elapsed_time": "1:26:38", "remaining_time": "0:51:02", "throughput": 5174.74, "total_tokens": 26902016} {"current_steps": 26700, "total_steps": 42420, "loss": 0.2715, "lr": 0.0003631941114572871, "epoch": 12.58840169731259, "percentage": 62.94, "elapsed_time": "1:26:39", "remaining_time": "0:51:01", "throughput": 5174.79, "total_tokens": 26907072} {"current_steps": 26705, "total_steps": 42420, "loss": 0.3676, "lr": 0.00036299625318617643, "epoch": 12.590759075907592, "percentage": 62.95, "elapsed_time": "1:26:40", "remaining_time": "0:51:00", "throughput": 5174.84, "total_tokens": 26912096} {"current_steps": 26710, "total_steps": 42420, "loss": 0.3577, "lr": 0.0003627984181074737, "epoch": 12.593116454502592, "percentage": 62.97, "elapsed_time": "1:26:41", "remaining_time": "0:50:59", "throughput": 5174.95, "total_tokens": 26917888} {"current_steps": 26715, "total_steps": 42420, "loss": 0.3202, "lr": 0.0003626006062546694, "epoch": 12.595473833097596, "percentage": 62.98, "elapsed_time": "1:26:42", "remaining_time": "0:50:58", "throughput": 5174.96, "total_tokens": 26922368} {"current_steps": 26720, "total_steps": 42420, "loss": 0.3531, "lr": 0.0003624028176612494, "epoch": 12.597831211692597, "percentage": 62.99, "elapsed_time": "1:26:43", "remaining_time": "0:50:57", "throughput": 5175.0, "total_tokens": 26927136} {"current_steps": 26725, "total_steps": 42420, "loss": 0.2987, "lr": 0.00036220505236069634, "epoch": 12.6001885902876, "percentage": 63.0, "elapsed_time": "1:26:44", "remaining_time": "0:50:56", "throughput": 5175.06, "total_tokens": 26932384} {"current_steps": 26730, "total_steps": 42420, "loss": 0.35, "lr": 0.00036200731038648807, "epoch": 12.602545968882602, "percentage": 63.01, "elapsed_time": "1:26:45", "remaining_time": "0:50:55", "throughput": 5175.15, "total_tokens": 26938208} {"current_steps": 26735, "total_steps": 42420, "loss": 0.3115, "lr": 0.0003618095917720994, "epoch": 12.604903347477604, "percentage": 63.02, "elapsed_time": "1:26:46", "remaining_time": "0:50:54", "throughput": 5175.14, "total_tokens": 26942432} {"current_steps": 26740, "total_steps": 42420, "loss": 0.3215, "lr": 0.00036161189655100034, "epoch": 12.607260726072607, "percentage": 63.04, "elapsed_time": "1:26:47", "remaining_time": "0:50:53", "throughput": 5175.17, "total_tokens": 26947200} {"current_steps": 26745, "total_steps": 42420, "loss": 0.3753, "lr": 0.0003614142247566575, "epoch": 12.609618104667609, "percentage": 63.05, "elapsed_time": "1:26:47", "remaining_time": "0:50:52", "throughput": 5175.22, "total_tokens": 26952128} {"current_steps": 26750, "total_steps": 42420, "loss": 0.3307, "lr": 0.0003612165764225334, "epoch": 12.611975483262611, "percentage": 63.06, "elapsed_time": "1:26:48", "remaining_time": "0:50:51", "throughput": 5175.28, "total_tokens": 26957344} {"current_steps": 26755, "total_steps": 42420, "loss": 0.3506, "lr": 0.00036101895158208646, "epoch": 12.614332861857614, "percentage": 63.07, "elapsed_time": "1:26:49", "remaining_time": "0:50:50", "throughput": 5175.31, "total_tokens": 26962240} {"current_steps": 26760, "total_steps": 42420, "loss": 0.3271, "lr": 0.0003608213502687713, "epoch": 12.616690240452616, "percentage": 63.08, "elapsed_time": "1:26:50", "remaining_time": "0:50:49", "throughput": 5175.42, "total_tokens": 26968480} {"current_steps": 26765, "total_steps": 42420, "loss": 0.3256, "lr": 0.0003606237725160383, "epoch": 12.619047619047619, "percentage": 63.1, "elapsed_time": "1:26:51", "remaining_time": "0:50:48", "throughput": 5175.46, "total_tokens": 26973248} {"current_steps": 26770, "total_steps": 42420, "loss": 0.3104, "lr": 0.0003604262183573342, "epoch": 12.621404997642621, "percentage": 63.11, "elapsed_time": "1:26:52", "remaining_time": "0:50:47", "throughput": 5175.47, "total_tokens": 26977664} {"current_steps": 26775, "total_steps": 42420, "loss": 0.2523, "lr": 0.00036022868782610156, "epoch": 12.623762376237623, "percentage": 63.12, "elapsed_time": "1:26:53", "remaining_time": "0:50:46", "throughput": 5175.5, "total_tokens": 26982400} {"current_steps": 26780, "total_steps": 42420, "loss": 0.3507, "lr": 0.00036003118095577887, "epoch": 12.626119754832626, "percentage": 63.13, "elapsed_time": "1:26:54", "remaining_time": "0:50:45", "throughput": 5175.51, "total_tokens": 26987136} {"current_steps": 26785, "total_steps": 42420, "loss": 0.2492, "lr": 0.0003598336977798006, "epoch": 12.628477133427628, "percentage": 63.14, "elapsed_time": "1:26:55", "remaining_time": "0:50:44", "throughput": 5175.56, "total_tokens": 26992064} {"current_steps": 26790, "total_steps": 42420, "loss": 0.4213, "lr": 0.00035963623833159755, "epoch": 12.63083451202263, "percentage": 63.15, "elapsed_time": "1:26:56", "remaining_time": "0:50:43", "throughput": 5175.66, "total_tokens": 26998016} {"current_steps": 26795, "total_steps": 42420, "loss": 0.3595, "lr": 0.00035943880264459603, "epoch": 12.633191890617633, "percentage": 63.17, "elapsed_time": "1:26:57", "remaining_time": "0:50:42", "throughput": 5175.75, "total_tokens": 27003392} {"current_steps": 26800, "total_steps": 42420, "loss": 0.2764, "lr": 0.0003592413907522187, "epoch": 12.635549269212635, "percentage": 63.18, "elapsed_time": "1:26:58", "remaining_time": "0:50:41", "throughput": 5175.85, "total_tokens": 27009376} {"current_steps": 26805, "total_steps": 42420, "loss": 0.2511, "lr": 0.0003590440026878839, "epoch": 12.637906647807638, "percentage": 63.19, "elapsed_time": "1:26:59", "remaining_time": "0:50:40", "throughput": 5175.88, "total_tokens": 27014272} {"current_steps": 26810, "total_steps": 42420, "loss": 0.2731, "lr": 0.00035884663848500614, "epoch": 12.64026402640264, "percentage": 63.2, "elapsed_time": "1:27:00", "remaining_time": "0:50:39", "throughput": 5175.9, "total_tokens": 27018944} {"current_steps": 26815, "total_steps": 42420, "loss": 0.3226, "lr": 0.000358649298176996, "epoch": 12.642621404997643, "percentage": 63.21, "elapsed_time": "1:27:01", "remaining_time": "0:50:38", "throughput": 5175.96, "total_tokens": 27023936} {"current_steps": 26820, "total_steps": 42420, "loss": 0.3723, "lr": 0.00035845198179725946, "epoch": 12.644978783592645, "percentage": 63.22, "elapsed_time": "1:27:02", "remaining_time": "0:50:37", "throughput": 5176.1, "total_tokens": 27030368} {"current_steps": 26825, "total_steps": 42420, "loss": 0.3985, "lr": 0.00035825468937919914, "epoch": 12.647336162187647, "percentage": 63.24, "elapsed_time": "1:27:03", "remaining_time": "0:50:36", "throughput": 5176.14, "total_tokens": 27035136} {"current_steps": 26830, "total_steps": 42420, "loss": 0.3563, "lr": 0.0003580574209562132, "epoch": 12.64969354078265, "percentage": 63.25, "elapsed_time": "1:27:03", "remaining_time": "0:50:35", "throughput": 5176.23, "total_tokens": 27040480} {"current_steps": 26835, "total_steps": 42420, "loss": 0.2972, "lr": 0.0003578601765616959, "epoch": 12.652050919377652, "percentage": 63.26, "elapsed_time": "1:27:04", "remaining_time": "0:50:34", "throughput": 5176.23, "total_tokens": 27045184} {"current_steps": 26840, "total_steps": 42420, "loss": 0.33, "lr": 0.0003576629562290372, "epoch": 12.654408297972655, "percentage": 63.27, "elapsed_time": "1:27:05", "remaining_time": "0:50:33", "throughput": 5176.32, "total_tokens": 27050656} {"current_steps": 26845, "total_steps": 42420, "loss": 0.339, "lr": 0.00035746575999162335, "epoch": 12.656765676567657, "percentage": 63.28, "elapsed_time": "1:27:06", "remaining_time": "0:50:32", "throughput": 5176.42, "total_tokens": 27056352} {"current_steps": 26850, "total_steps": 42420, "loss": 0.3474, "lr": 0.0003572685878828362, "epoch": 12.65912305516266, "percentage": 63.3, "elapsed_time": "1:27:07", "remaining_time": "0:50:31", "throughput": 5176.43, "total_tokens": 27060640} {"current_steps": 26855, "total_steps": 42420, "loss": 0.3743, "lr": 0.00035707143993605377, "epoch": 12.661480433757662, "percentage": 63.31, "elapsed_time": "1:27:08", "remaining_time": "0:50:30", "throughput": 5176.42, "total_tokens": 27064672} {"current_steps": 26860, "total_steps": 42420, "loss": 0.341, "lr": 0.0003568743161846497, "epoch": 12.663837812352664, "percentage": 63.32, "elapsed_time": "1:27:09", "remaining_time": "0:50:29", "throughput": 5176.47, "total_tokens": 27069952} {"current_steps": 26865, "total_steps": 42420, "loss": 0.3372, "lr": 0.0003566772166619937, "epoch": 12.666195190947667, "percentage": 63.33, "elapsed_time": "1:27:10", "remaining_time": "0:50:28", "throughput": 5176.57, "total_tokens": 27075552} {"current_steps": 26870, "total_steps": 42420, "loss": 0.3308, "lr": 0.00035648014140145153, "epoch": 12.668552569542669, "percentage": 63.34, "elapsed_time": "1:27:11", "remaining_time": "0:50:27", "throughput": 5176.67, "total_tokens": 27080896} {"current_steps": 26875, "total_steps": 42420, "loss": 0.3358, "lr": 0.0003562830904363846, "epoch": 12.670909948137671, "percentage": 63.35, "elapsed_time": "1:27:12", "remaining_time": "0:50:26", "throughput": 5176.65, "total_tokens": 27084960} {"current_steps": 26880, "total_steps": 42420, "loss": 0.25, "lr": 0.0003560860638001503, "epoch": 12.673267326732674, "percentage": 63.37, "elapsed_time": "1:27:13", "remaining_time": "0:50:25", "throughput": 5176.71, "total_tokens": 27090144} {"current_steps": 26885, "total_steps": 42420, "loss": 0.3633, "lr": 0.0003558890615261018, "epoch": 12.675624705327676, "percentage": 63.38, "elapsed_time": "1:27:13", "remaining_time": "0:50:24", "throughput": 5176.69, "total_tokens": 27094144} {"current_steps": 26890, "total_steps": 42420, "loss": 0.3512, "lr": 0.0003556920836475883, "epoch": 12.677982083922679, "percentage": 63.39, "elapsed_time": "1:27:14", "remaining_time": "0:50:23", "throughput": 5176.7, "total_tokens": 27098720} {"current_steps": 26895, "total_steps": 42420, "loss": 0.3487, "lr": 0.000355495130197955, "epoch": 12.680339462517681, "percentage": 63.4, "elapsed_time": "1:27:15", "remaining_time": "0:50:22", "throughput": 5176.74, "total_tokens": 27103840} {"current_steps": 26900, "total_steps": 42420, "loss": 0.3222, "lr": 0.0003552982012105424, "epoch": 12.682696841112683, "percentage": 63.41, "elapsed_time": "1:27:16", "remaining_time": "0:50:21", "throughput": 5176.8, "total_tokens": 27108736} {"current_steps": 26905, "total_steps": 42420, "loss": 0.3532, "lr": 0.0003551012967186875, "epoch": 12.685054219707686, "percentage": 63.43, "elapsed_time": "1:27:17", "remaining_time": "0:50:20", "throughput": 5176.88, "total_tokens": 27114528} {"current_steps": 26910, "total_steps": 42420, "loss": 0.333, "lr": 0.00035490441675572283, "epoch": 12.687411598302688, "percentage": 63.44, "elapsed_time": "1:27:18", "remaining_time": "0:50:19", "throughput": 5176.91, "total_tokens": 27119808} {"current_steps": 26915, "total_steps": 42420, "loss": 0.3347, "lr": 0.00035470756135497694, "epoch": 12.689768976897689, "percentage": 63.45, "elapsed_time": "1:27:19", "remaining_time": "0:50:18", "throughput": 5177.0, "total_tokens": 27125440} {"current_steps": 26920, "total_steps": 42420, "loss": 0.3091, "lr": 0.0003545107305497738, "epoch": 12.692126355492691, "percentage": 63.46, "elapsed_time": "1:27:20", "remaining_time": "0:50:17", "throughput": 5177.1, "total_tokens": 27131488} {"current_steps": 26925, "total_steps": 42420, "loss": 0.3198, "lr": 0.0003543139243734338, "epoch": 12.694483734087694, "percentage": 63.47, "elapsed_time": "1:27:21", "remaining_time": "0:50:16", "throughput": 5177.06, "total_tokens": 27135456} {"current_steps": 26930, "total_steps": 42420, "loss": 0.3761, "lr": 0.0003541171428592728, "epoch": 12.696841112682696, "percentage": 63.48, "elapsed_time": "1:27:22", "remaining_time": "0:50:15", "throughput": 5177.18, "total_tokens": 27141216} {"current_steps": 26935, "total_steps": 42420, "loss": 0.3434, "lr": 0.0003539203860406026, "epoch": 12.699198491277699, "percentage": 63.5, "elapsed_time": "1:27:23", "remaining_time": "0:50:14", "throughput": 5177.24, "total_tokens": 27146880} {"current_steps": 26940, "total_steps": 42420, "loss": 0.3697, "lr": 0.0003537236539507307, "epoch": 12.701555869872701, "percentage": 63.51, "elapsed_time": "1:27:24", "remaining_time": "0:50:13", "throughput": 5177.22, "total_tokens": 27151200} {"current_steps": 26945, "total_steps": 42420, "loss": 0.3557, "lr": 0.0003535269466229605, "epoch": 12.703913248467703, "percentage": 63.52, "elapsed_time": "1:27:25", "remaining_time": "0:50:12", "throughput": 5177.28, "total_tokens": 27156480} {"current_steps": 26950, "total_steps": 42420, "loss": 0.3329, "lr": 0.0003533302640905913, "epoch": 12.706270627062706, "percentage": 63.53, "elapsed_time": "1:27:26", "remaining_time": "0:50:11", "throughput": 5177.3, "total_tokens": 27161280} {"current_steps": 26955, "total_steps": 42420, "loss": 0.3495, "lr": 0.0003531336063869182, "epoch": 12.708628005657708, "percentage": 63.54, "elapsed_time": "1:27:27", "remaining_time": "0:50:10", "throughput": 5177.4, "total_tokens": 27167008} {"current_steps": 26960, "total_steps": 42420, "loss": 0.38, "lr": 0.00035293697354523194, "epoch": 12.71098538425271, "percentage": 63.55, "elapsed_time": "1:27:28", "remaining_time": "0:50:09", "throughput": 5177.43, "total_tokens": 27171744} {"current_steps": 26965, "total_steps": 42420, "loss": 0.3651, "lr": 0.00035274036559881895, "epoch": 12.713342762847713, "percentage": 63.57, "elapsed_time": "1:27:29", "remaining_time": "0:50:08", "throughput": 5177.44, "total_tokens": 27176384} {"current_steps": 26970, "total_steps": 42420, "loss": 0.324, "lr": 0.0003525437825809617, "epoch": 12.715700141442715, "percentage": 63.58, "elapsed_time": "1:27:29", "remaining_time": "0:50:07", "throughput": 5177.51, "total_tokens": 27181728} {"current_steps": 26975, "total_steps": 42420, "loss": 0.2917, "lr": 0.0003523472245249384, "epoch": 12.718057520037718, "percentage": 63.59, "elapsed_time": "1:27:30", "remaining_time": "0:50:06", "throughput": 5177.57, "total_tokens": 27186912} {"current_steps": 26980, "total_steps": 42420, "loss": 0.2683, "lr": 0.000352150691464023, "epoch": 12.72041489863272, "percentage": 63.6, "elapsed_time": "1:27:32", "remaining_time": "0:50:05", "throughput": 5177.75, "total_tokens": 27194080} {"current_steps": 26985, "total_steps": 42420, "loss": 0.4675, "lr": 0.0003519541834314852, "epoch": 12.722772277227723, "percentage": 63.61, "elapsed_time": "1:27:33", "remaining_time": "0:50:04", "throughput": 5177.85, "total_tokens": 27200352} {"current_steps": 26990, "total_steps": 42420, "loss": 0.4386, "lr": 0.0003517577004605905, "epoch": 12.725129655822725, "percentage": 63.63, "elapsed_time": "1:27:34", "remaining_time": "0:50:03", "throughput": 5177.87, "total_tokens": 27205056} {"current_steps": 26995, "total_steps": 42420, "loss": 0.2738, "lr": 0.0003515612425846, "epoch": 12.727487034417727, "percentage": 63.64, "elapsed_time": "1:27:34", "remaining_time": "0:50:02", "throughput": 5177.86, "total_tokens": 27209344} {"current_steps": 27000, "total_steps": 42420, "loss": 0.337, "lr": 0.0003513648098367709, "epoch": 12.72984441301273, "percentage": 63.65, "elapsed_time": "1:27:36", "remaining_time": "0:50:01", "throughput": 5177.99, "total_tokens": 27215680} {"current_steps": 27005, "total_steps": 42420, "loss": 0.3673, "lr": 0.0003511684022503558, "epoch": 12.732201791607732, "percentage": 63.66, "elapsed_time": "1:27:36", "remaining_time": "0:50:00", "throughput": 5178.0, "total_tokens": 27220256} {"current_steps": 27010, "total_steps": 42420, "loss": 0.3385, "lr": 0.0003509720198586031, "epoch": 12.734559170202735, "percentage": 63.67, "elapsed_time": "1:27:37", "remaining_time": "0:49:59", "throughput": 5178.1, "total_tokens": 27226304} {"current_steps": 27015, "total_steps": 42420, "loss": 0.3153, "lr": 0.00035077566269475714, "epoch": 12.736916548797737, "percentage": 63.68, "elapsed_time": "1:27:38", "remaining_time": "0:49:58", "throughput": 5178.15, "total_tokens": 27231104} {"current_steps": 27020, "total_steps": 42420, "loss": 0.3499, "lr": 0.0003505793307920577, "epoch": 12.73927392739274, "percentage": 63.7, "elapsed_time": "1:27:39", "remaining_time": "0:49:57", "throughput": 5178.22, "total_tokens": 27236256} {"current_steps": 27025, "total_steps": 42420, "loss": 0.3147, "lr": 0.00035038302418374074, "epoch": 12.741631305987742, "percentage": 63.71, "elapsed_time": "1:27:40", "remaining_time": "0:49:56", "throughput": 5178.39, "total_tokens": 27243328} {"current_steps": 27030, "total_steps": 42420, "loss": 0.3238, "lr": 0.0003501867429030373, "epoch": 12.743988684582744, "percentage": 63.72, "elapsed_time": "1:27:41", "remaining_time": "0:49:55", "throughput": 5178.47, "total_tokens": 27248768} {"current_steps": 27035, "total_steps": 42420, "loss": 0.319, "lr": 0.0003499904869831746, "epoch": 12.746346063177747, "percentage": 63.73, "elapsed_time": "1:27:43", "remaining_time": "0:49:55", "throughput": 5178.64, "total_tokens": 27255488} {"current_steps": 27040, "total_steps": 42420, "loss": 0.3132, "lr": 0.00034979425645737547, "epoch": 12.748703441772749, "percentage": 63.74, "elapsed_time": "1:27:43", "remaining_time": "0:49:54", "throughput": 5178.63, "total_tokens": 27259936} {"current_steps": 27045, "total_steps": 42420, "loss": 0.4028, "lr": 0.0003495980513588584, "epoch": 12.751060820367751, "percentage": 63.76, "elapsed_time": "1:27:44", "remaining_time": "0:49:52", "throughput": 5178.62, "total_tokens": 27264096} {"current_steps": 27050, "total_steps": 42420, "loss": 0.2898, "lr": 0.0003494018717208375, "epoch": 12.753418198962754, "percentage": 63.77, "elapsed_time": "1:27:45", "remaining_time": "0:49:51", "throughput": 5178.62, "total_tokens": 27268448} {"current_steps": 27055, "total_steps": 42420, "loss": 0.2686, "lr": 0.0003492057175765227, "epoch": 12.755775577557756, "percentage": 63.78, "elapsed_time": "1:27:46", "remaining_time": "0:49:50", "throughput": 5178.65, "total_tokens": 27272992} {"current_steps": 27060, "total_steps": 42420, "loss": 0.2471, "lr": 0.0003490095889591196, "epoch": 12.758132956152759, "percentage": 63.79, "elapsed_time": "1:27:47", "remaining_time": "0:49:49", "throughput": 5178.7, "total_tokens": 27278112} {"current_steps": 27065, "total_steps": 42420, "loss": 0.3397, "lr": 0.00034881348590182947, "epoch": 12.760490334747761, "percentage": 63.8, "elapsed_time": "1:27:48", "remaining_time": "0:49:48", "throughput": 5178.77, "total_tokens": 27283456} {"current_steps": 27070, "total_steps": 42420, "loss": 0.2596, "lr": 0.00034861740843784915, "epoch": 12.762847713342763, "percentage": 63.81, "elapsed_time": "1:27:49", "remaining_time": "0:49:47", "throughput": 5178.79, "total_tokens": 27288160} {"current_steps": 27075, "total_steps": 42420, "loss": 0.2277, "lr": 0.0003484213566003712, "epoch": 12.765205091937766, "percentage": 63.83, "elapsed_time": "1:27:50", "remaining_time": "0:49:46", "throughput": 5178.77, "total_tokens": 27292480} {"current_steps": 27080, "total_steps": 42420, "loss": 0.3761, "lr": 0.00034822533042258385, "epoch": 12.767562470532768, "percentage": 63.84, "elapsed_time": "1:27:50", "remaining_time": "0:49:45", "throughput": 5178.83, "total_tokens": 27297600} {"current_steps": 27085, "total_steps": 42420, "loss": 0.3792, "lr": 0.00034802932993767134, "epoch": 12.76991984912777, "percentage": 63.85, "elapsed_time": "1:27:51", "remaining_time": "0:49:44", "throughput": 5178.87, "total_tokens": 27302624} {"current_steps": 27090, "total_steps": 42420, "loss": 0.3448, "lr": 0.00034783335517881277, "epoch": 12.772277227722773, "percentage": 63.86, "elapsed_time": "1:27:52", "remaining_time": "0:49:43", "throughput": 5178.86, "total_tokens": 27307072} {"current_steps": 27095, "total_steps": 42420, "loss": 0.3181, "lr": 0.0003476374061791835, "epoch": 12.774634606317775, "percentage": 63.87, "elapsed_time": "1:27:53", "remaining_time": "0:49:42", "throughput": 5178.9, "total_tokens": 27312064} {"current_steps": 27100, "total_steps": 42420, "loss": 0.3574, "lr": 0.0003474414829719544, "epoch": 12.776991984912778, "percentage": 63.88, "elapsed_time": "1:27:54", "remaining_time": "0:49:41", "throughput": 5178.98, "total_tokens": 27317408} {"current_steps": 27105, "total_steps": 42420, "loss": 0.3818, "lr": 0.00034724558559029206, "epoch": 12.77934936350778, "percentage": 63.9, "elapsed_time": "1:27:55", "remaining_time": "0:49:40", "throughput": 5179.07, "total_tokens": 27323232} {"current_steps": 27110, "total_steps": 42420, "loss": 0.3645, "lr": 0.0003470497140673583, "epoch": 12.78170674210278, "percentage": 63.91, "elapsed_time": "1:27:56", "remaining_time": "0:49:39", "throughput": 5179.06, "total_tokens": 27327744} {"current_steps": 27115, "total_steps": 42420, "loss": 0.3488, "lr": 0.0003468538684363111, "epoch": 12.784064120697785, "percentage": 63.92, "elapsed_time": "1:27:57", "remaining_time": "0:49:38", "throughput": 5179.07, "total_tokens": 27332288} {"current_steps": 27120, "total_steps": 42420, "loss": 0.3535, "lr": 0.0003466580487303036, "epoch": 12.786421499292786, "percentage": 63.93, "elapsed_time": "1:27:58", "remaining_time": "0:49:37", "throughput": 5179.13, "total_tokens": 27337504} {"current_steps": 27125, "total_steps": 42420, "loss": 0.3443, "lr": 0.000346462254982485, "epoch": 12.788778877887788, "percentage": 63.94, "elapsed_time": "1:27:59", "remaining_time": "0:49:36", "throughput": 5179.12, "total_tokens": 27341728} {"current_steps": 27130, "total_steps": 42420, "loss": 0.3471, "lr": 0.00034626648722599964, "epoch": 12.79113625648279, "percentage": 63.96, "elapsed_time": "1:28:00", "remaining_time": "0:49:35", "throughput": 5179.2, "total_tokens": 27347168} {"current_steps": 27135, "total_steps": 42420, "loss": 0.3515, "lr": 0.00034607074549398764, "epoch": 12.793493635077793, "percentage": 63.97, "elapsed_time": "1:28:01", "remaining_time": "0:49:35", "throughput": 5179.36, "total_tokens": 27354752} {"current_steps": 27140, "total_steps": 42420, "loss": 0.2524, "lr": 0.00034587502981958494, "epoch": 12.795851013672795, "percentage": 63.98, "elapsed_time": "1:28:02", "remaining_time": "0:49:34", "throughput": 5179.38, "total_tokens": 27359616} {"current_steps": 27145, "total_steps": 42420, "loss": 0.3338, "lr": 0.000345679340235923, "epoch": 12.798208392267798, "percentage": 63.99, "elapsed_time": "1:28:03", "remaining_time": "0:49:33", "throughput": 5179.41, "total_tokens": 27364480} {"current_steps": 27150, "total_steps": 42420, "loss": 0.3703, "lr": 0.00034548367677612825, "epoch": 12.8005657708628, "percentage": 64.0, "elapsed_time": "1:28:04", "remaining_time": "0:49:32", "throughput": 5179.49, "total_tokens": 27369728} {"current_steps": 27155, "total_steps": 42420, "loss": 0.3299, "lr": 0.00034528803947332355, "epoch": 12.802923149457802, "percentage": 64.01, "elapsed_time": "1:28:05", "remaining_time": "0:49:31", "throughput": 5179.63, "total_tokens": 27376160} {"current_steps": 27160, "total_steps": 42420, "loss": 0.3546, "lr": 0.00034509242836062694, "epoch": 12.805280528052805, "percentage": 64.03, "elapsed_time": "1:28:06", "remaining_time": "0:49:30", "throughput": 5179.66, "total_tokens": 27380960} {"current_steps": 27165, "total_steps": 42420, "loss": 0.3416, "lr": 0.0003448968434711521, "epoch": 12.807637906647807, "percentage": 64.04, "elapsed_time": "1:28:07", "remaining_time": "0:49:29", "throughput": 5179.63, "total_tokens": 27384992} {"current_steps": 27170, "total_steps": 42420, "loss": 0.3241, "lr": 0.0003447012848380081, "epoch": 12.80999528524281, "percentage": 64.05, "elapsed_time": "1:28:08", "remaining_time": "0:49:28", "throughput": 5179.75, "total_tokens": 27390880} {"current_steps": 27175, "total_steps": 42420, "loss": 0.3469, "lr": 0.0003445057524942997, "epoch": 12.812352663837812, "percentage": 64.06, "elapsed_time": "1:28:08", "remaining_time": "0:49:27", "throughput": 5179.81, "total_tokens": 27395968} {"current_steps": 27180, "total_steps": 42420, "loss": 0.3117, "lr": 0.00034431024647312724, "epoch": 12.814710042432814, "percentage": 64.07, "elapsed_time": "1:28:09", "remaining_time": "0:49:26", "throughput": 5179.85, "total_tokens": 27400928} {"current_steps": 27185, "total_steps": 42420, "loss": 0.3538, "lr": 0.0003441147668075869, "epoch": 12.817067421027817, "percentage": 64.09, "elapsed_time": "1:28:10", "remaining_time": "0:49:25", "throughput": 5179.88, "total_tokens": 27405760} {"current_steps": 27190, "total_steps": 42420, "loss": 0.3054, "lr": 0.0003439193135307695, "epoch": 12.81942479962282, "percentage": 64.1, "elapsed_time": "1:28:11", "remaining_time": "0:49:24", "throughput": 5179.85, "total_tokens": 27409760} {"current_steps": 27195, "total_steps": 42420, "loss": 0.3236, "lr": 0.00034372388667576226, "epoch": 12.821782178217822, "percentage": 64.11, "elapsed_time": "1:28:12", "remaining_time": "0:49:22", "throughput": 5179.87, "total_tokens": 27414368} {"current_steps": 27200, "total_steps": 42420, "loss": 0.378, "lr": 0.0003435284862756476, "epoch": 12.824139556812824, "percentage": 64.12, "elapsed_time": "1:28:13", "remaining_time": "0:49:21", "throughput": 5179.9, "total_tokens": 27418976} {"current_steps": 27205, "total_steps": 42420, "loss": 0.3676, "lr": 0.0003433331123635034, "epoch": 12.826496935407826, "percentage": 64.13, "elapsed_time": "1:28:14", "remaining_time": "0:49:20", "throughput": 5179.97, "total_tokens": 27424160} {"current_steps": 27210, "total_steps": 42420, "loss": 0.3265, "lr": 0.0003431377649724032, "epoch": 12.828854314002829, "percentage": 64.14, "elapsed_time": "1:28:15", "remaining_time": "0:49:19", "throughput": 5180.08, "total_tokens": 27429856} {"current_steps": 27215, "total_steps": 42420, "loss": 0.3379, "lr": 0.0003429424441354162, "epoch": 12.831211692597831, "percentage": 64.16, "elapsed_time": "1:28:16", "remaining_time": "0:49:18", "throughput": 5180.1, "total_tokens": 27434336} {"current_steps": 27220, "total_steps": 42420, "loss": 0.3288, "lr": 0.0003427471498856066, "epoch": 12.833569071192834, "percentage": 64.17, "elapsed_time": "1:28:16", "remaining_time": "0:49:17", "throughput": 5180.08, "total_tokens": 27438592} {"current_steps": 27225, "total_steps": 42420, "loss": 0.336, "lr": 0.0003425518822560345, "epoch": 12.835926449787836, "percentage": 64.18, "elapsed_time": "1:28:17", "remaining_time": "0:49:16", "throughput": 5180.09, "total_tokens": 27442944} {"current_steps": 27230, "total_steps": 42420, "loss": 0.3467, "lr": 0.0003423566412797554, "epoch": 12.838283828382838, "percentage": 64.19, "elapsed_time": "1:28:18", "remaining_time": "0:49:15", "throughput": 5180.11, "total_tokens": 27447648} {"current_steps": 27235, "total_steps": 42420, "loss": 0.4024, "lr": 0.00034216142698982046, "epoch": 12.84064120697784, "percentage": 64.2, "elapsed_time": "1:28:19", "remaining_time": "0:49:14", "throughput": 5180.14, "total_tokens": 27452672} {"current_steps": 27240, "total_steps": 42420, "loss": 0.322, "lr": 0.0003419662394192758, "epoch": 12.842998585572843, "percentage": 64.21, "elapsed_time": "1:28:20", "remaining_time": "0:49:13", "throughput": 5180.15, "total_tokens": 27457376} {"current_steps": 27245, "total_steps": 42420, "loss": 0.3089, "lr": 0.00034177107860116365, "epoch": 12.845355964167846, "percentage": 64.23, "elapsed_time": "1:28:21", "remaining_time": "0:49:12", "throughput": 5180.12, "total_tokens": 27461408} {"current_steps": 27250, "total_steps": 42420, "loss": 0.3276, "lr": 0.00034157594456852127, "epoch": 12.847713342762848, "percentage": 64.24, "elapsed_time": "1:28:22", "remaining_time": "0:49:11", "throughput": 5180.08, "total_tokens": 27465536} {"current_steps": 27255, "total_steps": 42420, "loss": 0.3477, "lr": 0.0003413808373543817, "epoch": 12.85007072135785, "percentage": 64.25, "elapsed_time": "1:28:23", "remaining_time": "0:49:10", "throughput": 5180.11, "total_tokens": 27470528} {"current_steps": 27260, "total_steps": 42420, "loss": 0.3379, "lr": 0.000341185756991773, "epoch": 12.852428099952853, "percentage": 64.26, "elapsed_time": "1:28:23", "remaining_time": "0:49:09", "throughput": 5180.14, "total_tokens": 27475264} {"current_steps": 27265, "total_steps": 42420, "loss": 0.2209, "lr": 0.0003409907035137192, "epoch": 12.854785478547855, "percentage": 64.27, "elapsed_time": "1:28:24", "remaining_time": "0:49:08", "throughput": 5180.09, "total_tokens": 27478784} {"current_steps": 27270, "total_steps": 42420, "loss": 0.3721, "lr": 0.00034079567695323947, "epoch": 12.857142857142858, "percentage": 64.29, "elapsed_time": "1:28:25", "remaining_time": "0:49:07", "throughput": 5180.15, "total_tokens": 27484000} {"current_steps": 27275, "total_steps": 42420, "loss": 0.4414, "lr": 0.0003406006773433486, "epoch": 12.85950023573786, "percentage": 64.3, "elapsed_time": "1:28:26", "remaining_time": "0:49:06", "throughput": 5180.27, "total_tokens": 27490016} {"current_steps": 27280, "total_steps": 42420, "loss": 0.3081, "lr": 0.00034040570471705647, "epoch": 12.861857614332862, "percentage": 64.31, "elapsed_time": "1:28:27", "remaining_time": "0:49:05", "throughput": 5180.29, "total_tokens": 27494720} {"current_steps": 27285, "total_steps": 42420, "loss": 0.3206, "lr": 0.0003402107591073688, "epoch": 12.864214992927865, "percentage": 64.32, "elapsed_time": "1:28:28", "remaining_time": "0:49:04", "throughput": 5180.31, "total_tokens": 27499136} {"current_steps": 27290, "total_steps": 42420, "loss": 0.299, "lr": 0.0003400158405472865, "epoch": 12.866572371522867, "percentage": 64.33, "elapsed_time": "1:28:29", "remaining_time": "0:49:03", "throughput": 5180.38, "total_tokens": 27504608} {"current_steps": 27295, "total_steps": 42420, "loss": 0.3125, "lr": 0.0003398209490698062, "epoch": 12.86892975011787, "percentage": 64.34, "elapsed_time": "1:28:30", "remaining_time": "0:49:02", "throughput": 5180.41, "total_tokens": 27509504} {"current_steps": 27300, "total_steps": 42420, "loss": 0.3619, "lr": 0.0003396260847079194, "epoch": 12.871287128712872, "percentage": 64.36, "elapsed_time": "1:28:31", "remaining_time": "0:49:01", "throughput": 5180.52, "total_tokens": 27515520} {"current_steps": 27305, "total_steps": 42420, "loss": 0.3543, "lr": 0.0003394312474946134, "epoch": 12.873644507307874, "percentage": 64.37, "elapsed_time": "1:28:32", "remaining_time": "0:49:00", "throughput": 5180.61, "total_tokens": 27521120} {"current_steps": 27310, "total_steps": 42420, "loss": 0.3222, "lr": 0.000339236437462871, "epoch": 12.876001885902877, "percentage": 64.38, "elapsed_time": "1:28:33", "remaining_time": "0:48:59", "throughput": 5180.64, "total_tokens": 27525920} {"current_steps": 27315, "total_steps": 42420, "loss": 0.3484, "lr": 0.0003390416546456702, "epoch": 12.878359264497877, "percentage": 64.39, "elapsed_time": "1:28:34", "remaining_time": "0:48:58", "throughput": 5180.67, "total_tokens": 27530656} {"current_steps": 27320, "total_steps": 42420, "loss": 0.3469, "lr": 0.0003388468990759842, "epoch": 12.88071664309288, "percentage": 64.4, "elapsed_time": "1:28:34", "remaining_time": "0:48:57", "throughput": 5180.64, "total_tokens": 27534816} {"current_steps": 27325, "total_steps": 42420, "loss": 0.2937, "lr": 0.000338652170786782, "epoch": 12.883074021687882, "percentage": 64.42, "elapsed_time": "1:28:35", "remaining_time": "0:48:56", "throughput": 5180.74, "total_tokens": 27540384} {"current_steps": 27330, "total_steps": 42420, "loss": 0.3264, "lr": 0.00033845746981102777, "epoch": 12.885431400282885, "percentage": 64.43, "elapsed_time": "1:28:36", "remaining_time": "0:48:55", "throughput": 5180.79, "total_tokens": 27545888} {"current_steps": 27335, "total_steps": 42420, "loss": 0.3646, "lr": 0.0003382627961816811, "epoch": 12.887788778877887, "percentage": 64.44, "elapsed_time": "1:28:38", "remaining_time": "0:48:54", "throughput": 5180.94, "total_tokens": 27552992} {"current_steps": 27340, "total_steps": 42420, "loss": 0.4218, "lr": 0.0003380681499316969, "epoch": 12.89014615747289, "percentage": 64.45, "elapsed_time": "1:28:39", "remaining_time": "0:48:53", "throughput": 5181.04, "total_tokens": 27558880} {"current_steps": 27345, "total_steps": 42420, "loss": 0.3738, "lr": 0.0003378735310940254, "epoch": 12.892503536067892, "percentage": 64.46, "elapsed_time": "1:28:39", "remaining_time": "0:48:52", "throughput": 5181.04, "total_tokens": 27563136} {"current_steps": 27350, "total_steps": 42420, "loss": 0.3626, "lr": 0.0003376789397016123, "epoch": 12.894860914662894, "percentage": 64.47, "elapsed_time": "1:28:40", "remaining_time": "0:48:51", "throughput": 5181.04, "total_tokens": 27567680} {"current_steps": 27355, "total_steps": 42420, "loss": 0.3532, "lr": 0.00033748437578739885, "epoch": 12.897218293257897, "percentage": 64.49, "elapsed_time": "1:28:41", "remaining_time": "0:48:50", "throughput": 5181.01, "total_tokens": 27571584} {"current_steps": 27360, "total_steps": 42420, "loss": 0.3414, "lr": 0.00033728983938432106, "epoch": 12.899575671852899, "percentage": 64.5, "elapsed_time": "1:28:42", "remaining_time": "0:48:49", "throughput": 5181.04, "total_tokens": 27576352} {"current_steps": 27365, "total_steps": 42420, "loss": 0.3345, "lr": 0.0003370953305253108, "epoch": 12.901933050447902, "percentage": 64.51, "elapsed_time": "1:28:43", "remaining_time": "0:48:48", "throughput": 5181.08, "total_tokens": 27581248} {"current_steps": 27370, "total_steps": 42420, "loss": 0.329, "lr": 0.0003369008492432951, "epoch": 12.904290429042904, "percentage": 64.52, "elapsed_time": "1:28:44", "remaining_time": "0:48:47", "throughput": 5181.1, "total_tokens": 27586016} {"current_steps": 27375, "total_steps": 42420, "loss": 0.3123, "lr": 0.00033670639557119655, "epoch": 12.906647807637906, "percentage": 64.53, "elapsed_time": "1:28:45", "remaining_time": "0:48:46", "throughput": 5181.1, "total_tokens": 27590336} {"current_steps": 27380, "total_steps": 42420, "loss": 0.3985, "lr": 0.0003365119695419325, "epoch": 12.909005186232909, "percentage": 64.55, "elapsed_time": "1:28:46", "remaining_time": "0:48:45", "throughput": 5181.18, "total_tokens": 27595680} {"current_steps": 27385, "total_steps": 42420, "loss": 0.3519, "lr": 0.0003363175711884161, "epoch": 12.911362564827911, "percentage": 64.56, "elapsed_time": "1:28:47", "remaining_time": "0:48:44", "throughput": 5181.25, "total_tokens": 27600960} {"current_steps": 27390, "total_steps": 42420, "loss": 0.3658, "lr": 0.0003361232005435558, "epoch": 12.913719943422914, "percentage": 64.57, "elapsed_time": "1:28:47", "remaining_time": "0:48:43", "throughput": 5181.25, "total_tokens": 27605440} {"current_steps": 27395, "total_steps": 42420, "loss": 0.3028, "lr": 0.00033592885764025524, "epoch": 12.916077322017916, "percentage": 64.58, "elapsed_time": "1:28:48", "remaining_time": "0:48:42", "throughput": 5181.32, "total_tokens": 27610816} {"current_steps": 27400, "total_steps": 42420, "loss": 0.3499, "lr": 0.0003357345425114132, "epoch": 12.918434700612918, "percentage": 64.59, "elapsed_time": "1:28:49", "remaining_time": "0:48:41", "throughput": 5181.43, "total_tokens": 27616672} {"current_steps": 27405, "total_steps": 42420, "loss": 0.3094, "lr": 0.00033554025518992405, "epoch": 12.92079207920792, "percentage": 64.6, "elapsed_time": "1:28:50", "remaining_time": "0:48:40", "throughput": 5181.46, "total_tokens": 27621472} {"current_steps": 27410, "total_steps": 42420, "loss": 0.3291, "lr": 0.00033534599570867753, "epoch": 12.923149457802923, "percentage": 64.62, "elapsed_time": "1:28:51", "remaining_time": "0:48:39", "throughput": 5181.5, "total_tokens": 27626208} {"current_steps": 27415, "total_steps": 42420, "loss": 0.36, "lr": 0.00033515176410055817, "epoch": 12.925506836397926, "percentage": 64.63, "elapsed_time": "1:28:52", "remaining_time": "0:48:38", "throughput": 5181.64, "total_tokens": 27632448} {"current_steps": 27420, "total_steps": 42420, "loss": 0.3688, "lr": 0.00033495756039844617, "epoch": 12.927864214992928, "percentage": 64.64, "elapsed_time": "1:28:53", "remaining_time": "0:48:37", "throughput": 5181.69, "total_tokens": 27637472} {"current_steps": 27425, "total_steps": 42420, "loss": 0.3081, "lr": 0.0003347633846352168, "epoch": 12.93022159358793, "percentage": 64.65, "elapsed_time": "1:28:54", "remaining_time": "0:48:36", "throughput": 5181.69, "total_tokens": 27641856} {"current_steps": 27430, "total_steps": 42420, "loss": 0.3197, "lr": 0.00033456923684374096, "epoch": 12.932578972182933, "percentage": 64.66, "elapsed_time": "1:28:55", "remaining_time": "0:48:35", "throughput": 5181.7, "total_tokens": 27646496} {"current_steps": 27435, "total_steps": 42420, "loss": 0.3364, "lr": 0.0003343751170568844, "epoch": 12.934936350777935, "percentage": 64.67, "elapsed_time": "1:28:56", "remaining_time": "0:48:34", "throughput": 5181.67, "total_tokens": 27650496} {"current_steps": 27440, "total_steps": 42420, "loss": 0.3197, "lr": 0.0003341810253075085, "epoch": 12.937293729372938, "percentage": 64.69, "elapsed_time": "1:28:57", "remaining_time": "0:48:33", "throughput": 5181.67, "total_tokens": 27654816} {"current_steps": 27445, "total_steps": 42420, "loss": 0.3501, "lr": 0.00033398696162846944, "epoch": 12.93965110796794, "percentage": 64.7, "elapsed_time": "1:28:57", "remaining_time": "0:48:32", "throughput": 5181.66, "total_tokens": 27659040} {"current_steps": 27450, "total_steps": 42420, "loss": 0.2791, "lr": 0.000333792926052619, "epoch": 12.942008486562942, "percentage": 64.71, "elapsed_time": "1:28:58", "remaining_time": "0:48:31", "throughput": 5181.76, "total_tokens": 27664960} {"current_steps": 27455, "total_steps": 42420, "loss": 0.3673, "lr": 0.0003335989186128042, "epoch": 12.944365865157945, "percentage": 64.72, "elapsed_time": "1:28:59", "remaining_time": "0:48:30", "throughput": 5181.75, "total_tokens": 27669216} {"current_steps": 27460, "total_steps": 42420, "loss": 0.3081, "lr": 0.00033340493934186735, "epoch": 12.946723243752947, "percentage": 64.73, "elapsed_time": "1:29:00", "remaining_time": "0:48:29", "throughput": 5181.84, "total_tokens": 27674720} {"current_steps": 27465, "total_steps": 42420, "loss": 0.2979, "lr": 0.0003332109882726454, "epoch": 12.94908062234795, "percentage": 64.75, "elapsed_time": "1:29:01", "remaining_time": "0:48:28", "throughput": 5181.91, "total_tokens": 27680384} {"current_steps": 27470, "total_steps": 42420, "loss": 0.3264, "lr": 0.0003330170654379713, "epoch": 12.951438000942952, "percentage": 64.76, "elapsed_time": "1:29:02", "remaining_time": "0:48:27", "throughput": 5181.97, "total_tokens": 27685568} {"current_steps": 27475, "total_steps": 42420, "loss": 0.3583, "lr": 0.0003328231708706727, "epoch": 12.953795379537954, "percentage": 64.77, "elapsed_time": "1:29:03", "remaining_time": "0:48:26", "throughput": 5182.07, "total_tokens": 27691456} {"current_steps": 27480, "total_steps": 42420, "loss": 0.2279, "lr": 0.00033262930460357297, "epoch": 12.956152758132957, "percentage": 64.78, "elapsed_time": "1:29:04", "remaining_time": "0:48:25", "throughput": 5182.11, "total_tokens": 27696576} {"current_steps": 27485, "total_steps": 42420, "loss": 0.3254, "lr": 0.0003324354666694901, "epoch": 12.95851013672796, "percentage": 64.79, "elapsed_time": "1:29:05", "remaining_time": "0:48:24", "throughput": 5182.13, "total_tokens": 27701216} {"current_steps": 27490, "total_steps": 42420, "loss": 0.3406, "lr": 0.00033224165710123756, "epoch": 12.960867515322962, "percentage": 64.8, "elapsed_time": "1:29:06", "remaining_time": "0:48:23", "throughput": 5182.17, "total_tokens": 27706240} {"current_steps": 27495, "total_steps": 42420, "loss": 0.3064, "lr": 0.00033204787593162414, "epoch": 12.963224893917964, "percentage": 64.82, "elapsed_time": "1:29:07", "remaining_time": "0:48:22", "throughput": 5182.25, "total_tokens": 27711872} {"current_steps": 27500, "total_steps": 42420, "loss": 0.4015, "lr": 0.0003318541231934538, "epoch": 12.965582272512966, "percentage": 64.83, "elapsed_time": "1:29:08", "remaining_time": "0:48:21", "throughput": 5182.25, "total_tokens": 27716512} {"current_steps": 27505, "total_steps": 42420, "loss": 0.3426, "lr": 0.00033166039891952524, "epoch": 12.967939651107969, "percentage": 64.84, "elapsed_time": "1:29:09", "remaining_time": "0:48:20", "throughput": 5182.39, "total_tokens": 27722752} {"current_steps": 27510, "total_steps": 42420, "loss": 0.3344, "lr": 0.0003314667031426329, "epoch": 12.97029702970297, "percentage": 64.85, "elapsed_time": "1:29:10", "remaining_time": "0:48:19", "throughput": 5182.52, "total_tokens": 27729216} {"current_steps": 27515, "total_steps": 42420, "loss": 0.3491, "lr": 0.00033127303589556634, "epoch": 12.972654408297974, "percentage": 64.86, "elapsed_time": "1:29:11", "remaining_time": "0:48:18", "throughput": 5182.6, "total_tokens": 27734592} {"current_steps": 27520, "total_steps": 42420, "loss": 0.3367, "lr": 0.00033107939721110994, "epoch": 12.975011786892974, "percentage": 64.88, "elapsed_time": "1:29:12", "remaining_time": "0:48:17", "throughput": 5182.62, "total_tokens": 27739680} {"current_steps": 27525, "total_steps": 42420, "loss": 0.2982, "lr": 0.00033088578712204343, "epoch": 12.977369165487977, "percentage": 64.89, "elapsed_time": "1:29:13", "remaining_time": "0:48:16", "throughput": 5182.66, "total_tokens": 27744672} {"current_steps": 27530, "total_steps": 42420, "loss": 0.3329, "lr": 0.00033069220566114176, "epoch": 12.979726544082979, "percentage": 64.9, "elapsed_time": "1:29:14", "remaining_time": "0:48:15", "throughput": 5182.73, "total_tokens": 27750144} {"current_steps": 27535, "total_steps": 42420, "loss": 0.2719, "lr": 0.00033049865286117504, "epoch": 12.982083922677981, "percentage": 64.91, "elapsed_time": "1:29:15", "remaining_time": "0:48:15", "throughput": 5182.81, "total_tokens": 27755712} {"current_steps": 27540, "total_steps": 42420, "loss": 0.4583, "lr": 0.00033030512875490846, "epoch": 12.984441301272984, "percentage": 64.92, "elapsed_time": "1:29:16", "remaining_time": "0:48:13", "throughput": 5182.84, "total_tokens": 27760256} {"current_steps": 27545, "total_steps": 42420, "loss": 0.3131, "lr": 0.00033011163337510227, "epoch": 12.986798679867986, "percentage": 64.93, "elapsed_time": "1:29:17", "remaining_time": "0:48:12", "throughput": 5182.86, "total_tokens": 27765056} {"current_steps": 27550, "total_steps": 42420, "loss": 0.3545, "lr": 0.00032991816675451195, "epoch": 12.989156058462989, "percentage": 64.95, "elapsed_time": "1:29:17", "remaining_time": "0:48:11", "throughput": 5182.85, "total_tokens": 27769376} {"current_steps": 27555, "total_steps": 42420, "loss": 0.3288, "lr": 0.0003297247289258882, "epoch": 12.991513437057991, "percentage": 64.96, "elapsed_time": "1:29:18", "remaining_time": "0:48:10", "throughput": 5182.89, "total_tokens": 27774368} {"current_steps": 27560, "total_steps": 42420, "loss": 0.3247, "lr": 0.00032953131992197686, "epoch": 12.993870815652993, "percentage": 64.97, "elapsed_time": "1:29:19", "remaining_time": "0:48:09", "throughput": 5182.96, "total_tokens": 27779680} {"current_steps": 27565, "total_steps": 42420, "loss": 0.3595, "lr": 0.0003293379397755185, "epoch": 12.996228194247996, "percentage": 64.98, "elapsed_time": "1:29:20", "remaining_time": "0:48:08", "throughput": 5182.99, "total_tokens": 27784448} {"current_steps": 27570, "total_steps": 42420, "loss": 0.3379, "lr": 0.00032914458851924936, "epoch": 12.998585572842998, "percentage": 64.99, "elapsed_time": "1:29:21", "remaining_time": "0:48:07", "throughput": 5183.02, "total_tokens": 27789344} {"current_steps": 27573, "total_steps": 42420, "eval_loss": 0.33856454491615295, "epoch": 13.0, "percentage": 65.0, "elapsed_time": "1:29:41", "remaining_time": "0:48:17", "throughput": 5164.72, "total_tokens": 27792544} {"current_steps": 27575, "total_steps": 42420, "loss": 0.3958, "lr": 0.0003289512661859003, "epoch": 13.000942951438, "percentage": 65.0, "elapsed_time": "1:29:43", "remaining_time": "0:48:18", "throughput": 5162.77, "total_tokens": 27794496} {"current_steps": 27580, "total_steps": 42420, "loss": 0.3404, "lr": 0.00032875797280819787, "epoch": 13.003300330033003, "percentage": 65.02, "elapsed_time": "1:29:44", "remaining_time": "0:48:17", "throughput": 5162.83, "total_tokens": 27799648} {"current_steps": 27585, "total_steps": 42420, "loss": 0.338, "lr": 0.00032856470841886295, "epoch": 13.005657708628005, "percentage": 65.03, "elapsed_time": "1:29:45", "remaining_time": "0:48:16", "throughput": 5162.81, "total_tokens": 27803840} {"current_steps": 27590, "total_steps": 42420, "loss": 0.3337, "lr": 0.00032837147305061213, "epoch": 13.008015087223008, "percentage": 65.04, "elapsed_time": "1:29:46", "remaining_time": "0:48:15", "throughput": 5162.83, "total_tokens": 27808320} {"current_steps": 27595, "total_steps": 42420, "loss": 0.3282, "lr": 0.0003281782667361567, "epoch": 13.01037246581801, "percentage": 65.05, "elapsed_time": "1:29:47", "remaining_time": "0:48:14", "throughput": 5162.92, "total_tokens": 27813824} {"current_steps": 27600, "total_steps": 42420, "loss": 0.3489, "lr": 0.0003279850895082036, "epoch": 13.012729844413013, "percentage": 65.06, "elapsed_time": "1:29:48", "remaining_time": "0:48:13", "throughput": 5162.94, "total_tokens": 27818656} {"current_steps": 27605, "total_steps": 42420, "loss": 0.3405, "lr": 0.000327791941399454, "epoch": 13.015087223008015, "percentage": 65.08, "elapsed_time": "1:29:48", "remaining_time": "0:48:12", "throughput": 5162.93, "total_tokens": 27822784} {"current_steps": 27610, "total_steps": 42420, "loss": 0.4033, "lr": 0.00032759882244260487, "epoch": 13.017444601603017, "percentage": 65.09, "elapsed_time": "1:29:49", "remaining_time": "0:48:11", "throughput": 5162.95, "total_tokens": 27827360} {"current_steps": 27615, "total_steps": 42420, "loss": 0.332, "lr": 0.0003274057326703478, "epoch": 13.01980198019802, "percentage": 65.1, "elapsed_time": "1:29:50", "remaining_time": "0:48:10", "throughput": 5163.04, "total_tokens": 27833184} {"current_steps": 27620, "total_steps": 42420, "loss": 0.3105, "lr": 0.0003272126721153699, "epoch": 13.022159358793022, "percentage": 65.11, "elapsed_time": "1:29:51", "remaining_time": "0:48:09", "throughput": 5163.15, "total_tokens": 27839168} {"current_steps": 27625, "total_steps": 42420, "loss": 0.3188, "lr": 0.00032701964081035276, "epoch": 13.024516737388025, "percentage": 65.12, "elapsed_time": "1:29:52", "remaining_time": "0:48:08", "throughput": 5163.18, "total_tokens": 27843840} {"current_steps": 27630, "total_steps": 42420, "loss": 0.3335, "lr": 0.00032682663878797326, "epoch": 13.026874115983027, "percentage": 65.13, "elapsed_time": "1:29:53", "remaining_time": "0:48:07", "throughput": 5163.17, "total_tokens": 27848096} {"current_steps": 27635, "total_steps": 42420, "loss": 0.415, "lr": 0.0003266336660809038, "epoch": 13.02923149457803, "percentage": 65.15, "elapsed_time": "1:29:54", "remaining_time": "0:48:06", "throughput": 5163.2, "total_tokens": 27852832} {"current_steps": 27640, "total_steps": 42420, "loss": 0.3518, "lr": 0.0003264407227218109, "epoch": 13.031588873173032, "percentage": 65.16, "elapsed_time": "1:29:55", "remaining_time": "0:48:05", "throughput": 5163.28, "total_tokens": 27858176} {"current_steps": 27645, "total_steps": 42420, "loss": 0.2916, "lr": 0.0003262478087433567, "epoch": 13.033946251768034, "percentage": 65.17, "elapsed_time": "1:29:56", "remaining_time": "0:48:04", "throughput": 5163.29, "total_tokens": 27862688} {"current_steps": 27650, "total_steps": 42420, "loss": 0.2825, "lr": 0.0003260549241781985, "epoch": 13.036303630363037, "percentage": 65.18, "elapsed_time": "1:29:57", "remaining_time": "0:48:03", "throughput": 5163.43, "total_tokens": 27868896} {"current_steps": 27655, "total_steps": 42420, "loss": 0.2582, "lr": 0.0003258620690589881, "epoch": 13.038661008958039, "percentage": 65.19, "elapsed_time": "1:29:58", "remaining_time": "0:48:02", "throughput": 5163.49, "total_tokens": 27874368} {"current_steps": 27660, "total_steps": 42420, "loss": 0.3097, "lr": 0.0003256692434183727, "epoch": 13.041018387553041, "percentage": 65.21, "elapsed_time": "1:29:59", "remaining_time": "0:48:01", "throughput": 5163.53, "total_tokens": 27879392} {"current_steps": 27665, "total_steps": 42420, "loss": 0.344, "lr": 0.0003254764472889944, "epoch": 13.043375766148044, "percentage": 65.22, "elapsed_time": "1:30:00", "remaining_time": "0:48:00", "throughput": 5163.51, "total_tokens": 27883712} {"current_steps": 27670, "total_steps": 42420, "loss": 0.4014, "lr": 0.0003252836807034905, "epoch": 13.045733144743046, "percentage": 65.23, "elapsed_time": "1:30:00", "remaining_time": "0:47:59", "throughput": 5163.5, "total_tokens": 27888000} {"current_steps": 27675, "total_steps": 42420, "loss": 0.342, "lr": 0.00032509094369449277, "epoch": 13.048090523338049, "percentage": 65.24, "elapsed_time": "1:30:01", "remaining_time": "0:47:58", "throughput": 5163.53, "total_tokens": 27892480} {"current_steps": 27680, "total_steps": 42420, "loss": 0.3566, "lr": 0.00032489823629462845, "epoch": 13.050447901933051, "percentage": 65.25, "elapsed_time": "1:30:02", "remaining_time": "0:47:56", "throughput": 5163.49, "total_tokens": 27896544} {"current_steps": 27685, "total_steps": 42420, "loss": 0.3707, "lr": 0.0003247055585365197, "epoch": 13.052805280528053, "percentage": 65.26, "elapsed_time": "1:30:03", "remaining_time": "0:47:56", "throughput": 5163.55, "total_tokens": 27901792} {"current_steps": 27690, "total_steps": 42420, "loss": 0.3618, "lr": 0.00032451291045278375, "epoch": 13.055162659123056, "percentage": 65.28, "elapsed_time": "1:30:04", "remaining_time": "0:47:55", "throughput": 5163.64, "total_tokens": 27907744} {"current_steps": 27695, "total_steps": 42420, "loss": 0.3305, "lr": 0.0003243202920760323, "epoch": 13.057520037718058, "percentage": 65.29, "elapsed_time": "1:30:05", "remaining_time": "0:47:54", "throughput": 5163.67, "total_tokens": 27912736} {"current_steps": 27700, "total_steps": 42420, "loss": 0.3316, "lr": 0.00032412770343887254, "epoch": 13.05987741631306, "percentage": 65.3, "elapsed_time": "1:30:06", "remaining_time": "0:47:52", "throughput": 5163.66, "total_tokens": 27916736} {"current_steps": 27705, "total_steps": 42420, "loss": 0.3495, "lr": 0.0003239351445739064, "epoch": 13.062234794908063, "percentage": 65.31, "elapsed_time": "1:30:07", "remaining_time": "0:47:52", "throughput": 5163.73, "total_tokens": 27922144} {"current_steps": 27710, "total_steps": 42420, "loss": 0.3302, "lr": 0.0003237426155137311, "epoch": 13.064592173503065, "percentage": 65.32, "elapsed_time": "1:30:08", "remaining_time": "0:47:51", "throughput": 5163.77, "total_tokens": 27927360} {"current_steps": 27715, "total_steps": 42420, "loss": 0.3483, "lr": 0.00032355011629093806, "epoch": 13.066949552098066, "percentage": 65.33, "elapsed_time": "1:30:09", "remaining_time": "0:47:50", "throughput": 5163.85, "total_tokens": 27932960} {"current_steps": 27720, "total_steps": 42420, "loss": 0.3049, "lr": 0.00032335764693811444, "epoch": 13.069306930693068, "percentage": 65.35, "elapsed_time": "1:30:10", "remaining_time": "0:47:49", "throughput": 5163.9, "total_tokens": 27938112} {"current_steps": 27725, "total_steps": 42420, "loss": 0.3304, "lr": 0.00032316520748784194, "epoch": 13.07166430928807, "percentage": 65.36, "elapsed_time": "1:30:11", "remaining_time": "0:47:48", "throughput": 5163.97, "total_tokens": 27943328} {"current_steps": 27730, "total_steps": 42420, "loss": 0.367, "lr": 0.0003229727979726975, "epoch": 13.074021687883073, "percentage": 65.37, "elapsed_time": "1:30:12", "remaining_time": "0:47:47", "throughput": 5164.02, "total_tokens": 27948448} {"current_steps": 27735, "total_steps": 42420, "loss": 0.3057, "lr": 0.0003227804184252525, "epoch": 13.076379066478076, "percentage": 65.38, "elapsed_time": "1:30:12", "remaining_time": "0:47:46", "throughput": 5163.98, "total_tokens": 27952320} {"current_steps": 27740, "total_steps": 42420, "loss": 0.3263, "lr": 0.0003225880688780736, "epoch": 13.078736445073078, "percentage": 65.39, "elapsed_time": "1:30:13", "remaining_time": "0:47:45", "throughput": 5164.08, "total_tokens": 27958048} {"current_steps": 27745, "total_steps": 42420, "loss": 0.3047, "lr": 0.0003223957493637224, "epoch": 13.08109382366808, "percentage": 65.41, "elapsed_time": "1:30:14", "remaining_time": "0:47:44", "throughput": 5164.12, "total_tokens": 27963168} {"current_steps": 27750, "total_steps": 42420, "loss": 0.3411, "lr": 0.00032220345991475544, "epoch": 13.083451202263083, "percentage": 65.42, "elapsed_time": "1:30:15", "remaining_time": "0:47:43", "throughput": 5164.12, "total_tokens": 27967360} {"current_steps": 27755, "total_steps": 42420, "loss": 0.3212, "lr": 0.0003220112005637237, "epoch": 13.085808580858085, "percentage": 65.43, "elapsed_time": "1:30:16", "remaining_time": "0:47:42", "throughput": 5164.26, "total_tokens": 27973664} {"current_steps": 27760, "total_steps": 42420, "loss": 0.2862, "lr": 0.0003218189713431736, "epoch": 13.088165959453088, "percentage": 65.44, "elapsed_time": "1:30:17", "remaining_time": "0:47:41", "throughput": 5164.34, "total_tokens": 27979488} {"current_steps": 27765, "total_steps": 42420, "loss": 0.2813, "lr": 0.0003216267722856464, "epoch": 13.09052333804809, "percentage": 65.45, "elapsed_time": "1:30:18", "remaining_time": "0:47:40", "throughput": 5164.42, "total_tokens": 27984992} {"current_steps": 27770, "total_steps": 42420, "loss": 0.367, "lr": 0.00032143460342367816, "epoch": 13.092880716643092, "percentage": 65.46, "elapsed_time": "1:30:19", "remaining_time": "0:47:39", "throughput": 5164.53, "total_tokens": 27991104} {"current_steps": 27775, "total_steps": 42420, "loss": 0.278, "lr": 0.00032124246478979943, "epoch": 13.095238095238095, "percentage": 65.48, "elapsed_time": "1:30:20", "remaining_time": "0:47:38", "throughput": 5164.65, "total_tokens": 27996896} {"current_steps": 27780, "total_steps": 42420, "loss": 0.3101, "lr": 0.00032105035641653645, "epoch": 13.097595473833097, "percentage": 65.49, "elapsed_time": "1:30:21", "remaining_time": "0:47:37", "throughput": 5164.65, "total_tokens": 28001312} {"current_steps": 27785, "total_steps": 42420, "loss": 0.2418, "lr": 0.0003208582783364097, "epoch": 13.0999528524281, "percentage": 65.5, "elapsed_time": "1:30:22", "remaining_time": "0:47:36", "throughput": 5164.73, "total_tokens": 28006688} {"current_steps": 27790, "total_steps": 42420, "loss": 0.2625, "lr": 0.00032066623058193483, "epoch": 13.102310231023102, "percentage": 65.51, "elapsed_time": "1:30:23", "remaining_time": "0:47:35", "throughput": 5164.79, "total_tokens": 28011968} {"current_steps": 27795, "total_steps": 42420, "loss": 0.3246, "lr": 0.00032047421318562216, "epoch": 13.104667609618105, "percentage": 65.52, "elapsed_time": "1:30:24", "remaining_time": "0:47:34", "throughput": 5164.82, "total_tokens": 28016544} {"current_steps": 27800, "total_steps": 42420, "loss": 0.3062, "lr": 0.000320282226179977, "epoch": 13.107024988213107, "percentage": 65.54, "elapsed_time": "1:30:25", "remaining_time": "0:47:33", "throughput": 5164.84, "total_tokens": 28020992} {"current_steps": 27805, "total_steps": 42420, "loss": 0.2922, "lr": 0.00032009026959749947, "epoch": 13.10938236680811, "percentage": 65.55, "elapsed_time": "1:30:26", "remaining_time": "0:47:32", "throughput": 5164.83, "total_tokens": 28025248} {"current_steps": 27810, "total_steps": 42420, "loss": 0.3386, "lr": 0.00031989834347068483, "epoch": 13.111739745403112, "percentage": 65.56, "elapsed_time": "1:30:27", "remaining_time": "0:47:31", "throughput": 5164.94, "total_tokens": 28031200} {"current_steps": 27815, "total_steps": 42420, "loss": 0.3295, "lr": 0.00031970644783202245, "epoch": 13.114097123998114, "percentage": 65.57, "elapsed_time": "1:30:28", "remaining_time": "0:47:30", "throughput": 5164.96, "total_tokens": 28035808} {"current_steps": 27820, "total_steps": 42420, "loss": 0.2839, "lr": 0.0003195145827139973, "epoch": 13.116454502593117, "percentage": 65.58, "elapsed_time": "1:30:28", "remaining_time": "0:47:29", "throughput": 5164.96, "total_tokens": 28040224} {"current_steps": 27825, "total_steps": 42420, "loss": 0.3295, "lr": 0.00031932274814908877, "epoch": 13.118811881188119, "percentage": 65.59, "elapsed_time": "1:30:29", "remaining_time": "0:47:28", "throughput": 5164.96, "total_tokens": 28044384} {"current_steps": 27830, "total_steps": 42420, "loss": 0.3463, "lr": 0.00031913094416977143, "epoch": 13.121169259783121, "percentage": 65.61, "elapsed_time": "1:30:30", "remaining_time": "0:47:27", "throughput": 5165.08, "total_tokens": 28050592} {"current_steps": 27835, "total_steps": 42420, "loss": 0.3639, "lr": 0.0003189391708085142, "epoch": 13.123526638378124, "percentage": 65.62, "elapsed_time": "1:30:31", "remaining_time": "0:47:26", "throughput": 5165.05, "total_tokens": 28054432} {"current_steps": 27840, "total_steps": 42420, "loss": 0.3447, "lr": 0.000318747428097781, "epoch": 13.125884016973126, "percentage": 65.63, "elapsed_time": "1:30:32", "remaining_time": "0:47:25", "throughput": 5165.1, "total_tokens": 28059456} {"current_steps": 27845, "total_steps": 42420, "loss": 0.3509, "lr": 0.0003185557160700308, "epoch": 13.128241395568129, "percentage": 65.64, "elapsed_time": "1:30:33", "remaining_time": "0:47:24", "throughput": 5165.14, "total_tokens": 28064320} {"current_steps": 27850, "total_steps": 42420, "loss": 0.3094, "lr": 0.00031836403475771723, "epoch": 13.130598774163131, "percentage": 65.65, "elapsed_time": "1:30:34", "remaining_time": "0:47:23", "throughput": 5165.19, "total_tokens": 28069568} {"current_steps": 27855, "total_steps": 42420, "loss": 0.3243, "lr": 0.00031817238419328843, "epoch": 13.132956152758133, "percentage": 65.66, "elapsed_time": "1:30:35", "remaining_time": "0:47:22", "throughput": 5165.21, "total_tokens": 28074240} {"current_steps": 27860, "total_steps": 42420, "loss": 0.3093, "lr": 0.00031798076440918795, "epoch": 13.135313531353136, "percentage": 65.68, "elapsed_time": "1:30:36", "remaining_time": "0:47:21", "throughput": 5165.24, "total_tokens": 28079232} {"current_steps": 27865, "total_steps": 42420, "loss": 0.2963, "lr": 0.00031778917543785333, "epoch": 13.137670909948138, "percentage": 65.69, "elapsed_time": "1:30:37", "remaining_time": "0:47:19", "throughput": 5165.25, "total_tokens": 28083808} {"current_steps": 27870, "total_steps": 42420, "loss": 0.3941, "lr": 0.00031759761731171756, "epoch": 13.14002828854314, "percentage": 65.7, "elapsed_time": "1:30:37", "remaining_time": "0:47:18", "throughput": 5165.28, "total_tokens": 28088640} {"current_steps": 27875, "total_steps": 42420, "loss": 0.3254, "lr": 0.00031740609006320814, "epoch": 13.142385667138143, "percentage": 65.71, "elapsed_time": "1:30:38", "remaining_time": "0:47:17", "throughput": 5165.3, "total_tokens": 28093280} {"current_steps": 27880, "total_steps": 42420, "loss": 0.3157, "lr": 0.00031721459372474755, "epoch": 13.144743045733145, "percentage": 65.72, "elapsed_time": "1:30:39", "remaining_time": "0:47:16", "throughput": 5165.36, "total_tokens": 28098720} {"current_steps": 27885, "total_steps": 42420, "loss": 0.3739, "lr": 0.00031702312832875256, "epoch": 13.147100424328148, "percentage": 65.74, "elapsed_time": "1:30:40", "remaining_time": "0:47:15", "throughput": 5165.39, "total_tokens": 28103712} {"current_steps": 27890, "total_steps": 42420, "loss": 0.3245, "lr": 0.00031683169390763524, "epoch": 13.14945780292315, "percentage": 65.75, "elapsed_time": "1:30:41", "remaining_time": "0:47:15", "throughput": 5165.5, "total_tokens": 28109504} {"current_steps": 27895, "total_steps": 42420, "loss": 0.3252, "lr": 0.00031664029049380206, "epoch": 13.151815181518153, "percentage": 65.76, "elapsed_time": "1:30:42", "remaining_time": "0:47:13", "throughput": 5165.5, "total_tokens": 28113920} {"current_steps": 27900, "total_steps": 42420, "loss": 0.3003, "lr": 0.0003164489181196546, "epoch": 13.154172560113155, "percentage": 65.77, "elapsed_time": "1:30:43", "remaining_time": "0:47:12", "throughput": 5165.5, "total_tokens": 28118144} {"current_steps": 27905, "total_steps": 42420, "loss": 0.3996, "lr": 0.0003162575768175886, "epoch": 13.156529938708157, "percentage": 65.78, "elapsed_time": "1:30:44", "remaining_time": "0:47:11", "throughput": 5165.54, "total_tokens": 28123040} {"current_steps": 27910, "total_steps": 42420, "loss": 0.3476, "lr": 0.00031606626661999517, "epoch": 13.15888731730316, "percentage": 65.79, "elapsed_time": "1:30:45", "remaining_time": "0:47:11", "throughput": 5165.67, "total_tokens": 28129376} {"current_steps": 27915, "total_steps": 42420, "loss": 0.3478, "lr": 0.0003158749875592597, "epoch": 13.16124469589816, "percentage": 65.81, "elapsed_time": "1:30:46", "remaining_time": "0:47:10", "throughput": 5165.78, "total_tokens": 28135296} {"current_steps": 27920, "total_steps": 42420, "loss": 0.3372, "lr": 0.0003156837396677627, "epoch": 13.163602074493163, "percentage": 65.82, "elapsed_time": "1:30:47", "remaining_time": "0:47:09", "throughput": 5165.8, "total_tokens": 28139968} {"current_steps": 27925, "total_steps": 42420, "loss": 0.3505, "lr": 0.00031549252297787895, "epoch": 13.165959453088165, "percentage": 65.83, "elapsed_time": "1:30:48", "remaining_time": "0:47:08", "throughput": 5165.86, "total_tokens": 28145312} {"current_steps": 27930, "total_steps": 42420, "loss": 0.3178, "lr": 0.00031530133752197834, "epoch": 13.168316831683168, "percentage": 65.84, "elapsed_time": "1:30:49", "remaining_time": "0:47:07", "throughput": 5165.93, "total_tokens": 28151040} {"current_steps": 27935, "total_steps": 42420, "loss": 0.3168, "lr": 0.00031511018333242526, "epoch": 13.17067421027817, "percentage": 65.85, "elapsed_time": "1:30:50", "remaining_time": "0:47:06", "throughput": 5165.96, "total_tokens": 28155808} {"current_steps": 27940, "total_steps": 42420, "loss": 0.2889, "lr": 0.00031491906044157897, "epoch": 13.173031588873172, "percentage": 65.87, "elapsed_time": "1:30:51", "remaining_time": "0:47:05", "throughput": 5165.96, "total_tokens": 28160256} {"current_steps": 27945, "total_steps": 42420, "loss": 0.4078, "lr": 0.00031472796888179315, "epoch": 13.175388967468175, "percentage": 65.88, "elapsed_time": "1:30:52", "remaining_time": "0:47:04", "throughput": 5166.02, "total_tokens": 28165312} {"current_steps": 27950, "total_steps": 42420, "loss": 0.2344, "lr": 0.0003145369086854165, "epoch": 13.177746346063177, "percentage": 65.89, "elapsed_time": "1:30:52", "remaining_time": "0:47:03", "throughput": 5166.08, "total_tokens": 28170336} {"current_steps": 27955, "total_steps": 42420, "loss": 0.3148, "lr": 0.0003143458798847922, "epoch": 13.18010372465818, "percentage": 65.9, "elapsed_time": "1:30:53", "remaining_time": "0:47:01", "throughput": 5166.09, "total_tokens": 28174688} {"current_steps": 27960, "total_steps": 42420, "loss": 0.3314, "lr": 0.00031415488251225834, "epoch": 13.182461103253182, "percentage": 65.91, "elapsed_time": "1:30:54", "remaining_time": "0:47:00", "throughput": 5166.12, "total_tokens": 28179552} {"current_steps": 27965, "total_steps": 42420, "loss": 0.4044, "lr": 0.00031396391660014724, "epoch": 13.184818481848184, "percentage": 65.92, "elapsed_time": "1:30:55", "remaining_time": "0:46:59", "throughput": 5166.14, "total_tokens": 28184544} {"current_steps": 27970, "total_steps": 42420, "loss": 0.3295, "lr": 0.0003137729821807864, "epoch": 13.187175860443187, "percentage": 65.94, "elapsed_time": "1:30:56", "remaining_time": "0:46:58", "throughput": 5166.17, "total_tokens": 28189440} {"current_steps": 27975, "total_steps": 42420, "loss": 0.3296, "lr": 0.0003135820792864977, "epoch": 13.18953323903819, "percentage": 65.95, "elapsed_time": "1:30:57", "remaining_time": "0:46:57", "throughput": 5166.17, "total_tokens": 28193888} {"current_steps": 27980, "total_steps": 42420, "loss": 0.3979, "lr": 0.00031339120794959786, "epoch": 13.191890617633192, "percentage": 65.96, "elapsed_time": "1:30:58", "remaining_time": "0:46:56", "throughput": 5166.25, "total_tokens": 28199424} {"current_steps": 27985, "total_steps": 42420, "loss": 0.3378, "lr": 0.000313200368202398, "epoch": 13.194247996228194, "percentage": 65.97, "elapsed_time": "1:30:59", "remaining_time": "0:46:55", "throughput": 5166.29, "total_tokens": 28204512} {"current_steps": 27990, "total_steps": 42420, "loss": 0.3561, "lr": 0.0003130095600772042, "epoch": 13.196605374823196, "percentage": 65.98, "elapsed_time": "1:31:00", "remaining_time": "0:46:54", "throughput": 5166.34, "total_tokens": 28209440} {"current_steps": 27995, "total_steps": 42420, "loss": 0.3477, "lr": 0.00031281878360631685, "epoch": 13.198962753418199, "percentage": 65.99, "elapsed_time": "1:31:01", "remaining_time": "0:46:53", "throughput": 5166.35, "total_tokens": 28213824} {"current_steps": 28000, "total_steps": 42420, "loss": 0.3273, "lr": 0.0003126280388220314, "epoch": 13.201320132013201, "percentage": 66.01, "elapsed_time": "1:31:01", "remaining_time": "0:46:52", "throughput": 5166.38, "total_tokens": 28218624} {"current_steps": 28005, "total_steps": 42420, "loss": 0.3402, "lr": 0.0003124373257566375, "epoch": 13.203677510608204, "percentage": 66.02, "elapsed_time": "1:31:03", "remaining_time": "0:46:52", "throughput": 5166.55, "total_tokens": 28225472} {"current_steps": 28010, "total_steps": 42420, "loss": 0.2919, "lr": 0.0003122466444424197, "epoch": 13.206034889203206, "percentage": 66.03, "elapsed_time": "1:31:04", "remaining_time": "0:46:51", "throughput": 5166.63, "total_tokens": 28230848} {"current_steps": 28015, "total_steps": 42420, "loss": 0.3403, "lr": 0.0003120559949116571, "epoch": 13.208392267798208, "percentage": 66.04, "elapsed_time": "1:31:04", "remaining_time": "0:46:50", "throughput": 5166.68, "total_tokens": 28235872} {"current_steps": 28020, "total_steps": 42420, "loss": 0.3067, "lr": 0.0003118653771966237, "epoch": 13.21074964639321, "percentage": 66.05, "elapsed_time": "1:31:05", "remaining_time": "0:46:49", "throughput": 5166.71, "total_tokens": 28240544} {"current_steps": 28025, "total_steps": 42420, "loss": 0.3561, "lr": 0.0003116747913295873, "epoch": 13.213107024988213, "percentage": 66.07, "elapsed_time": "1:31:06", "remaining_time": "0:46:47", "throughput": 5166.75, "total_tokens": 28245440} {"current_steps": 28030, "total_steps": 42420, "loss": 0.3039, "lr": 0.00031148423734281126, "epoch": 13.215464403583216, "percentage": 66.08, "elapsed_time": "1:31:07", "remaining_time": "0:46:46", "throughput": 5166.71, "total_tokens": 28249120} {"current_steps": 28035, "total_steps": 42420, "loss": 0.3399, "lr": 0.00031129371526855297, "epoch": 13.217821782178218, "percentage": 66.09, "elapsed_time": "1:31:08", "remaining_time": "0:46:45", "throughput": 5166.78, "total_tokens": 28254368} {"current_steps": 28040, "total_steps": 42420, "loss": 0.2884, "lr": 0.0003111032251390648, "epoch": 13.22017916077322, "percentage": 66.1, "elapsed_time": "1:31:09", "remaining_time": "0:46:44", "throughput": 5166.87, "total_tokens": 28259968} {"current_steps": 28045, "total_steps": 42420, "loss": 0.3575, "lr": 0.0003109127669865932, "epoch": 13.222536539368223, "percentage": 66.11, "elapsed_time": "1:31:10", "remaining_time": "0:46:44", "throughput": 5167.01, "total_tokens": 28266240} {"current_steps": 28050, "total_steps": 42420, "loss": 0.2903, "lr": 0.00031072234084337965, "epoch": 13.224893917963225, "percentage": 66.12, "elapsed_time": "1:31:11", "remaining_time": "0:46:43", "throughput": 5167.08, "total_tokens": 28271456} {"current_steps": 28055, "total_steps": 42420, "loss": 0.3153, "lr": 0.00031053194674165996, "epoch": 13.227251296558228, "percentage": 66.14, "elapsed_time": "1:31:12", "remaining_time": "0:46:42", "throughput": 5167.15, "total_tokens": 28276736} {"current_steps": 28060, "total_steps": 42420, "loss": 0.2346, "lr": 0.0003103415847136649, "epoch": 13.22960867515323, "percentage": 66.15, "elapsed_time": "1:31:13", "remaining_time": "0:46:41", "throughput": 5167.17, "total_tokens": 28281664} {"current_steps": 28065, "total_steps": 42420, "loss": 0.3709, "lr": 0.0003101512547916192, "epoch": 13.231966053748232, "percentage": 66.16, "elapsed_time": "1:31:14", "remaining_time": "0:46:40", "throughput": 5167.24, "total_tokens": 28287072} {"current_steps": 28070, "total_steps": 42420, "loss": 0.3855, "lr": 0.0003099609570077426, "epoch": 13.234323432343235, "percentage": 66.17, "elapsed_time": "1:31:15", "remaining_time": "0:46:39", "throughput": 5167.25, "total_tokens": 28291520} {"current_steps": 28075, "total_steps": 42420, "loss": 0.3359, "lr": 0.0003097706913942493, "epoch": 13.236680810938237, "percentage": 66.18, "elapsed_time": "1:31:16", "remaining_time": "0:46:37", "throughput": 5167.25, "total_tokens": 28295936} {"current_steps": 28080, "total_steps": 42420, "loss": 0.3484, "lr": 0.0003095804579833482, "epoch": 13.23903818953324, "percentage": 66.2, "elapsed_time": "1:31:16", "remaining_time": "0:46:36", "throughput": 5167.29, "total_tokens": 28300992} {"current_steps": 28085, "total_steps": 42420, "loss": 0.3176, "lr": 0.0003093902568072423, "epoch": 13.241395568128242, "percentage": 66.21, "elapsed_time": "1:31:17", "remaining_time": "0:46:36", "throughput": 5167.4, "total_tokens": 28307008} {"current_steps": 28090, "total_steps": 42420, "loss": 0.3179, "lr": 0.0003092000878981296, "epoch": 13.243752946723244, "percentage": 66.22, "elapsed_time": "1:31:18", "remaining_time": "0:46:35", "throughput": 5167.41, "total_tokens": 28311520} {"current_steps": 28095, "total_steps": 42420, "loss": 0.3338, "lr": 0.0003090099512882024, "epoch": 13.246110325318247, "percentage": 66.23, "elapsed_time": "1:31:19", "remaining_time": "0:46:34", "throughput": 5167.45, "total_tokens": 28316672} {"current_steps": 28100, "total_steps": 42420, "loss": 0.3394, "lr": 0.00030881984700964753, "epoch": 13.24846770391325, "percentage": 66.24, "elapsed_time": "1:31:20", "remaining_time": "0:46:33", "throughput": 5167.5, "total_tokens": 28321728} {"current_steps": 28105, "total_steps": 42420, "loss": 0.3806, "lr": 0.00030862977509464655, "epoch": 13.250825082508252, "percentage": 66.25, "elapsed_time": "1:31:21", "remaining_time": "0:46:32", "throughput": 5167.63, "total_tokens": 28327968} {"current_steps": 28110, "total_steps": 42420, "loss": 0.3085, "lr": 0.0003084397355753755, "epoch": 13.253182461103254, "percentage": 66.27, "elapsed_time": "1:31:22", "remaining_time": "0:46:31", "throughput": 5167.65, "total_tokens": 28332576} {"current_steps": 28115, "total_steps": 42420, "loss": 0.3594, "lr": 0.0003082497284840046, "epoch": 13.255539839698255, "percentage": 66.28, "elapsed_time": "1:31:23", "remaining_time": "0:46:30", "throughput": 5167.78, "total_tokens": 28339104} {"current_steps": 28120, "total_steps": 42420, "loss": 0.3493, "lr": 0.0003080597538526988, "epoch": 13.257897218293257, "percentage": 66.29, "elapsed_time": "1:31:24", "remaining_time": "0:46:29", "throughput": 5167.88, "total_tokens": 28344864} {"current_steps": 28125, "total_steps": 42420, "loss": 0.3404, "lr": 0.0003078698117136178, "epoch": 13.26025459688826, "percentage": 66.3, "elapsed_time": "1:31:25", "remaining_time": "0:46:28", "throughput": 5167.87, "total_tokens": 28349280} {"current_steps": 28130, "total_steps": 42420, "loss": 0.3496, "lr": 0.00030767990209891563, "epoch": 13.262611975483262, "percentage": 66.31, "elapsed_time": "1:31:26", "remaining_time": "0:46:27", "throughput": 5167.93, "total_tokens": 28354528} {"current_steps": 28135, "total_steps": 42420, "loss": 0.3078, "lr": 0.00030749002504074053, "epoch": 13.264969354078264, "percentage": 66.32, "elapsed_time": "1:31:27", "remaining_time": "0:46:26", "throughput": 5168.03, "total_tokens": 28360416} {"current_steps": 28140, "total_steps": 42420, "loss": 0.3172, "lr": 0.0003073001805712355, "epoch": 13.267326732673267, "percentage": 66.34, "elapsed_time": "1:31:28", "remaining_time": "0:46:25", "throughput": 5168.22, "total_tokens": 28367680} {"current_steps": 28145, "total_steps": 42420, "loss": 0.3644, "lr": 0.00030711036872253805, "epoch": 13.269684111268269, "percentage": 66.35, "elapsed_time": "1:31:29", "remaining_time": "0:46:24", "throughput": 5168.23, "total_tokens": 28372160} {"current_steps": 28150, "total_steps": 42420, "loss": 0.3144, "lr": 0.0003069205895267803, "epoch": 13.272041489863271, "percentage": 66.36, "elapsed_time": "1:31:30", "remaining_time": "0:46:23", "throughput": 5168.28, "total_tokens": 28377376} {"current_steps": 28155, "total_steps": 42420, "loss": 0.2968, "lr": 0.0003067308430160882, "epoch": 13.274398868458274, "percentage": 66.37, "elapsed_time": "1:31:31", "remaining_time": "0:46:22", "throughput": 5168.29, "total_tokens": 28381984} {"current_steps": 28160, "total_steps": 42420, "loss": 0.3582, "lr": 0.0003065411292225829, "epoch": 13.276756247053276, "percentage": 66.38, "elapsed_time": "1:31:32", "remaining_time": "0:46:21", "throughput": 5168.33, "total_tokens": 28386848} {"current_steps": 28165, "total_steps": 42420, "loss": 0.3869, "lr": 0.00030635144817837967, "epoch": 13.279113625648279, "percentage": 66.4, "elapsed_time": "1:31:33", "remaining_time": "0:46:20", "throughput": 5168.36, "total_tokens": 28391648} {"current_steps": 28170, "total_steps": 42420, "loss": 0.3555, "lr": 0.00030616179991558846, "epoch": 13.281471004243281, "percentage": 66.41, "elapsed_time": "1:31:34", "remaining_time": "0:46:19", "throughput": 5168.38, "total_tokens": 28396320} {"current_steps": 28175, "total_steps": 42420, "loss": 0.3354, "lr": 0.00030597218446631316, "epoch": 13.283828382838283, "percentage": 66.42, "elapsed_time": "1:31:35", "remaining_time": "0:46:18", "throughput": 5168.48, "total_tokens": 28401952} {"current_steps": 28180, "total_steps": 42420, "loss": 0.3479, "lr": 0.00030578260186265265, "epoch": 13.286185761433286, "percentage": 66.43, "elapsed_time": "1:31:36", "remaining_time": "0:46:17", "throughput": 5168.6, "total_tokens": 28408064} {"current_steps": 28185, "total_steps": 42420, "loss": 0.3236, "lr": 0.00030559305213670007, "epoch": 13.288543140028288, "percentage": 66.44, "elapsed_time": "1:31:37", "remaining_time": "0:46:16", "throughput": 5168.58, "total_tokens": 28412064} {"current_steps": 28190, "total_steps": 42420, "loss": 0.3449, "lr": 0.000305403535320543, "epoch": 13.29090051862329, "percentage": 66.45, "elapsed_time": "1:31:37", "remaining_time": "0:46:15", "throughput": 5168.58, "total_tokens": 28416512} {"current_steps": 28195, "total_steps": 42420, "loss": 0.3445, "lr": 0.0003052140514462632, "epoch": 13.293257897218293, "percentage": 66.47, "elapsed_time": "1:31:38", "remaining_time": "0:46:14", "throughput": 5168.66, "total_tokens": 28422144} {"current_steps": 28200, "total_steps": 42420, "loss": 0.3105, "lr": 0.00030502460054593724, "epoch": 13.295615275813295, "percentage": 66.48, "elapsed_time": "1:31:39", "remaining_time": "0:46:13", "throughput": 5168.7, "total_tokens": 28426880} {"current_steps": 28205, "total_steps": 42420, "loss": 0.3281, "lr": 0.0003048351826516359, "epoch": 13.297972654408298, "percentage": 66.49, "elapsed_time": "1:31:40", "remaining_time": "0:46:12", "throughput": 5168.77, "total_tokens": 28432416} {"current_steps": 28210, "total_steps": 42420, "loss": 0.2444, "lr": 0.00030464579779542454, "epoch": 13.3003300330033, "percentage": 66.5, "elapsed_time": "1:31:41", "remaining_time": "0:46:11", "throughput": 5168.82, "total_tokens": 28437408} {"current_steps": 28215, "total_steps": 42420, "loss": 0.3085, "lr": 0.00030445644600936257, "epoch": 13.302687411598303, "percentage": 66.51, "elapsed_time": "1:31:42", "remaining_time": "0:46:10", "throughput": 5168.89, "total_tokens": 28442912} {"current_steps": 28220, "total_steps": 42420, "loss": 0.4088, "lr": 0.00030426712732550414, "epoch": 13.305044790193305, "percentage": 66.53, "elapsed_time": "1:31:43", "remaining_time": "0:46:09", "throughput": 5168.91, "total_tokens": 28447552} {"current_steps": 28225, "total_steps": 42420, "loss": 0.4339, "lr": 0.0003040778417758978, "epoch": 13.307402168788308, "percentage": 66.54, "elapsed_time": "1:31:44", "remaining_time": "0:46:08", "throughput": 5168.92, "total_tokens": 28451936} {"current_steps": 28230, "total_steps": 42420, "loss": 0.3589, "lr": 0.0003038885893925862, "epoch": 13.30975954738331, "percentage": 66.55, "elapsed_time": "1:31:45", "remaining_time": "0:46:07", "throughput": 5168.95, "total_tokens": 28456864} {"current_steps": 28235, "total_steps": 42420, "loss": 0.3467, "lr": 0.0003036993702076067, "epoch": 13.312116925978312, "percentage": 66.56, "elapsed_time": "1:31:46", "remaining_time": "0:46:06", "throughput": 5168.98, "total_tokens": 28461568} {"current_steps": 28240, "total_steps": 42420, "loss": 0.3388, "lr": 0.0003035101842529907, "epoch": 13.314474304573315, "percentage": 66.57, "elapsed_time": "1:31:47", "remaining_time": "0:46:05", "throughput": 5169.05, "total_tokens": 28467328} {"current_steps": 28245, "total_steps": 42420, "loss": 0.328, "lr": 0.0003033210315607644, "epoch": 13.316831683168317, "percentage": 66.58, "elapsed_time": "1:31:48", "remaining_time": "0:46:04", "throughput": 5169.11, "total_tokens": 28472608} {"current_steps": 28250, "total_steps": 42420, "loss": 0.3302, "lr": 0.0003031319121629481, "epoch": 13.31918906176332, "percentage": 66.6, "elapsed_time": "1:31:49", "remaining_time": "0:46:03", "throughput": 5169.18, "total_tokens": 28478016} {"current_steps": 28255, "total_steps": 42420, "loss": 0.3392, "lr": 0.0003029428260915563, "epoch": 13.321546440358322, "percentage": 66.61, "elapsed_time": "1:31:50", "remaining_time": "0:46:02", "throughput": 5169.23, "total_tokens": 28482816} {"current_steps": 28260, "total_steps": 42420, "loss": 0.2965, "lr": 0.0003027537733785983, "epoch": 13.323903818953324, "percentage": 66.62, "elapsed_time": "1:31:51", "remaining_time": "0:46:01", "throughput": 5169.27, "total_tokens": 28487968} {"current_steps": 28265, "total_steps": 42420, "loss": 0.2945, "lr": 0.00030256475405607736, "epoch": 13.326261197548327, "percentage": 66.63, "elapsed_time": "1:31:51", "remaining_time": "0:46:00", "throughput": 5169.31, "total_tokens": 28492960} {"current_steps": 28270, "total_steps": 42420, "loss": 0.3731, "lr": 0.00030237576815599136, "epoch": 13.32861857614333, "percentage": 66.64, "elapsed_time": "1:31:52", "remaining_time": "0:45:59", "throughput": 5169.3, "total_tokens": 28497312} {"current_steps": 28275, "total_steps": 42420, "loss": 0.3333, "lr": 0.0003021868157103323, "epoch": 13.330975954738332, "percentage": 66.65, "elapsed_time": "1:31:53", "remaining_time": "0:45:58", "throughput": 5169.35, "total_tokens": 28502240} {"current_steps": 28280, "total_steps": 42420, "loss": 0.3738, "lr": 0.00030199789675108667, "epoch": 13.333333333333334, "percentage": 66.67, "elapsed_time": "1:31:54", "remaining_time": "0:45:57", "throughput": 5169.34, "total_tokens": 28506464} {"current_steps": 28285, "total_steps": 42420, "loss": 0.3198, "lr": 0.0003018090113102352, "epoch": 13.335690711928336, "percentage": 66.68, "elapsed_time": "1:31:55", "remaining_time": "0:45:56", "throughput": 5169.42, "total_tokens": 28511872} {"current_steps": 28290, "total_steps": 42420, "loss": 0.3607, "lr": 0.0003016201594197531, "epoch": 13.338048090523339, "percentage": 66.69, "elapsed_time": "1:31:56", "remaining_time": "0:45:55", "throughput": 5169.57, "total_tokens": 28518400} {"current_steps": 28295, "total_steps": 42420, "loss": 0.351, "lr": 0.0003014313411116096, "epoch": 13.340405469118341, "percentage": 66.7, "elapsed_time": "1:31:57", "remaining_time": "0:45:54", "throughput": 5169.7, "total_tokens": 28524640} {"current_steps": 28300, "total_steps": 42420, "loss": 0.3593, "lr": 0.00030124255641776864, "epoch": 13.342762847713344, "percentage": 66.71, "elapsed_time": "1:31:58", "remaining_time": "0:45:53", "throughput": 5169.76, "total_tokens": 28530240} {"current_steps": 28305, "total_steps": 42420, "loss": 0.3576, "lr": 0.0003010538053701881, "epoch": 13.345120226308346, "percentage": 66.73, "elapsed_time": "1:31:59", "remaining_time": "0:45:52", "throughput": 5169.91, "total_tokens": 28537056} {"current_steps": 28310, "total_steps": 42420, "loss": 0.3422, "lr": 0.0003008650880008207, "epoch": 13.347477604903348, "percentage": 66.74, "elapsed_time": "1:32:00", "remaining_time": "0:45:51", "throughput": 5169.97, "total_tokens": 28542304} {"current_steps": 28315, "total_steps": 42420, "loss": 0.3465, "lr": 0.0003006764043416126, "epoch": 13.34983498349835, "percentage": 66.75, "elapsed_time": "1:32:01", "remaining_time": "0:45:50", "throughput": 5170.0, "total_tokens": 28546976} {"current_steps": 28320, "total_steps": 42420, "loss": 0.353, "lr": 0.0003004877544245048, "epoch": 13.352192362093351, "percentage": 66.76, "elapsed_time": "1:32:02", "remaining_time": "0:45:49", "throughput": 5170.01, "total_tokens": 28551584} {"current_steps": 28325, "total_steps": 42420, "loss": 0.3299, "lr": 0.00030029913828143266, "epoch": 13.354549740688354, "percentage": 66.77, "elapsed_time": "1:32:03", "remaining_time": "0:45:48", "throughput": 5170.08, "total_tokens": 28556960} {"current_steps": 28330, "total_steps": 42420, "loss": 0.3112, "lr": 0.0003001105559443257, "epoch": 13.356907119283356, "percentage": 66.78, "elapsed_time": "1:32:04", "remaining_time": "0:45:47", "throughput": 5170.07, "total_tokens": 28561440} {"current_steps": 28335, "total_steps": 42420, "loss": 0.3556, "lr": 0.0002999220074451078, "epoch": 13.359264497878359, "percentage": 66.8, "elapsed_time": "1:32:05", "remaining_time": "0:45:46", "throughput": 5170.14, "total_tokens": 28566752} {"current_steps": 28340, "total_steps": 42420, "loss": 0.3096, "lr": 0.0002997334928156968, "epoch": 13.361621876473361, "percentage": 66.81, "elapsed_time": "1:32:06", "remaining_time": "0:45:45", "throughput": 5170.13, "total_tokens": 28570880} {"current_steps": 28345, "total_steps": 42420, "loss": 0.3097, "lr": 0.0002995450120880051, "epoch": 13.363979255068363, "percentage": 66.82, "elapsed_time": "1:32:07", "remaining_time": "0:45:44", "throughput": 5170.18, "total_tokens": 28576288} {"current_steps": 28350, "total_steps": 42420, "loss": 0.2619, "lr": 0.0002993565652939394, "epoch": 13.366336633663366, "percentage": 66.83, "elapsed_time": "1:32:07", "remaining_time": "0:45:43", "throughput": 5170.17, "total_tokens": 28580544} {"current_steps": 28355, "total_steps": 42420, "loss": 0.2711, "lr": 0.00029916815246540054, "epoch": 13.368694012258368, "percentage": 66.84, "elapsed_time": "1:32:08", "remaining_time": "0:45:42", "throughput": 5170.19, "total_tokens": 28585376} {"current_steps": 28360, "total_steps": 42420, "loss": 0.3372, "lr": 0.00029897977363428337, "epoch": 13.37105139085337, "percentage": 66.86, "elapsed_time": "1:32:09", "remaining_time": "0:45:41", "throughput": 5170.3, "total_tokens": 28591488} {"current_steps": 28365, "total_steps": 42420, "loss": 0.2798, "lr": 0.00029879142883247747, "epoch": 13.373408769448373, "percentage": 66.87, "elapsed_time": "1:32:10", "remaining_time": "0:45:40", "throughput": 5170.38, "total_tokens": 28596800} {"current_steps": 28370, "total_steps": 42420, "loss": 0.2914, "lr": 0.00029860311809186624, "epoch": 13.375766148043375, "percentage": 66.88, "elapsed_time": "1:32:11", "remaining_time": "0:45:39", "throughput": 5170.35, "total_tokens": 28600928} {"current_steps": 28375, "total_steps": 42420, "loss": 0.3457, "lr": 0.0002984148414443278, "epoch": 13.378123526638378, "percentage": 66.89, "elapsed_time": "1:32:12", "remaining_time": "0:45:38", "throughput": 5170.4, "total_tokens": 28605952} {"current_steps": 28380, "total_steps": 42420, "loss": 0.3208, "lr": 0.00029822659892173384, "epoch": 13.38048090523338, "percentage": 66.9, "elapsed_time": "1:32:13", "remaining_time": "0:45:37", "throughput": 5170.39, "total_tokens": 28610144} {"current_steps": 28385, "total_steps": 42420, "loss": 0.371, "lr": 0.00029803839055595073, "epoch": 13.382838283828383, "percentage": 66.91, "elapsed_time": "1:32:14", "remaining_time": "0:45:36", "throughput": 5170.39, "total_tokens": 28614496} {"current_steps": 28390, "total_steps": 42420, "loss": 0.3246, "lr": 0.000297850216378839, "epoch": 13.385195662423385, "percentage": 66.93, "elapsed_time": "1:32:15", "remaining_time": "0:45:35", "throughput": 5170.38, "total_tokens": 28618624} {"current_steps": 28395, "total_steps": 42420, "loss": 0.3395, "lr": 0.00029766207642225353, "epoch": 13.387553041018387, "percentage": 66.94, "elapsed_time": "1:32:16", "remaining_time": "0:45:34", "throughput": 5170.41, "total_tokens": 28623648} {"current_steps": 28400, "total_steps": 42420, "loss": 0.3111, "lr": 0.0002974739707180428, "epoch": 13.38991041961339, "percentage": 66.95, "elapsed_time": "1:32:16", "remaining_time": "0:45:33", "throughput": 5170.44, "total_tokens": 28628672} {"current_steps": 28405, "total_steps": 42420, "loss": 0.3419, "lr": 0.00029728589929805014, "epoch": 13.392267798208392, "percentage": 66.96, "elapsed_time": "1:32:17", "remaining_time": "0:45:32", "throughput": 5170.45, "total_tokens": 28633440} {"current_steps": 28410, "total_steps": 42420, "loss": 0.3504, "lr": 0.00029709786219411294, "epoch": 13.394625176803395, "percentage": 66.97, "elapsed_time": "1:32:18", "remaining_time": "0:45:31", "throughput": 5170.52, "total_tokens": 28638912} {"current_steps": 28415, "total_steps": 42420, "loss": 0.3771, "lr": 0.0002969098594380626, "epoch": 13.396982555398397, "percentage": 66.98, "elapsed_time": "1:32:19", "remaining_time": "0:45:30", "throughput": 5170.53, "total_tokens": 28643552} {"current_steps": 28420, "total_steps": 42420, "loss": 0.305, "lr": 0.0002967218910617247, "epoch": 13.3993399339934, "percentage": 67.0, "elapsed_time": "1:32:20", "remaining_time": "0:45:29", "throughput": 5170.61, "total_tokens": 28648864} {"current_steps": 28425, "total_steps": 42420, "loss": 0.2837, "lr": 0.0002965339570969192, "epoch": 13.401697312588402, "percentage": 67.01, "elapsed_time": "1:32:21", "remaining_time": "0:45:28", "throughput": 5170.66, "total_tokens": 28654240} {"current_steps": 28430, "total_steps": 42420, "loss": 0.3523, "lr": 0.0002963460575754601, "epoch": 13.404054691183404, "percentage": 67.02, "elapsed_time": "1:32:22", "remaining_time": "0:45:27", "throughput": 5170.68, "total_tokens": 28659168} {"current_steps": 28435, "total_steps": 42420, "loss": 0.264, "lr": 0.0002961581925291557, "epoch": 13.406412069778407, "percentage": 67.03, "elapsed_time": "1:32:23", "remaining_time": "0:45:26", "throughput": 5170.76, "total_tokens": 28664480} {"current_steps": 28440, "total_steps": 42420, "loss": 0.3336, "lr": 0.0002959703619898081, "epoch": 13.408769448373409, "percentage": 67.04, "elapsed_time": "1:32:24", "remaining_time": "0:45:25", "throughput": 5170.78, "total_tokens": 28669024} {"current_steps": 28445, "total_steps": 42420, "loss": 0.3086, "lr": 0.00029578256598921407, "epoch": 13.411126826968411, "percentage": 67.06, "elapsed_time": "1:32:25", "remaining_time": "0:45:24", "throughput": 5170.83, "total_tokens": 28674080} {"current_steps": 28450, "total_steps": 42420, "loss": 0.3486, "lr": 0.00029559480455916423, "epoch": 13.413484205563414, "percentage": 67.07, "elapsed_time": "1:32:26", "remaining_time": "0:45:23", "throughput": 5170.84, "total_tokens": 28678688} {"current_steps": 28455, "total_steps": 42420, "loss": 0.3698, "lr": 0.0002954070777314434, "epoch": 13.415841584158416, "percentage": 67.08, "elapsed_time": "1:32:27", "remaining_time": "0:45:22", "throughput": 5170.92, "total_tokens": 28684256} {"current_steps": 28460, "total_steps": 42420, "loss": 0.3418, "lr": 0.00029521938553783044, "epoch": 13.418198962753419, "percentage": 67.09, "elapsed_time": "1:32:28", "remaining_time": "0:45:21", "throughput": 5171.0, "total_tokens": 28690016} {"current_steps": 28465, "total_steps": 42420, "loss": 0.3247, "lr": 0.0002950317280100985, "epoch": 13.420556341348421, "percentage": 67.1, "elapsed_time": "1:32:29", "remaining_time": "0:45:20", "throughput": 5171.0, "total_tokens": 28694528} {"current_steps": 28470, "total_steps": 42420, "loss": 0.3661, "lr": 0.0002948441051800149, "epoch": 13.422913719943423, "percentage": 67.11, "elapsed_time": "1:32:30", "remaining_time": "0:45:19", "throughput": 5171.05, "total_tokens": 28699648} {"current_steps": 28475, "total_steps": 42420, "loss": 0.345, "lr": 0.00029465651707934106, "epoch": 13.425271098538426, "percentage": 67.13, "elapsed_time": "1:32:30", "remaining_time": "0:45:18", "throughput": 5171.07, "total_tokens": 28704384} {"current_steps": 28480, "total_steps": 42420, "loss": 0.3506, "lr": 0.00029446896373983223, "epoch": 13.427628477133428, "percentage": 67.14, "elapsed_time": "1:32:31", "remaining_time": "0:45:17", "throughput": 5171.18, "total_tokens": 28710272} {"current_steps": 28485, "total_steps": 42420, "loss": 0.3443, "lr": 0.0002942814451932381, "epoch": 13.42998585572843, "percentage": 67.15, "elapsed_time": "1:32:32", "remaining_time": "0:45:16", "throughput": 5171.16, "total_tokens": 28714400} {"current_steps": 28490, "total_steps": 42420, "loss": 0.3429, "lr": 0.0002940939614713024, "epoch": 13.432343234323433, "percentage": 67.16, "elapsed_time": "1:32:34", "remaining_time": "0:45:15", "throughput": 5171.34, "total_tokens": 28721664} {"current_steps": 28495, "total_steps": 42420, "loss": 0.3382, "lr": 0.000293906512605763, "epoch": 13.434700612918435, "percentage": 67.17, "elapsed_time": "1:32:34", "remaining_time": "0:45:14", "throughput": 5171.42, "total_tokens": 28727104} {"current_steps": 28500, "total_steps": 42420, "loss": 0.3513, "lr": 0.0002937190986283517, "epoch": 13.437057991513438, "percentage": 67.19, "elapsed_time": "1:32:36", "remaining_time": "0:45:13", "throughput": 5171.48, "total_tokens": 28733088} {"current_steps": 28505, "total_steps": 42420, "loss": 0.3225, "lr": 0.00029353171957079453, "epoch": 13.43941537010844, "percentage": 67.2, "elapsed_time": "1:32:37", "remaining_time": "0:45:12", "throughput": 5171.56, "total_tokens": 28738656} {"current_steps": 28510, "total_steps": 42420, "loss": 0.3278, "lr": 0.00029334437546481166, "epoch": 13.441772748703443, "percentage": 67.21, "elapsed_time": "1:32:37", "remaining_time": "0:45:11", "throughput": 5171.56, "total_tokens": 28743040} {"current_steps": 28515, "total_steps": 42420, "loss": 0.3486, "lr": 0.00029315706634211724, "epoch": 13.444130127298443, "percentage": 67.22, "elapsed_time": "1:32:38", "remaining_time": "0:45:10", "throughput": 5171.66, "total_tokens": 28749216} {"current_steps": 28520, "total_steps": 42420, "loss": 0.3579, "lr": 0.0002929697922344194, "epoch": 13.446487505893446, "percentage": 67.23, "elapsed_time": "1:32:40", "remaining_time": "0:45:09", "throughput": 5171.76, "total_tokens": 28755360} {"current_steps": 28525, "total_steps": 42420, "loss": 0.2868, "lr": 0.0002927825531734206, "epoch": 13.448844884488448, "percentage": 67.24, "elapsed_time": "1:32:40", "remaining_time": "0:45:08", "throughput": 5171.76, "total_tokens": 28759872} {"current_steps": 28530, "total_steps": 42420, "loss": 0.3882, "lr": 0.0002925953491908172, "epoch": 13.45120226308345, "percentage": 67.26, "elapsed_time": "1:32:41", "remaining_time": "0:45:07", "throughput": 5171.78, "total_tokens": 28764448} {"current_steps": 28535, "total_steps": 42420, "loss": 0.3418, "lr": 0.0002924081803182998, "epoch": 13.453559641678453, "percentage": 67.27, "elapsed_time": "1:32:42", "remaining_time": "0:45:06", "throughput": 5171.81, "total_tokens": 28769280} {"current_steps": 28540, "total_steps": 42420, "loss": 0.3354, "lr": 0.00029222104658755256, "epoch": 13.455917020273455, "percentage": 67.28, "elapsed_time": "1:32:43", "remaining_time": "0:45:05", "throughput": 5171.85, "total_tokens": 28774080} {"current_steps": 28545, "total_steps": 42420, "loss": 0.3595, "lr": 0.0002920339480302544, "epoch": 13.458274398868458, "percentage": 67.29, "elapsed_time": "1:32:44", "remaining_time": "0:45:04", "throughput": 5171.93, "total_tokens": 28779552} {"current_steps": 28550, "total_steps": 42420, "loss": 0.3476, "lr": 0.00029184688467807755, "epoch": 13.46063177746346, "percentage": 67.3, "elapsed_time": "1:32:45", "remaining_time": "0:45:03", "throughput": 5172.02, "total_tokens": 28785600} {"current_steps": 28555, "total_steps": 42420, "loss": 0.3276, "lr": 0.0002916598565626892, "epoch": 13.462989156058462, "percentage": 67.31, "elapsed_time": "1:32:46", "remaining_time": "0:45:02", "throughput": 5172.07, "total_tokens": 28790432} {"current_steps": 28560, "total_steps": 42420, "loss": 0.3309, "lr": 0.00029147286371574956, "epoch": 13.465346534653465, "percentage": 67.33, "elapsed_time": "1:32:47", "remaining_time": "0:45:01", "throughput": 5172.13, "total_tokens": 28795680} {"current_steps": 28565, "total_steps": 42420, "loss": 0.3029, "lr": 0.00029128590616891336, "epoch": 13.467703913248467, "percentage": 67.34, "elapsed_time": "1:32:48", "remaining_time": "0:45:00", "throughput": 5172.17, "total_tokens": 28800672} {"current_steps": 28570, "total_steps": 42420, "loss": 0.3039, "lr": 0.0002910989839538296, "epoch": 13.47006129184347, "percentage": 67.35, "elapsed_time": "1:32:49", "remaining_time": "0:44:59", "throughput": 5172.18, "total_tokens": 28805280} {"current_steps": 28575, "total_steps": 42420, "loss": 0.355, "lr": 0.00029091209710214094, "epoch": 13.472418670438472, "percentage": 67.36, "elapsed_time": "1:32:50", "remaining_time": "0:44:58", "throughput": 5172.33, "total_tokens": 28812096} {"current_steps": 28580, "total_steps": 42420, "loss": 0.3548, "lr": 0.0002907252456454841, "epoch": 13.474776049033474, "percentage": 67.37, "elapsed_time": "1:32:51", "remaining_time": "0:44:57", "throughput": 5172.39, "total_tokens": 28817696} {"current_steps": 28585, "total_steps": 42420, "loss": 0.3979, "lr": 0.0002905384296154898, "epoch": 13.477133427628477, "percentage": 67.39, "elapsed_time": "1:32:52", "remaining_time": "0:44:57", "throughput": 5172.5, "total_tokens": 28823584} {"current_steps": 28590, "total_steps": 42420, "loss": 0.3, "lr": 0.000290351649043783, "epoch": 13.47949080622348, "percentage": 67.4, "elapsed_time": "1:32:53", "remaining_time": "0:44:56", "throughput": 5172.52, "total_tokens": 28828352} {"current_steps": 28595, "total_steps": 42420, "loss": 0.3667, "lr": 0.00029016490396198256, "epoch": 13.481848184818482, "percentage": 67.41, "elapsed_time": "1:32:54", "remaining_time": "0:44:54", "throughput": 5172.5, "total_tokens": 28832384} {"current_steps": 28600, "total_steps": 42420, "loss": 0.3478, "lr": 0.00028997819440170093, "epoch": 13.484205563413484, "percentage": 67.42, "elapsed_time": "1:32:54", "remaining_time": "0:44:53", "throughput": 5172.49, "total_tokens": 28836480} {"current_steps": 28605, "total_steps": 42420, "loss": 0.3453, "lr": 0.00028979152039454497, "epoch": 13.486562942008486, "percentage": 67.43, "elapsed_time": "1:32:55", "remaining_time": "0:44:52", "throughput": 5172.5, "total_tokens": 28841024} {"current_steps": 28610, "total_steps": 42420, "loss": 0.3348, "lr": 0.00028960488197211557, "epoch": 13.488920320603489, "percentage": 67.44, "elapsed_time": "1:32:56", "remaining_time": "0:44:51", "throughput": 5172.52, "total_tokens": 28845760} {"current_steps": 28615, "total_steps": 42420, "loss": 0.3283, "lr": 0.0002894182791660072, "epoch": 13.491277699198491, "percentage": 67.46, "elapsed_time": "1:32:57", "remaining_time": "0:44:50", "throughput": 5172.66, "total_tokens": 28852448} {"current_steps": 28620, "total_steps": 42420, "loss": 0.2886, "lr": 0.00028923171200780876, "epoch": 13.493635077793494, "percentage": 67.47, "elapsed_time": "1:32:58", "remaining_time": "0:44:49", "throughput": 5172.71, "total_tokens": 28857568} {"current_steps": 28625, "total_steps": 42420, "loss": 0.341, "lr": 0.0002890451805291028, "epoch": 13.495992456388496, "percentage": 67.48, "elapsed_time": "1:32:59", "remaining_time": "0:44:49", "throughput": 5172.8, "total_tokens": 28863232} {"current_steps": 28630, "total_steps": 42420, "loss": 0.3551, "lr": 0.000288858684761466, "epoch": 13.498349834983498, "percentage": 67.49, "elapsed_time": "1:33:00", "remaining_time": "0:44:47", "throughput": 5172.79, "total_tokens": 28867520} {"current_steps": 28635, "total_steps": 42420, "loss": 0.4344, "lr": 0.00028867222473646914, "epoch": 13.500707213578501, "percentage": 67.5, "elapsed_time": "1:33:01", "remaining_time": "0:44:46", "throughput": 5172.79, "total_tokens": 28871776} {"current_steps": 28640, "total_steps": 42420, "loss": 0.3887, "lr": 0.0002884858004856762, "epoch": 13.503064592173503, "percentage": 67.52, "elapsed_time": "1:33:02", "remaining_time": "0:44:45", "throughput": 5172.78, "total_tokens": 28876128} {"current_steps": 28645, "total_steps": 42420, "loss": 0.3514, "lr": 0.000288299412040646, "epoch": 13.505421970768506, "percentage": 67.53, "elapsed_time": "1:33:03", "remaining_time": "0:44:44", "throughput": 5172.78, "total_tokens": 28880192} {"current_steps": 28650, "total_steps": 42420, "loss": 0.3461, "lr": 0.00028811305943293085, "epoch": 13.507779349363508, "percentage": 67.54, "elapsed_time": "1:33:03", "remaining_time": "0:44:43", "throughput": 5172.78, "total_tokens": 28884768} {"current_steps": 28655, "total_steps": 42420, "loss": 0.3565, "lr": 0.00028792674269407717, "epoch": 13.51013672795851, "percentage": 67.55, "elapsed_time": "1:33:04", "remaining_time": "0:44:42", "throughput": 5172.83, "total_tokens": 28889856} {"current_steps": 28660, "total_steps": 42420, "loss": 0.3264, "lr": 0.000287740461855625, "epoch": 13.512494106553513, "percentage": 67.56, "elapsed_time": "1:33:05", "remaining_time": "0:44:41", "throughput": 5172.89, "total_tokens": 28895008} {"current_steps": 28665, "total_steps": 42420, "loss": 0.3294, "lr": 0.0002875542169491088, "epoch": 13.514851485148515, "percentage": 67.57, "elapsed_time": "1:33:06", "remaining_time": "0:44:40", "throughput": 5172.88, "total_tokens": 28899424} {"current_steps": 28670, "total_steps": 42420, "loss": 0.3004, "lr": 0.0002873680080060565, "epoch": 13.517208863743518, "percentage": 67.59, "elapsed_time": "1:33:07", "remaining_time": "0:44:39", "throughput": 5172.86, "total_tokens": 28903520} {"current_steps": 28675, "total_steps": 42420, "loss": 0.3806, "lr": 0.0002871818350579903, "epoch": 13.51956624233852, "percentage": 67.6, "elapsed_time": "1:33:08", "remaining_time": "0:44:38", "throughput": 5172.87, "total_tokens": 28908192} {"current_steps": 28680, "total_steps": 42420, "loss": 0.3385, "lr": 0.0002869956981364259, "epoch": 13.521923620933523, "percentage": 67.61, "elapsed_time": "1:33:09", "remaining_time": "0:44:37", "throughput": 5172.91, "total_tokens": 28913440} {"current_steps": 28685, "total_steps": 42420, "loss": 0.3856, "lr": 0.00028680959727287315, "epoch": 13.524280999528525, "percentage": 67.62, "elapsed_time": "1:33:10", "remaining_time": "0:44:36", "throughput": 5172.95, "total_tokens": 28918592} {"current_steps": 28690, "total_steps": 42420, "loss": 0.319, "lr": 0.0002866235324988358, "epoch": 13.526638378123527, "percentage": 67.63, "elapsed_time": "1:33:11", "remaining_time": "0:44:35", "throughput": 5172.96, "total_tokens": 28923008} {"current_steps": 28695, "total_steps": 42420, "loss": 0.3396, "lr": 0.0002864375038458115, "epoch": 13.52899575671853, "percentage": 67.64, "elapsed_time": "1:33:12", "remaining_time": "0:44:34", "throughput": 5173.01, "total_tokens": 28928128} {"current_steps": 28700, "total_steps": 42420, "loss": 0.3493, "lr": 0.0002862515113452917, "epoch": 13.531353135313532, "percentage": 67.66, "elapsed_time": "1:33:13", "remaining_time": "0:44:33", "throughput": 5173.04, "total_tokens": 28933056} {"current_steps": 28705, "total_steps": 42420, "loss": 0.3337, "lr": 0.0002860655550287617, "epoch": 13.533710513908535, "percentage": 67.67, "elapsed_time": "1:33:13", "remaining_time": "0:44:32", "throughput": 5173.06, "total_tokens": 28937760} {"current_steps": 28710, "total_steps": 42420, "loss": 0.3373, "lr": 0.00028587963492770097, "epoch": 13.536067892503535, "percentage": 67.68, "elapsed_time": "1:33:14", "remaining_time": "0:44:31", "throughput": 5173.04, "total_tokens": 28941760} {"current_steps": 28715, "total_steps": 42420, "loss": 0.3492, "lr": 0.0002856937510735826, "epoch": 13.53842527109854, "percentage": 67.69, "elapsed_time": "1:33:15", "remaining_time": "0:44:30", "throughput": 5173.11, "total_tokens": 28947200} {"current_steps": 28720, "total_steps": 42420, "loss": 0.3568, "lr": 0.0002855079034978733, "epoch": 13.54078264969354, "percentage": 67.7, "elapsed_time": "1:33:16", "remaining_time": "0:44:29", "throughput": 5173.09, "total_tokens": 28951488} {"current_steps": 28725, "total_steps": 42420, "loss": 0.3501, "lr": 0.0002853220922320341, "epoch": 13.543140028288542, "percentage": 67.72, "elapsed_time": "1:33:17", "remaining_time": "0:44:28", "throughput": 5173.15, "total_tokens": 28956864} {"current_steps": 28730, "total_steps": 42420, "loss": 0.3363, "lr": 0.0002851363173075196, "epoch": 13.545497406883545, "percentage": 67.73, "elapsed_time": "1:33:18", "remaining_time": "0:44:27", "throughput": 5173.18, "total_tokens": 28961632} {"current_steps": 28735, "total_steps": 42420, "loss": 0.3427, "lr": 0.0002849505787557784, "epoch": 13.547854785478547, "percentage": 67.74, "elapsed_time": "1:33:19", "remaining_time": "0:44:26", "throughput": 5173.11, "total_tokens": 28964864} {"current_steps": 28740, "total_steps": 42420, "loss": 0.3086, "lr": 0.0002847648766082529, "epoch": 13.55021216407355, "percentage": 67.75, "elapsed_time": "1:33:19", "remaining_time": "0:44:25", "throughput": 5173.09, "total_tokens": 28969024} {"current_steps": 28745, "total_steps": 42420, "loss": 0.302, "lr": 0.0002845792108963793, "epoch": 13.552569542668552, "percentage": 67.76, "elapsed_time": "1:33:20", "remaining_time": "0:44:24", "throughput": 5173.16, "total_tokens": 28974272} {"current_steps": 28750, "total_steps": 42420, "loss": 0.4032, "lr": 0.0002843935816515877, "epoch": 13.554926921263554, "percentage": 67.77, "elapsed_time": "1:33:21", "remaining_time": "0:44:23", "throughput": 5173.19, "total_tokens": 28979104} {"current_steps": 28755, "total_steps": 42420, "loss": 0.3197, "lr": 0.00028420798890530207, "epoch": 13.557284299858557, "percentage": 67.79, "elapsed_time": "1:33:23", "remaining_time": "0:44:22", "throughput": 5173.35, "total_tokens": 28986432} {"current_steps": 28760, "total_steps": 42420, "loss": 0.3383, "lr": 0.0002840224326889399, "epoch": 13.55964167845356, "percentage": 67.8, "elapsed_time": "1:33:23", "remaining_time": "0:44:21", "throughput": 5173.36, "total_tokens": 28991008} {"current_steps": 28765, "total_steps": 42420, "loss": 0.2752, "lr": 0.0002838369130339129, "epoch": 13.561999057048562, "percentage": 67.81, "elapsed_time": "1:33:24", "remaining_time": "0:44:20", "throughput": 5173.4, "total_tokens": 28996256} {"current_steps": 28770, "total_steps": 42420, "loss": 0.3219, "lr": 0.00028365142997162606, "epoch": 13.564356435643564, "percentage": 67.82, "elapsed_time": "1:33:25", "remaining_time": "0:44:19", "throughput": 5173.39, "total_tokens": 29000544} {"current_steps": 28775, "total_steps": 42420, "loss": 0.3904, "lr": 0.0002834659835334787, "epoch": 13.566713814238566, "percentage": 67.83, "elapsed_time": "1:33:26", "remaining_time": "0:44:18", "throughput": 5173.45, "total_tokens": 29005664} {"current_steps": 28780, "total_steps": 42420, "loss": 0.3248, "lr": 0.0002832805737508639, "epoch": 13.569071192833569, "percentage": 67.85, "elapsed_time": "1:33:27", "remaining_time": "0:44:17", "throughput": 5173.49, "total_tokens": 29010816} {"current_steps": 28785, "total_steps": 42420, "loss": 0.3606, "lr": 0.00028309520065516823, "epoch": 13.571428571428571, "percentage": 67.86, "elapsed_time": "1:33:28", "remaining_time": "0:44:16", "throughput": 5173.63, "total_tokens": 29018208} {"current_steps": 28790, "total_steps": 42420, "loss": 0.3149, "lr": 0.0002829098642777723, "epoch": 13.573785950023574, "percentage": 67.87, "elapsed_time": "1:33:29", "remaining_time": "0:44:15", "throughput": 5173.67, "total_tokens": 29023488} {"current_steps": 28795, "total_steps": 42420, "loss": 0.3614, "lr": 0.00028272456465005033, "epoch": 13.576143328618576, "percentage": 67.88, "elapsed_time": "1:33:30", "remaining_time": "0:44:14", "throughput": 5173.74, "total_tokens": 29029024} {"current_steps": 28800, "total_steps": 42420, "loss": 0.3449, "lr": 0.0002825393018033704, "epoch": 13.578500707213578, "percentage": 67.89, "elapsed_time": "1:33:31", "remaining_time": "0:44:13", "throughput": 5173.78, "total_tokens": 29034240} {"current_steps": 28805, "total_steps": 42420, "loss": 0.3296, "lr": 0.0002823540757690947, "epoch": 13.58085808580858, "percentage": 67.9, "elapsed_time": "1:33:32", "remaining_time": "0:44:12", "throughput": 5173.8, "total_tokens": 29039136} {"current_steps": 28810, "total_steps": 42420, "loss": 0.3084, "lr": 0.0002821688865785782, "epoch": 13.583215464403583, "percentage": 67.92, "elapsed_time": "1:33:33", "remaining_time": "0:44:11", "throughput": 5173.82, "total_tokens": 29043808} {"current_steps": 28815, "total_steps": 42420, "loss": 0.3384, "lr": 0.00028198373426317075, "epoch": 13.585572842998586, "percentage": 67.93, "elapsed_time": "1:33:34", "remaining_time": "0:44:10", "throughput": 5173.83, "total_tokens": 29048544} {"current_steps": 28820, "total_steps": 42420, "loss": 0.361, "lr": 0.00028179861885421536, "epoch": 13.587930221593588, "percentage": 67.94, "elapsed_time": "1:33:35", "remaining_time": "0:44:09", "throughput": 5173.9, "total_tokens": 29053824} {"current_steps": 28825, "total_steps": 42420, "loss": 0.3073, "lr": 0.00028161354038304897, "epoch": 13.59028760018859, "percentage": 67.95, "elapsed_time": "1:33:36", "remaining_time": "0:44:08", "throughput": 5173.94, "total_tokens": 29058688} {"current_steps": 28830, "total_steps": 42420, "loss": 0.3112, "lr": 0.00028142849888100217, "epoch": 13.592644978783593, "percentage": 67.96, "elapsed_time": "1:33:37", "remaining_time": "0:44:07", "throughput": 5173.94, "total_tokens": 29063040} {"current_steps": 28835, "total_steps": 42420, "loss": 0.3095, "lr": 0.00028124349437939943, "epoch": 13.595002357378595, "percentage": 67.98, "elapsed_time": "1:33:38", "remaining_time": "0:44:06", "throughput": 5173.92, "total_tokens": 29067168} {"current_steps": 28840, "total_steps": 42420, "loss": 0.3633, "lr": 0.0002810585269095588, "epoch": 13.597359735973598, "percentage": 67.99, "elapsed_time": "1:33:38", "remaining_time": "0:44:05", "throughput": 5173.95, "total_tokens": 29071744} {"current_steps": 28845, "total_steps": 42420, "loss": 0.3157, "lr": 0.0002808735965027923, "epoch": 13.5997171145686, "percentage": 68.0, "elapsed_time": "1:33:39", "remaining_time": "0:44:04", "throughput": 5173.9, "total_tokens": 29075296} {"current_steps": 28850, "total_steps": 42420, "loss": 0.3342, "lr": 0.0002806887031904052, "epoch": 13.602074493163602, "percentage": 68.01, "elapsed_time": "1:33:40", "remaining_time": "0:44:03", "throughput": 5173.94, "total_tokens": 29080256} {"current_steps": 28855, "total_steps": 42420, "loss": 0.3647, "lr": 0.0002805038470036969, "epoch": 13.604431871758605, "percentage": 68.02, "elapsed_time": "1:33:41", "remaining_time": "0:44:02", "throughput": 5173.98, "total_tokens": 29085184} {"current_steps": 28860, "total_steps": 42420, "loss": 0.3374, "lr": 0.0002803190279739605, "epoch": 13.606789250353607, "percentage": 68.03, "elapsed_time": "1:33:42", "remaining_time": "0:44:01", "throughput": 5173.95, "total_tokens": 29089280} {"current_steps": 28865, "total_steps": 42420, "loss": 0.3346, "lr": 0.00028013424613248275, "epoch": 13.60914662894861, "percentage": 68.05, "elapsed_time": "1:33:43", "remaining_time": "0:44:00", "throughput": 5174.0, "total_tokens": 29094304} {"current_steps": 28870, "total_steps": 42420, "loss": 0.3231, "lr": 0.000279949501510544, "epoch": 13.611504007543612, "percentage": 68.06, "elapsed_time": "1:33:44", "remaining_time": "0:43:59", "throughput": 5174.02, "total_tokens": 29099008} {"current_steps": 28875, "total_steps": 42420, "loss": 0.3516, "lr": 0.00027976479413941834, "epoch": 13.613861386138614, "percentage": 68.07, "elapsed_time": "1:33:45", "remaining_time": "0:43:58", "throughput": 5174.12, "total_tokens": 29105024} {"current_steps": 28880, "total_steps": 42420, "loss": 0.3454, "lr": 0.0002795801240503737, "epoch": 13.616218764733617, "percentage": 68.08, "elapsed_time": "1:33:46", "remaining_time": "0:43:57", "throughput": 5174.19, "total_tokens": 29110496} {"current_steps": 28885, "total_steps": 42420, "loss": 0.3541, "lr": 0.00027939549127467165, "epoch": 13.61857614332862, "percentage": 68.09, "elapsed_time": "1:33:47", "remaining_time": "0:43:56", "throughput": 5174.25, "total_tokens": 29115712} {"current_steps": 28890, "total_steps": 42420, "loss": 0.3412, "lr": 0.0002792108958435672, "epoch": 13.620933521923622, "percentage": 68.1, "elapsed_time": "1:33:47", "remaining_time": "0:43:55", "throughput": 5174.27, "total_tokens": 29120320} {"current_steps": 28895, "total_steps": 42420, "loss": 0.3489, "lr": 0.0002790263377883093, "epoch": 13.623290900518624, "percentage": 68.12, "elapsed_time": "1:33:48", "remaining_time": "0:43:54", "throughput": 5174.31, "total_tokens": 29125312} {"current_steps": 28900, "total_steps": 42420, "loss": 0.3264, "lr": 0.00027884181714014043, "epoch": 13.625648279113626, "percentage": 68.13, "elapsed_time": "1:33:49", "remaining_time": "0:43:53", "throughput": 5174.33, "total_tokens": 29130176} {"current_steps": 28905, "total_steps": 42420, "loss": 0.349, "lr": 0.0002786573339302969, "epoch": 13.628005657708629, "percentage": 68.14, "elapsed_time": "1:33:50", "remaining_time": "0:43:52", "throughput": 5174.39, "total_tokens": 29135200} {"current_steps": 28910, "total_steps": 42420, "loss": 0.3425, "lr": 0.0002784728881900085, "epoch": 13.630363036303631, "percentage": 68.15, "elapsed_time": "1:33:51", "remaining_time": "0:43:51", "throughput": 5174.44, "total_tokens": 29140384} {"current_steps": 28915, "total_steps": 42420, "loss": 0.3324, "lr": 0.0002782884799504989, "epoch": 13.632720414898632, "percentage": 68.16, "elapsed_time": "1:33:52", "remaining_time": "0:43:50", "throughput": 5174.48, "total_tokens": 29145184} {"current_steps": 28920, "total_steps": 42420, "loss": 0.3395, "lr": 0.00027810410924298515, "epoch": 13.635077793493634, "percentage": 68.18, "elapsed_time": "1:33:53", "remaining_time": "0:43:49", "throughput": 5174.53, "total_tokens": 29150016} {"current_steps": 28925, "total_steps": 42420, "loss": 0.3557, "lr": 0.0002779197760986782, "epoch": 13.637435172088637, "percentage": 68.19, "elapsed_time": "1:33:54", "remaining_time": "0:43:48", "throughput": 5174.6, "total_tokens": 29155328} {"current_steps": 28930, "total_steps": 42420, "loss": 0.3273, "lr": 0.00027773548054878244, "epoch": 13.639792550683639, "percentage": 68.2, "elapsed_time": "1:33:55", "remaining_time": "0:43:47", "throughput": 5174.66, "total_tokens": 29160768} {"current_steps": 28935, "total_steps": 42420, "loss": 0.3552, "lr": 0.00027755122262449595, "epoch": 13.642149929278641, "percentage": 68.21, "elapsed_time": "1:33:56", "remaining_time": "0:43:46", "throughput": 5174.66, "total_tokens": 29165056} {"current_steps": 28940, "total_steps": 42420, "loss": 0.3283, "lr": 0.0002773670023570104, "epoch": 13.644507307873644, "percentage": 68.22, "elapsed_time": "1:33:56", "remaining_time": "0:43:45", "throughput": 5174.67, "total_tokens": 29169472} {"current_steps": 28945, "total_steps": 42420, "loss": 0.3273, "lr": 0.00027718281977751134, "epoch": 13.646864686468646, "percentage": 68.23, "elapsed_time": "1:33:57", "remaining_time": "0:43:44", "throughput": 5174.75, "total_tokens": 29174848} {"current_steps": 28950, "total_steps": 42420, "loss": 0.3439, "lr": 0.0002769986749171777, "epoch": 13.649222065063649, "percentage": 68.25, "elapsed_time": "1:33:58", "remaining_time": "0:43:43", "throughput": 5174.78, "total_tokens": 29179840} {"current_steps": 28955, "total_steps": 42420, "loss": 0.3381, "lr": 0.000276814567807182, "epoch": 13.651579443658651, "percentage": 68.26, "elapsed_time": "1:33:59", "remaining_time": "0:43:42", "throughput": 5174.82, "total_tokens": 29185120} {"current_steps": 28960, "total_steps": 42420, "loss": 0.3266, "lr": 0.0002766304984786905, "epoch": 13.653936822253653, "percentage": 68.27, "elapsed_time": "1:34:00", "remaining_time": "0:43:41", "throughput": 5174.84, "total_tokens": 29189664} {"current_steps": 28965, "total_steps": 42420, "loss": 0.3362, "lr": 0.00027644646696286325, "epoch": 13.656294200848656, "percentage": 68.28, "elapsed_time": "1:34:01", "remaining_time": "0:43:40", "throughput": 5174.86, "total_tokens": 29194208} {"current_steps": 28970, "total_steps": 42420, "loss": 0.3007, "lr": 0.00027626247329085317, "epoch": 13.658651579443658, "percentage": 68.29, "elapsed_time": "1:34:02", "remaining_time": "0:43:39", "throughput": 5174.93, "total_tokens": 29199552} {"current_steps": 28975, "total_steps": 42420, "loss": 0.3413, "lr": 0.00027607851749380754, "epoch": 13.66100895803866, "percentage": 68.31, "elapsed_time": "1:34:03", "remaining_time": "0:43:38", "throughput": 5174.93, "total_tokens": 29204096} {"current_steps": 28980, "total_steps": 42420, "loss": 0.3757, "lr": 0.00027589459960286687, "epoch": 13.663366336633663, "percentage": 68.32, "elapsed_time": "1:34:04", "remaining_time": "0:43:37", "throughput": 5174.91, "total_tokens": 29208192} {"current_steps": 28985, "total_steps": 42420, "loss": 0.3296, "lr": 0.00027571071964916565, "epoch": 13.665723715228665, "percentage": 68.33, "elapsed_time": "1:34:05", "remaining_time": "0:43:36", "throughput": 5174.95, "total_tokens": 29213056} {"current_steps": 28990, "total_steps": 42420, "loss": 0.3475, "lr": 0.0002755268776638312, "epoch": 13.668081093823668, "percentage": 68.34, "elapsed_time": "1:34:05", "remaining_time": "0:43:35", "throughput": 5174.95, "total_tokens": 29217696} {"current_steps": 28995, "total_steps": 42420, "loss": 0.3418, "lr": 0.000275343073677985, "epoch": 13.67043847241867, "percentage": 68.35, "elapsed_time": "1:34:06", "remaining_time": "0:43:34", "throughput": 5174.98, "total_tokens": 29222528} {"current_steps": 29000, "total_steps": 42420, "loss": 0.3287, "lr": 0.00027515930772274195, "epoch": 13.672795851013673, "percentage": 68.36, "elapsed_time": "1:34:07", "remaining_time": "0:43:33", "throughput": 5175.02, "total_tokens": 29227200} {"current_steps": 29005, "total_steps": 42420, "loss": 0.3685, "lr": 0.0002749755798292105, "epoch": 13.675153229608675, "percentage": 68.38, "elapsed_time": "1:34:08", "remaining_time": "0:43:32", "throughput": 5175.1, "total_tokens": 29232384} {"current_steps": 29010, "total_steps": 42420, "loss": 0.3246, "lr": 0.00027479189002849283, "epoch": 13.677510608203677, "percentage": 68.39, "elapsed_time": "1:34:09", "remaining_time": "0:43:31", "throughput": 5175.12, "total_tokens": 29237056} {"current_steps": 29015, "total_steps": 42420, "loss": 0.3693, "lr": 0.0002746082383516844, "epoch": 13.67986798679868, "percentage": 68.4, "elapsed_time": "1:34:10", "remaining_time": "0:43:30", "throughput": 5175.09, "total_tokens": 29241024} {"current_steps": 29020, "total_steps": 42420, "loss": 0.3182, "lr": 0.0002744246248298742, "epoch": 13.682225365393682, "percentage": 68.41, "elapsed_time": "1:34:11", "remaining_time": "0:43:29", "throughput": 5175.13, "total_tokens": 29246048} {"current_steps": 29025, "total_steps": 42420, "loss": 0.3542, "lr": 0.00027424104949414483, "epoch": 13.684582743988685, "percentage": 68.42, "elapsed_time": "1:34:12", "remaining_time": "0:43:28", "throughput": 5175.12, "total_tokens": 29250144} {"current_steps": 29030, "total_steps": 42420, "loss": 0.3421, "lr": 0.00027405751237557276, "epoch": 13.686940122583687, "percentage": 68.43, "elapsed_time": "1:34:12", "remaining_time": "0:43:27", "throughput": 5175.14, "total_tokens": 29254752} {"current_steps": 29035, "total_steps": 42420, "loss": 0.3513, "lr": 0.00027387401350522744, "epoch": 13.68929750117869, "percentage": 68.45, "elapsed_time": "1:34:13", "remaining_time": "0:43:26", "throughput": 5175.2, "total_tokens": 29260160} {"current_steps": 29040, "total_steps": 42420, "loss": 0.3516, "lr": 0.00027369055291417234, "epoch": 13.691654879773692, "percentage": 68.46, "elapsed_time": "1:34:14", "remaining_time": "0:43:25", "throughput": 5175.33, "total_tokens": 29266400} {"current_steps": 29045, "total_steps": 42420, "loss": 0.2965, "lr": 0.000273507130633464, "epoch": 13.694012258368694, "percentage": 68.47, "elapsed_time": "1:34:15", "remaining_time": "0:43:24", "throughput": 5175.37, "total_tokens": 29271296} {"current_steps": 29050, "total_steps": 42420, "loss": 0.3078, "lr": 0.00027332374669415286, "epoch": 13.696369636963697, "percentage": 68.48, "elapsed_time": "1:34:17", "remaining_time": "0:43:23", "throughput": 5175.47, "total_tokens": 29277696} {"current_steps": 29055, "total_steps": 42420, "loss": 0.3657, "lr": 0.00027314040112728287, "epoch": 13.698727015558699, "percentage": 68.49, "elapsed_time": "1:34:17", "remaining_time": "0:43:22", "throughput": 5175.51, "total_tokens": 29282944} {"current_steps": 29060, "total_steps": 42420, "loss": 0.3183, "lr": 0.0002729570939638909, "epoch": 13.701084394153701, "percentage": 68.51, "elapsed_time": "1:34:18", "remaining_time": "0:43:21", "throughput": 5175.57, "total_tokens": 29287968} {"current_steps": 29065, "total_steps": 42420, "loss": 0.3249, "lr": 0.00027277382523500807, "epoch": 13.703441772748704, "percentage": 68.52, "elapsed_time": "1:34:19", "remaining_time": "0:43:20", "throughput": 5175.66, "total_tokens": 29293472} {"current_steps": 29070, "total_steps": 42420, "loss": 0.2605, "lr": 0.0002725905949716584, "epoch": 13.705799151343706, "percentage": 68.53, "elapsed_time": "1:34:20", "remaining_time": "0:43:19", "throughput": 5175.68, "total_tokens": 29298144} {"current_steps": 29075, "total_steps": 42420, "loss": 0.2896, "lr": 0.00027240740320485986, "epoch": 13.708156529938709, "percentage": 68.54, "elapsed_time": "1:34:21", "remaining_time": "0:43:18", "throughput": 5175.7, "total_tokens": 29302848} {"current_steps": 29080, "total_steps": 42420, "loss": 0.3074, "lr": 0.0002722242499656236, "epoch": 13.710513908533711, "percentage": 68.55, "elapsed_time": "1:34:22", "remaining_time": "0:43:17", "throughput": 5175.78, "total_tokens": 29308416} {"current_steps": 29085, "total_steps": 42420, "loss": 0.4695, "lr": 0.00027204113528495446, "epoch": 13.712871287128714, "percentage": 68.56, "elapsed_time": "1:34:23", "remaining_time": "0:43:16", "throughput": 5175.75, "total_tokens": 29312640} {"current_steps": 29090, "total_steps": 42420, "loss": 0.3272, "lr": 0.00027185805919385065, "epoch": 13.715228665723716, "percentage": 68.58, "elapsed_time": "1:34:24", "remaining_time": "0:43:15", "throughput": 5175.78, "total_tokens": 29317440} {"current_steps": 29095, "total_steps": 42420, "loss": 0.3153, "lr": 0.00027167502172330396, "epoch": 13.717586044318718, "percentage": 68.59, "elapsed_time": "1:34:25", "remaining_time": "0:43:14", "throughput": 5175.85, "total_tokens": 29322784} {"current_steps": 29100, "total_steps": 42420, "loss": 0.3481, "lr": 0.0002714920229042991, "epoch": 13.71994342291372, "percentage": 68.6, "elapsed_time": "1:34:26", "remaining_time": "0:43:13", "throughput": 5176.04, "total_tokens": 29329984} {"current_steps": 29105, "total_steps": 42420, "loss": 0.2987, "lr": 0.00027130906276781494, "epoch": 13.722300801508723, "percentage": 68.61, "elapsed_time": "1:34:27", "remaining_time": "0:43:12", "throughput": 5176.07, "total_tokens": 29335008} {"current_steps": 29110, "total_steps": 42420, "loss": 0.3002, "lr": 0.0002711261413448235, "epoch": 13.724658180103724, "percentage": 68.62, "elapsed_time": "1:34:28", "remaining_time": "0:43:11", "throughput": 5176.12, "total_tokens": 29340320} {"current_steps": 29115, "total_steps": 42420, "loss": 0.2571, "lr": 0.00027094325866629026, "epoch": 13.727015558698728, "percentage": 68.64, "elapsed_time": "1:34:29", "remaining_time": "0:43:10", "throughput": 5176.13, "total_tokens": 29345056} {"current_steps": 29120, "total_steps": 42420, "loss": 0.3798, "lr": 0.00027076041476317415, "epoch": 13.729372937293729, "percentage": 68.65, "elapsed_time": "1:34:30", "remaining_time": "0:43:09", "throughput": 5176.11, "total_tokens": 29349248} {"current_steps": 29125, "total_steps": 42420, "loss": 0.3981, "lr": 0.0002705776096664276, "epoch": 13.731730315888731, "percentage": 68.66, "elapsed_time": "1:34:30", "remaining_time": "0:43:08", "throughput": 5176.09, "total_tokens": 29353376} {"current_steps": 29130, "total_steps": 42420, "loss": 0.3763, "lr": 0.0002703948434069964, "epoch": 13.734087694483733, "percentage": 68.67, "elapsed_time": "1:34:31", "remaining_time": "0:43:07", "throughput": 5176.16, "total_tokens": 29358656} {"current_steps": 29135, "total_steps": 42420, "loss": 0.33, "lr": 0.00027021211601581987, "epoch": 13.736445073078736, "percentage": 68.68, "elapsed_time": "1:34:32", "remaining_time": "0:43:06", "throughput": 5176.18, "total_tokens": 29363328} {"current_steps": 29140, "total_steps": 42420, "loss": 0.3166, "lr": 0.00027002942752383033, "epoch": 13.738802451673738, "percentage": 68.69, "elapsed_time": "1:34:33", "remaining_time": "0:43:05", "throughput": 5176.31, "total_tokens": 29370016} {"current_steps": 29145, "total_steps": 42420, "loss": 0.3473, "lr": 0.00026984677796195404, "epoch": 13.74115983026874, "percentage": 68.71, "elapsed_time": "1:34:34", "remaining_time": "0:43:04", "throughput": 5176.35, "total_tokens": 29375392} {"current_steps": 29150, "total_steps": 42420, "loss": 0.3197, "lr": 0.0002696641673611105, "epoch": 13.743517208863743, "percentage": 68.72, "elapsed_time": "1:34:35", "remaining_time": "0:43:03", "throughput": 5176.44, "total_tokens": 29380960} {"current_steps": 29155, "total_steps": 42420, "loss": 0.3157, "lr": 0.0002694815957522125, "epoch": 13.745874587458745, "percentage": 68.73, "elapsed_time": "1:34:36", "remaining_time": "0:43:02", "throughput": 5176.5, "total_tokens": 29386432} {"current_steps": 29160, "total_steps": 42420, "loss": 0.3687, "lr": 0.0002692990631661664, "epoch": 13.748231966053748, "percentage": 68.74, "elapsed_time": "1:34:38", "remaining_time": "0:43:02", "throughput": 5176.64, "total_tokens": 29393632} {"current_steps": 29165, "total_steps": 42420, "loss": 0.3075, "lr": 0.00026911656963387177, "epoch": 13.75058934464875, "percentage": 68.75, "elapsed_time": "1:34:39", "remaining_time": "0:43:01", "throughput": 5176.68, "total_tokens": 29398720} {"current_steps": 29170, "total_steps": 42420, "loss": 0.3455, "lr": 0.00026893411518622175, "epoch": 13.752946723243753, "percentage": 68.76, "elapsed_time": "1:34:40", "remaining_time": "0:43:00", "throughput": 5176.81, "total_tokens": 29405344} {"current_steps": 29175, "total_steps": 42420, "loss": 0.3207, "lr": 0.00026875169985410296, "epoch": 13.755304101838755, "percentage": 68.78, "elapsed_time": "1:34:40", "remaining_time": "0:42:59", "throughput": 5176.74, "total_tokens": 29408672} {"current_steps": 29180, "total_steps": 42420, "loss": 0.368, "lr": 0.0002685693236683948, "epoch": 13.757661480433757, "percentage": 68.79, "elapsed_time": "1:34:41", "remaining_time": "0:42:58", "throughput": 5176.79, "total_tokens": 29413792} {"current_steps": 29185, "total_steps": 42420, "loss": 0.2917, "lr": 0.0002683869866599707, "epoch": 13.76001885902876, "percentage": 68.8, "elapsed_time": "1:34:42", "remaining_time": "0:42:57", "throughput": 5176.8, "total_tokens": 29418528} {"current_steps": 29190, "total_steps": 42420, "loss": 0.2771, "lr": 0.00026820468885969717, "epoch": 13.762376237623762, "percentage": 68.81, "elapsed_time": "1:34:43", "remaining_time": "0:42:56", "throughput": 5176.85, "total_tokens": 29423584} {"current_steps": 29195, "total_steps": 42420, "loss": 0.373, "lr": 0.0002680224302984342, "epoch": 13.764733616218765, "percentage": 68.82, "elapsed_time": "1:34:44", "remaining_time": "0:42:55", "throughput": 5176.91, "total_tokens": 29428896} {"current_steps": 29200, "total_steps": 42420, "loss": 0.2687, "lr": 0.0002678402110070349, "epoch": 13.767090994813767, "percentage": 68.84, "elapsed_time": "1:34:45", "remaining_time": "0:42:54", "throughput": 5176.98, "total_tokens": 29434240} {"current_steps": 29205, "total_steps": 42420, "loss": 0.433, "lr": 0.00026765803101634617, "epoch": 13.76944837340877, "percentage": 68.85, "elapsed_time": "1:34:46", "remaining_time": "0:42:53", "throughput": 5177.0, "total_tokens": 29438720} {"current_steps": 29210, "total_steps": 42420, "loss": 0.333, "lr": 0.000267475890357208, "epoch": 13.771805752003772, "percentage": 68.86, "elapsed_time": "1:34:47", "remaining_time": "0:42:52", "throughput": 5177.0, "total_tokens": 29443072} {"current_steps": 29215, "total_steps": 42420, "loss": 0.3475, "lr": 0.0002672937890604533, "epoch": 13.774163130598774, "percentage": 68.87, "elapsed_time": "1:34:48", "remaining_time": "0:42:51", "throughput": 5177.11, "total_tokens": 29449344} {"current_steps": 29220, "total_steps": 42420, "loss": 0.3268, "lr": 0.00026711172715690907, "epoch": 13.776520509193777, "percentage": 68.88, "elapsed_time": "1:34:49", "remaining_time": "0:42:50", "throughput": 5177.13, "total_tokens": 29453952} {"current_steps": 29225, "total_steps": 42420, "loss": 0.3038, "lr": 0.0002669297046773954, "epoch": 13.778877887788779, "percentage": 68.89, "elapsed_time": "1:34:50", "remaining_time": "0:42:49", "throughput": 5177.08, "total_tokens": 29457664} {"current_steps": 29230, "total_steps": 42420, "loss": 0.2979, "lr": 0.0002667477216527252, "epoch": 13.781235266383781, "percentage": 68.91, "elapsed_time": "1:34:50", "remaining_time": "0:42:48", "throughput": 5177.14, "total_tokens": 29463072} {"current_steps": 29235, "total_steps": 42420, "loss": 0.3542, "lr": 0.0002665657781137053, "epoch": 13.783592644978784, "percentage": 68.92, "elapsed_time": "1:34:51", "remaining_time": "0:42:47", "throughput": 5177.19, "total_tokens": 29468192} {"current_steps": 29240, "total_steps": 42420, "loss": 0.3547, "lr": 0.00026638387409113576, "epoch": 13.785950023573786, "percentage": 68.93, "elapsed_time": "1:34:52", "remaining_time": "0:42:46", "throughput": 5177.23, "total_tokens": 29473152} {"current_steps": 29245, "total_steps": 42420, "loss": 0.3451, "lr": 0.00026620200961580966, "epoch": 13.788307402168789, "percentage": 68.94, "elapsed_time": "1:34:53", "remaining_time": "0:42:45", "throughput": 5177.22, "total_tokens": 29477344} {"current_steps": 29250, "total_steps": 42420, "loss": 0.3212, "lr": 0.0002660201847185137, "epoch": 13.790664780763791, "percentage": 68.95, "elapsed_time": "1:34:54", "remaining_time": "0:42:44", "throughput": 5177.22, "total_tokens": 29481856} {"current_steps": 29255, "total_steps": 42420, "loss": 0.3738, "lr": 0.00026583839943002775, "epoch": 13.793022159358793, "percentage": 68.97, "elapsed_time": "1:34:55", "remaining_time": "0:42:43", "throughput": 5177.26, "total_tokens": 29486848} {"current_steps": 29260, "total_steps": 42420, "loss": 0.3456, "lr": 0.0002656566537811249, "epoch": 13.795379537953796, "percentage": 68.98, "elapsed_time": "1:34:56", "remaining_time": "0:42:42", "throughput": 5177.33, "total_tokens": 29492224} {"current_steps": 29265, "total_steps": 42420, "loss": 0.3092, "lr": 0.00026547494780257186, "epoch": 13.797736916548798, "percentage": 68.99, "elapsed_time": "1:34:57", "remaining_time": "0:42:41", "throughput": 5177.42, "total_tokens": 29497632} {"current_steps": 29270, "total_steps": 42420, "loss": 0.3482, "lr": 0.0002652932815251279, "epoch": 13.8000942951438, "percentage": 69.0, "elapsed_time": "1:34:58", "remaining_time": "0:42:40", "throughput": 5177.45, "total_tokens": 29502176} {"current_steps": 29275, "total_steps": 42420, "loss": 0.3612, "lr": 0.00026511165497954635, "epoch": 13.802451673738803, "percentage": 69.01, "elapsed_time": "1:34:59", "remaining_time": "0:42:38", "throughput": 5177.46, "total_tokens": 29506784} {"current_steps": 29280, "total_steps": 42420, "loss": 0.3186, "lr": 0.0002649300681965734, "epoch": 13.804809052333805, "percentage": 69.02, "elapsed_time": "1:35:00", "remaining_time": "0:42:38", "throughput": 5177.53, "total_tokens": 29512064} {"current_steps": 29285, "total_steps": 42420, "loss": 0.3515, "lr": 0.0002647485212069487, "epoch": 13.807166430928808, "percentage": 69.04, "elapsed_time": "1:35:00", "remaining_time": "0:42:36", "throughput": 5177.51, "total_tokens": 29516192} {"current_steps": 29290, "total_steps": 42420, "loss": 0.349, "lr": 0.0002645670140414049, "epoch": 13.80952380952381, "percentage": 69.05, "elapsed_time": "1:35:01", "remaining_time": "0:42:35", "throughput": 5177.52, "total_tokens": 29520576} {"current_steps": 29295, "total_steps": 42420, "loss": 0.3563, "lr": 0.00026438554673066834, "epoch": 13.811881188118813, "percentage": 69.06, "elapsed_time": "1:35:02", "remaining_time": "0:42:34", "throughput": 5177.52, "total_tokens": 29524768} {"current_steps": 29300, "total_steps": 42420, "loss": 0.315, "lr": 0.0002642041193054582, "epoch": 13.814238566713815, "percentage": 69.07, "elapsed_time": "1:35:03", "remaining_time": "0:42:33", "throughput": 5177.55, "total_tokens": 29529664} {"current_steps": 29305, "total_steps": 42420, "loss": 0.2853, "lr": 0.00026402273179648716, "epoch": 13.816595945308817, "percentage": 69.08, "elapsed_time": "1:35:04", "remaining_time": "0:42:32", "throughput": 5177.58, "total_tokens": 29534176} {"current_steps": 29310, "total_steps": 42420, "loss": 0.3612, "lr": 0.00026384138423446083, "epoch": 13.81895332390382, "percentage": 69.09, "elapsed_time": "1:35:05", "remaining_time": "0:42:31", "throughput": 5177.7, "total_tokens": 29540608} {"current_steps": 29315, "total_steps": 42420, "loss": 0.3302, "lr": 0.0002636600766500784, "epoch": 13.82131070249882, "percentage": 69.11, "elapsed_time": "1:35:06", "remaining_time": "0:42:30", "throughput": 5177.74, "total_tokens": 29545472} {"current_steps": 29320, "total_steps": 42420, "loss": 0.3064, "lr": 0.00026347880907403215, "epoch": 13.823668081093825, "percentage": 69.12, "elapsed_time": "1:35:07", "remaining_time": "0:42:29", "throughput": 5177.83, "total_tokens": 29550976} {"current_steps": 29325, "total_steps": 42420, "loss": 0.3506, "lr": 0.0002632975815370077, "epoch": 13.826025459688825, "percentage": 69.13, "elapsed_time": "1:35:08", "remaining_time": "0:42:28", "throughput": 5177.86, "total_tokens": 29555808} {"current_steps": 29330, "total_steps": 42420, "loss": 0.3395, "lr": 0.00026311639406968355, "epoch": 13.828382838283828, "percentage": 69.14, "elapsed_time": "1:35:08", "remaining_time": "0:42:27", "throughput": 5177.89, "total_tokens": 29560384} {"current_steps": 29335, "total_steps": 42420, "loss": 0.335, "lr": 0.000262935246702732, "epoch": 13.83074021687883, "percentage": 69.15, "elapsed_time": "1:35:09", "remaining_time": "0:42:26", "throughput": 5177.89, "total_tokens": 29564896} {"current_steps": 29340, "total_steps": 42420, "loss": 0.3448, "lr": 0.000262754139466818, "epoch": 13.833097595473832, "percentage": 69.17, "elapsed_time": "1:35:10", "remaining_time": "0:42:25", "throughput": 5177.92, "total_tokens": 29569536} {"current_steps": 29345, "total_steps": 42420, "loss": 0.3644, "lr": 0.00026257307239260016, "epoch": 13.835454974068835, "percentage": 69.18, "elapsed_time": "1:35:11", "remaining_time": "0:42:24", "throughput": 5177.97, "total_tokens": 29574752} {"current_steps": 29350, "total_steps": 42420, "loss": 0.323, "lr": 0.0002623920455107297, "epoch": 13.837812352663837, "percentage": 69.19, "elapsed_time": "1:35:12", "remaining_time": "0:42:23", "throughput": 5178.02, "total_tokens": 29579808} {"current_steps": 29355, "total_steps": 42420, "loss": 0.3072, "lr": 0.0002622110588518516, "epoch": 13.84016973125884, "percentage": 69.2, "elapsed_time": "1:35:13", "remaining_time": "0:42:22", "throughput": 5178.12, "total_tokens": 29585664} {"current_steps": 29360, "total_steps": 42420, "loss": 0.3065, "lr": 0.0002620301124466038, "epoch": 13.842527109853842, "percentage": 69.21, "elapsed_time": "1:35:14", "remaining_time": "0:42:21", "throughput": 5178.2, "total_tokens": 29591008} {"current_steps": 29365, "total_steps": 42420, "loss": 0.3694, "lr": 0.0002618492063256174, "epoch": 13.844884488448844, "percentage": 69.22, "elapsed_time": "1:35:15", "remaining_time": "0:42:20", "throughput": 5178.2, "total_tokens": 29595296} {"current_steps": 29370, "total_steps": 42420, "loss": 0.2741, "lr": 0.00026166834051951683, "epoch": 13.847241867043847, "percentage": 69.24, "elapsed_time": "1:35:16", "remaining_time": "0:42:19", "throughput": 5178.23, "total_tokens": 29600288} {"current_steps": 29375, "total_steps": 42420, "loss": 0.285, "lr": 0.00026148751505891964, "epoch": 13.84959924563885, "percentage": 69.25, "elapsed_time": "1:35:17", "remaining_time": "0:42:18", "throughput": 5178.24, "total_tokens": 29604672} {"current_steps": 29380, "total_steps": 42420, "loss": 0.3501, "lr": 0.00026130672997443636, "epoch": 13.851956624233852, "percentage": 69.26, "elapsed_time": "1:35:18", "remaining_time": "0:42:17", "throughput": 5178.36, "total_tokens": 29610912} {"current_steps": 29385, "total_steps": 42420, "loss": 0.2971, "lr": 0.00026112598529667115, "epoch": 13.854314002828854, "percentage": 69.27, "elapsed_time": "1:35:19", "remaining_time": "0:42:16", "throughput": 5178.37, "total_tokens": 29615616} {"current_steps": 29390, "total_steps": 42420, "loss": 0.3457, "lr": 0.00026094528105622056, "epoch": 13.856671381423856, "percentage": 69.28, "elapsed_time": "1:35:20", "remaining_time": "0:42:15", "throughput": 5178.44, "total_tokens": 29621184} {"current_steps": 29395, "total_steps": 42420, "loss": 0.3543, "lr": 0.000260764617283675, "epoch": 13.859028760018859, "percentage": 69.3, "elapsed_time": "1:35:21", "remaining_time": "0:42:15", "throughput": 5178.49, "total_tokens": 29626656} {"current_steps": 29400, "total_steps": 42420, "loss": 0.3902, "lr": 0.0002605839940096177, "epoch": 13.861386138613861, "percentage": 69.31, "elapsed_time": "1:35:21", "remaining_time": "0:42:14", "throughput": 5178.5, "total_tokens": 29631104} {"current_steps": 29405, "total_steps": 42420, "loss": 0.3331, "lr": 0.0002604034112646252, "epoch": 13.863743517208864, "percentage": 69.32, "elapsed_time": "1:35:22", "remaining_time": "0:42:12", "throughput": 5178.52, "total_tokens": 29635712} {"current_steps": 29410, "total_steps": 42420, "loss": 0.3368, "lr": 0.00026022286907926705, "epoch": 13.866100895803866, "percentage": 69.33, "elapsed_time": "1:35:23", "remaining_time": "0:42:11", "throughput": 5178.54, "total_tokens": 29640448} {"current_steps": 29415, "total_steps": 42420, "loss": 0.352, "lr": 0.000260042367484106, "epoch": 13.868458274398868, "percentage": 69.34, "elapsed_time": "1:35:24", "remaining_time": "0:42:10", "throughput": 5178.61, "total_tokens": 29645792} {"current_steps": 29420, "total_steps": 42420, "loss": 0.3417, "lr": 0.0002598619065096979, "epoch": 13.87081565299387, "percentage": 69.35, "elapsed_time": "1:35:25", "remaining_time": "0:42:10", "throughput": 5178.67, "total_tokens": 29650976} {"current_steps": 29425, "total_steps": 42420, "loss": 0.3504, "lr": 0.0002596814861865918, "epoch": 13.873173031588873, "percentage": 69.37, "elapsed_time": "1:35:26", "remaining_time": "0:42:08", "throughput": 5178.71, "total_tokens": 29655808} {"current_steps": 29430, "total_steps": 42420, "loss": 0.3436, "lr": 0.00025950110654532967, "epoch": 13.875530410183876, "percentage": 69.38, "elapsed_time": "1:35:27", "remaining_time": "0:42:07", "throughput": 5178.73, "total_tokens": 29660416} {"current_steps": 29435, "total_steps": 42420, "loss": 0.3421, "lr": 0.00025932076761644686, "epoch": 13.877887788778878, "percentage": 69.39, "elapsed_time": "1:35:28", "remaining_time": "0:42:06", "throughput": 5178.71, "total_tokens": 29664288} {"current_steps": 29440, "total_steps": 42420, "loss": 0.341, "lr": 0.0002591404694304714, "epoch": 13.88024516737388, "percentage": 69.4, "elapsed_time": "1:35:29", "remaining_time": "0:42:05", "throughput": 5178.77, "total_tokens": 29669600} {"current_steps": 29445, "total_steps": 42420, "loss": 0.3364, "lr": 0.0002589602120179249, "epoch": 13.882602545968883, "percentage": 69.41, "elapsed_time": "1:35:29", "remaining_time": "0:42:04", "throughput": 5178.78, "total_tokens": 29674400} {"current_steps": 29450, "total_steps": 42420, "loss": 0.3358, "lr": 0.00025877999540932197, "epoch": 13.884959924563885, "percentage": 69.42, "elapsed_time": "1:35:30", "remaining_time": "0:42:03", "throughput": 5178.82, "total_tokens": 29679168} {"current_steps": 29455, "total_steps": 42420, "loss": 0.3817, "lr": 0.00025859981963517, "epoch": 13.887317303158888, "percentage": 69.44, "elapsed_time": "1:35:31", "remaining_time": "0:42:02", "throughput": 5178.84, "total_tokens": 29683776} {"current_steps": 29460, "total_steps": 42420, "loss": 0.3271, "lr": 0.0002584196847259699, "epoch": 13.88967468175389, "percentage": 69.45, "elapsed_time": "1:35:32", "remaining_time": "0:42:01", "throughput": 5178.88, "total_tokens": 29688736} {"current_steps": 29465, "total_steps": 42420, "loss": 0.329, "lr": 0.00025823959071221537, "epoch": 13.892032060348892, "percentage": 69.46, "elapsed_time": "1:35:33", "remaining_time": "0:42:00", "throughput": 5178.92, "total_tokens": 29693792} {"current_steps": 29470, "total_steps": 42420, "loss": 0.3381, "lr": 0.0002580595376243935, "epoch": 13.894389438943895, "percentage": 69.47, "elapsed_time": "1:35:34", "remaining_time": "0:41:59", "throughput": 5178.9, "total_tokens": 29697920} {"current_steps": 29475, "total_steps": 42420, "loss": 0.3522, "lr": 0.0002578795254929838, "epoch": 13.896746817538897, "percentage": 69.48, "elapsed_time": "1:35:35", "remaining_time": "0:41:58", "throughput": 5178.92, "total_tokens": 29702720} {"current_steps": 29480, "total_steps": 42420, "loss": 0.3347, "lr": 0.00025769955434845935, "epoch": 13.8991041961339, "percentage": 69.5, "elapsed_time": "1:35:36", "remaining_time": "0:41:57", "throughput": 5178.92, "total_tokens": 29706944} {"current_steps": 29485, "total_steps": 42420, "loss": 0.3455, "lr": 0.00025751962422128637, "epoch": 13.901461574728902, "percentage": 69.51, "elapsed_time": "1:35:37", "remaining_time": "0:41:56", "throughput": 5179.0, "total_tokens": 29712608} {"current_steps": 29490, "total_steps": 42420, "loss": 0.3517, "lr": 0.000257339735141924, "epoch": 13.903818953323904, "percentage": 69.52, "elapsed_time": "1:35:37", "remaining_time": "0:41:55", "throughput": 5178.98, "total_tokens": 29716928} {"current_steps": 29495, "total_steps": 42420, "loss": 0.3294, "lr": 0.00025715988714082426, "epoch": 13.906176331918907, "percentage": 69.53, "elapsed_time": "1:35:38", "remaining_time": "0:41:54", "throughput": 5179.04, "total_tokens": 29721952} {"current_steps": 29500, "total_steps": 42420, "loss": 0.3288, "lr": 0.0002569800802484325, "epoch": 13.90853371051391, "percentage": 69.54, "elapsed_time": "1:35:39", "remaining_time": "0:41:53", "throughput": 5179.06, "total_tokens": 29726624} {"current_steps": 29505, "total_steps": 42420, "loss": 0.3074, "lr": 0.00025680031449518685, "epoch": 13.910891089108912, "percentage": 69.55, "elapsed_time": "1:35:40", "remaining_time": "0:41:52", "throughput": 5179.11, "total_tokens": 29731840} {"current_steps": 29510, "total_steps": 42420, "loss": 0.3116, "lr": 0.00025662058991151893, "epoch": 13.913248467703912, "percentage": 69.57, "elapsed_time": "1:35:41", "remaining_time": "0:41:51", "throughput": 5179.2, "total_tokens": 29737312} {"current_steps": 29515, "total_steps": 42420, "loss": 0.295, "lr": 0.0002564409065278525, "epoch": 13.915605846298917, "percentage": 69.58, "elapsed_time": "1:35:42", "remaining_time": "0:41:50", "throughput": 5179.24, "total_tokens": 29742496} {"current_steps": 29520, "total_steps": 42420, "loss": 0.3154, "lr": 0.0002562612643746053, "epoch": 13.917963224893917, "percentage": 69.59, "elapsed_time": "1:35:43", "remaining_time": "0:41:49", "throughput": 5179.28, "total_tokens": 29747392} {"current_steps": 29525, "total_steps": 42420, "loss": 0.3771, "lr": 0.00025608166348218756, "epoch": 13.92032060348892, "percentage": 69.6, "elapsed_time": "1:35:44", "remaining_time": "0:41:48", "throughput": 5179.33, "total_tokens": 29752640} {"current_steps": 29530, "total_steps": 42420, "loss": 0.3545, "lr": 0.00025590210388100267, "epoch": 13.922677982083922, "percentage": 69.61, "elapsed_time": "1:35:45", "remaining_time": "0:41:47", "throughput": 5179.35, "total_tokens": 29756992} {"current_steps": 29535, "total_steps": 42420, "loss": 0.3565, "lr": 0.000255722585601447, "epoch": 13.925035360678924, "percentage": 69.63, "elapsed_time": "1:35:46", "remaining_time": "0:41:46", "throughput": 5179.34, "total_tokens": 29761408} {"current_steps": 29540, "total_steps": 42420, "loss": 0.3935, "lr": 0.00025554310867391004, "epoch": 13.927392739273927, "percentage": 69.64, "elapsed_time": "1:35:47", "remaining_time": "0:41:45", "throughput": 5179.4, "total_tokens": 29766560} {"current_steps": 29545, "total_steps": 42420, "loss": 0.33, "lr": 0.00025536367312877417, "epoch": 13.92975011786893, "percentage": 69.65, "elapsed_time": "1:35:48", "remaining_time": "0:41:44", "throughput": 5179.45, "total_tokens": 29772032} {"current_steps": 29550, "total_steps": 42420, "loss": 0.3523, "lr": 0.00025518427899641493, "epoch": 13.932107496463932, "percentage": 69.66, "elapsed_time": "1:35:49", "remaining_time": "0:41:43", "throughput": 5179.48, "total_tokens": 29776896} {"current_steps": 29555, "total_steps": 42420, "loss": 0.3078, "lr": 0.0002550049263072003, "epoch": 13.934464875058934, "percentage": 69.67, "elapsed_time": "1:35:49", "remaining_time": "0:41:42", "throughput": 5179.53, "total_tokens": 29782048} {"current_steps": 29560, "total_steps": 42420, "loss": 0.3629, "lr": 0.0002548256150914919, "epoch": 13.936822253653936, "percentage": 69.68, "elapsed_time": "1:35:50", "remaining_time": "0:41:41", "throughput": 5179.53, "total_tokens": 29786432} {"current_steps": 29565, "total_steps": 42420, "loss": 0.338, "lr": 0.00025464634537964397, "epoch": 13.939179632248939, "percentage": 69.7, "elapsed_time": "1:35:51", "remaining_time": "0:41:40", "throughput": 5179.58, "total_tokens": 29791712} {"current_steps": 29570, "total_steps": 42420, "loss": 0.3513, "lr": 0.000254467117202004, "epoch": 13.941537010843941, "percentage": 69.71, "elapsed_time": "1:35:52", "remaining_time": "0:41:39", "throughput": 5179.65, "total_tokens": 29797376} {"current_steps": 29575, "total_steps": 42420, "loss": 0.3399, "lr": 0.0002542879305889121, "epoch": 13.943894389438944, "percentage": 69.72, "elapsed_time": "1:35:53", "remaining_time": "0:41:38", "throughput": 5179.68, "total_tokens": 29802336} {"current_steps": 29580, "total_steps": 42420, "loss": 0.3597, "lr": 0.0002541087855707016, "epoch": 13.946251768033946, "percentage": 69.73, "elapsed_time": "1:35:54", "remaining_time": "0:41:37", "throughput": 5179.73, "total_tokens": 29807552} {"current_steps": 29585, "total_steps": 42420, "loss": 0.3513, "lr": 0.00025392968217769873, "epoch": 13.948609146628948, "percentage": 69.74, "elapsed_time": "1:35:55", "remaining_time": "0:41:36", "throughput": 5179.78, "total_tokens": 29812768} {"current_steps": 29590, "total_steps": 42420, "loss": 0.3554, "lr": 0.0002537506204402229, "epoch": 13.95096652522395, "percentage": 69.75, "elapsed_time": "1:35:56", "remaining_time": "0:41:35", "throughput": 5179.8, "total_tokens": 29817536} {"current_steps": 29595, "total_steps": 42420, "loss": 0.3492, "lr": 0.0002535716003885856, "epoch": 13.953323903818953, "percentage": 69.77, "elapsed_time": "1:35:57", "remaining_time": "0:41:34", "throughput": 5179.85, "total_tokens": 29822752} {"current_steps": 29600, "total_steps": 42420, "loss": 0.3343, "lr": 0.0002533926220530924, "epoch": 13.955681282413956, "percentage": 69.78, "elapsed_time": "1:35:58", "remaining_time": "0:41:33", "throughput": 5179.88, "total_tokens": 29827552} {"current_steps": 29605, "total_steps": 42420, "loss": 0.2992, "lr": 0.00025321368546404107, "epoch": 13.958038661008958, "percentage": 69.79, "elapsed_time": "1:35:59", "remaining_time": "0:41:32", "throughput": 5179.89, "total_tokens": 29832160} {"current_steps": 29610, "total_steps": 42420, "loss": 0.3761, "lr": 0.0002530347906517226, "epoch": 13.96039603960396, "percentage": 69.8, "elapsed_time": "1:36:00", "remaining_time": "0:41:31", "throughput": 5179.94, "total_tokens": 29836928} {"current_steps": 29615, "total_steps": 42420, "loss": 0.436, "lr": 0.0002528559376464208, "epoch": 13.962753418198963, "percentage": 69.81, "elapsed_time": "1:36:00", "remaining_time": "0:41:30", "throughput": 5179.95, "total_tokens": 29841440} {"current_steps": 29620, "total_steps": 42420, "loss": 0.3408, "lr": 0.0002526771264784125, "epoch": 13.965110796793965, "percentage": 69.83, "elapsed_time": "1:36:01", "remaining_time": "0:41:29", "throughput": 5180.0, "total_tokens": 29846528} {"current_steps": 29625, "total_steps": 42420, "loss": 0.3111, "lr": 0.0002524983571779674, "epoch": 13.967468175388968, "percentage": 69.84, "elapsed_time": "1:36:02", "remaining_time": "0:41:28", "throughput": 5180.04, "total_tokens": 29851456} {"current_steps": 29630, "total_steps": 42420, "loss": 0.3281, "lr": 0.0002523196297753483, "epoch": 13.96982555398397, "percentage": 69.85, "elapsed_time": "1:36:03", "remaining_time": "0:41:28", "throughput": 5180.15, "total_tokens": 29857824} {"current_steps": 29635, "total_steps": 42420, "loss": 0.3348, "lr": 0.0002521409443008103, "epoch": 13.972182932578972, "percentage": 69.86, "elapsed_time": "1:36:04", "remaining_time": "0:41:27", "throughput": 5180.28, "total_tokens": 29864032} {"current_steps": 29640, "total_steps": 42420, "loss": 0.3447, "lr": 0.00025196230078460203, "epoch": 13.974540311173975, "percentage": 69.87, "elapsed_time": "1:36:05", "remaining_time": "0:41:26", "throughput": 5180.35, "total_tokens": 29869472} {"current_steps": 29645, "total_steps": 42420, "loss": 0.3414, "lr": 0.0002517836992569648, "epoch": 13.976897689768977, "percentage": 69.88, "elapsed_time": "1:36:06", "remaining_time": "0:41:25", "throughput": 5180.4, "total_tokens": 29874624} {"current_steps": 29650, "total_steps": 42420, "loss": 0.3339, "lr": 0.00025160513974813284, "epoch": 13.97925506836398, "percentage": 69.9, "elapsed_time": "1:36:07", "remaining_time": "0:41:24", "throughput": 5180.45, "total_tokens": 29879904} {"current_steps": 29655, "total_steps": 42420, "loss": 0.3217, "lr": 0.00025142662228833313, "epoch": 13.981612446958982, "percentage": 69.91, "elapsed_time": "1:36:08", "remaining_time": "0:41:23", "throughput": 5180.54, "total_tokens": 29885664} {"current_steps": 29660, "total_steps": 42420, "loss": 0.3695, "lr": 0.000251248146907786, "epoch": 13.983969825553984, "percentage": 69.92, "elapsed_time": "1:36:09", "remaining_time": "0:41:22", "throughput": 5180.59, "total_tokens": 29890752} {"current_steps": 29665, "total_steps": 42420, "loss": 0.3588, "lr": 0.0002510697136367038, "epoch": 13.986327204148987, "percentage": 69.93, "elapsed_time": "1:36:10", "remaining_time": "0:41:21", "throughput": 5180.6, "total_tokens": 29895296} {"current_steps": 29670, "total_steps": 42420, "loss": 0.3647, "lr": 0.00025089132250529243, "epoch": 13.98868458274399, "percentage": 69.94, "elapsed_time": "1:36:11", "remaining_time": "0:41:20", "throughput": 5180.65, "total_tokens": 29900512} {"current_steps": 29675, "total_steps": 42420, "loss": 0.3181, "lr": 0.00025071297354375066, "epoch": 13.991041961338992, "percentage": 69.96, "elapsed_time": "1:36:12", "remaining_time": "0:41:19", "throughput": 5180.77, "total_tokens": 29907104} {"current_steps": 29680, "total_steps": 42420, "loss": 0.3711, "lr": 0.0002505346667822699, "epoch": 13.993399339933994, "percentage": 69.97, "elapsed_time": "1:36:13", "remaining_time": "0:41:18", "throughput": 5180.81, "total_tokens": 29911968} {"current_steps": 29685, "total_steps": 42420, "loss": 0.3538, "lr": 0.00025035640225103427, "epoch": 13.995756718528996, "percentage": 69.98, "elapsed_time": "1:36:14", "remaining_time": "0:41:17", "throughput": 5180.86, "total_tokens": 29917088} {"current_steps": 29690, "total_steps": 42420, "loss": 0.3317, "lr": 0.000250178179980221, "epoch": 13.998114097123999, "percentage": 69.99, "elapsed_time": "1:36:15", "remaining_time": "0:41:16", "throughput": 5180.89, "total_tokens": 29921920} {"current_steps": 29694, "total_steps": 42420, "eval_loss": 0.33954769372940063, "epoch": 14.0, "percentage": 70.0, "elapsed_time": "1:36:35", "remaining_time": "0:41:23", "throughput": 5163.92, "total_tokens": 29926112} {"current_steps": 29695, "total_steps": 42420, "loss": 0.3326, "lr": 0.0002500000000000001, "epoch": 14.000471475719001, "percentage": 70.0, "elapsed_time": "1:36:37", "remaining_time": "0:41:24", "throughput": 5162.05, "total_tokens": 29926880} {"current_steps": 29700, "total_steps": 42420, "loss": 0.3335, "lr": 0.00024982186234053443, "epoch": 14.002828854314004, "percentage": 70.01, "elapsed_time": "1:36:38", "remaining_time": "0:41:23", "throughput": 5162.12, "total_tokens": 29932256} {"current_steps": 29705, "total_steps": 42420, "loss": 0.2778, "lr": 0.0002496437670319798, "epoch": 14.005186232909006, "percentage": 70.03, "elapsed_time": "1:36:39", "remaining_time": "0:41:22", "throughput": 5162.12, "total_tokens": 29936512} {"current_steps": 29710, "total_steps": 42420, "loss": 0.3534, "lr": 0.0002494657141044845, "epoch": 14.007543611504008, "percentage": 70.04, "elapsed_time": "1:36:40", "remaining_time": "0:41:21", "throughput": 5162.21, "total_tokens": 29942144} {"current_steps": 29715, "total_steps": 42420, "loss": 0.3344, "lr": 0.00024928770358819, "epoch": 14.009900990099009, "percentage": 70.05, "elapsed_time": "1:36:41", "remaining_time": "0:41:20", "throughput": 5162.37, "total_tokens": 29948960} {"current_steps": 29720, "total_steps": 42420, "loss": 0.3565, "lr": 0.0002491097355132306, "epoch": 14.012258368694011, "percentage": 70.06, "elapsed_time": "1:36:42", "remaining_time": "0:41:19", "throughput": 5162.34, "total_tokens": 29952928} {"current_steps": 29725, "total_steps": 42420, "loss": 0.2851, "lr": 0.0002489318099097329, "epoch": 14.014615747289014, "percentage": 70.07, "elapsed_time": "1:36:43", "remaining_time": "0:41:18", "throughput": 5162.37, "total_tokens": 29957632} {"current_steps": 29730, "total_steps": 42420, "loss": 0.429, "lr": 0.0002487539268078169, "epoch": 14.016973125884016, "percentage": 70.08, "elapsed_time": "1:36:44", "remaining_time": "0:41:17", "throughput": 5162.45, "total_tokens": 29963232} {"current_steps": 29735, "total_steps": 42420, "loss": 0.3071, "lr": 0.0002485760862375952, "epoch": 14.019330504479019, "percentage": 70.1, "elapsed_time": "1:36:44", "remaining_time": "0:41:16", "throughput": 5162.46, "total_tokens": 29967872} {"current_steps": 29740, "total_steps": 42420, "loss": 0.3234, "lr": 0.00024839828822917316, "epoch": 14.021687883074021, "percentage": 70.11, "elapsed_time": "1:36:45", "remaining_time": "0:41:15", "throughput": 5162.4, "total_tokens": 29971648} {"current_steps": 29745, "total_steps": 42420, "loss": 0.3393, "lr": 0.00024822053281264893, "epoch": 14.024045261669023, "percentage": 70.12, "elapsed_time": "1:36:46", "remaining_time": "0:41:14", "throughput": 5162.49, "total_tokens": 29977792} {"current_steps": 29750, "total_steps": 42420, "loss": 0.3202, "lr": 0.00024804282001811355, "epoch": 14.026402640264026, "percentage": 70.13, "elapsed_time": "1:36:47", "remaining_time": "0:41:13", "throughput": 5162.5, "total_tokens": 29982560} {"current_steps": 29755, "total_steps": 42420, "loss": 0.3265, "lr": 0.0002478651498756508, "epoch": 14.028760018859028, "percentage": 70.14, "elapsed_time": "1:36:48", "remaining_time": "0:41:12", "throughput": 5162.53, "total_tokens": 29987552} {"current_steps": 29760, "total_steps": 42420, "loss": 0.3265, "lr": 0.0002476875224153373, "epoch": 14.03111739745403, "percentage": 70.16, "elapsed_time": "1:36:49", "remaining_time": "0:41:11", "throughput": 5162.5, "total_tokens": 29991648} {"current_steps": 29765, "total_steps": 42420, "loss": 0.2833, "lr": 0.00024750993766724206, "epoch": 14.033474776049033, "percentage": 70.17, "elapsed_time": "1:36:50", "remaining_time": "0:41:10", "throughput": 5162.54, "total_tokens": 29997024} {"current_steps": 29770, "total_steps": 42420, "loss": 0.3802, "lr": 0.00024733239566142735, "epoch": 14.035832154644035, "percentage": 70.18, "elapsed_time": "1:36:51", "remaining_time": "0:41:09", "throughput": 5162.58, "total_tokens": 30002304} {"current_steps": 29775, "total_steps": 42420, "loss": 0.3162, "lr": 0.00024715489642794796, "epoch": 14.038189533239038, "percentage": 70.19, "elapsed_time": "1:36:52", "remaining_time": "0:41:08", "throughput": 5162.63, "total_tokens": 30007648} {"current_steps": 29780, "total_steps": 42420, "loss": 0.3226, "lr": 0.00024697743999685157, "epoch": 14.04054691183404, "percentage": 70.2, "elapsed_time": "1:36:53", "remaining_time": "0:41:07", "throughput": 5162.73, "total_tokens": 30013696} {"current_steps": 29785, "total_steps": 42420, "loss": 0.3317, "lr": 0.0002468000263981785, "epoch": 14.042904290429043, "percentage": 70.21, "elapsed_time": "1:36:54", "remaining_time": "0:41:06", "throughput": 5162.78, "total_tokens": 30019296} {"current_steps": 29790, "total_steps": 42420, "loss": 0.2883, "lr": 0.0002466226556619619, "epoch": 14.045261669024045, "percentage": 70.23, "elapsed_time": "1:36:55", "remaining_time": "0:41:05", "throughput": 5162.75, "total_tokens": 30023264} {"current_steps": 29795, "total_steps": 42420, "loss": 0.3514, "lr": 0.00024644532781822756, "epoch": 14.047619047619047, "percentage": 70.24, "elapsed_time": "1:36:56", "remaining_time": "0:41:04", "throughput": 5162.89, "total_tokens": 30029984} {"current_steps": 29800, "total_steps": 42420, "loss": 0.3278, "lr": 0.0002462680428969943, "epoch": 14.04997642621405, "percentage": 70.25, "elapsed_time": "1:36:57", "remaining_time": "0:41:03", "throughput": 5162.93, "total_tokens": 30035040} {"current_steps": 29805, "total_steps": 42420, "loss": 0.3161, "lr": 0.00024609080092827305, "epoch": 14.052333804809052, "percentage": 70.26, "elapsed_time": "1:36:58", "remaining_time": "0:41:02", "throughput": 5162.93, "total_tokens": 30039328} {"current_steps": 29810, "total_steps": 42420, "loss": 0.3573, "lr": 0.00024591360194206813, "epoch": 14.054691183404055, "percentage": 70.27, "elapsed_time": "1:36:59", "remaining_time": "0:41:01", "throughput": 5162.98, "total_tokens": 30044736} {"current_steps": 29815, "total_steps": 42420, "loss": 0.3657, "lr": 0.0002457364459683762, "epoch": 14.057048561999057, "percentage": 70.29, "elapsed_time": "1:37:00", "remaining_time": "0:41:00", "throughput": 5163.08, "total_tokens": 30050688} {"current_steps": 29820, "total_steps": 42420, "loss": 0.3293, "lr": 0.00024555933303718687, "epoch": 14.05940594059406, "percentage": 70.3, "elapsed_time": "1:37:01", "remaining_time": "0:40:59", "throughput": 5163.18, "total_tokens": 30056672} {"current_steps": 29825, "total_steps": 42420, "loss": 0.3627, "lr": 0.0002453822631784822, "epoch": 14.061763319189062, "percentage": 70.31, "elapsed_time": "1:37:02", "remaining_time": "0:40:58", "throughput": 5163.21, "total_tokens": 30061472} {"current_steps": 29830, "total_steps": 42420, "loss": 0.3506, "lr": 0.0002452052364222373, "epoch": 14.064120697784064, "percentage": 70.32, "elapsed_time": "1:37:03", "remaining_time": "0:40:57", "throughput": 5163.3, "total_tokens": 30067264} {"current_steps": 29835, "total_steps": 42420, "loss": 0.3609, "lr": 0.00024502825279841973, "epoch": 14.066478076379067, "percentage": 70.33, "elapsed_time": "1:37:04", "remaining_time": "0:40:56", "throughput": 5163.36, "total_tokens": 30072576} {"current_steps": 29840, "total_steps": 42420, "loss": 0.3531, "lr": 0.0002448513123369899, "epoch": 14.068835454974069, "percentage": 70.34, "elapsed_time": "1:37:05", "remaining_time": "0:40:55", "throughput": 5163.4, "total_tokens": 30077504} {"current_steps": 29845, "total_steps": 42420, "loss": 0.3404, "lr": 0.0002446744150679006, "epoch": 14.071192833569071, "percentage": 70.36, "elapsed_time": "1:37:06", "remaining_time": "0:40:54", "throughput": 5163.42, "total_tokens": 30082176} {"current_steps": 29850, "total_steps": 42420, "loss": 0.2907, "lr": 0.00024449756102109757, "epoch": 14.073550212164074, "percentage": 70.37, "elapsed_time": "1:37:06", "remaining_time": "0:40:53", "throughput": 5163.46, "total_tokens": 30086880} {"current_steps": 29855, "total_steps": 42420, "loss": 0.3664, "lr": 0.00024432075022651933, "epoch": 14.075907590759076, "percentage": 70.38, "elapsed_time": "1:37:08", "remaining_time": "0:40:52", "throughput": 5163.62, "total_tokens": 30093760} {"current_steps": 29860, "total_steps": 42420, "loss": 0.2963, "lr": 0.00024414398271409683, "epoch": 14.078264969354079, "percentage": 70.39, "elapsed_time": "1:37:08", "remaining_time": "0:40:51", "throughput": 5163.64, "total_tokens": 30098528} {"current_steps": 29865, "total_steps": 42420, "loss": 0.3827, "lr": 0.00024396725851375396, "epoch": 14.080622347949081, "percentage": 70.4, "elapsed_time": "1:37:09", "remaining_time": "0:40:50", "throughput": 5163.62, "total_tokens": 30102752} {"current_steps": 29870, "total_steps": 42420, "loss": 0.4078, "lr": 0.00024379057765540696, "epoch": 14.082979726544083, "percentage": 70.41, "elapsed_time": "1:37:10", "remaining_time": "0:40:49", "throughput": 5163.65, "total_tokens": 30107488} {"current_steps": 29875, "total_steps": 42420, "loss": 0.3174, "lr": 0.00024361394016896498, "epoch": 14.085337105139086, "percentage": 70.43, "elapsed_time": "1:37:11", "remaining_time": "0:40:48", "throughput": 5163.66, "total_tokens": 30111968} {"current_steps": 29880, "total_steps": 42420, "loss": 0.3203, "lr": 0.00024343734608432994, "epoch": 14.087694483734088, "percentage": 70.44, "elapsed_time": "1:37:12", "remaining_time": "0:40:47", "throughput": 5163.7, "total_tokens": 30117184} {"current_steps": 29885, "total_steps": 42420, "loss": 0.3445, "lr": 0.00024326079543139585, "epoch": 14.09005186232909, "percentage": 70.45, "elapsed_time": "1:37:13", "remaining_time": "0:40:46", "throughput": 5163.69, "total_tokens": 30121376} {"current_steps": 29890, "total_steps": 42420, "loss": 0.2972, "lr": 0.00024308428824005007, "epoch": 14.092409240924093, "percentage": 70.46, "elapsed_time": "1:37:14", "remaining_time": "0:40:45", "throughput": 5163.73, "total_tokens": 30126464} {"current_steps": 29895, "total_steps": 42420, "loss": 0.2869, "lr": 0.00024290782454017195, "epoch": 14.094766619519095, "percentage": 70.47, "elapsed_time": "1:37:15", "remaining_time": "0:40:44", "throughput": 5163.92, "total_tokens": 30134336} {"current_steps": 29900, "total_steps": 42420, "loss": 0.3419, "lr": 0.00024273140436163398, "epoch": 14.097123998114098, "percentage": 70.49, "elapsed_time": "1:37:16", "remaining_time": "0:40:43", "throughput": 5163.95, "total_tokens": 30139040} {"current_steps": 29905, "total_steps": 42420, "loss": 0.3665, "lr": 0.00024255502773430106, "epoch": 14.0994813767091, "percentage": 70.5, "elapsed_time": "1:37:17", "remaining_time": "0:40:42", "throughput": 5164.0, "total_tokens": 30144192} {"current_steps": 29910, "total_steps": 42420, "loss": 0.4055, "lr": 0.00024237869468803075, "epoch": 14.101838755304103, "percentage": 70.51, "elapsed_time": "1:37:18", "remaining_time": "0:40:41", "throughput": 5164.05, "total_tokens": 30149312} {"current_steps": 29915, "total_steps": 42420, "loss": 0.3349, "lr": 0.00024220240525267328, "epoch": 14.104196133899103, "percentage": 70.52, "elapsed_time": "1:37:19", "remaining_time": "0:40:40", "throughput": 5164.16, "total_tokens": 30155520} {"current_steps": 29920, "total_steps": 42420, "loss": 0.3321, "lr": 0.00024202615945807142, "epoch": 14.106553512494106, "percentage": 70.53, "elapsed_time": "1:37:20", "remaining_time": "0:40:39", "throughput": 5164.13, "total_tokens": 30159552} {"current_steps": 29925, "total_steps": 42420, "loss": 0.3261, "lr": 0.00024184995733406067, "epoch": 14.108910891089108, "percentage": 70.54, "elapsed_time": "1:37:21", "remaining_time": "0:40:38", "throughput": 5164.16, "total_tokens": 30164160} {"current_steps": 29930, "total_steps": 42420, "loss": 0.2719, "lr": 0.0002416737989104692, "epoch": 14.11126826968411, "percentage": 70.56, "elapsed_time": "1:37:21", "remaining_time": "0:40:37", "throughput": 5164.2, "total_tokens": 30169184} {"current_steps": 29935, "total_steps": 42420, "loss": 0.3044, "lr": 0.0002414976842171172, "epoch": 14.113625648279113, "percentage": 70.57, "elapsed_time": "1:37:22", "remaining_time": "0:40:36", "throughput": 5164.26, "total_tokens": 30174368} {"current_steps": 29940, "total_steps": 42420, "loss": 0.3129, "lr": 0.0002413216132838182, "epoch": 14.115983026874115, "percentage": 70.58, "elapsed_time": "1:37:23", "remaining_time": "0:40:35", "throughput": 5164.26, "total_tokens": 30178720} {"current_steps": 29945, "total_steps": 42420, "loss": 0.3605, "lr": 0.00024114558614037792, "epoch": 14.118340405469118, "percentage": 70.59, "elapsed_time": "1:37:24", "remaining_time": "0:40:34", "throughput": 5164.25, "total_tokens": 30183040} {"current_steps": 29950, "total_steps": 42420, "loss": 0.2886, "lr": 0.0002409696028165948, "epoch": 14.12069778406412, "percentage": 70.6, "elapsed_time": "1:37:25", "remaining_time": "0:40:33", "throughput": 5164.28, "total_tokens": 30187616} {"current_steps": 29955, "total_steps": 42420, "loss": 0.3191, "lr": 0.0002407936633422599, "epoch": 14.123055162659123, "percentage": 70.62, "elapsed_time": "1:37:26", "remaining_time": "0:40:32", "throughput": 5164.39, "total_tokens": 30193920} {"current_steps": 29960, "total_steps": 42420, "loss": 0.388, "lr": 0.00024061776774715667, "epoch": 14.125412541254125, "percentage": 70.63, "elapsed_time": "1:37:27", "remaining_time": "0:40:31", "throughput": 5164.36, "total_tokens": 30197792} {"current_steps": 29965, "total_steps": 42420, "loss": 0.312, "lr": 0.0002404419160610613, "epoch": 14.127769919849127, "percentage": 70.64, "elapsed_time": "1:37:28", "remaining_time": "0:40:30", "throughput": 5164.36, "total_tokens": 30202144} {"current_steps": 29970, "total_steps": 42420, "loss": 0.3203, "lr": 0.00024026610831374273, "epoch": 14.13012729844413, "percentage": 70.65, "elapsed_time": "1:37:29", "remaining_time": "0:40:29", "throughput": 5164.46, "total_tokens": 30208256} {"current_steps": 29975, "total_steps": 42420, "loss": 0.3144, "lr": 0.00024009034453496174, "epoch": 14.132484677039132, "percentage": 70.66, "elapsed_time": "1:37:30", "remaining_time": "0:40:28", "throughput": 5164.49, "total_tokens": 30212960} {"current_steps": 29980, "total_steps": 42420, "loss": 0.423, "lr": 0.00023991462475447245, "epoch": 14.134842055634135, "percentage": 70.67, "elapsed_time": "1:37:31", "remaining_time": "0:40:27", "throughput": 5164.52, "total_tokens": 30217632} {"current_steps": 29985, "total_steps": 42420, "loss": 0.327, "lr": 0.0002397389490020212, "epoch": 14.137199434229137, "percentage": 70.69, "elapsed_time": "1:37:31", "remaining_time": "0:40:26", "throughput": 5164.6, "total_tokens": 30223040} {"current_steps": 29990, "total_steps": 42420, "loss": 0.3481, "lr": 0.00023956331730734688, "epoch": 14.13955681282414, "percentage": 70.7, "elapsed_time": "1:37:32", "remaining_time": "0:40:25", "throughput": 5164.58, "total_tokens": 30227104} {"current_steps": 29995, "total_steps": 42420, "loss": 0.3267, "lr": 0.00023938772970018098, "epoch": 14.141914191419142, "percentage": 70.71, "elapsed_time": "1:37:33", "remaining_time": "0:40:24", "throughput": 5164.68, "total_tokens": 30232512} {"current_steps": 30000, "total_steps": 42420, "loss": 0.3232, "lr": 0.00023921218621024748, "epoch": 14.144271570014144, "percentage": 70.72, "elapsed_time": "1:37:34", "remaining_time": "0:40:23", "throughput": 5164.7, "total_tokens": 30237216} {"current_steps": 30005, "total_steps": 42420, "loss": 0.3415, "lr": 0.00023903668686726294, "epoch": 14.146628948609147, "percentage": 70.73, "elapsed_time": "1:37:35", "remaining_time": "0:40:22", "throughput": 5164.74, "total_tokens": 30242176} {"current_steps": 30010, "total_steps": 42420, "loss": 0.2485, "lr": 0.0002388612317009366, "epoch": 14.148986327204149, "percentage": 70.74, "elapsed_time": "1:37:36", "remaining_time": "0:40:21", "throughput": 5164.82, "total_tokens": 30247520} {"current_steps": 30015, "total_steps": 42420, "loss": 0.3613, "lr": 0.00023868582074096966, "epoch": 14.151343705799151, "percentage": 70.76, "elapsed_time": "1:37:37", "remaining_time": "0:40:20", "throughput": 5164.84, "total_tokens": 30252288} {"current_steps": 30020, "total_steps": 42420, "loss": 0.3616, "lr": 0.0002385104540170564, "epoch": 14.153701084394154, "percentage": 70.77, "elapsed_time": "1:37:38", "remaining_time": "0:40:19", "throughput": 5164.89, "total_tokens": 30257568} {"current_steps": 30025, "total_steps": 42420, "loss": 0.278, "lr": 0.00023833513155888347, "epoch": 14.156058462989156, "percentage": 70.78, "elapsed_time": "1:37:39", "remaining_time": "0:40:18", "throughput": 5164.93, "total_tokens": 30262368} {"current_steps": 30030, "total_steps": 42420, "loss": 0.4019, "lr": 0.00023815985339612988, "epoch": 14.158415841584159, "percentage": 70.79, "elapsed_time": "1:37:40", "remaining_time": "0:40:17", "throughput": 5164.96, "total_tokens": 30267008} {"current_steps": 30035, "total_steps": 42420, "loss": 0.3283, "lr": 0.00023798461955846739, "epoch": 14.160773220179161, "percentage": 70.8, "elapsed_time": "1:37:41", "remaining_time": "0:40:16", "throughput": 5165.08, "total_tokens": 30273184} {"current_steps": 30040, "total_steps": 42420, "loss": 0.3809, "lr": 0.00023780943007556, "epoch": 14.163130598774163, "percentage": 70.82, "elapsed_time": "1:37:42", "remaining_time": "0:40:15", "throughput": 5165.17, "total_tokens": 30278784} {"current_steps": 30045, "total_steps": 42420, "loss": 0.3233, "lr": 0.00023763428497706435, "epoch": 14.165487977369166, "percentage": 70.83, "elapsed_time": "1:37:42", "remaining_time": "0:40:14", "throughput": 5165.18, "total_tokens": 30283360} {"current_steps": 30050, "total_steps": 42420, "loss": 0.3313, "lr": 0.00023745918429262975, "epoch": 14.167845355964168, "percentage": 70.84, "elapsed_time": "1:37:43", "remaining_time": "0:40:13", "throughput": 5165.2, "total_tokens": 30287936} {"current_steps": 30055, "total_steps": 42420, "loss": 0.326, "lr": 0.00023728412805189736, "epoch": 14.17020273455917, "percentage": 70.85, "elapsed_time": "1:37:44", "remaining_time": "0:40:12", "throughput": 5165.22, "total_tokens": 30292864} {"current_steps": 30060, "total_steps": 42420, "loss": 0.3728, "lr": 0.00023710911628450145, "epoch": 14.172560113154173, "percentage": 70.86, "elapsed_time": "1:37:45", "remaining_time": "0:40:11", "throughput": 5165.22, "total_tokens": 30297120} {"current_steps": 30065, "total_steps": 42420, "loss": 0.3456, "lr": 0.00023693414902006848, "epoch": 14.174917491749175, "percentage": 70.87, "elapsed_time": "1:37:46", "remaining_time": "0:40:10", "throughput": 5165.28, "total_tokens": 30302304} {"current_steps": 30070, "total_steps": 42420, "loss": 0.3215, "lr": 0.00023675922628821756, "epoch": 14.177274870344178, "percentage": 70.89, "elapsed_time": "1:37:47", "remaining_time": "0:40:09", "throughput": 5165.33, "total_tokens": 30307392} {"current_steps": 30075, "total_steps": 42420, "loss": 0.3556, "lr": 0.00023658434811856, "epoch": 14.17963224893918, "percentage": 70.9, "elapsed_time": "1:37:48", "remaining_time": "0:40:08", "throughput": 5165.38, "total_tokens": 30312352} {"current_steps": 30080, "total_steps": 42420, "loss": 0.3629, "lr": 0.00023640951454069977, "epoch": 14.181989627534183, "percentage": 70.91, "elapsed_time": "1:37:49", "remaining_time": "0:40:07", "throughput": 5165.43, "total_tokens": 30317440} {"current_steps": 30085, "total_steps": 42420, "loss": 0.3701, "lr": 0.00023623472558423315, "epoch": 14.184347006129185, "percentage": 70.92, "elapsed_time": "1:37:50", "remaining_time": "0:40:06", "throughput": 5165.53, "total_tokens": 30323264} {"current_steps": 30090, "total_steps": 42420, "loss": 0.3466, "lr": 0.00023605998127874928, "epoch": 14.186704384724187, "percentage": 70.93, "elapsed_time": "1:37:51", "remaining_time": "0:40:05", "throughput": 5165.56, "total_tokens": 30327872} {"current_steps": 30095, "total_steps": 42420, "loss": 0.3553, "lr": 0.0002358852816538289, "epoch": 14.18906176331919, "percentage": 70.95, "elapsed_time": "1:37:52", "remaining_time": "0:40:04", "throughput": 5165.59, "total_tokens": 30332576} {"current_steps": 30100, "total_steps": 42420, "loss": 0.3604, "lr": 0.00023571062673904592, "epoch": 14.191419141914192, "percentage": 70.96, "elapsed_time": "1:37:53", "remaining_time": "0:40:03", "throughput": 5165.64, "total_tokens": 30337984} {"current_steps": 30105, "total_steps": 42420, "loss": 0.3635, "lr": 0.0002355360165639665, "epoch": 14.193776520509195, "percentage": 70.97, "elapsed_time": "1:37:53", "remaining_time": "0:40:02", "throughput": 5165.7, "total_tokens": 30343104} {"current_steps": 30110, "total_steps": 42420, "loss": 0.3592, "lr": 0.00023536145115814926, "epoch": 14.196133899104197, "percentage": 70.98, "elapsed_time": "1:37:54", "remaining_time": "0:40:01", "throughput": 5165.7, "total_tokens": 30347392} {"current_steps": 30115, "total_steps": 42420, "loss": 0.3366, "lr": 0.00023518693055114492, "epoch": 14.198491277699198, "percentage": 70.99, "elapsed_time": "1:37:55", "remaining_time": "0:40:00", "throughput": 5165.79, "total_tokens": 30353216} {"current_steps": 30120, "total_steps": 42420, "loss": 0.338, "lr": 0.00023501245477249694, "epoch": 14.2008486562942, "percentage": 71.0, "elapsed_time": "1:37:56", "remaining_time": "0:39:59", "throughput": 5165.85, "total_tokens": 30358336} {"current_steps": 30125, "total_steps": 42420, "loss": 0.3239, "lr": 0.00023483802385174107, "epoch": 14.203206034889202, "percentage": 71.02, "elapsed_time": "1:37:57", "remaining_time": "0:39:58", "throughput": 5165.85, "total_tokens": 30362560} {"current_steps": 30130, "total_steps": 42420, "loss": 0.3292, "lr": 0.00023466363781840568, "epoch": 14.205563413484205, "percentage": 71.03, "elapsed_time": "1:37:58", "remaining_time": "0:39:57", "throughput": 5165.84, "total_tokens": 30366720} {"current_steps": 30135, "total_steps": 42420, "loss": 0.3328, "lr": 0.00023448929670201118, "epoch": 14.207920792079207, "percentage": 71.04, "elapsed_time": "1:37:59", "remaining_time": "0:39:56", "throughput": 5165.93, "total_tokens": 30372352} {"current_steps": 30140, "total_steps": 42420, "loss": 0.2697, "lr": 0.00023431500053207083, "epoch": 14.21027817067421, "percentage": 71.05, "elapsed_time": "1:38:00", "remaining_time": "0:39:55", "throughput": 5165.94, "total_tokens": 30376864} {"current_steps": 30145, "total_steps": 42420, "loss": 0.3299, "lr": 0.0002341407493380896, "epoch": 14.212635549269212, "percentage": 71.06, "elapsed_time": "1:38:01", "remaining_time": "0:39:54", "throughput": 5165.96, "total_tokens": 30381312} {"current_steps": 30150, "total_steps": 42420, "loss": 0.3103, "lr": 0.0002339665431495655, "epoch": 14.214992927864214, "percentage": 71.07, "elapsed_time": "1:38:01", "remaining_time": "0:39:53", "throughput": 5165.98, "total_tokens": 30385792} {"current_steps": 30155, "total_steps": 42420, "loss": 0.3532, "lr": 0.0002337923819959887, "epoch": 14.217350306459217, "percentage": 71.09, "elapsed_time": "1:38:02", "remaining_time": "0:39:52", "throughput": 5166.04, "total_tokens": 30391136} {"current_steps": 30160, "total_steps": 42420, "loss": 0.3073, "lr": 0.0002336182659068416, "epoch": 14.21970768505422, "percentage": 71.1, "elapsed_time": "1:38:03", "remaining_time": "0:39:51", "throughput": 5166.11, "total_tokens": 30396480} {"current_steps": 30165, "total_steps": 42420, "loss": 0.3504, "lr": 0.00023344419491159923, "epoch": 14.222065063649222, "percentage": 71.11, "elapsed_time": "1:38:04", "remaining_time": "0:39:50", "throughput": 5166.13, "total_tokens": 30401120} {"current_steps": 30170, "total_steps": 42420, "loss": 0.3302, "lr": 0.0002332701690397288, "epoch": 14.224422442244224, "percentage": 71.12, "elapsed_time": "1:38:05", "remaining_time": "0:39:49", "throughput": 5166.23, "total_tokens": 30406944} {"current_steps": 30175, "total_steps": 42420, "loss": 0.3343, "lr": 0.0002330961883206899, "epoch": 14.226779820839226, "percentage": 71.13, "elapsed_time": "1:38:06", "remaining_time": "0:39:48", "throughput": 5166.29, "total_tokens": 30412640} {"current_steps": 30180, "total_steps": 42420, "loss": 0.2833, "lr": 0.00023292225278393475, "epoch": 14.229137199434229, "percentage": 71.15, "elapsed_time": "1:38:07", "remaining_time": "0:39:47", "throughput": 5166.29, "total_tokens": 30416864} {"current_steps": 30185, "total_steps": 42420, "loss": 0.3626, "lr": 0.00023274836245890736, "epoch": 14.231494578029231, "percentage": 71.16, "elapsed_time": "1:38:08", "remaining_time": "0:39:46", "throughput": 5166.32, "total_tokens": 30421792} {"current_steps": 30190, "total_steps": 42420, "loss": 0.3288, "lr": 0.00023257451737504443, "epoch": 14.233851956624234, "percentage": 71.17, "elapsed_time": "1:38:09", "remaining_time": "0:39:45", "throughput": 5166.36, "total_tokens": 30426528} {"current_steps": 30195, "total_steps": 42420, "loss": 0.3608, "lr": 0.00023240071756177517, "epoch": 14.236209335219236, "percentage": 71.18, "elapsed_time": "1:38:10", "remaining_time": "0:39:44", "throughput": 5166.37, "total_tokens": 30431008} {"current_steps": 30200, "total_steps": 42420, "loss": 0.2891, "lr": 0.00023222696304852082, "epoch": 14.238566713814238, "percentage": 71.19, "elapsed_time": "1:38:11", "remaining_time": "0:39:43", "throughput": 5166.44, "total_tokens": 30436448} {"current_steps": 30205, "total_steps": 42420, "loss": 0.2901, "lr": 0.00023205325386469505, "epoch": 14.24092409240924, "percentage": 71.2, "elapsed_time": "1:38:12", "remaining_time": "0:39:42", "throughput": 5166.53, "total_tokens": 30442272} {"current_steps": 30210, "total_steps": 42420, "loss": 0.324, "lr": 0.00023187959003970394, "epoch": 14.243281471004243, "percentage": 71.22, "elapsed_time": "1:38:12", "remaining_time": "0:39:41", "throughput": 5166.53, "total_tokens": 30446272} {"current_steps": 30215, "total_steps": 42420, "loss": 0.3949, "lr": 0.00023170597160294588, "epoch": 14.245638849599246, "percentage": 71.23, "elapsed_time": "1:38:13", "remaining_time": "0:39:40", "throughput": 5166.51, "total_tokens": 30450400} {"current_steps": 30220, "total_steps": 42420, "loss": 0.3214, "lr": 0.00023153239858381158, "epoch": 14.247996228194248, "percentage": 71.24, "elapsed_time": "1:38:14", "remaining_time": "0:39:39", "throughput": 5166.56, "total_tokens": 30455392} {"current_steps": 30225, "total_steps": 42420, "loss": 0.2846, "lr": 0.00023135887101168368, "epoch": 14.25035360678925, "percentage": 71.25, "elapsed_time": "1:38:15", "remaining_time": "0:39:38", "throughput": 5166.56, "total_tokens": 30459776} {"current_steps": 30230, "total_steps": 42420, "loss": 0.3591, "lr": 0.00023118538891593772, "epoch": 14.252710985384253, "percentage": 71.26, "elapsed_time": "1:38:16", "remaining_time": "0:39:37", "throughput": 5166.57, "total_tokens": 30464448} {"current_steps": 30235, "total_steps": 42420, "loss": 0.351, "lr": 0.00023101195232594113, "epoch": 14.255068363979255, "percentage": 71.28, "elapsed_time": "1:38:17", "remaining_time": "0:39:36", "throughput": 5166.64, "total_tokens": 30469856} {"current_steps": 30240, "total_steps": 42420, "loss": 0.3114, "lr": 0.0002308385612710539, "epoch": 14.257425742574258, "percentage": 71.29, "elapsed_time": "1:38:18", "remaining_time": "0:39:35", "throughput": 5166.65, "total_tokens": 30474272} {"current_steps": 30245, "total_steps": 42420, "loss": 0.311, "lr": 0.0002306652157806281, "epoch": 14.25978312116926, "percentage": 71.3, "elapsed_time": "1:38:19", "remaining_time": "0:39:34", "throughput": 5166.67, "total_tokens": 30478816} {"current_steps": 30250, "total_steps": 42420, "loss": 0.3355, "lr": 0.00023049191588400824, "epoch": 14.262140499764262, "percentage": 71.31, "elapsed_time": "1:38:20", "remaining_time": "0:39:33", "throughput": 5166.71, "total_tokens": 30483744} {"current_steps": 30255, "total_steps": 42420, "loss": 0.3803, "lr": 0.00023031866161053106, "epoch": 14.264497878359265, "percentage": 71.32, "elapsed_time": "1:38:20", "remaining_time": "0:39:32", "throughput": 5166.73, "total_tokens": 30488224} {"current_steps": 30260, "total_steps": 42420, "loss": 0.3624, "lr": 0.0002301454529895257, "epoch": 14.266855256954267, "percentage": 71.33, "elapsed_time": "1:38:21", "remaining_time": "0:39:31", "throughput": 5166.79, "total_tokens": 30493664} {"current_steps": 30265, "total_steps": 42420, "loss": 0.3486, "lr": 0.0002299722900503131, "epoch": 14.26921263554927, "percentage": 71.35, "elapsed_time": "1:38:22", "remaining_time": "0:39:30", "throughput": 5166.81, "total_tokens": 30498560} {"current_steps": 30270, "total_steps": 42420, "loss": 0.318, "lr": 0.00022979917282220703, "epoch": 14.271570014144272, "percentage": 71.36, "elapsed_time": "1:38:23", "remaining_time": "0:39:29", "throughput": 5166.84, "total_tokens": 30503136} {"current_steps": 30275, "total_steps": 42420, "loss": 0.3177, "lr": 0.00022962610133451324, "epoch": 14.273927392739274, "percentage": 71.37, "elapsed_time": "1:38:24", "remaining_time": "0:39:28", "throughput": 5166.91, "total_tokens": 30508544} {"current_steps": 30280, "total_steps": 42420, "loss": 0.3344, "lr": 0.00022945307561652973, "epoch": 14.276284771334277, "percentage": 71.38, "elapsed_time": "1:38:25", "remaining_time": "0:39:27", "throughput": 5166.96, "total_tokens": 30513824} {"current_steps": 30285, "total_steps": 42420, "loss": 0.3302, "lr": 0.00022928009569754697, "epoch": 14.27864214992928, "percentage": 71.39, "elapsed_time": "1:38:26", "remaining_time": "0:39:26", "throughput": 5166.99, "total_tokens": 30518720} {"current_steps": 30290, "total_steps": 42420, "loss": 0.2964, "lr": 0.0002291071616068474, "epoch": 14.280999528524282, "percentage": 71.4, "elapsed_time": "1:38:27", "remaining_time": "0:39:25", "throughput": 5166.99, "total_tokens": 30523136} {"current_steps": 30295, "total_steps": 42420, "loss": 0.3965, "lr": 0.00022893427337370592, "epoch": 14.283356907119284, "percentage": 71.42, "elapsed_time": "1:38:28", "remaining_time": "0:39:24", "throughput": 5167.01, "total_tokens": 30527776} {"current_steps": 30300, "total_steps": 42420, "loss": 0.3004, "lr": 0.00022876143102738966, "epoch": 14.285714285714286, "percentage": 71.43, "elapsed_time": "1:38:29", "remaining_time": "0:39:23", "throughput": 5167.12, "total_tokens": 30533920} {"current_steps": 30305, "total_steps": 42420, "loss": 0.3269, "lr": 0.00022858863459715762, "epoch": 14.288071664309289, "percentage": 71.44, "elapsed_time": "1:38:30", "remaining_time": "0:39:22", "throughput": 5167.18, "total_tokens": 30539296} {"current_steps": 30310, "total_steps": 42420, "loss": 0.2981, "lr": 0.0002284158841122614, "epoch": 14.290429042904291, "percentage": 71.45, "elapsed_time": "1:38:31", "remaining_time": "0:39:21", "throughput": 5167.19, "total_tokens": 30543712} {"current_steps": 30315, "total_steps": 42420, "loss": 0.2851, "lr": 0.0002282431796019448, "epoch": 14.292786421499294, "percentage": 71.46, "elapsed_time": "1:38:32", "remaining_time": "0:39:20", "throughput": 5167.25, "total_tokens": 30548864} {"current_steps": 30320, "total_steps": 42420, "loss": 0.3391, "lr": 0.00022807052109544368, "epoch": 14.295143800094294, "percentage": 71.48, "elapsed_time": "1:38:33", "remaining_time": "0:39:19", "throughput": 5167.31, "total_tokens": 30554560} {"current_steps": 30325, "total_steps": 42420, "loss": 0.3621, "lr": 0.0002278979086219863, "epoch": 14.297501178689297, "percentage": 71.49, "elapsed_time": "1:38:33", "remaining_time": "0:39:18", "throughput": 5167.36, "total_tokens": 30559520} {"current_steps": 30330, "total_steps": 42420, "loss": 0.4021, "lr": 0.0002277253422107929, "epoch": 14.299858557284299, "percentage": 71.5, "elapsed_time": "1:38:34", "remaining_time": "0:39:17", "throughput": 5167.36, "total_tokens": 30563872} {"current_steps": 30335, "total_steps": 42420, "loss": 0.3957, "lr": 0.00022755282189107628, "epoch": 14.302215935879302, "percentage": 71.51, "elapsed_time": "1:38:35", "remaining_time": "0:39:16", "throughput": 5167.46, "total_tokens": 30569536} {"current_steps": 30340, "total_steps": 42420, "loss": 0.3109, "lr": 0.0002273803476920409, "epoch": 14.304573314474304, "percentage": 71.52, "elapsed_time": "1:38:36", "remaining_time": "0:39:15", "throughput": 5167.54, "total_tokens": 30575040} {"current_steps": 30345, "total_steps": 42420, "loss": 0.3181, "lr": 0.000227207919642884, "epoch": 14.306930693069306, "percentage": 71.53, "elapsed_time": "1:38:37", "remaining_time": "0:39:14", "throughput": 5167.63, "total_tokens": 30581024} {"current_steps": 30350, "total_steps": 42420, "loss": 0.3434, "lr": 0.00022703553777279434, "epoch": 14.309288071664309, "percentage": 71.55, "elapsed_time": "1:38:38", "remaining_time": "0:39:13", "throughput": 5167.7, "total_tokens": 30586464} {"current_steps": 30355, "total_steps": 42420, "loss": 0.3185, "lr": 0.00022686320211095352, "epoch": 14.311645450259311, "percentage": 71.56, "elapsed_time": "1:38:39", "remaining_time": "0:39:12", "throughput": 5167.73, "total_tokens": 30591168} {"current_steps": 30360, "total_steps": 42420, "loss": 0.365, "lr": 0.00022669091268653492, "epoch": 14.314002828854314, "percentage": 71.57, "elapsed_time": "1:38:40", "remaining_time": "0:39:11", "throughput": 5167.79, "total_tokens": 30596416} {"current_steps": 30365, "total_steps": 42420, "loss": 0.324, "lr": 0.0002265186695287043, "epoch": 14.316360207449316, "percentage": 71.58, "elapsed_time": "1:38:41", "remaining_time": "0:39:10", "throughput": 5167.84, "total_tokens": 30601664} {"current_steps": 30370, "total_steps": 42420, "loss": 0.3416, "lr": 0.0002263464726666195, "epoch": 14.318717586044318, "percentage": 71.59, "elapsed_time": "1:38:42", "remaining_time": "0:39:09", "throughput": 5167.92, "total_tokens": 30607360} {"current_steps": 30375, "total_steps": 42420, "loss": 0.3362, "lr": 0.00022617432212943052, "epoch": 14.32107496463932, "percentage": 71.61, "elapsed_time": "1:38:43", "remaining_time": "0:39:08", "throughput": 5167.96, "total_tokens": 30612256} {"current_steps": 30380, "total_steps": 42420, "loss": 0.3369, "lr": 0.00022600221794627945, "epoch": 14.323432343234323, "percentage": 71.62, "elapsed_time": "1:38:44", "remaining_time": "0:39:07", "throughput": 5168.03, "total_tokens": 30617568} {"current_steps": 30385, "total_steps": 42420, "loss": 0.3312, "lr": 0.00022583016014630092, "epoch": 14.325789721829326, "percentage": 71.63, "elapsed_time": "1:38:45", "remaining_time": "0:39:06", "throughput": 5168.03, "total_tokens": 30621952} {"current_steps": 30390, "total_steps": 42420, "loss": 0.3479, "lr": 0.0002256581487586209, "epoch": 14.328147100424328, "percentage": 71.64, "elapsed_time": "1:38:46", "remaining_time": "0:39:05", "throughput": 5168.01, "total_tokens": 30626144} {"current_steps": 30395, "total_steps": 42420, "loss": 0.3379, "lr": 0.00022548618381235824, "epoch": 14.33050447901933, "percentage": 71.65, "elapsed_time": "1:38:47", "remaining_time": "0:39:04", "throughput": 5168.13, "total_tokens": 30632608} {"current_steps": 30400, "total_steps": 42420, "loss": 0.3136, "lr": 0.0002253142653366237, "epoch": 14.332861857614333, "percentage": 71.66, "elapsed_time": "1:38:48", "remaining_time": "0:39:03", "throughput": 5168.19, "total_tokens": 30637984} {"current_steps": 30405, "total_steps": 42420, "loss": 0.331, "lr": 0.0002251423933605201, "epoch": 14.335219236209335, "percentage": 71.68, "elapsed_time": "1:38:48", "remaining_time": "0:39:02", "throughput": 5168.18, "total_tokens": 30642112} {"current_steps": 30410, "total_steps": 42420, "loss": 0.3209, "lr": 0.00022497056791314257, "epoch": 14.337576614804338, "percentage": 71.69, "elapsed_time": "1:38:49", "remaining_time": "0:39:01", "throughput": 5168.2, "total_tokens": 30646528} {"current_steps": 30415, "total_steps": 42420, "loss": 0.4057, "lr": 0.0002247987890235781, "epoch": 14.33993399339934, "percentage": 71.7, "elapsed_time": "1:38:50", "remaining_time": "0:39:00", "throughput": 5168.25, "total_tokens": 30651680} {"current_steps": 30420, "total_steps": 42420, "loss": 0.359, "lr": 0.000224627056720906, "epoch": 14.342291371994342, "percentage": 71.71, "elapsed_time": "1:38:51", "remaining_time": "0:38:59", "throughput": 5168.29, "total_tokens": 30656704} {"current_steps": 30425, "total_steps": 42420, "loss": 0.31, "lr": 0.00022445537103419784, "epoch": 14.344648750589345, "percentage": 71.72, "elapsed_time": "1:38:52", "remaining_time": "0:38:58", "throughput": 5168.26, "total_tokens": 30660960} {"current_steps": 30430, "total_steps": 42420, "loss": 0.3495, "lr": 0.00022428373199251668, "epoch": 14.347006129184347, "percentage": 71.74, "elapsed_time": "1:38:53", "remaining_time": "0:38:57", "throughput": 5168.33, "total_tokens": 30666112} {"current_steps": 30435, "total_steps": 42420, "loss": 0.342, "lr": 0.0002241121396249183, "epoch": 14.34936350777935, "percentage": 71.75, "elapsed_time": "1:38:54", "remaining_time": "0:38:56", "throughput": 5168.35, "total_tokens": 30670688} {"current_steps": 30440, "total_steps": 42420, "loss": 0.3362, "lr": 0.00022394059396045035, "epoch": 14.351720886374352, "percentage": 71.76, "elapsed_time": "1:38:55", "remaining_time": "0:38:55", "throughput": 5168.39, "total_tokens": 30675680} {"current_steps": 30445, "total_steps": 42420, "loss": 0.3426, "lr": 0.00022376909502815263, "epoch": 14.354078264969354, "percentage": 71.77, "elapsed_time": "1:38:56", "remaining_time": "0:38:54", "throughput": 5168.39, "total_tokens": 30679840} {"current_steps": 30450, "total_steps": 42420, "loss": 0.319, "lr": 0.00022359764285705702, "epoch": 14.356435643564357, "percentage": 71.78, "elapsed_time": "1:38:57", "remaining_time": "0:38:53", "throughput": 5168.46, "total_tokens": 30685376} {"current_steps": 30455, "total_steps": 42420, "loss": 0.3266, "lr": 0.0002234262374761874, "epoch": 14.35879302215936, "percentage": 71.79, "elapsed_time": "1:38:58", "remaining_time": "0:38:52", "throughput": 5168.58, "total_tokens": 30691648} {"current_steps": 30460, "total_steps": 42420, "loss": 0.3456, "lr": 0.0002232548789145598, "epoch": 14.361150400754362, "percentage": 71.81, "elapsed_time": "1:38:58", "remaining_time": "0:38:51", "throughput": 5168.56, "total_tokens": 30695584} {"current_steps": 30465, "total_steps": 42420, "loss": 0.2742, "lr": 0.00022308356720118257, "epoch": 14.363507779349364, "percentage": 71.82, "elapsed_time": "1:38:59", "remaining_time": "0:38:50", "throughput": 5168.63, "total_tokens": 30701120} {"current_steps": 30470, "total_steps": 42420, "loss": 0.299, "lr": 0.00022291230236505543, "epoch": 14.365865157944366, "percentage": 71.83, "elapsed_time": "1:39:00", "remaining_time": "0:38:49", "throughput": 5168.68, "total_tokens": 30706304} {"current_steps": 30475, "total_steps": 42420, "loss": 0.378, "lr": 0.00022274108443517084, "epoch": 14.368222536539369, "percentage": 71.84, "elapsed_time": "1:39:01", "remaining_time": "0:38:48", "throughput": 5168.71, "total_tokens": 30711232} {"current_steps": 30480, "total_steps": 42420, "loss": 0.275, "lr": 0.00022256991344051302, "epoch": 14.370579915134371, "percentage": 71.85, "elapsed_time": "1:39:02", "remaining_time": "0:38:47", "throughput": 5168.73, "total_tokens": 30715936} {"current_steps": 30485, "total_steps": 42420, "loss": 0.3193, "lr": 0.00022239878941005843, "epoch": 14.372937293729374, "percentage": 71.86, "elapsed_time": "1:39:03", "remaining_time": "0:38:46", "throughput": 5168.82, "total_tokens": 30721600} {"current_steps": 30490, "total_steps": 42420, "loss": 0.3211, "lr": 0.00022222771237277532, "epoch": 14.375294672324376, "percentage": 71.88, "elapsed_time": "1:39:04", "remaining_time": "0:38:45", "throughput": 5168.83, "total_tokens": 30726240} {"current_steps": 30495, "total_steps": 42420, "loss": 0.3309, "lr": 0.00022205668235762422, "epoch": 14.377652050919378, "percentage": 71.89, "elapsed_time": "1:39:05", "remaining_time": "0:38:44", "throughput": 5168.91, "total_tokens": 30731680} {"current_steps": 30500, "total_steps": 42420, "loss": 0.3028, "lr": 0.00022188569939355762, "epoch": 14.38000942951438, "percentage": 71.9, "elapsed_time": "1:39:06", "remaining_time": "0:38:43", "throughput": 5168.97, "total_tokens": 30736768} {"current_steps": 30505, "total_steps": 42420, "loss": 0.34, "lr": 0.00022171476350952015, "epoch": 14.382366808109383, "percentage": 71.91, "elapsed_time": "1:39:07", "remaining_time": "0:38:42", "throughput": 5169.02, "total_tokens": 30741632} {"current_steps": 30510, "total_steps": 42420, "loss": 0.3667, "lr": 0.00022154387473444803, "epoch": 14.384724186704386, "percentage": 71.92, "elapsed_time": "1:39:08", "remaining_time": "0:38:41", "throughput": 5169.03, "total_tokens": 30746144} {"current_steps": 30515, "total_steps": 42420, "loss": 0.3377, "lr": 0.00022137303309727004, "epoch": 14.387081565299386, "percentage": 71.94, "elapsed_time": "1:39:09", "remaining_time": "0:38:40", "throughput": 5169.1, "total_tokens": 30751200} {"current_steps": 30520, "total_steps": 42420, "loss": 0.3361, "lr": 0.0002212022386269067, "epoch": 14.389438943894389, "percentage": 71.95, "elapsed_time": "1:39:09", "remaining_time": "0:38:39", "throughput": 5169.14, "total_tokens": 30756032} {"current_steps": 30525, "total_steps": 42420, "loss": 0.2907, "lr": 0.00022103149135227064, "epoch": 14.391796322489391, "percentage": 71.96, "elapsed_time": "1:39:10", "remaining_time": "0:38:38", "throughput": 5169.12, "total_tokens": 30760224} {"current_steps": 30530, "total_steps": 42420, "loss": 0.3015, "lr": 0.00022086079130226644, "epoch": 14.394153701084393, "percentage": 71.97, "elapsed_time": "1:39:11", "remaining_time": "0:38:37", "throughput": 5169.19, "total_tokens": 30765760} {"current_steps": 30535, "total_steps": 42420, "loss": 0.4819, "lr": 0.00022069013850579077, "epoch": 14.396511079679396, "percentage": 71.98, "elapsed_time": "1:39:12", "remaining_time": "0:38:36", "throughput": 5169.26, "total_tokens": 30771168} {"current_steps": 30540, "total_steps": 42420, "loss": 0.3571, "lr": 0.00022051953299173229, "epoch": 14.398868458274398, "percentage": 71.99, "elapsed_time": "1:39:13", "remaining_time": "0:38:36", "throughput": 5169.34, "total_tokens": 30777088} {"current_steps": 30545, "total_steps": 42420, "loss": 0.3487, "lr": 0.0002203489747889717, "epoch": 14.4012258368694, "percentage": 72.01, "elapsed_time": "1:39:14", "remaining_time": "0:38:35", "throughput": 5169.41, "total_tokens": 30782624} {"current_steps": 30550, "total_steps": 42420, "loss": 0.3393, "lr": 0.00022017846392638135, "epoch": 14.403583215464403, "percentage": 72.02, "elapsed_time": "1:39:15", "remaining_time": "0:38:34", "throughput": 5169.48, "total_tokens": 30787968} {"current_steps": 30555, "total_steps": 42420, "loss": 0.337, "lr": 0.00022000800043282592, "epoch": 14.405940594059405, "percentage": 72.03, "elapsed_time": "1:39:16", "remaining_time": "0:38:33", "throughput": 5169.52, "total_tokens": 30792864} {"current_steps": 30560, "total_steps": 42420, "loss": 0.3364, "lr": 0.00021983758433716222, "epoch": 14.408297972654408, "percentage": 72.04, "elapsed_time": "1:39:17", "remaining_time": "0:38:32", "throughput": 5169.52, "total_tokens": 30797312} {"current_steps": 30565, "total_steps": 42420, "loss": 0.3602, "lr": 0.0002196672156682384, "epoch": 14.41065535124941, "percentage": 72.05, "elapsed_time": "1:39:18", "remaining_time": "0:38:31", "throughput": 5169.54, "total_tokens": 30801952} {"current_steps": 30570, "total_steps": 42420, "loss": 0.3347, "lr": 0.00021949689445489517, "epoch": 14.413012729844413, "percentage": 72.07, "elapsed_time": "1:39:19", "remaining_time": "0:38:30", "throughput": 5169.57, "total_tokens": 30807072} {"current_steps": 30575, "total_steps": 42420, "loss": 0.3548, "lr": 0.0002193266207259651, "epoch": 14.415370108439415, "percentage": 72.08, "elapsed_time": "1:39:20", "remaining_time": "0:38:29", "throughput": 5169.61, "total_tokens": 30812000} {"current_steps": 30580, "total_steps": 42420, "loss": 0.3456, "lr": 0.00021915639451027254, "epoch": 14.417727487034417, "percentage": 72.09, "elapsed_time": "1:39:21", "remaining_time": "0:38:27", "throughput": 5169.59, "total_tokens": 30815936} {"current_steps": 30585, "total_steps": 42420, "loss": 0.333, "lr": 0.00021898621583663392, "epoch": 14.42008486562942, "percentage": 72.1, "elapsed_time": "1:39:21", "remaining_time": "0:38:27", "throughput": 5169.65, "total_tokens": 30821472} {"current_steps": 30590, "total_steps": 42420, "loss": 0.3302, "lr": 0.00021881608473385767, "epoch": 14.422442244224422, "percentage": 72.11, "elapsed_time": "1:39:22", "remaining_time": "0:38:26", "throughput": 5169.7, "total_tokens": 30826592} {"current_steps": 30595, "total_steps": 42420, "loss": 0.3003, "lr": 0.00021864600123074418, "epoch": 14.424799622819425, "percentage": 72.12, "elapsed_time": "1:39:23", "remaining_time": "0:38:25", "throughput": 5169.7, "total_tokens": 30830880} {"current_steps": 30600, "total_steps": 42420, "loss": 0.3732, "lr": 0.0002184759653560854, "epoch": 14.427157001414427, "percentage": 72.14, "elapsed_time": "1:39:24", "remaining_time": "0:38:24", "throughput": 5169.72, "total_tokens": 30835712} {"current_steps": 30605, "total_steps": 42420, "loss": 0.3124, "lr": 0.00021830597713866568, "epoch": 14.42951438000943, "percentage": 72.15, "elapsed_time": "1:39:25", "remaining_time": "0:38:23", "throughput": 5169.75, "total_tokens": 30840672} {"current_steps": 30610, "total_steps": 42420, "loss": 0.367, "lr": 0.0002181360366072611, "epoch": 14.431871758604432, "percentage": 72.16, "elapsed_time": "1:39:26", "remaining_time": "0:38:22", "throughput": 5169.79, "total_tokens": 30845568} {"current_steps": 30615, "total_steps": 42420, "loss": 0.3726, "lr": 0.0002179661437906398, "epoch": 14.434229137199434, "percentage": 72.17, "elapsed_time": "1:39:27", "remaining_time": "0:38:20", "throughput": 5169.79, "total_tokens": 30849856} {"current_steps": 30620, "total_steps": 42420, "loss": 0.3132, "lr": 0.00021779629871756167, "epoch": 14.436586515794437, "percentage": 72.18, "elapsed_time": "1:39:28", "remaining_time": "0:38:19", "throughput": 5169.86, "total_tokens": 30855232} {"current_steps": 30625, "total_steps": 42420, "loss": 0.3239, "lr": 0.0002176265014167786, "epoch": 14.438943894389439, "percentage": 72.19, "elapsed_time": "1:39:29", "remaining_time": "0:38:19", "throughput": 5170.05, "total_tokens": 30863168} {"current_steps": 30630, "total_steps": 42420, "loss": 0.3229, "lr": 0.00021745675191703445, "epoch": 14.441301272984441, "percentage": 72.21, "elapsed_time": "1:39:30", "remaining_time": "0:38:18", "throughput": 5170.1, "total_tokens": 30867904} {"current_steps": 30635, "total_steps": 42420, "loss": 0.3459, "lr": 0.000217287050247065, "epoch": 14.443658651579444, "percentage": 72.22, "elapsed_time": "1:39:31", "remaining_time": "0:38:17", "throughput": 5170.19, "total_tokens": 30873856} {"current_steps": 30640, "total_steps": 42420, "loss": 0.3446, "lr": 0.00021711739643559764, "epoch": 14.446016030174446, "percentage": 72.23, "elapsed_time": "1:39:32", "remaining_time": "0:38:16", "throughput": 5170.29, "total_tokens": 30879808} {"current_steps": 30645, "total_steps": 42420, "loss": 0.34, "lr": 0.00021694779051135198, "epoch": 14.448373408769449, "percentage": 72.24, "elapsed_time": "1:39:33", "remaining_time": "0:38:15", "throughput": 5170.31, "total_tokens": 30884352} {"current_steps": 30650, "total_steps": 42420, "loss": 0.3529, "lr": 0.00021677823250303946, "epoch": 14.450730787364451, "percentage": 72.25, "elapsed_time": "1:39:34", "remaining_time": "0:38:14", "throughput": 5170.34, "total_tokens": 30889088} {"current_steps": 30655, "total_steps": 42420, "loss": 0.3424, "lr": 0.0002166087224393633, "epoch": 14.453088165959453, "percentage": 72.27, "elapsed_time": "1:39:35", "remaining_time": "0:38:13", "throughput": 5170.34, "total_tokens": 30893504} {"current_steps": 30660, "total_steps": 42420, "loss": 0.3208, "lr": 0.00021643926034901873, "epoch": 14.455445544554456, "percentage": 72.28, "elapsed_time": "1:39:36", "remaining_time": "0:38:12", "throughput": 5170.41, "total_tokens": 30899008} {"current_steps": 30665, "total_steps": 42420, "loss": 0.3384, "lr": 0.0002162698462606928, "epoch": 14.457802923149458, "percentage": 72.29, "elapsed_time": "1:39:37", "remaining_time": "0:38:11", "throughput": 5170.46, "total_tokens": 30904288} {"current_steps": 30670, "total_steps": 42420, "loss": 0.338, "lr": 0.0002161004802030644, "epoch": 14.46016030174446, "percentage": 72.3, "elapsed_time": "1:39:37", "remaining_time": "0:38:10", "throughput": 5170.48, "total_tokens": 30908640} {"current_steps": 30675, "total_steps": 42420, "loss": 0.377, "lr": 0.00021593116220480456, "epoch": 14.462517680339463, "percentage": 72.31, "elapsed_time": "1:39:38", "remaining_time": "0:38:09", "throughput": 5170.44, "total_tokens": 30912320} {"current_steps": 30680, "total_steps": 42420, "loss": 0.3613, "lr": 0.0002157618922945756, "epoch": 14.464875058934465, "percentage": 72.32, "elapsed_time": "1:39:39", "remaining_time": "0:38:08", "throughput": 5170.42, "total_tokens": 30916640} {"current_steps": 30685, "total_steps": 42420, "loss": 0.3502, "lr": 0.00021559267050103216, "epoch": 14.467232437529468, "percentage": 72.34, "elapsed_time": "1:39:40", "remaining_time": "0:38:07", "throughput": 5170.52, "total_tokens": 30922368} {"current_steps": 30690, "total_steps": 42420, "loss": 0.3198, "lr": 0.00021542349685282065, "epoch": 14.46958981612447, "percentage": 72.35, "elapsed_time": "1:39:41", "remaining_time": "0:38:06", "throughput": 5170.6, "total_tokens": 30928160} {"current_steps": 30695, "total_steps": 42420, "loss": 0.3256, "lr": 0.0002152543713785793, "epoch": 14.471947194719473, "percentage": 72.36, "elapsed_time": "1:39:42", "remaining_time": "0:38:05", "throughput": 5170.73, "total_tokens": 30934496} {"current_steps": 30700, "total_steps": 42420, "loss": 0.3356, "lr": 0.00021508529410693816, "epoch": 14.474304573314475, "percentage": 72.37, "elapsed_time": "1:39:43", "remaining_time": "0:38:04", "throughput": 5170.72, "total_tokens": 30938656} {"current_steps": 30705, "total_steps": 42420, "loss": 0.3343, "lr": 0.00021491626506651914, "epoch": 14.476661951909477, "percentage": 72.38, "elapsed_time": "1:39:44", "remaining_time": "0:38:03", "throughput": 5170.74, "total_tokens": 30943520} {"current_steps": 30710, "total_steps": 42420, "loss": 0.3344, "lr": 0.000214747284285936, "epoch": 14.47901933050448, "percentage": 72.4, "elapsed_time": "1:39:45", "remaining_time": "0:38:02", "throughput": 5170.77, "total_tokens": 30948448} {"current_steps": 30715, "total_steps": 42420, "loss": 0.2938, "lr": 0.00021457835179379447, "epoch": 14.481376709099482, "percentage": 72.41, "elapsed_time": "1:39:46", "remaining_time": "0:38:01", "throughput": 5170.77, "total_tokens": 30952832} {"current_steps": 30720, "total_steps": 42420, "loss": 0.2711, "lr": 0.00021440946761869167, "epoch": 14.483734087694483, "percentage": 72.42, "elapsed_time": "1:39:46", "remaining_time": "0:38:00", "throughput": 5170.77, "total_tokens": 30957088} {"current_steps": 30725, "total_steps": 42420, "loss": 0.3, "lr": 0.00021424063178921693, "epoch": 14.486091466289485, "percentage": 72.43, "elapsed_time": "1:39:47", "remaining_time": "0:37:59", "throughput": 5170.75, "total_tokens": 30961216} {"current_steps": 30730, "total_steps": 42420, "loss": 0.3968, "lr": 0.00021407184433395134, "epoch": 14.488448844884488, "percentage": 72.44, "elapsed_time": "1:39:48", "remaining_time": "0:37:58", "throughput": 5170.78, "total_tokens": 30966272} {"current_steps": 30735, "total_steps": 42420, "loss": 0.2469, "lr": 0.00021390310528146767, "epoch": 14.49080622347949, "percentage": 72.45, "elapsed_time": "1:39:49", "remaining_time": "0:37:57", "throughput": 5170.83, "total_tokens": 30971424} {"current_steps": 30740, "total_steps": 42420, "loss": 0.3695, "lr": 0.0002137344146603306, "epoch": 14.493163602074493, "percentage": 72.47, "elapsed_time": "1:39:50", "remaining_time": "0:37:56", "throughput": 5170.85, "total_tokens": 30976192} {"current_steps": 30745, "total_steps": 42420, "loss": 0.2701, "lr": 0.0002135657724990967, "epoch": 14.495520980669495, "percentage": 72.48, "elapsed_time": "1:39:51", "remaining_time": "0:37:55", "throughput": 5170.92, "total_tokens": 30981312} {"current_steps": 30750, "total_steps": 42420, "loss": 0.3134, "lr": 0.0002133971788263141, "epoch": 14.497878359264497, "percentage": 72.49, "elapsed_time": "1:39:52", "remaining_time": "0:37:54", "throughput": 5170.99, "total_tokens": 30987264} {"current_steps": 30755, "total_steps": 42420, "loss": 0.2878, "lr": 0.00021322863367052308, "epoch": 14.5002357378595, "percentage": 72.5, "elapsed_time": "1:39:53", "remaining_time": "0:37:53", "throughput": 5171.08, "total_tokens": 30993024} {"current_steps": 30760, "total_steps": 42420, "loss": 0.3986, "lr": 0.00021306013706025507, "epoch": 14.502593116454502, "percentage": 72.51, "elapsed_time": "1:39:54", "remaining_time": "0:37:52", "throughput": 5171.17, "total_tokens": 30998912} {"current_steps": 30765, "total_steps": 42420, "loss": 0.3844, "lr": 0.0002128916890240339, "epoch": 14.504950495049505, "percentage": 72.52, "elapsed_time": "1:39:55", "remaining_time": "0:37:51", "throughput": 5171.17, "total_tokens": 31003072} {"current_steps": 30770, "total_steps": 42420, "loss": 0.2741, "lr": 0.00021272328959037496, "epoch": 14.507307873644507, "percentage": 72.54, "elapsed_time": "1:39:56", "remaining_time": "0:37:50", "throughput": 5171.23, "total_tokens": 31008352} {"current_steps": 30775, "total_steps": 42420, "loss": 0.3425, "lr": 0.00021255493878778544, "epoch": 14.50966525223951, "percentage": 72.55, "elapsed_time": "1:39:57", "remaining_time": "0:37:49", "throughput": 5171.25, "total_tokens": 31013408} {"current_steps": 30780, "total_steps": 42420, "loss": 0.3382, "lr": 0.0002123866366447642, "epoch": 14.512022630834512, "percentage": 72.56, "elapsed_time": "1:39:58", "remaining_time": "0:37:48", "throughput": 5171.23, "total_tokens": 31017920} {"current_steps": 30785, "total_steps": 42420, "loss": 0.3161, "lr": 0.0002122183831898022, "epoch": 14.514380009429514, "percentage": 72.57, "elapsed_time": "1:39:59", "remaining_time": "0:37:47", "throughput": 5171.29, "total_tokens": 31023104} {"current_steps": 30790, "total_steps": 42420, "loss": 0.3146, "lr": 0.00021205017845138147, "epoch": 14.516737388024517, "percentage": 72.58, "elapsed_time": "1:39:59", "remaining_time": "0:37:46", "throughput": 5171.3, "total_tokens": 31027712} {"current_steps": 30795, "total_steps": 42420, "loss": 0.3022, "lr": 0.00021188202245797644, "epoch": 14.519094766619519, "percentage": 72.6, "elapsed_time": "1:40:00", "remaining_time": "0:37:45", "throughput": 5171.29, "total_tokens": 31031776} {"current_steps": 30800, "total_steps": 42420, "loss": 0.3271, "lr": 0.000211713915238053, "epoch": 14.521452145214521, "percentage": 72.61, "elapsed_time": "1:40:01", "remaining_time": "0:37:44", "throughput": 5171.38, "total_tokens": 31037440} {"current_steps": 30805, "total_steps": 42420, "loss": 0.3372, "lr": 0.0002115458568200691, "epoch": 14.523809523809524, "percentage": 72.62, "elapsed_time": "1:40:02", "remaining_time": "0:37:43", "throughput": 5171.42, "total_tokens": 31042720} {"current_steps": 30810, "total_steps": 42420, "loss": 0.3104, "lr": 0.00021137784723247377, "epoch": 14.526166902404526, "percentage": 72.63, "elapsed_time": "1:40:03", "remaining_time": "0:37:42", "throughput": 5171.45, "total_tokens": 31047424} {"current_steps": 30815, "total_steps": 42420, "loss": 0.3063, "lr": 0.00021120988650370843, "epoch": 14.528524280999529, "percentage": 72.64, "elapsed_time": "1:40:04", "remaining_time": "0:37:41", "throughput": 5171.51, "total_tokens": 31052704} {"current_steps": 30820, "total_steps": 42420, "loss": 0.3279, "lr": 0.00021104197466220587, "epoch": 14.530881659594531, "percentage": 72.65, "elapsed_time": "1:40:05", "remaining_time": "0:37:40", "throughput": 5171.53, "total_tokens": 31057280} {"current_steps": 30825, "total_steps": 42420, "loss": 0.385, "lr": 0.0002108741117363908, "epoch": 14.533239038189533, "percentage": 72.67, "elapsed_time": "1:40:06", "remaining_time": "0:37:39", "throughput": 5171.63, "total_tokens": 31063392} {"current_steps": 30830, "total_steps": 42420, "loss": 0.3197, "lr": 0.00021070629775467954, "epoch": 14.535596416784536, "percentage": 72.68, "elapsed_time": "1:40:07", "remaining_time": "0:37:38", "throughput": 5171.69, "total_tokens": 31068576} {"current_steps": 30835, "total_steps": 42420, "loss": 0.3012, "lr": 0.00021053853274548013, "epoch": 14.537953795379538, "percentage": 72.69, "elapsed_time": "1:40:08", "remaining_time": "0:37:37", "throughput": 5171.74, "total_tokens": 31074080} {"current_steps": 30840, "total_steps": 42420, "loss": 0.3564, "lr": 0.00021037081673719232, "epoch": 14.54031117397454, "percentage": 72.7, "elapsed_time": "1:40:09", "remaining_time": "0:37:36", "throughput": 5171.79, "total_tokens": 31079136} {"current_steps": 30845, "total_steps": 42420, "loss": 0.3999, "lr": 0.00021020314975820782, "epoch": 14.542668552569543, "percentage": 72.71, "elapsed_time": "1:40:10", "remaining_time": "0:37:35", "throughput": 5171.82, "total_tokens": 31084384} {"current_steps": 30850, "total_steps": 42420, "loss": 0.2967, "lr": 0.00021003553183690944, "epoch": 14.545025931164545, "percentage": 72.73, "elapsed_time": "1:40:11", "remaining_time": "0:37:34", "throughput": 5171.9, "total_tokens": 31089856} {"current_steps": 30855, "total_steps": 42420, "loss": 0.3228, "lr": 0.00020986796300167217, "epoch": 14.547383309759548, "percentage": 72.74, "elapsed_time": "1:40:12", "remaining_time": "0:37:33", "throughput": 5171.89, "total_tokens": 31094240} {"current_steps": 30860, "total_steps": 42420, "loss": 0.3048, "lr": 0.00020970044328086262, "epoch": 14.54974068835455, "percentage": 72.75, "elapsed_time": "1:40:13", "remaining_time": "0:37:32", "throughput": 5171.88, "total_tokens": 31098656} {"current_steps": 30865, "total_steps": 42420, "loss": 0.3289, "lr": 0.000209532972702839, "epoch": 14.552098066949553, "percentage": 72.76, "elapsed_time": "1:40:13", "remaining_time": "0:37:31", "throughput": 5171.92, "total_tokens": 31103712} {"current_steps": 30870, "total_steps": 42420, "loss": 0.3057, "lr": 0.00020936555129595126, "epoch": 14.554455445544555, "percentage": 72.77, "elapsed_time": "1:40:14", "remaining_time": "0:37:30", "throughput": 5171.91, "total_tokens": 31107936} {"current_steps": 30875, "total_steps": 42420, "loss": 0.4227, "lr": 0.000209198179088541, "epoch": 14.556812824139557, "percentage": 72.78, "elapsed_time": "1:40:15", "remaining_time": "0:37:29", "throughput": 5171.94, "total_tokens": 31112704} {"current_steps": 30880, "total_steps": 42420, "loss": 0.3659, "lr": 0.00020903085610894147, "epoch": 14.55917020273456, "percentage": 72.8, "elapsed_time": "1:40:16", "remaining_time": "0:37:28", "throughput": 5172.04, "total_tokens": 31118656} {"current_steps": 30885, "total_steps": 42420, "loss": 0.301, "lr": 0.0002088635823854778, "epoch": 14.561527581329562, "percentage": 72.81, "elapsed_time": "1:40:17", "remaining_time": "0:37:27", "throughput": 5172.02, "total_tokens": 31122848} {"current_steps": 30890, "total_steps": 42420, "loss": 0.3069, "lr": 0.0002086963579464663, "epoch": 14.563884959924565, "percentage": 72.82, "elapsed_time": "1:40:18", "remaining_time": "0:37:26", "throughput": 5172.05, "total_tokens": 31127456} {"current_steps": 30895, "total_steps": 42420, "loss": 0.3393, "lr": 0.00020852918282021526, "epoch": 14.566242338519567, "percentage": 72.83, "elapsed_time": "1:40:19", "remaining_time": "0:37:25", "throughput": 5172.05, "total_tokens": 31131840} {"current_steps": 30900, "total_steps": 42420, "loss": 0.3837, "lr": 0.00020836205703502476, "epoch": 14.56859971711457, "percentage": 72.84, "elapsed_time": "1:40:20", "remaining_time": "0:37:24", "throughput": 5172.09, "total_tokens": 31136736} {"current_steps": 30905, "total_steps": 42420, "loss": 0.3176, "lr": 0.00020819498061918617, "epoch": 14.570957095709572, "percentage": 72.85, "elapsed_time": "1:40:21", "remaining_time": "0:37:23", "throughput": 5172.15, "total_tokens": 31141888} {"current_steps": 30910, "total_steps": 42420, "loss": 0.3397, "lr": 0.00020802795360098288, "epoch": 14.573314474304574, "percentage": 72.87, "elapsed_time": "1:40:21", "remaining_time": "0:37:22", "throughput": 5172.19, "total_tokens": 31146880} {"current_steps": 30915, "total_steps": 42420, "loss": 0.3272, "lr": 0.00020786097600868965, "epoch": 14.575671852899575, "percentage": 72.88, "elapsed_time": "1:40:22", "remaining_time": "0:37:21", "throughput": 5172.21, "total_tokens": 31151584} {"current_steps": 30920, "total_steps": 42420, "loss": 0.3299, "lr": 0.0002076940478705729, "epoch": 14.578029231494579, "percentage": 72.89, "elapsed_time": "1:40:23", "remaining_time": "0:37:20", "throughput": 5172.28, "total_tokens": 31156832} {"current_steps": 30925, "total_steps": 42420, "loss": 0.3131, "lr": 0.00020752716921489095, "epoch": 14.58038661008958, "percentage": 72.9, "elapsed_time": "1:40:24", "remaining_time": "0:37:19", "throughput": 5172.34, "total_tokens": 31162080} {"current_steps": 30930, "total_steps": 42420, "loss": 0.3572, "lr": 0.0002073603400698932, "epoch": 14.582743988684582, "percentage": 72.91, "elapsed_time": "1:40:25", "remaining_time": "0:37:18", "throughput": 5172.37, "total_tokens": 31167040} {"current_steps": 30935, "total_steps": 42420, "loss": 0.3164, "lr": 0.0002071935604638211, "epoch": 14.585101367279584, "percentage": 72.93, "elapsed_time": "1:40:26", "remaining_time": "0:37:17", "throughput": 5172.39, "total_tokens": 31171936} {"current_steps": 30940, "total_steps": 42420, "loss": 0.383, "lr": 0.00020702683042490765, "epoch": 14.587458745874587, "percentage": 72.94, "elapsed_time": "1:40:27", "remaining_time": "0:37:16", "throughput": 5172.4, "total_tokens": 31176352} {"current_steps": 30945, "total_steps": 42420, "loss": 0.3366, "lr": 0.00020686014998137742, "epoch": 14.58981612446959, "percentage": 72.95, "elapsed_time": "1:40:28", "remaining_time": "0:37:15", "throughput": 5172.41, "total_tokens": 31180736} {"current_steps": 30950, "total_steps": 42420, "loss": 0.335, "lr": 0.0002066935191614466, "epoch": 14.592173503064592, "percentage": 72.96, "elapsed_time": "1:40:29", "remaining_time": "0:37:14", "throughput": 5172.44, "total_tokens": 31185664} {"current_steps": 30955, "total_steps": 42420, "loss": 0.3113, "lr": 0.00020652693799332283, "epoch": 14.594530881659594, "percentage": 72.97, "elapsed_time": "1:40:30", "remaining_time": "0:37:13", "throughput": 5172.5, "total_tokens": 31191072} {"current_steps": 30960, "total_steps": 42420, "loss": 0.325, "lr": 0.00020636040650520566, "epoch": 14.596888260254596, "percentage": 72.98, "elapsed_time": "1:40:31", "remaining_time": "0:37:12", "throughput": 5172.55, "total_tokens": 31196032} {"current_steps": 30965, "total_steps": 42420, "loss": 0.3687, "lr": 0.00020619392472528609, "epoch": 14.599245638849599, "percentage": 73.0, "elapsed_time": "1:40:31", "remaining_time": "0:37:11", "throughput": 5172.6, "total_tokens": 31201056} {"current_steps": 30970, "total_steps": 42420, "loss": 0.34, "lr": 0.00020602749268174638, "epoch": 14.601603017444601, "percentage": 73.01, "elapsed_time": "1:40:32", "remaining_time": "0:37:10", "throughput": 5172.65, "total_tokens": 31206208} {"current_steps": 30975, "total_steps": 42420, "loss": 0.2984, "lr": 0.00020586111040276075, "epoch": 14.603960396039604, "percentage": 73.02, "elapsed_time": "1:40:33", "remaining_time": "0:37:09", "throughput": 5172.72, "total_tokens": 31211872} {"current_steps": 30980, "total_steps": 42420, "loss": 0.3651, "lr": 0.00020569477791649495, "epoch": 14.606317774634606, "percentage": 73.03, "elapsed_time": "1:40:34", "remaining_time": "0:37:08", "throughput": 5172.79, "total_tokens": 31217184} {"current_steps": 30985, "total_steps": 42420, "loss": 0.3178, "lr": 0.00020552849525110617, "epoch": 14.608675153229608, "percentage": 73.04, "elapsed_time": "1:40:35", "remaining_time": "0:37:07", "throughput": 5172.8, "total_tokens": 31221792} {"current_steps": 30990, "total_steps": 42420, "loss": 0.2994, "lr": 0.00020536226243474337, "epoch": 14.61103253182461, "percentage": 73.06, "elapsed_time": "1:40:36", "remaining_time": "0:37:06", "throughput": 5172.82, "total_tokens": 31226560} {"current_steps": 30995, "total_steps": 42420, "loss": 0.3402, "lr": 0.00020519607949554686, "epoch": 14.613389910419613, "percentage": 73.07, "elapsed_time": "1:40:37", "remaining_time": "0:37:05", "throughput": 5172.9, "total_tokens": 31232416} {"current_steps": 31000, "total_steps": 42420, "loss": 0.3396, "lr": 0.00020502994646164863, "epoch": 14.615747289014616, "percentage": 73.08, "elapsed_time": "1:40:38", "remaining_time": "0:37:04", "throughput": 5173.0, "total_tokens": 31238528} {"current_steps": 31005, "total_steps": 42420, "loss": 0.3447, "lr": 0.0002048638633611723, "epoch": 14.618104667609618, "percentage": 73.09, "elapsed_time": "1:40:39", "remaining_time": "0:37:03", "throughput": 5172.99, "total_tokens": 31242656} {"current_steps": 31010, "total_steps": 42420, "loss": 0.4032, "lr": 0.00020469783022223265, "epoch": 14.62046204620462, "percentage": 73.1, "elapsed_time": "1:40:40", "remaining_time": "0:37:02", "throughput": 5172.99, "total_tokens": 31247040} {"current_steps": 31015, "total_steps": 42420, "loss": 0.3114, "lr": 0.00020453184707293654, "epoch": 14.622819424799623, "percentage": 73.11, "elapsed_time": "1:40:41", "remaining_time": "0:37:01", "throughput": 5173.05, "total_tokens": 31252288} {"current_steps": 31020, "total_steps": 42420, "loss": 0.3598, "lr": 0.00020436591394138178, "epoch": 14.625176803394625, "percentage": 73.13, "elapsed_time": "1:40:42", "remaining_time": "0:37:00", "throughput": 5173.05, "total_tokens": 31256448} {"current_steps": 31025, "total_steps": 42420, "loss": 0.3461, "lr": 0.0002042000308556582, "epoch": 14.627534181989628, "percentage": 73.14, "elapsed_time": "1:40:43", "remaining_time": "0:36:59", "throughput": 5173.13, "total_tokens": 31262080} {"current_steps": 31030, "total_steps": 42420, "loss": 0.3251, "lr": 0.00020403419784384704, "epoch": 14.62989156058463, "percentage": 73.15, "elapsed_time": "1:40:44", "remaining_time": "0:36:58", "throughput": 5173.29, "total_tokens": 31268992} {"current_steps": 31035, "total_steps": 42420, "loss": 0.352, "lr": 0.0002038684149340209, "epoch": 14.632248939179632, "percentage": 73.16, "elapsed_time": "1:40:45", "remaining_time": "0:36:57", "throughput": 5173.3, "total_tokens": 31273600} {"current_steps": 31040, "total_steps": 42420, "loss": 0.3169, "lr": 0.00020370268215424414, "epoch": 14.634606317774635, "percentage": 73.17, "elapsed_time": "1:40:46", "remaining_time": "0:36:56", "throughput": 5173.36, "total_tokens": 31278720} {"current_steps": 31045, "total_steps": 42420, "loss": 0.3205, "lr": 0.00020353699953257244, "epoch": 14.636963696369637, "percentage": 73.18, "elapsed_time": "1:40:47", "remaining_time": "0:36:55", "throughput": 5173.43, "total_tokens": 31284256} {"current_steps": 31050, "total_steps": 42420, "loss": 0.321, "lr": 0.000203371367097053, "epoch": 14.63932107496464, "percentage": 73.2, "elapsed_time": "1:40:48", "remaining_time": "0:36:54", "throughput": 5173.49, "total_tokens": 31289408} {"current_steps": 31055, "total_steps": 42420, "loss": 0.3419, "lr": 0.00020320578487572488, "epoch": 14.641678453559642, "percentage": 73.21, "elapsed_time": "1:40:49", "remaining_time": "0:36:53", "throughput": 5173.58, "total_tokens": 31295328} {"current_steps": 31060, "total_steps": 42420, "loss": 0.3888, "lr": 0.00020304025289661793, "epoch": 14.644035832154644, "percentage": 73.22, "elapsed_time": "1:40:49", "remaining_time": "0:36:52", "throughput": 5173.61, "total_tokens": 31300064} {"current_steps": 31065, "total_steps": 42420, "loss": 0.3496, "lr": 0.00020287477118775406, "epoch": 14.646393210749647, "percentage": 73.23, "elapsed_time": "1:40:51", "remaining_time": "0:36:51", "throughput": 5173.75, "total_tokens": 31307136} {"current_steps": 31070, "total_steps": 42420, "loss": 0.3308, "lr": 0.00020270933977714655, "epoch": 14.64875058934465, "percentage": 73.24, "elapsed_time": "1:40:52", "remaining_time": "0:36:50", "throughput": 5173.81, "total_tokens": 31312384} {"current_steps": 31075, "total_steps": 42420, "loss": 0.3279, "lr": 0.00020254395869280007, "epoch": 14.651107967939652, "percentage": 73.26, "elapsed_time": "1:40:53", "remaining_time": "0:36:49", "throughput": 5173.89, "total_tokens": 31318208} {"current_steps": 31080, "total_steps": 42420, "loss": 0.3571, "lr": 0.0002023786279627109, "epoch": 14.653465346534654, "percentage": 73.27, "elapsed_time": "1:40:54", "remaining_time": "0:36:48", "throughput": 5173.95, "total_tokens": 31323712} {"current_steps": 31085, "total_steps": 42420, "loss": 0.3427, "lr": 0.0002022133476148667, "epoch": 14.655822725129656, "percentage": 73.28, "elapsed_time": "1:40:55", "remaining_time": "0:36:47", "throughput": 5173.99, "total_tokens": 31328768} {"current_steps": 31090, "total_steps": 42420, "loss": 0.3209, "lr": 0.00020204811767724656, "epoch": 14.658180103724659, "percentage": 73.29, "elapsed_time": "1:40:55", "remaining_time": "0:36:46", "throughput": 5174.01, "total_tokens": 31333568} {"current_steps": 31095, "total_steps": 42420, "loss": 0.3532, "lr": 0.0002018829381778214, "epoch": 14.660537482319661, "percentage": 73.3, "elapsed_time": "1:40:56", "remaining_time": "0:36:45", "throughput": 5174.03, "total_tokens": 31338400} {"current_steps": 31100, "total_steps": 42420, "loss": 0.3182, "lr": 0.0002017178091445529, "epoch": 14.662894860914664, "percentage": 73.31, "elapsed_time": "1:40:57", "remaining_time": "0:36:44", "throughput": 5174.01, "total_tokens": 31342272} {"current_steps": 31105, "total_steps": 42420, "loss": 0.3505, "lr": 0.0002015527306053947, "epoch": 14.665252239509666, "percentage": 73.33, "elapsed_time": "1:40:58", "remaining_time": "0:36:43", "throughput": 5174.07, "total_tokens": 31347680} {"current_steps": 31110, "total_steps": 42420, "loss": 0.3434, "lr": 0.0002013877025882919, "epoch": 14.667609618104667, "percentage": 73.34, "elapsed_time": "1:40:59", "remaining_time": "0:36:42", "throughput": 5174.07, "total_tokens": 31352064} {"current_steps": 31115, "total_steps": 42420, "loss": 0.2936, "lr": 0.0002012227251211809, "epoch": 14.66996699669967, "percentage": 73.35, "elapsed_time": "1:41:00", "remaining_time": "0:36:41", "throughput": 5174.1, "total_tokens": 31357024} {"current_steps": 31120, "total_steps": 42420, "loss": 0.3497, "lr": 0.0002010577982319896, "epoch": 14.672324375294671, "percentage": 73.36, "elapsed_time": "1:41:01", "remaining_time": "0:36:40", "throughput": 5174.14, "total_tokens": 31361920} {"current_steps": 31125, "total_steps": 42420, "loss": 0.3393, "lr": 0.0002008929219486373, "epoch": 14.674681753889674, "percentage": 73.37, "elapsed_time": "1:41:02", "remaining_time": "0:36:39", "throughput": 5174.16, "total_tokens": 31366432} {"current_steps": 31130, "total_steps": 42420, "loss": 0.3097, "lr": 0.00020072809629903478, "epoch": 14.677039132484676, "percentage": 73.39, "elapsed_time": "1:41:03", "remaining_time": "0:36:38", "throughput": 5174.22, "total_tokens": 31371936} {"current_steps": 31135, "total_steps": 42420, "loss": 0.3965, "lr": 0.00020056332131108436, "epoch": 14.679396511079679, "percentage": 73.4, "elapsed_time": "1:41:03", "remaining_time": "0:36:37", "throughput": 5174.22, "total_tokens": 31376288} {"current_steps": 31140, "total_steps": 42420, "loss": 0.3787, "lr": 0.00020039859701267933, "epoch": 14.681753889674681, "percentage": 73.41, "elapsed_time": "1:41:04", "remaining_time": "0:36:36", "throughput": 5174.23, "total_tokens": 31380832} {"current_steps": 31145, "total_steps": 42420, "loss": 0.31, "lr": 0.00020023392343170487, "epoch": 14.684111268269683, "percentage": 73.42, "elapsed_time": "1:41:05", "remaining_time": "0:36:35", "throughput": 5174.33, "total_tokens": 31386848} {"current_steps": 31150, "total_steps": 42420, "loss": 0.3254, "lr": 0.00020006930059603744, "epoch": 14.686468646864686, "percentage": 73.43, "elapsed_time": "1:41:06", "remaining_time": "0:36:34", "throughput": 5174.37, "total_tokens": 31391744} {"current_steps": 31155, "total_steps": 42420, "loss": 0.3066, "lr": 0.00019990472853354492, "epoch": 14.688826025459688, "percentage": 73.44, "elapsed_time": "1:41:07", "remaining_time": "0:36:33", "throughput": 5174.41, "total_tokens": 31396768} {"current_steps": 31160, "total_steps": 42420, "loss": 0.3311, "lr": 0.00019974020727208652, "epoch": 14.69118340405469, "percentage": 73.46, "elapsed_time": "1:41:08", "remaining_time": "0:36:32", "throughput": 5174.47, "total_tokens": 31402112} {"current_steps": 31165, "total_steps": 42420, "loss": 0.3321, "lr": 0.00019957573683951292, "epoch": 14.693540782649693, "percentage": 73.47, "elapsed_time": "1:41:09", "remaining_time": "0:36:31", "throughput": 5174.5, "total_tokens": 31407168} {"current_steps": 31170, "total_steps": 42420, "loss": 0.3017, "lr": 0.00019941131726366614, "epoch": 14.695898161244696, "percentage": 73.48, "elapsed_time": "1:41:10", "remaining_time": "0:36:31", "throughput": 5174.59, "total_tokens": 31412960} {"current_steps": 31175, "total_steps": 42420, "loss": 0.3644, "lr": 0.0001992469485723798, "epoch": 14.698255539839698, "percentage": 73.49, "elapsed_time": "1:41:11", "remaining_time": "0:36:30", "throughput": 5174.59, "total_tokens": 31417440} {"current_steps": 31180, "total_steps": 42420, "loss": 0.2852, "lr": 0.00019908263079347838, "epoch": 14.7006129184347, "percentage": 73.5, "elapsed_time": "1:41:12", "remaining_time": "0:36:28", "throughput": 5174.58, "total_tokens": 31421728} {"current_steps": 31185, "total_steps": 42420, "loss": 0.3067, "lr": 0.0001989183639547783, "epoch": 14.702970297029703, "percentage": 73.51, "elapsed_time": "1:41:13", "remaining_time": "0:36:27", "throughput": 5174.6, "total_tokens": 31426496} {"current_steps": 31190, "total_steps": 42420, "loss": 0.3082, "lr": 0.00019875414808408716, "epoch": 14.705327675624705, "percentage": 73.53, "elapsed_time": "1:41:14", "remaining_time": "0:36:26", "throughput": 5174.6, "total_tokens": 31430880} {"current_steps": 31195, "total_steps": 42420, "loss": 0.2912, "lr": 0.00019858998320920385, "epoch": 14.707685054219708, "percentage": 73.54, "elapsed_time": "1:41:15", "remaining_time": "0:36:25", "throughput": 5174.65, "total_tokens": 31436192} {"current_steps": 31200, "total_steps": 42420, "loss": 0.3453, "lr": 0.00019842586935791873, "epoch": 14.71004243281471, "percentage": 73.55, "elapsed_time": "1:41:16", "remaining_time": "0:36:25", "throughput": 5174.74, "total_tokens": 31441728} {"current_steps": 31205, "total_steps": 42420, "loss": 0.427, "lr": 0.00019826180655801352, "epoch": 14.712399811409712, "percentage": 73.56, "elapsed_time": "1:41:16", "remaining_time": "0:36:24", "throughput": 5174.78, "total_tokens": 31446880} {"current_steps": 31210, "total_steps": 42420, "loss": 0.317, "lr": 0.0001980977948372612, "epoch": 14.714757190004715, "percentage": 73.57, "elapsed_time": "1:41:17", "remaining_time": "0:36:23", "throughput": 5174.83, "total_tokens": 31452608} {"current_steps": 31215, "total_steps": 42420, "loss": 0.283, "lr": 0.0001979338342234264, "epoch": 14.717114568599717, "percentage": 73.59, "elapsed_time": "1:41:18", "remaining_time": "0:36:22", "throughput": 5174.83, "total_tokens": 31456864} {"current_steps": 31220, "total_steps": 42420, "loss": 0.3374, "lr": 0.00019776992474426452, "epoch": 14.71947194719472, "percentage": 73.6, "elapsed_time": "1:41:19", "remaining_time": "0:36:21", "throughput": 5174.88, "total_tokens": 31461856} {"current_steps": 31225, "total_steps": 42420, "loss": 0.3105, "lr": 0.00019760606642752276, "epoch": 14.721829325789722, "percentage": 73.61, "elapsed_time": "1:41:20", "remaining_time": "0:36:20", "throughput": 5174.9, "total_tokens": 31466656} {"current_steps": 31230, "total_steps": 42420, "loss": 0.3987, "lr": 0.00019744225930093967, "epoch": 14.724186704384724, "percentage": 73.62, "elapsed_time": "1:41:21", "remaining_time": "0:36:19", "throughput": 5174.94, "total_tokens": 31471392} {"current_steps": 31235, "total_steps": 42420, "loss": 0.2543, "lr": 0.00019727850339224507, "epoch": 14.726544082979727, "percentage": 73.63, "elapsed_time": "1:41:22", "remaining_time": "0:36:18", "throughput": 5174.96, "total_tokens": 31476320} {"current_steps": 31240, "total_steps": 42420, "loss": 0.3554, "lr": 0.0001971147987291597, "epoch": 14.72890146157473, "percentage": 73.64, "elapsed_time": "1:41:23", "remaining_time": "0:36:17", "throughput": 5174.97, "total_tokens": 31480736} {"current_steps": 31245, "total_steps": 42420, "loss": 0.3274, "lr": 0.00019695114533939629, "epoch": 14.731258840169732, "percentage": 73.66, "elapsed_time": "1:41:24", "remaining_time": "0:36:16", "throughput": 5175.02, "total_tokens": 31485760} {"current_steps": 31250, "total_steps": 42420, "loss": 0.3665, "lr": 0.00019678754325065846, "epoch": 14.733616218764734, "percentage": 73.67, "elapsed_time": "1:41:25", "remaining_time": "0:36:15", "throughput": 5175.02, "total_tokens": 31490048} {"current_steps": 31255, "total_steps": 42420, "loss": 0.3907, "lr": 0.00019662399249064134, "epoch": 14.735973597359736, "percentage": 73.68, "elapsed_time": "1:41:25", "remaining_time": "0:36:14", "throughput": 5175.07, "total_tokens": 31495232} {"current_steps": 31260, "total_steps": 42420, "loss": 0.3274, "lr": 0.00019646049308703122, "epoch": 14.738330975954739, "percentage": 73.69, "elapsed_time": "1:41:26", "remaining_time": "0:36:13", "throughput": 5175.08, "total_tokens": 31499680} {"current_steps": 31265, "total_steps": 42420, "loss": 0.3267, "lr": 0.000196297045067506, "epoch": 14.740688354549741, "percentage": 73.7, "elapsed_time": "1:41:27", "remaining_time": "0:36:12", "throughput": 5175.09, "total_tokens": 31504192} {"current_steps": 31270, "total_steps": 42420, "loss": 0.3488, "lr": 0.00019613364845973431, "epoch": 14.743045733144744, "percentage": 73.72, "elapsed_time": "1:41:28", "remaining_time": "0:36:10", "throughput": 5175.08, "total_tokens": 31508384} {"current_steps": 31275, "total_steps": 42420, "loss": 0.3252, "lr": 0.00019597030329137656, "epoch": 14.745403111739746, "percentage": 73.73, "elapsed_time": "1:41:29", "remaining_time": "0:36:09", "throughput": 5175.06, "total_tokens": 31512448} {"current_steps": 31280, "total_steps": 42420, "loss": 0.3035, "lr": 0.0001958070095900843, "epoch": 14.747760490334748, "percentage": 73.74, "elapsed_time": "1:41:30", "remaining_time": "0:36:08", "throughput": 5175.15, "total_tokens": 31518368} {"current_steps": 31285, "total_steps": 42420, "loss": 0.3288, "lr": 0.00019564376738350048, "epoch": 14.75011786892975, "percentage": 73.75, "elapsed_time": "1:41:31", "remaining_time": "0:36:07", "throughput": 5175.15, "total_tokens": 31522720} {"current_steps": 31290, "total_steps": 42420, "loss": 0.3259, "lr": 0.0001954805766992591, "epoch": 14.752475247524753, "percentage": 73.76, "elapsed_time": "1:41:32", "remaining_time": "0:36:06", "throughput": 5175.22, "total_tokens": 31528000} {"current_steps": 31295, "total_steps": 42420, "loss": 0.355, "lr": 0.0001953174375649857, "epoch": 14.754832626119756, "percentage": 73.77, "elapsed_time": "1:41:33", "remaining_time": "0:36:06", "throughput": 5175.27, "total_tokens": 31533344} {"current_steps": 31300, "total_steps": 42420, "loss": 0.364, "lr": 0.00019515435000829685, "epoch": 14.757190004714758, "percentage": 73.79, "elapsed_time": "1:41:33", "remaining_time": "0:36:05", "throughput": 5175.28, "total_tokens": 31537888} {"current_steps": 31305, "total_steps": 42420, "loss": 0.3082, "lr": 0.00019499131405680077, "epoch": 14.75954738330976, "percentage": 73.8, "elapsed_time": "1:41:34", "remaining_time": "0:36:03", "throughput": 5175.29, "total_tokens": 31542368} {"current_steps": 31310, "total_steps": 42420, "loss": 0.3414, "lr": 0.00019482832973809628, "epoch": 14.761904761904763, "percentage": 73.81, "elapsed_time": "1:41:35", "remaining_time": "0:36:02", "throughput": 5175.28, "total_tokens": 31546496} {"current_steps": 31315, "total_steps": 42420, "loss": 0.3471, "lr": 0.00019466539707977404, "epoch": 14.764262140499763, "percentage": 73.82, "elapsed_time": "1:41:36", "remaining_time": "0:36:01", "throughput": 5175.29, "total_tokens": 31550944} {"current_steps": 31320, "total_steps": 42420, "loss": 0.3501, "lr": 0.00019450251610941583, "epoch": 14.766619519094768, "percentage": 73.83, "elapsed_time": "1:41:37", "remaining_time": "0:36:00", "throughput": 5175.37, "total_tokens": 31556352} {"current_steps": 31325, "total_steps": 42420, "loss": 0.3595, "lr": 0.00019433968685459452, "epoch": 14.768976897689768, "percentage": 73.84, "elapsed_time": "1:41:38", "remaining_time": "0:35:59", "throughput": 5175.32, "total_tokens": 31560032} {"current_steps": 31330, "total_steps": 42420, "loss": 0.319, "lr": 0.00019417690934287447, "epoch": 14.77133427628477, "percentage": 73.86, "elapsed_time": "1:41:39", "remaining_time": "0:35:58", "throughput": 5175.34, "total_tokens": 31564768} {"current_steps": 31335, "total_steps": 42420, "loss": 0.3481, "lr": 0.00019401418360181106, "epoch": 14.773691654879773, "percentage": 73.87, "elapsed_time": "1:41:39", "remaining_time": "0:35:57", "throughput": 5175.38, "total_tokens": 31569536} {"current_steps": 31340, "total_steps": 42420, "loss": 0.3469, "lr": 0.00019385150965895104, "epoch": 14.776049033474775, "percentage": 73.88, "elapsed_time": "1:41:40", "remaining_time": "0:35:56", "throughput": 5175.41, "total_tokens": 31574560} {"current_steps": 31345, "total_steps": 42420, "loss": 0.3553, "lr": 0.0001936888875418325, "epoch": 14.778406412069778, "percentage": 73.89, "elapsed_time": "1:41:41", "remaining_time": "0:35:55", "throughput": 5175.47, "total_tokens": 31579584} {"current_steps": 31350, "total_steps": 42420, "loss": 0.3598, "lr": 0.00019352631727798426, "epoch": 14.78076379066478, "percentage": 73.9, "elapsed_time": "1:41:42", "remaining_time": "0:35:54", "throughput": 5175.52, "total_tokens": 31584800} {"current_steps": 31355, "total_steps": 42420, "loss": 0.3702, "lr": 0.00019336379889492683, "epoch": 14.783121169259783, "percentage": 73.92, "elapsed_time": "1:41:43", "remaining_time": "0:35:53", "throughput": 5175.48, "total_tokens": 31588736} {"current_steps": 31360, "total_steps": 42420, "loss": 0.334, "lr": 0.0001932013324201719, "epoch": 14.785478547854785, "percentage": 73.93, "elapsed_time": "1:41:44", "remaining_time": "0:35:52", "throughput": 5175.51, "total_tokens": 31593632} {"current_steps": 31365, "total_steps": 42420, "loss": 0.3454, "lr": 0.00019303891788122223, "epoch": 14.787835926449787, "percentage": 73.94, "elapsed_time": "1:41:45", "remaining_time": "0:35:51", "throughput": 5175.57, "total_tokens": 31599232} {"current_steps": 31370, "total_steps": 42420, "loss": 0.3208, "lr": 0.0001928765553055719, "epoch": 14.79019330504479, "percentage": 73.95, "elapsed_time": "1:41:46", "remaining_time": "0:35:50", "throughput": 5175.59, "total_tokens": 31603744} {"current_steps": 31375, "total_steps": 42420, "loss": 0.3783, "lr": 0.0001927142447207061, "epoch": 14.792550683639792, "percentage": 73.96, "elapsed_time": "1:41:47", "remaining_time": "0:35:49", "throughput": 5175.66, "total_tokens": 31609344} {"current_steps": 31380, "total_steps": 42420, "loss": 0.305, "lr": 0.00019255198615410118, "epoch": 14.794908062234795, "percentage": 73.97, "elapsed_time": "1:41:48", "remaining_time": "0:35:49", "throughput": 5175.77, "total_tokens": 31615552} {"current_steps": 31385, "total_steps": 42420, "loss": 0.3108, "lr": 0.00019238977963322506, "epoch": 14.797265440829797, "percentage": 73.99, "elapsed_time": "1:41:49", "remaining_time": "0:35:48", "throughput": 5175.77, "total_tokens": 31619936} {"current_steps": 31390, "total_steps": 42420, "loss": 0.3694, "lr": 0.00019222762518553615, "epoch": 14.7996228194248, "percentage": 74.0, "elapsed_time": "1:41:50", "remaining_time": "0:35:47", "throughput": 5175.87, "total_tokens": 31626112} {"current_steps": 31395, "total_steps": 42420, "loss": 0.3121, "lr": 0.0001920655228384846, "epoch": 14.801980198019802, "percentage": 74.01, "elapsed_time": "1:41:51", "remaining_time": "0:35:46", "throughput": 5175.89, "total_tokens": 31630976} {"current_steps": 31400, "total_steps": 42420, "loss": 0.3421, "lr": 0.00019190347261951164, "epoch": 14.804337576614804, "percentage": 74.02, "elapsed_time": "1:41:52", "remaining_time": "0:35:45", "throughput": 5175.96, "total_tokens": 31636736} {"current_steps": 31405, "total_steps": 42420, "loss": 0.347, "lr": 0.00019174147455604956, "epoch": 14.806694955209807, "percentage": 74.03, "elapsed_time": "1:41:53", "remaining_time": "0:35:44", "throughput": 5175.98, "total_tokens": 31641312} {"current_steps": 31410, "total_steps": 42420, "loss": 0.398, "lr": 0.00019157952867552192, "epoch": 14.809052333804809, "percentage": 74.05, "elapsed_time": "1:41:54", "remaining_time": "0:35:43", "throughput": 5176.02, "total_tokens": 31646784} {"current_steps": 31415, "total_steps": 42420, "loss": 0.2774, "lr": 0.00019141763500534338, "epoch": 14.811409712399811, "percentage": 74.06, "elapsed_time": "1:41:55", "remaining_time": "0:35:42", "throughput": 5176.09, "total_tokens": 31652288} {"current_steps": 31420, "total_steps": 42420, "loss": 0.3012, "lr": 0.0001912557935729198, "epoch": 14.813767090994814, "percentage": 74.07, "elapsed_time": "1:41:56", "remaining_time": "0:35:41", "throughput": 5176.19, "total_tokens": 31658176} {"current_steps": 31425, "total_steps": 42420, "loss": 0.3022, "lr": 0.00019109400440564834, "epoch": 14.816124469589816, "percentage": 74.08, "elapsed_time": "1:41:57", "remaining_time": "0:35:40", "throughput": 5176.23, "total_tokens": 31663424} {"current_steps": 31430, "total_steps": 42420, "loss": 0.3679, "lr": 0.00019093226753091685, "epoch": 14.818481848184819, "percentage": 74.09, "elapsed_time": "1:41:57", "remaining_time": "0:35:39", "throughput": 5176.26, "total_tokens": 31668160} {"current_steps": 31435, "total_steps": 42420, "loss": 0.3146, "lr": 0.00019077058297610483, "epoch": 14.820839226779821, "percentage": 74.1, "elapsed_time": "1:41:59", "remaining_time": "0:35:38", "throughput": 5176.4, "total_tokens": 31675232} {"current_steps": 31440, "total_steps": 42420, "loss": 0.3321, "lr": 0.0001906089507685827, "epoch": 14.823196605374823, "percentage": 74.12, "elapsed_time": "1:42:00", "remaining_time": "0:35:37", "throughput": 5176.41, "total_tokens": 31679680} {"current_steps": 31445, "total_steps": 42420, "loss": 0.3582, "lr": 0.00019044737093571202, "epoch": 14.825553983969826, "percentage": 74.13, "elapsed_time": "1:42:01", "remaining_time": "0:35:36", "throughput": 5176.49, "total_tokens": 31685696} {"current_steps": 31450, "total_steps": 42420, "loss": 0.2963, "lr": 0.0001902858435048455, "epoch": 14.827911362564828, "percentage": 74.14, "elapsed_time": "1:42:01", "remaining_time": "0:35:35", "throughput": 5176.5, "total_tokens": 31690496} {"current_steps": 31455, "total_steps": 42420, "loss": 0.3633, "lr": 0.0001901243685033271, "epoch": 14.83026874115983, "percentage": 74.15, "elapsed_time": "1:42:02", "remaining_time": "0:35:34", "throughput": 5176.53, "total_tokens": 31695552} {"current_steps": 31460, "total_steps": 42420, "loss": 0.376, "lr": 0.00018996294595849183, "epoch": 14.832626119754833, "percentage": 74.16, "elapsed_time": "1:42:03", "remaining_time": "0:35:33", "throughput": 5176.54, "total_tokens": 31700032} {"current_steps": 31465, "total_steps": 42420, "loss": 0.2916, "lr": 0.00018980157589766557, "epoch": 14.834983498349835, "percentage": 74.17, "elapsed_time": "1:42:04", "remaining_time": "0:35:32", "throughput": 5176.52, "total_tokens": 31704288} {"current_steps": 31470, "total_steps": 42420, "loss": 0.3078, "lr": 0.00018964025834816573, "epoch": 14.837340876944838, "percentage": 74.19, "elapsed_time": "1:42:05", "remaining_time": "0:35:31", "throughput": 5176.57, "total_tokens": 31709152} {"current_steps": 31475, "total_steps": 42420, "loss": 0.2962, "lr": 0.00018947899333730047, "epoch": 14.83969825553984, "percentage": 74.2, "elapsed_time": "1:42:06", "remaining_time": "0:35:30", "throughput": 5176.56, "total_tokens": 31713472} {"current_steps": 31480, "total_steps": 42420, "loss": 0.278, "lr": 0.00018931778089236923, "epoch": 14.842055634134843, "percentage": 74.21, "elapsed_time": "1:42:07", "remaining_time": "0:35:29", "throughput": 5176.57, "total_tokens": 31717984} {"current_steps": 31485, "total_steps": 42420, "loss": 0.4135, "lr": 0.0001891566210406626, "epoch": 14.844413012729845, "percentage": 74.22, "elapsed_time": "1:42:08", "remaining_time": "0:35:28", "throughput": 5176.59, "total_tokens": 31722912} {"current_steps": 31490, "total_steps": 42420, "loss": 0.3509, "lr": 0.00018899551380946222, "epoch": 14.846770391324847, "percentage": 74.23, "elapsed_time": "1:42:09", "remaining_time": "0:35:27", "throughput": 5176.73, "total_tokens": 31730176} {"current_steps": 31495, "total_steps": 42420, "loss": 0.3909, "lr": 0.00018883445922604077, "epoch": 14.84912776991985, "percentage": 74.25, "elapsed_time": "1:42:10", "remaining_time": "0:35:26", "throughput": 5176.88, "total_tokens": 31737184} {"current_steps": 31500, "total_steps": 42420, "loss": 0.3248, "lr": 0.0001886734573176621, "epoch": 14.851485148514852, "percentage": 74.26, "elapsed_time": "1:42:11", "remaining_time": "0:35:25", "throughput": 5176.91, "total_tokens": 31741952} {"current_steps": 31505, "total_steps": 42420, "loss": 0.3344, "lr": 0.0001885125081115811, "epoch": 14.853842527109855, "percentage": 74.27, "elapsed_time": "1:42:12", "remaining_time": "0:35:24", "throughput": 5176.92, "total_tokens": 31746432} {"current_steps": 31510, "total_steps": 42420, "loss": 0.328, "lr": 0.0001883516116350439, "epoch": 14.856199905704855, "percentage": 74.28, "elapsed_time": "1:42:13", "remaining_time": "0:35:23", "throughput": 5176.89, "total_tokens": 31750336} {"current_steps": 31515, "total_steps": 42420, "loss": 0.3612, "lr": 0.00018819076791528716, "epoch": 14.85855728429986, "percentage": 74.29, "elapsed_time": "1:42:14", "remaining_time": "0:35:22", "throughput": 5176.94, "total_tokens": 31755744} {"current_steps": 31520, "total_steps": 42420, "loss": 0.3696, "lr": 0.00018802997697953922, "epoch": 14.86091466289486, "percentage": 74.3, "elapsed_time": "1:42:15", "remaining_time": "0:35:21", "throughput": 5177.05, "total_tokens": 31762080} {"current_steps": 31525, "total_steps": 42420, "loss": 0.3386, "lr": 0.00018786923885501922, "epoch": 14.863272041489862, "percentage": 74.32, "elapsed_time": "1:42:16", "remaining_time": "0:35:20", "throughput": 5177.1, "total_tokens": 31767456} {"current_steps": 31530, "total_steps": 42420, "loss": 0.3615, "lr": 0.0001877085535689374, "epoch": 14.865629420084865, "percentage": 74.33, "elapsed_time": "1:42:17", "remaining_time": "0:35:19", "throughput": 5177.15, "total_tokens": 31772576} {"current_steps": 31535, "total_steps": 42420, "loss": 0.3494, "lr": 0.000187547921148495, "epoch": 14.867986798679867, "percentage": 74.34, "elapsed_time": "1:42:17", "remaining_time": "0:35:18", "throughput": 5177.15, "total_tokens": 31777088} {"current_steps": 31540, "total_steps": 42420, "loss": 0.3386, "lr": 0.0001873873416208845, "epoch": 14.87034417727487, "percentage": 74.35, "elapsed_time": "1:42:18", "remaining_time": "0:35:17", "throughput": 5177.19, "total_tokens": 31782080} {"current_steps": 31545, "total_steps": 42420, "loss": 0.3485, "lr": 0.00018722681501328914, "epoch": 14.872701555869872, "percentage": 74.36, "elapsed_time": "1:42:19", "remaining_time": "0:35:16", "throughput": 5177.3, "total_tokens": 31788224} {"current_steps": 31550, "total_steps": 42420, "loss": 0.3543, "lr": 0.00018706634135288365, "epoch": 14.875058934464874, "percentage": 74.38, "elapsed_time": "1:42:20", "remaining_time": "0:35:15", "throughput": 5177.35, "total_tokens": 31793312} {"current_steps": 31555, "total_steps": 42420, "loss": 0.3575, "lr": 0.00018690592066683304, "epoch": 14.877416313059877, "percentage": 74.39, "elapsed_time": "1:42:21", "remaining_time": "0:35:14", "throughput": 5177.39, "total_tokens": 31798144} {"current_steps": 31560, "total_steps": 42420, "loss": 0.3631, "lr": 0.000186745552982294, "epoch": 14.87977369165488, "percentage": 74.4, "elapsed_time": "1:42:22", "remaining_time": "0:35:13", "throughput": 5177.46, "total_tokens": 31803616} {"current_steps": 31565, "total_steps": 42420, "loss": 0.303, "lr": 0.00018658523832641418, "epoch": 14.882131070249882, "percentage": 74.41, "elapsed_time": "1:42:23", "remaining_time": "0:35:12", "throughput": 5177.44, "total_tokens": 31807808} {"current_steps": 31570, "total_steps": 42420, "loss": 0.3113, "lr": 0.0001864249767263319, "epoch": 14.884488448844884, "percentage": 74.42, "elapsed_time": "1:42:24", "remaining_time": "0:35:11", "throughput": 5177.51, "total_tokens": 31813216} {"current_steps": 31575, "total_steps": 42420, "loss": 0.3363, "lr": 0.000186264768209177, "epoch": 14.886845827439886, "percentage": 74.43, "elapsed_time": "1:42:25", "remaining_time": "0:35:10", "throughput": 5177.63, "total_tokens": 31819648} {"current_steps": 31580, "total_steps": 42420, "loss": 0.3785, "lr": 0.0001861046128020698, "epoch": 14.889203206034889, "percentage": 74.45, "elapsed_time": "1:42:26", "remaining_time": "0:35:09", "throughput": 5177.64, "total_tokens": 31824000} {"current_steps": 31585, "total_steps": 42420, "loss": 0.3153, "lr": 0.0001859445105321221, "epoch": 14.891560584629891, "percentage": 74.46, "elapsed_time": "1:42:27", "remaining_time": "0:35:08", "throughput": 5177.68, "total_tokens": 31828928} {"current_steps": 31590, "total_steps": 42420, "loss": 0.3867, "lr": 0.0001857844614264365, "epoch": 14.893917963224894, "percentage": 74.47, "elapsed_time": "1:42:28", "remaining_time": "0:35:07", "throughput": 5177.72, "total_tokens": 31833728} {"current_steps": 31595, "total_steps": 42420, "loss": 0.3268, "lr": 0.00018562446551210638, "epoch": 14.896275341819896, "percentage": 74.48, "elapsed_time": "1:42:29", "remaining_time": "0:35:06", "throughput": 5177.76, "total_tokens": 31838880} {"current_steps": 31600, "total_steps": 42420, "loss": 0.3546, "lr": 0.00018546452281621647, "epoch": 14.898632720414899, "percentage": 74.49, "elapsed_time": "1:42:29", "remaining_time": "0:35:05", "throughput": 5177.72, "total_tokens": 31842656} {"current_steps": 31605, "total_steps": 42420, "loss": 0.3515, "lr": 0.0001853046333658423, "epoch": 14.900990099009901, "percentage": 74.5, "elapsed_time": "1:42:30", "remaining_time": "0:35:04", "throughput": 5177.77, "total_tokens": 31847840} {"current_steps": 31610, "total_steps": 42420, "loss": 0.3196, "lr": 0.0001851447971880505, "epoch": 14.903347477604903, "percentage": 74.52, "elapsed_time": "1:42:31", "remaining_time": "0:35:03", "throughput": 5177.78, "total_tokens": 31852256} {"current_steps": 31615, "total_steps": 42420, "loss": 0.3814, "lr": 0.00018498501430989862, "epoch": 14.905704856199906, "percentage": 74.53, "elapsed_time": "1:42:32", "remaining_time": "0:35:02", "throughput": 5177.85, "total_tokens": 31857856} {"current_steps": 31620, "total_steps": 42420, "loss": 0.3374, "lr": 0.00018482528475843513, "epoch": 14.908062234794908, "percentage": 74.54, "elapsed_time": "1:42:33", "remaining_time": "0:35:01", "throughput": 5177.85, "total_tokens": 31862176} {"current_steps": 31625, "total_steps": 42420, "loss": 0.31, "lr": 0.0001846656085606996, "epoch": 14.91041961338991, "percentage": 74.55, "elapsed_time": "1:42:34", "remaining_time": "0:35:00", "throughput": 5177.9, "total_tokens": 31867264} {"current_steps": 31630, "total_steps": 42420, "loss": 0.3077, "lr": 0.00018450598574372258, "epoch": 14.912776991984913, "percentage": 74.56, "elapsed_time": "1:42:35", "remaining_time": "0:34:59", "throughput": 5177.96, "total_tokens": 31872736} {"current_steps": 31635, "total_steps": 42420, "loss": 0.3031, "lr": 0.00018434641633452526, "epoch": 14.915134370579915, "percentage": 74.58, "elapsed_time": "1:42:36", "remaining_time": "0:34:58", "throughput": 5177.95, "total_tokens": 31877280} {"current_steps": 31640, "total_steps": 42420, "loss": 0.3571, "lr": 0.00018418690036012015, "epoch": 14.917491749174918, "percentage": 74.59, "elapsed_time": "1:42:37", "remaining_time": "0:34:57", "throughput": 5177.97, "total_tokens": 31881920} {"current_steps": 31645, "total_steps": 42420, "loss": 0.3407, "lr": 0.00018402743784751063, "epoch": 14.91984912776992, "percentage": 74.6, "elapsed_time": "1:42:38", "remaining_time": "0:34:56", "throughput": 5178.07, "total_tokens": 31888128} {"current_steps": 31650, "total_steps": 42420, "loss": 0.3446, "lr": 0.00018386802882369097, "epoch": 14.922206506364923, "percentage": 74.61, "elapsed_time": "1:42:39", "remaining_time": "0:34:55", "throughput": 5178.12, "total_tokens": 31893312} {"current_steps": 31655, "total_steps": 42420, "loss": 0.3581, "lr": 0.00018370867331564639, "epoch": 14.924563884959925, "percentage": 74.62, "elapsed_time": "1:42:40", "remaining_time": "0:34:54", "throughput": 5178.16, "total_tokens": 31898144} {"current_steps": 31660, "total_steps": 42420, "loss": 0.3381, "lr": 0.0001835493713503531, "epoch": 14.926921263554927, "percentage": 74.63, "elapsed_time": "1:42:41", "remaining_time": "0:34:53", "throughput": 5178.24, "total_tokens": 31903872} {"current_steps": 31665, "total_steps": 42420, "loss": 0.3629, "lr": 0.00018339012295477824, "epoch": 14.92927864214993, "percentage": 74.65, "elapsed_time": "1:42:42", "remaining_time": "0:34:53", "throughput": 5178.38, "total_tokens": 31910912} {"current_steps": 31670, "total_steps": 42420, "loss": 0.3268, "lr": 0.00018323092815588, "epoch": 14.931636020744932, "percentage": 74.66, "elapsed_time": "1:42:43", "remaining_time": "0:34:52", "throughput": 5178.39, "total_tokens": 31915456} {"current_steps": 31675, "total_steps": 42420, "loss": 0.3595, "lr": 0.000183071786980607, "epoch": 14.933993399339935, "percentage": 74.67, "elapsed_time": "1:42:44", "remaining_time": "0:34:51", "throughput": 5178.44, "total_tokens": 31920544} {"current_steps": 31680, "total_steps": 42420, "loss": 0.3032, "lr": 0.0001829126994558994, "epoch": 14.936350777934937, "percentage": 74.68, "elapsed_time": "1:42:45", "remaining_time": "0:34:50", "throughput": 5178.46, "total_tokens": 31925568} {"current_steps": 31685, "total_steps": 42420, "loss": 0.3166, "lr": 0.000182753665608688, "epoch": 14.93870815652994, "percentage": 74.69, "elapsed_time": "1:42:46", "remaining_time": "0:34:49", "throughput": 5178.5, "total_tokens": 31930816} {"current_steps": 31690, "total_steps": 42420, "loss": 0.337, "lr": 0.00018259468546589437, "epoch": 14.941065535124942, "percentage": 74.71, "elapsed_time": "1:42:46", "remaining_time": "0:34:48", "throughput": 5178.48, "total_tokens": 31934816} {"current_steps": 31695, "total_steps": 42420, "loss": 0.3483, "lr": 0.00018243575905443126, "epoch": 14.943422913719944, "percentage": 74.72, "elapsed_time": "1:42:47", "remaining_time": "0:34:47", "throughput": 5178.53, "total_tokens": 31940352} {"current_steps": 31700, "total_steps": 42420, "loss": 0.3189, "lr": 0.0001822768864012022, "epoch": 14.945780292314947, "percentage": 74.73, "elapsed_time": "1:42:48", "remaining_time": "0:34:46", "throughput": 5178.52, "total_tokens": 31944480} {"current_steps": 31705, "total_steps": 42420, "loss": 0.3216, "lr": 0.0001821180675331016, "epoch": 14.948137670909949, "percentage": 74.74, "elapsed_time": "1:42:49", "remaining_time": "0:34:45", "throughput": 5178.55, "total_tokens": 31949408} {"current_steps": 31710, "total_steps": 42420, "loss": 0.3467, "lr": 0.0001819593024770149, "epoch": 14.950495049504951, "percentage": 74.75, "elapsed_time": "1:42:50", "remaining_time": "0:34:44", "throughput": 5178.58, "total_tokens": 31954400} {"current_steps": 31715, "total_steps": 42420, "loss": 0.3471, "lr": 0.00018180059125981823, "epoch": 14.952852428099952, "percentage": 74.76, "elapsed_time": "1:42:51", "remaining_time": "0:34:43", "throughput": 5178.66, "total_tokens": 31960160} {"current_steps": 31720, "total_steps": 42420, "loss": 0.3291, "lr": 0.00018164193390837895, "epoch": 14.955209806694956, "percentage": 74.78, "elapsed_time": "1:42:52", "remaining_time": "0:34:42", "throughput": 5178.76, "total_tokens": 31965952} {"current_steps": 31725, "total_steps": 42420, "loss": 0.3216, "lr": 0.00018148333044955468, "epoch": 14.957567185289957, "percentage": 74.79, "elapsed_time": "1:42:53", "remaining_time": "0:34:41", "throughput": 5178.8, "total_tokens": 31970912} {"current_steps": 31730, "total_steps": 42420, "loss": 0.3348, "lr": 0.00018132478091019443, "epoch": 14.95992456388496, "percentage": 74.8, "elapsed_time": "1:42:54", "remaining_time": "0:34:40", "throughput": 5178.81, "total_tokens": 31975424} {"current_steps": 31735, "total_steps": 42420, "loss": 0.3072, "lr": 0.00018116628531713796, "epoch": 14.962281942479962, "percentage": 74.81, "elapsed_time": "1:42:55", "remaining_time": "0:34:39", "throughput": 5178.84, "total_tokens": 31980160} {"current_steps": 31740, "total_steps": 42420, "loss": 0.3258, "lr": 0.0001810078436972159, "epoch": 14.964639321074964, "percentage": 74.82, "elapsed_time": "1:42:56", "remaining_time": "0:34:38", "throughput": 5178.89, "total_tokens": 31985504} {"current_steps": 31745, "total_steps": 42420, "loss": 0.3142, "lr": 0.0001808494560772498, "epoch": 14.966996699669966, "percentage": 74.83, "elapsed_time": "1:42:57", "remaining_time": "0:34:37", "throughput": 5178.99, "total_tokens": 31991392} {"current_steps": 31750, "total_steps": 42420, "loss": 0.3073, "lr": 0.00018069112248405184, "epoch": 14.969354078264969, "percentage": 74.85, "elapsed_time": "1:42:58", "remaining_time": "0:34:36", "throughput": 5179.04, "total_tokens": 31996640} {"current_steps": 31755, "total_steps": 42420, "loss": 0.3509, "lr": 0.0001805328429444253, "epoch": 14.971711456859971, "percentage": 74.86, "elapsed_time": "1:42:58", "remaining_time": "0:34:35", "throughput": 5179.02, "total_tokens": 32000896} {"current_steps": 31760, "total_steps": 42420, "loss": 0.282, "lr": 0.00018037461748516436, "epoch": 14.974068835454974, "percentage": 74.87, "elapsed_time": "1:42:59", "remaining_time": "0:34:34", "throughput": 5179.09, "total_tokens": 32006720} {"current_steps": 31765, "total_steps": 42420, "loss": 0.3435, "lr": 0.0001802164461330536, "epoch": 14.976426214049976, "percentage": 74.88, "elapsed_time": "1:43:00", "remaining_time": "0:34:33", "throughput": 5179.08, "total_tokens": 32010976} {"current_steps": 31770, "total_steps": 42420, "loss": 0.2936, "lr": 0.0001800583289148689, "epoch": 14.978783592644978, "percentage": 74.89, "elapsed_time": "1:43:01", "remaining_time": "0:34:32", "throughput": 5179.06, "total_tokens": 32015040} {"current_steps": 31775, "total_steps": 42420, "loss": 0.2729, "lr": 0.00017990026585737679, "epoch": 14.98114097123998, "percentage": 74.91, "elapsed_time": "1:43:02", "remaining_time": "0:34:31", "throughput": 5179.09, "total_tokens": 32019872} {"current_steps": 31780, "total_steps": 42420, "loss": 0.3638, "lr": 0.0001797422569873347, "epoch": 14.983498349834983, "percentage": 74.92, "elapsed_time": "1:43:03", "remaining_time": "0:34:30", "throughput": 5179.12, "total_tokens": 32024864} {"current_steps": 31785, "total_steps": 42420, "loss": 0.3103, "lr": 0.00017958430233149086, "epoch": 14.985855728429986, "percentage": 74.93, "elapsed_time": "1:43:04", "remaining_time": "0:34:29", "throughput": 5179.12, "total_tokens": 32029344} {"current_steps": 31790, "total_steps": 42420, "loss": 0.2909, "lr": 0.00017942640191658427, "epoch": 14.988213107024988, "percentage": 74.94, "elapsed_time": "1:43:05", "remaining_time": "0:34:28", "throughput": 5179.16, "total_tokens": 32034336} {"current_steps": 31795, "total_steps": 42420, "loss": 0.33, "lr": 0.0001792685557693448, "epoch": 14.99057048561999, "percentage": 74.95, "elapsed_time": "1:43:06", "remaining_time": "0:34:27", "throughput": 5179.22, "total_tokens": 32039360} {"current_steps": 31800, "total_steps": 42420, "loss": 0.3483, "lr": 0.00017911076391649328, "epoch": 14.992927864214993, "percentage": 74.96, "elapsed_time": "1:43:07", "remaining_time": "0:34:26", "throughput": 5179.29, "total_tokens": 32045248} {"current_steps": 31805, "total_steps": 42420, "loss": 0.3096, "lr": 0.00017895302638474091, "epoch": 14.995285242809995, "percentage": 74.98, "elapsed_time": "1:43:08", "remaining_time": "0:34:25", "throughput": 5179.28, "total_tokens": 32049408} {"current_steps": 31810, "total_steps": 42420, "loss": 0.3817, "lr": 0.00017879534320079004, "epoch": 14.997642621404998, "percentage": 74.99, "elapsed_time": "1:43:09", "remaining_time": "0:34:24", "throughput": 5179.34, "total_tokens": 32054944} {"current_steps": 31815, "total_steps": 42420, "loss": 0.3598, "lr": 0.00017863771439133392, "epoch": 15.0, "percentage": 75.0, "elapsed_time": "1:43:09", "remaining_time": "0:34:23", "throughput": 5179.26, "total_tokens": 32059584} {"current_steps": 31815, "total_steps": 42420, "eval_loss": 0.33835846185684204, "epoch": 15.0, "percentage": 75.0, "elapsed_time": "1:43:28", "remaining_time": "0:34:29", "throughput": 5163.41, "total_tokens": 32059584} {"current_steps": 31820, "total_steps": 42420, "loss": 0.3632, "lr": 0.0001784801399830563, "epoch": 15.002357378595002, "percentage": 75.01, "elapsed_time": "1:43:32", "remaining_time": "0:34:29", "throughput": 5161.74, "total_tokens": 32065152} {"current_steps": 31825, "total_steps": 42420, "loss": 0.3322, "lr": 0.00017832262000263183, "epoch": 15.004714757190005, "percentage": 75.02, "elapsed_time": "1:43:32", "remaining_time": "0:34:28", "throughput": 5161.75, "total_tokens": 32069536} {"current_steps": 31830, "total_steps": 42420, "loss": 0.3087, "lr": 0.00017816515447672605, "epoch": 15.007072135785007, "percentage": 75.04, "elapsed_time": "1:43:33", "remaining_time": "0:34:27", "throughput": 5161.81, "total_tokens": 32074784} {"current_steps": 31835, "total_steps": 42420, "loss": 0.3299, "lr": 0.00017800774343199516, "epoch": 15.00942951438001, "percentage": 75.05, "elapsed_time": "1:43:34", "remaining_time": "0:34:26", "throughput": 5161.81, "total_tokens": 32079072} {"current_steps": 31840, "total_steps": 42420, "loss": 0.3322, "lr": 0.00017785038689508643, "epoch": 15.011786892975012, "percentage": 75.06, "elapsed_time": "1:43:35", "remaining_time": "0:34:25", "throughput": 5161.87, "total_tokens": 32084480} {"current_steps": 31845, "total_steps": 42420, "loss": 0.3761, "lr": 0.00017769308489263714, "epoch": 15.014144271570014, "percentage": 75.07, "elapsed_time": "1:43:36", "remaining_time": "0:34:24", "throughput": 5161.89, "total_tokens": 32088992} {"current_steps": 31850, "total_steps": 42420, "loss": 0.3199, "lr": 0.00017753583745127617, "epoch": 15.016501650165017, "percentage": 75.08, "elapsed_time": "1:43:37", "remaining_time": "0:34:23", "throughput": 5161.97, "total_tokens": 32094720} {"current_steps": 31855, "total_steps": 42420, "loss": 0.2717, "lr": 0.00017737864459762282, "epoch": 15.01885902876002, "percentage": 75.09, "elapsed_time": "1:43:38", "remaining_time": "0:34:22", "throughput": 5162.0, "total_tokens": 32099200} {"current_steps": 31860, "total_steps": 42420, "loss": 0.3397, "lr": 0.0001772215063582871, "epoch": 15.021216407355022, "percentage": 75.11, "elapsed_time": "1:43:39", "remaining_time": "0:34:21", "throughput": 5162.12, "total_tokens": 32105440} {"current_steps": 31865, "total_steps": 42420, "loss": 0.3653, "lr": 0.0001770644227598699, "epoch": 15.023573785950024, "percentage": 75.12, "elapsed_time": "1:43:40", "remaining_time": "0:34:20", "throughput": 5162.14, "total_tokens": 32110176} {"current_steps": 31870, "total_steps": 42420, "loss": 0.2435, "lr": 0.00017690739382896276, "epoch": 15.025931164545026, "percentage": 75.13, "elapsed_time": "1:43:41", "remaining_time": "0:34:19", "throughput": 5162.14, "total_tokens": 32114880} {"current_steps": 31875, "total_steps": 42420, "loss": 0.2918, "lr": 0.00017675041959214804, "epoch": 15.028288543140029, "percentage": 75.14, "elapsed_time": "1:43:42", "remaining_time": "0:34:18", "throughput": 5162.15, "total_tokens": 32119392} {"current_steps": 31880, "total_steps": 42420, "loss": 0.4168, "lr": 0.00017659350007599906, "epoch": 15.030645921735031, "percentage": 75.15, "elapsed_time": "1:43:42", "remaining_time": "0:34:17", "throughput": 5162.18, "total_tokens": 32124224} {"current_steps": 31885, "total_steps": 42420, "loss": 0.2873, "lr": 0.00017643663530707916, "epoch": 15.033003300330034, "percentage": 75.17, "elapsed_time": "1:43:43", "remaining_time": "0:34:16", "throughput": 5162.2, "total_tokens": 32128768} {"current_steps": 31890, "total_steps": 42420, "loss": 0.2967, "lr": 0.00017627982531194313, "epoch": 15.035360678925036, "percentage": 75.18, "elapsed_time": "1:43:44", "remaining_time": "0:34:15", "throughput": 5162.3, "total_tokens": 32134592} {"current_steps": 31895, "total_steps": 42420, "loss": 0.3261, "lr": 0.00017612307011713625, "epoch": 15.037718057520038, "percentage": 75.19, "elapsed_time": "1:43:45", "remaining_time": "0:34:14", "throughput": 5162.34, "total_tokens": 32140000} {"current_steps": 31900, "total_steps": 42420, "loss": 0.3657, "lr": 0.0001759663697491944, "epoch": 15.04007543611504, "percentage": 75.2, "elapsed_time": "1:43:46", "remaining_time": "0:34:13", "throughput": 5162.48, "total_tokens": 32146432} {"current_steps": 31905, "total_steps": 42420, "loss": 0.3298, "lr": 0.00017580972423464442, "epoch": 15.042432814710043, "percentage": 75.21, "elapsed_time": "1:43:47", "remaining_time": "0:34:12", "throughput": 5162.51, "total_tokens": 32151264} {"current_steps": 31910, "total_steps": 42420, "loss": 0.3873, "lr": 0.00017565313360000384, "epoch": 15.044790193305046, "percentage": 75.22, "elapsed_time": "1:43:48", "remaining_time": "0:34:11", "throughput": 5162.58, "total_tokens": 32156704} {"current_steps": 31915, "total_steps": 42420, "loss": 0.3452, "lr": 0.00017549659787178051, "epoch": 15.047147571900048, "percentage": 75.24, "elapsed_time": "1:43:49", "remaining_time": "0:34:10", "throughput": 5162.63, "total_tokens": 32161888} {"current_steps": 31920, "total_steps": 42420, "loss": 0.3113, "lr": 0.0001753401170764734, "epoch": 15.049504950495049, "percentage": 75.25, "elapsed_time": "1:43:50", "remaining_time": "0:34:09", "throughput": 5162.73, "total_tokens": 32167840} {"current_steps": 31925, "total_steps": 42420, "loss": 0.3009, "lr": 0.00017518369124057203, "epoch": 15.051862329090051, "percentage": 75.26, "elapsed_time": "1:43:51", "remaining_time": "0:34:08", "throughput": 5162.76, "total_tokens": 32172960} {"current_steps": 31930, "total_steps": 42420, "loss": 0.3586, "lr": 0.00017502732039055685, "epoch": 15.054219707685053, "percentage": 75.27, "elapsed_time": "1:43:52", "remaining_time": "0:34:07", "throughput": 5162.88, "total_tokens": 32179232} {"current_steps": 31935, "total_steps": 42420, "loss": 0.3826, "lr": 0.00017487100455289845, "epoch": 15.056577086280056, "percentage": 75.28, "elapsed_time": "1:43:53", "remaining_time": "0:34:06", "throughput": 5162.94, "total_tokens": 32184576} {"current_steps": 31940, "total_steps": 42420, "loss": 0.3278, "lr": 0.0001747147437540586, "epoch": 15.058934464875058, "percentage": 75.29, "elapsed_time": "1:43:54", "remaining_time": "0:34:05", "throughput": 5163.01, "total_tokens": 32190080} {"current_steps": 31945, "total_steps": 42420, "loss": 0.3473, "lr": 0.0001745585380204896, "epoch": 15.06129184347006, "percentage": 75.31, "elapsed_time": "1:43:55", "remaining_time": "0:34:04", "throughput": 5163.03, "total_tokens": 32195104} {"current_steps": 31950, "total_steps": 42420, "loss": 0.3329, "lr": 0.00017440238737863445, "epoch": 15.063649222065063, "percentage": 75.32, "elapsed_time": "1:43:56", "remaining_time": "0:34:03", "throughput": 5163.04, "total_tokens": 32199840} {"current_steps": 31955, "total_steps": 42420, "loss": 0.3514, "lr": 0.00017424629185492675, "epoch": 15.066006600660065, "percentage": 75.33, "elapsed_time": "1:43:57", "remaining_time": "0:34:02", "throughput": 5163.0, "total_tokens": 32203616} {"current_steps": 31960, "total_steps": 42420, "loss": 0.3478, "lr": 0.00017409025147579088, "epoch": 15.068363979255068, "percentage": 75.34, "elapsed_time": "1:43:58", "remaining_time": "0:34:01", "throughput": 5163.15, "total_tokens": 32210400} {"current_steps": 31965, "total_steps": 42420, "loss": 0.3436, "lr": 0.00017393426626764174, "epoch": 15.07072135785007, "percentage": 75.35, "elapsed_time": "1:43:59", "remaining_time": "0:34:00", "throughput": 5163.18, "total_tokens": 32215456} {"current_steps": 31970, "total_steps": 42420, "loss": 0.3444, "lr": 0.00017377833625688527, "epoch": 15.073078736445073, "percentage": 75.37, "elapsed_time": "1:44:00", "remaining_time": "0:33:59", "throughput": 5163.23, "total_tokens": 32220704} {"current_steps": 31975, "total_steps": 42420, "loss": 0.3557, "lr": 0.00017362246146991739, "epoch": 15.075436115040075, "percentage": 75.38, "elapsed_time": "1:44:01", "remaining_time": "0:33:58", "throughput": 5163.32, "total_tokens": 32226336} {"current_steps": 31980, "total_steps": 42420, "loss": 0.3543, "lr": 0.00017346664193312517, "epoch": 15.077793493635077, "percentage": 75.39, "elapsed_time": "1:44:02", "remaining_time": "0:33:57", "throughput": 5163.44, "total_tokens": 32232896} {"current_steps": 31985, "total_steps": 42420, "loss": 0.3054, "lr": 0.00017331087767288628, "epoch": 15.08015087223008, "percentage": 75.4, "elapsed_time": "1:44:03", "remaining_time": "0:33:56", "throughput": 5163.5, "total_tokens": 32238112} {"current_steps": 31990, "total_steps": 42420, "loss": 0.3317, "lr": 0.00017315516871556892, "epoch": 15.082508250825082, "percentage": 75.41, "elapsed_time": "1:44:04", "remaining_time": "0:33:55", "throughput": 5163.53, "total_tokens": 32242912} {"current_steps": 31995, "total_steps": 42420, "loss": 0.3435, "lr": 0.00017299951508753197, "epoch": 15.084865629420085, "percentage": 75.42, "elapsed_time": "1:44:05", "remaining_time": "0:33:54", "throughput": 5163.56, "total_tokens": 32248000} {"current_steps": 32000, "total_steps": 42420, "loss": 0.3542, "lr": 0.00017284391681512496, "epoch": 15.087223008015087, "percentage": 75.44, "elapsed_time": "1:44:06", "remaining_time": "0:33:53", "throughput": 5163.58, "total_tokens": 32252576} {"current_steps": 32005, "total_steps": 42420, "loss": 0.3278, "lr": 0.00017268837392468806, "epoch": 15.08958038661009, "percentage": 75.45, "elapsed_time": "1:44:07", "remaining_time": "0:33:52", "throughput": 5163.6, "total_tokens": 32257152} {"current_steps": 32010, "total_steps": 42420, "loss": 0.3739, "lr": 0.0001725328864425522, "epoch": 15.091937765205092, "percentage": 75.46, "elapsed_time": "1:44:08", "remaining_time": "0:33:51", "throughput": 5163.7, "total_tokens": 32263296} {"current_steps": 32015, "total_steps": 42420, "loss": 0.3269, "lr": 0.00017237745439503837, "epoch": 15.094295143800094, "percentage": 75.47, "elapsed_time": "1:44:08", "remaining_time": "0:33:50", "throughput": 5163.7, "total_tokens": 32267616} {"current_steps": 32020, "total_steps": 42420, "loss": 0.2841, "lr": 0.00017222207780845885, "epoch": 15.096652522395097, "percentage": 75.48, "elapsed_time": "1:44:09", "remaining_time": "0:33:49", "throughput": 5163.77, "total_tokens": 32273024} {"current_steps": 32025, "total_steps": 42420, "loss": 0.2828, "lr": 0.0001720667567091162, "epoch": 15.099009900990099, "percentage": 75.5, "elapsed_time": "1:44:10", "remaining_time": "0:33:48", "throughput": 5163.8, "total_tokens": 32277824} {"current_steps": 32030, "total_steps": 42420, "loss": 0.3698, "lr": 0.00017191149112330363, "epoch": 15.101367279585101, "percentage": 75.51, "elapsed_time": "1:44:11", "remaining_time": "0:33:47", "throughput": 5163.85, "total_tokens": 32282816} {"current_steps": 32035, "total_steps": 42420, "loss": 0.3414, "lr": 0.000171756281077305, "epoch": 15.103724658180104, "percentage": 75.52, "elapsed_time": "1:44:12", "remaining_time": "0:33:46", "throughput": 5163.89, "total_tokens": 32287872} {"current_steps": 32040, "total_steps": 42420, "loss": 0.3422, "lr": 0.0001716011265973947, "epoch": 15.106082036775106, "percentage": 75.53, "elapsed_time": "1:44:13", "remaining_time": "0:33:45", "throughput": 5163.91, "total_tokens": 32292320} {"current_steps": 32045, "total_steps": 42420, "loss": 0.2983, "lr": 0.0001714460277098378, "epoch": 15.108439415370109, "percentage": 75.54, "elapsed_time": "1:44:14", "remaining_time": "0:33:44", "throughput": 5163.95, "total_tokens": 32297536} {"current_steps": 32050, "total_steps": 42420, "loss": 0.3338, "lr": 0.00017129098444089002, "epoch": 15.110796793965111, "percentage": 75.55, "elapsed_time": "1:44:15", "remaining_time": "0:33:43", "throughput": 5163.97, "total_tokens": 32302080} {"current_steps": 32055, "total_steps": 42420, "loss": 0.2947, "lr": 0.00017113599681679726, "epoch": 15.113154172560114, "percentage": 75.57, "elapsed_time": "1:44:16", "remaining_time": "0:33:42", "throughput": 5163.98, "total_tokens": 32306560} {"current_steps": 32060, "total_steps": 42420, "loss": 0.2887, "lr": 0.00017098106486379644, "epoch": 15.115511551155116, "percentage": 75.58, "elapsed_time": "1:44:17", "remaining_time": "0:33:41", "throughput": 5164.03, "total_tokens": 32311744} {"current_steps": 32065, "total_steps": 42420, "loss": 0.3318, "lr": 0.00017082618860811495, "epoch": 15.117868929750118, "percentage": 75.59, "elapsed_time": "1:44:17", "remaining_time": "0:33:40", "throughput": 5164.02, "total_tokens": 32315968} {"current_steps": 32070, "total_steps": 42420, "loss": 0.329, "lr": 0.00017067136807597061, "epoch": 15.12022630834512, "percentage": 75.6, "elapsed_time": "1:44:18", "remaining_time": "0:33:39", "throughput": 5164.06, "total_tokens": 32320960} {"current_steps": 32075, "total_steps": 42420, "loss": 0.3338, "lr": 0.00017051660329357206, "epoch": 15.122583686940123, "percentage": 75.61, "elapsed_time": "1:44:19", "remaining_time": "0:33:38", "throughput": 5164.11, "total_tokens": 32326336} {"current_steps": 32080, "total_steps": 42420, "loss": 0.3603, "lr": 0.00017036189428711819, "epoch": 15.124941065535126, "percentage": 75.62, "elapsed_time": "1:44:20", "remaining_time": "0:33:37", "throughput": 5164.18, "total_tokens": 32331840} {"current_steps": 32085, "total_steps": 42420, "loss": 0.2713, "lr": 0.00017020724108279874, "epoch": 15.127298444130128, "percentage": 75.64, "elapsed_time": "1:44:21", "remaining_time": "0:33:36", "throughput": 5164.19, "total_tokens": 32336224} {"current_steps": 32090, "total_steps": 42420, "loss": 0.2803, "lr": 0.0001700526437067939, "epoch": 15.12965582272513, "percentage": 75.65, "elapsed_time": "1:44:22", "remaining_time": "0:33:35", "throughput": 5164.21, "total_tokens": 32340992} {"current_steps": 32095, "total_steps": 42420, "loss": 0.2866, "lr": 0.0001698981021852742, "epoch": 15.132013201320133, "percentage": 75.66, "elapsed_time": "1:44:23", "remaining_time": "0:33:34", "throughput": 5164.24, "total_tokens": 32345760} {"current_steps": 32100, "total_steps": 42420, "loss": 0.3395, "lr": 0.0001697436165444009, "epoch": 15.134370579915135, "percentage": 75.67, "elapsed_time": "1:44:24", "remaining_time": "0:33:33", "throughput": 5164.29, "total_tokens": 32351008} {"current_steps": 32105, "total_steps": 42420, "loss": 0.32, "lr": 0.000169589186810326, "epoch": 15.136727958510138, "percentage": 75.68, "elapsed_time": "1:44:25", "remaining_time": "0:33:33", "throughput": 5164.39, "total_tokens": 32356928} {"current_steps": 32110, "total_steps": 42420, "loss": 0.2739, "lr": 0.00016943481300919162, "epoch": 15.13908533710514, "percentage": 75.7, "elapsed_time": "1:44:26", "remaining_time": "0:33:31", "throughput": 5164.42, "total_tokens": 32361696} {"current_steps": 32115, "total_steps": 42420, "loss": 0.3137, "lr": 0.0001692804951671308, "epoch": 15.14144271570014, "percentage": 75.71, "elapsed_time": "1:44:27", "remaining_time": "0:33:30", "throughput": 5164.37, "total_tokens": 32365344} {"current_steps": 32120, "total_steps": 42420, "loss": 0.2862, "lr": 0.00016912623331026677, "epoch": 15.143800094295143, "percentage": 75.72, "elapsed_time": "1:44:27", "remaining_time": "0:33:29", "throughput": 5164.41, "total_tokens": 32370432} {"current_steps": 32125, "total_steps": 42420, "loss": 0.3742, "lr": 0.00016897202746471356, "epoch": 15.146157472890145, "percentage": 75.73, "elapsed_time": "1:44:28", "remaining_time": "0:33:28", "throughput": 5164.44, "total_tokens": 32375424} {"current_steps": 32130, "total_steps": 42420, "loss": 0.3463, "lr": 0.00016881787765657576, "epoch": 15.148514851485148, "percentage": 75.74, "elapsed_time": "1:44:29", "remaining_time": "0:33:27", "throughput": 5164.45, "total_tokens": 32379968} {"current_steps": 32135, "total_steps": 42420, "loss": 0.339, "lr": 0.0001686637839119479, "epoch": 15.15087223008015, "percentage": 75.75, "elapsed_time": "1:44:30", "remaining_time": "0:33:26", "throughput": 5164.49, "total_tokens": 32385088} {"current_steps": 32140, "total_steps": 42420, "loss": 0.3755, "lr": 0.00016850974625691584, "epoch": 15.153229608675153, "percentage": 75.77, "elapsed_time": "1:44:31", "remaining_time": "0:33:26", "throughput": 5164.62, "total_tokens": 32391616} {"current_steps": 32145, "total_steps": 42420, "loss": 0.3442, "lr": 0.0001683557647175552, "epoch": 15.155586987270155, "percentage": 75.78, "elapsed_time": "1:44:32", "remaining_time": "0:33:25", "throughput": 5164.61, "total_tokens": 32395872} {"current_steps": 32150, "total_steps": 42420, "loss": 0.3198, "lr": 0.0001682018393199326, "epoch": 15.157944365865157, "percentage": 75.79, "elapsed_time": "1:44:33", "remaining_time": "0:33:24", "throughput": 5164.66, "total_tokens": 32401088} {"current_steps": 32155, "total_steps": 42420, "loss": 0.334, "lr": 0.00016804797009010503, "epoch": 15.16030174446016, "percentage": 75.8, "elapsed_time": "1:44:34", "remaining_time": "0:33:23", "throughput": 5164.65, "total_tokens": 32405216} {"current_steps": 32160, "total_steps": 42420, "loss": 0.3369, "lr": 0.00016789415705411982, "epoch": 15.162659123055162, "percentage": 75.81, "elapsed_time": "1:44:35", "remaining_time": "0:33:22", "throughput": 5164.79, "total_tokens": 32412384} {"current_steps": 32165, "total_steps": 42420, "loss": 0.3649, "lr": 0.000167740400238015, "epoch": 15.165016501650165, "percentage": 75.83, "elapsed_time": "1:44:36", "remaining_time": "0:33:21", "throughput": 5164.78, "total_tokens": 32416800} {"current_steps": 32170, "total_steps": 42420, "loss": 0.3725, "lr": 0.00016758669966781902, "epoch": 15.167373880245167, "percentage": 75.84, "elapsed_time": "1:44:37", "remaining_time": "0:33:20", "throughput": 5164.84, "total_tokens": 32421984} {"current_steps": 32175, "total_steps": 42420, "loss": 0.3081, "lr": 0.00016743305536955072, "epoch": 15.16973125884017, "percentage": 75.85, "elapsed_time": "1:44:38", "remaining_time": "0:33:19", "throughput": 5164.87, "total_tokens": 32426976} {"current_steps": 32180, "total_steps": 42420, "loss": 0.3359, "lr": 0.00016727946736921963, "epoch": 15.172088637435172, "percentage": 75.86, "elapsed_time": "1:44:39", "remaining_time": "0:33:18", "throughput": 5164.88, "total_tokens": 32431232} {"current_steps": 32185, "total_steps": 42420, "loss": 0.3141, "lr": 0.00016712593569282526, "epoch": 15.174446016030174, "percentage": 75.87, "elapsed_time": "1:44:40", "remaining_time": "0:33:17", "throughput": 5164.97, "total_tokens": 32437120} {"current_steps": 32190, "total_steps": 42420, "loss": 0.3285, "lr": 0.0001669724603663581, "epoch": 15.176803394625177, "percentage": 75.88, "elapsed_time": "1:44:41", "remaining_time": "0:33:16", "throughput": 5165.0, "total_tokens": 32442080} {"current_steps": 32195, "total_steps": 42420, "loss": 0.3511, "lr": 0.0001668190414157989, "epoch": 15.179160773220179, "percentage": 75.9, "elapsed_time": "1:44:42", "remaining_time": "0:33:15", "throughput": 5165.02, "total_tokens": 32446944} {"current_steps": 32200, "total_steps": 42420, "loss": 0.3612, "lr": 0.0001666656788671189, "epoch": 15.181518151815181, "percentage": 75.91, "elapsed_time": "1:44:42", "remaining_time": "0:33:14", "throughput": 5165.06, "total_tokens": 32451840} {"current_steps": 32205, "total_steps": 42420, "loss": 0.3465, "lr": 0.00016651237274627984, "epoch": 15.183875530410184, "percentage": 75.92, "elapsed_time": "1:44:43", "remaining_time": "0:33:13", "throughput": 5165.13, "total_tokens": 32457152} {"current_steps": 32210, "total_steps": 42420, "loss": 0.3202, "lr": 0.00016635912307923372, "epoch": 15.186232909005186, "percentage": 75.93, "elapsed_time": "1:44:44", "remaining_time": "0:33:12", "throughput": 5165.12, "total_tokens": 32461472} {"current_steps": 32215, "total_steps": 42420, "loss": 0.3486, "lr": 0.00016620592989192317, "epoch": 15.188590287600189, "percentage": 75.94, "elapsed_time": "1:44:45", "remaining_time": "0:33:11", "throughput": 5165.13, "total_tokens": 32466144} {"current_steps": 32220, "total_steps": 42420, "loss": 0.331, "lr": 0.0001660527932102814, "epoch": 15.190947666195191, "percentage": 75.95, "elapsed_time": "1:44:46", "remaining_time": "0:33:10", "throughput": 5165.14, "total_tokens": 32470624} {"current_steps": 32225, "total_steps": 42420, "loss": 0.335, "lr": 0.0001658997130602314, "epoch": 15.193305044790193, "percentage": 75.97, "elapsed_time": "1:44:47", "remaining_time": "0:33:09", "throughput": 5165.15, "total_tokens": 32475200} {"current_steps": 32230, "total_steps": 42420, "loss": 0.3416, "lr": 0.00016574668946768735, "epoch": 15.195662423385196, "percentage": 75.98, "elapsed_time": "1:44:48", "remaining_time": "0:33:08", "throughput": 5165.16, "total_tokens": 32479584} {"current_steps": 32235, "total_steps": 42420, "loss": 0.3336, "lr": 0.00016559372245855342, "epoch": 15.198019801980198, "percentage": 75.99, "elapsed_time": "1:44:49", "remaining_time": "0:33:07", "throughput": 5165.13, "total_tokens": 32483584} {"current_steps": 32240, "total_steps": 42420, "loss": 0.3278, "lr": 0.00016544081205872434, "epoch": 15.2003771805752, "percentage": 76.0, "elapsed_time": "1:44:49", "remaining_time": "0:33:06", "throughput": 5165.09, "total_tokens": 32487552} {"current_steps": 32245, "total_steps": 42420, "loss": 0.3213, "lr": 0.00016528795829408538, "epoch": 15.202734559170203, "percentage": 76.01, "elapsed_time": "1:44:50", "remaining_time": "0:33:05", "throughput": 5165.16, "total_tokens": 32492832} {"current_steps": 32250, "total_steps": 42420, "loss": 0.3685, "lr": 0.00016513516119051198, "epoch": 15.205091937765205, "percentage": 76.03, "elapsed_time": "1:44:51", "remaining_time": "0:33:04", "throughput": 5165.22, "total_tokens": 32498336} {"current_steps": 32255, "total_steps": 42420, "loss": 0.307, "lr": 0.0001649824207738701, "epoch": 15.207449316360208, "percentage": 76.04, "elapsed_time": "1:44:52", "remaining_time": "0:33:03", "throughput": 5165.26, "total_tokens": 32503232} {"current_steps": 32260, "total_steps": 42420, "loss": 0.3864, "lr": 0.00016482973707001626, "epoch": 15.20980669495521, "percentage": 76.05, "elapsed_time": "1:44:53", "remaining_time": "0:33:02", "throughput": 5165.26, "total_tokens": 32507776} {"current_steps": 32265, "total_steps": 42420, "loss": 0.3446, "lr": 0.00016467711010479684, "epoch": 15.212164073550213, "percentage": 76.06, "elapsed_time": "1:44:54", "remaining_time": "0:33:01", "throughput": 5165.36, "total_tokens": 32514080} {"current_steps": 32270, "total_steps": 42420, "loss": 0.3512, "lr": 0.0001645245399040493, "epoch": 15.214521452145215, "percentage": 76.07, "elapsed_time": "1:44:55", "remaining_time": "0:33:00", "throughput": 5165.38, "total_tokens": 32518784} {"current_steps": 32275, "total_steps": 42420, "loss": 0.3534, "lr": 0.0001643720264936011, "epoch": 15.216878830740217, "percentage": 76.08, "elapsed_time": "1:44:56", "remaining_time": "0:32:59", "throughput": 5165.49, "total_tokens": 32525024} {"current_steps": 32280, "total_steps": 42420, "loss": 0.327, "lr": 0.00016421956989927017, "epoch": 15.21923620933522, "percentage": 76.1, "elapsed_time": "1:44:57", "remaining_time": "0:32:58", "throughput": 5165.53, "total_tokens": 32529824} {"current_steps": 32285, "total_steps": 42420, "loss": 0.3729, "lr": 0.00016406717014686483, "epoch": 15.221593587930222, "percentage": 76.11, "elapsed_time": "1:44:58", "remaining_time": "0:32:57", "throughput": 5165.52, "total_tokens": 32534176} {"current_steps": 32290, "total_steps": 42420, "loss": 0.3367, "lr": 0.00016391482726218375, "epoch": 15.223950966525225, "percentage": 76.12, "elapsed_time": "1:44:59", "remaining_time": "0:32:56", "throughput": 5165.58, "total_tokens": 32539296} {"current_steps": 32295, "total_steps": 42420, "loss": 0.3132, "lr": 0.00016376254127101604, "epoch": 15.226308345120227, "percentage": 76.13, "elapsed_time": "1:45:00", "remaining_time": "0:32:55", "throughput": 5165.58, "total_tokens": 32543744} {"current_steps": 32300, "total_steps": 42420, "loss": 0.327, "lr": 0.00016361031219914125, "epoch": 15.22866572371523, "percentage": 76.14, "elapsed_time": "1:45:01", "remaining_time": "0:32:54", "throughput": 5165.6, "total_tokens": 32548544} {"current_steps": 32305, "total_steps": 42420, "loss": 0.3437, "lr": 0.0001634581400723289, "epoch": 15.231023102310232, "percentage": 76.16, "elapsed_time": "1:45:01", "remaining_time": "0:32:53", "throughput": 5165.6, "total_tokens": 32552896} {"current_steps": 32310, "total_steps": 42420, "loss": 0.3323, "lr": 0.00016330602491633923, "epoch": 15.233380480905234, "percentage": 76.17, "elapsed_time": "1:45:02", "remaining_time": "0:32:52", "throughput": 5165.61, "total_tokens": 32557504} {"current_steps": 32315, "total_steps": 42420, "loss": 0.3855, "lr": 0.00016315396675692284, "epoch": 15.235737859500237, "percentage": 76.18, "elapsed_time": "1:45:03", "remaining_time": "0:32:51", "throughput": 5165.7, "total_tokens": 32563360} {"current_steps": 32320, "total_steps": 42420, "loss": 0.3245, "lr": 0.00016300196561982056, "epoch": 15.238095238095237, "percentage": 76.19, "elapsed_time": "1:45:04", "remaining_time": "0:32:50", "throughput": 5165.7, "total_tokens": 32567872} {"current_steps": 32325, "total_steps": 42420, "loss": 0.3463, "lr": 0.00016285002153076356, "epoch": 15.24045261669024, "percentage": 76.2, "elapsed_time": "1:45:05", "remaining_time": "0:32:49", "throughput": 5165.72, "total_tokens": 32572832} {"current_steps": 32330, "total_steps": 42420, "loss": 0.3051, "lr": 0.0001626981345154735, "epoch": 15.242809995285242, "percentage": 76.21, "elapsed_time": "1:45:06", "remaining_time": "0:32:48", "throughput": 5165.76, "total_tokens": 32577408} {"current_steps": 32335, "total_steps": 42420, "loss": 0.3331, "lr": 0.00016254630459966213, "epoch": 15.245167373880244, "percentage": 76.23, "elapsed_time": "1:45:07", "remaining_time": "0:32:47", "throughput": 5165.77, "total_tokens": 32581888} {"current_steps": 32340, "total_steps": 42420, "loss": 0.3088, "lr": 0.00016239453180903196, "epoch": 15.247524752475247, "percentage": 76.24, "elapsed_time": "1:45:08", "remaining_time": "0:32:46", "throughput": 5165.86, "total_tokens": 32587904} {"current_steps": 32345, "total_steps": 42420, "loss": 0.2946, "lr": 0.00016224281616927517, "epoch": 15.24988213107025, "percentage": 76.25, "elapsed_time": "1:45:09", "remaining_time": "0:32:45", "throughput": 5165.89, "total_tokens": 32592448} {"current_steps": 32350, "total_steps": 42420, "loss": 0.262, "lr": 0.0001620911577060748, "epoch": 15.252239509665252, "percentage": 76.26, "elapsed_time": "1:45:10", "remaining_time": "0:32:44", "throughput": 5165.98, "total_tokens": 32598528} {"current_steps": 32355, "total_steps": 42420, "loss": 0.3441, "lr": 0.0001619395564451041, "epoch": 15.254596888260254, "percentage": 76.27, "elapsed_time": "1:45:11", "remaining_time": "0:32:43", "throughput": 5166.13, "total_tokens": 32605184} {"current_steps": 32360, "total_steps": 42420, "loss": 0.3954, "lr": 0.00016178801241202668, "epoch": 15.256954266855256, "percentage": 76.28, "elapsed_time": "1:45:12", "remaining_time": "0:32:42", "throughput": 5166.2, "total_tokens": 32610624} {"current_steps": 32365, "total_steps": 42420, "loss": 0.3702, "lr": 0.0001616365256324961, "epoch": 15.259311645450259, "percentage": 76.3, "elapsed_time": "1:45:13", "remaining_time": "0:32:41", "throughput": 5166.2, "total_tokens": 32615104} {"current_steps": 32370, "total_steps": 42420, "loss": 0.3639, "lr": 0.00016148509613215668, "epoch": 15.261669024045261, "percentage": 76.31, "elapsed_time": "1:45:14", "remaining_time": "0:32:40", "throughput": 5166.24, "total_tokens": 32620064} {"current_steps": 32375, "total_steps": 42420, "loss": 0.3581, "lr": 0.00016133372393664275, "epoch": 15.264026402640264, "percentage": 76.32, "elapsed_time": "1:45:15", "remaining_time": "0:32:39", "throughput": 5166.31, "total_tokens": 32625568} {"current_steps": 32380, "total_steps": 42420, "loss": 0.3027, "lr": 0.0001611824090715792, "epoch": 15.266383781235266, "percentage": 76.33, "elapsed_time": "1:45:15", "remaining_time": "0:32:38", "throughput": 5166.3, "total_tokens": 32629888} {"current_steps": 32385, "total_steps": 42420, "loss": 0.3775, "lr": 0.00016103115156258113, "epoch": 15.268741159830268, "percentage": 76.34, "elapsed_time": "1:45:16", "remaining_time": "0:32:37", "throughput": 5166.31, "total_tokens": 32634368} {"current_steps": 32390, "total_steps": 42420, "loss": 0.3269, "lr": 0.00016087995143525358, "epoch": 15.27109853842527, "percentage": 76.36, "elapsed_time": "1:45:17", "remaining_time": "0:32:36", "throughput": 5166.31, "total_tokens": 32638720} {"current_steps": 32395, "total_steps": 42420, "loss": 0.3122, "lr": 0.0001607288087151923, "epoch": 15.273455917020273, "percentage": 76.37, "elapsed_time": "1:45:18", "remaining_time": "0:32:35", "throughput": 5166.32, "total_tokens": 32643136} {"current_steps": 32400, "total_steps": 42420, "loss": 0.3565, "lr": 0.00016057772342798328, "epoch": 15.275813295615276, "percentage": 76.38, "elapsed_time": "1:45:19", "remaining_time": "0:32:34", "throughput": 5166.38, "total_tokens": 32648832} {"current_steps": 32405, "total_steps": 42420, "loss": 0.3438, "lr": 0.00016042669559920253, "epoch": 15.278170674210278, "percentage": 76.39, "elapsed_time": "1:45:20", "remaining_time": "0:32:33", "throughput": 5166.47, "total_tokens": 32654752} {"current_steps": 32410, "total_steps": 42420, "loss": 0.3014, "lr": 0.00016027572525441664, "epoch": 15.28052805280528, "percentage": 76.4, "elapsed_time": "1:45:21", "remaining_time": "0:32:32", "throughput": 5166.5, "total_tokens": 32659808} {"current_steps": 32415, "total_steps": 42420, "loss": 0.3611, "lr": 0.00016012481241918226, "epoch": 15.282885431400283, "percentage": 76.41, "elapsed_time": "1:45:22", "remaining_time": "0:32:31", "throughput": 5166.5, "total_tokens": 32664032} {"current_steps": 32420, "total_steps": 42420, "loss": 0.3092, "lr": 0.0001599739571190464, "epoch": 15.285242809995285, "percentage": 76.43, "elapsed_time": "1:45:23", "remaining_time": "0:32:30", "throughput": 5166.51, "total_tokens": 32668352} {"current_steps": 32425, "total_steps": 42420, "loss": 0.2919, "lr": 0.00015982315937954644, "epoch": 15.287600188590288, "percentage": 76.44, "elapsed_time": "1:45:23", "remaining_time": "0:32:29", "throughput": 5166.53, "total_tokens": 32672992} {"current_steps": 32430, "total_steps": 42420, "loss": 0.318, "lr": 0.00015967241922620957, "epoch": 15.28995756718529, "percentage": 76.45, "elapsed_time": "1:45:25", "remaining_time": "0:32:28", "throughput": 5166.66, "total_tokens": 32679840} {"current_steps": 32435, "total_steps": 42420, "loss": 0.2705, "lr": 0.00015952173668455367, "epoch": 15.292314945780292, "percentage": 76.46, "elapsed_time": "1:45:26", "remaining_time": "0:32:27", "throughput": 5166.66, "total_tokens": 32684352} {"current_steps": 32440, "total_steps": 42420, "loss": 0.3468, "lr": 0.0001593711117800868, "epoch": 15.294672324375295, "percentage": 76.47, "elapsed_time": "1:45:27", "remaining_time": "0:32:26", "throughput": 5166.79, "total_tokens": 32691232} {"current_steps": 32445, "total_steps": 42420, "loss": 0.3821, "lr": 0.0001592205445383071, "epoch": 15.297029702970297, "percentage": 76.49, "elapsed_time": "1:45:28", "remaining_time": "0:32:25", "throughput": 5166.79, "total_tokens": 32695648} {"current_steps": 32450, "total_steps": 42420, "loss": 0.3714, "lr": 0.00015907003498470313, "epoch": 15.2993870815653, "percentage": 76.5, "elapsed_time": "1:45:28", "remaining_time": "0:32:24", "throughput": 5166.82, "total_tokens": 32700320} {"current_steps": 32455, "total_steps": 42420, "loss": 0.3056, "lr": 0.00015891958314475358, "epoch": 15.301744460160302, "percentage": 76.51, "elapsed_time": "1:45:29", "remaining_time": "0:32:23", "throughput": 5166.79, "total_tokens": 32704160} {"current_steps": 32460, "total_steps": 42420, "loss": 0.2421, "lr": 0.00015876918904392733, "epoch": 15.304101838755304, "percentage": 76.52, "elapsed_time": "1:45:30", "remaining_time": "0:32:22", "throughput": 5166.81, "total_tokens": 32708672} {"current_steps": 32465, "total_steps": 42420, "loss": 0.3661, "lr": 0.00015861885270768383, "epoch": 15.306459217350307, "percentage": 76.53, "elapsed_time": "1:45:31", "remaining_time": "0:32:21", "throughput": 5166.84, "total_tokens": 32713376} {"current_steps": 32470, "total_steps": 42420, "loss": 0.3512, "lr": 0.0001584685741614721, "epoch": 15.30881659594531, "percentage": 76.54, "elapsed_time": "1:45:32", "remaining_time": "0:32:20", "throughput": 5166.87, "total_tokens": 32718240} {"current_steps": 32475, "total_steps": 42420, "loss": 0.3477, "lr": 0.00015831835343073188, "epoch": 15.311173974540312, "percentage": 76.56, "elapsed_time": "1:45:33", "remaining_time": "0:32:19", "throughput": 5166.88, "total_tokens": 32722816} {"current_steps": 32480, "total_steps": 42420, "loss": 0.3392, "lr": 0.00015816819054089294, "epoch": 15.313531353135314, "percentage": 76.57, "elapsed_time": "1:45:33", "remaining_time": "0:32:18", "throughput": 5166.84, "total_tokens": 32726304} {"current_steps": 32485, "total_steps": 42420, "loss": 0.3506, "lr": 0.0001580180855173754, "epoch": 15.315888731730317, "percentage": 76.58, "elapsed_time": "1:45:34", "remaining_time": "0:32:17", "throughput": 5166.87, "total_tokens": 32731136} {"current_steps": 32490, "total_steps": 42420, "loss": 0.278, "lr": 0.0001578680383855895, "epoch": 15.318246110325319, "percentage": 76.59, "elapsed_time": "1:45:35", "remaining_time": "0:32:16", "throughput": 5166.94, "total_tokens": 32736640} {"current_steps": 32495, "total_steps": 42420, "loss": 0.3214, "lr": 0.00015771804917093558, "epoch": 15.320603488920321, "percentage": 76.6, "elapsed_time": "1:45:36", "remaining_time": "0:32:15", "throughput": 5166.95, "total_tokens": 32741152} {"current_steps": 32500, "total_steps": 42420, "loss": 0.3197, "lr": 0.00015756811789880437, "epoch": 15.322960867515324, "percentage": 76.61, "elapsed_time": "1:45:37", "remaining_time": "0:32:14", "throughput": 5167.0, "total_tokens": 32746336} {"current_steps": 32505, "total_steps": 42420, "loss": 0.3136, "lr": 0.00015741824459457676, "epoch": 15.325318246110326, "percentage": 76.63, "elapsed_time": "1:45:38", "remaining_time": "0:32:13", "throughput": 5167.09, "total_tokens": 32752128} {"current_steps": 32510, "total_steps": 42420, "loss": 0.3941, "lr": 0.00015726842928362345, "epoch": 15.327675624705329, "percentage": 76.64, "elapsed_time": "1:45:39", "remaining_time": "0:32:12", "throughput": 5167.14, "total_tokens": 32757248} {"current_steps": 32515, "total_steps": 42420, "loss": 0.3097, "lr": 0.00015711867199130587, "epoch": 15.33003300330033, "percentage": 76.65, "elapsed_time": "1:45:40", "remaining_time": "0:32:11", "throughput": 5167.2, "total_tokens": 32762464} {"current_steps": 32520, "total_steps": 42420, "loss": 0.3098, "lr": 0.00015696897274297529, "epoch": 15.332390381895332, "percentage": 76.66, "elapsed_time": "1:45:41", "remaining_time": "0:32:10", "throughput": 5167.26, "total_tokens": 32767680} {"current_steps": 32525, "total_steps": 42420, "loss": 0.3695, "lr": 0.00015681933156397332, "epoch": 15.334747760490334, "percentage": 76.67, "elapsed_time": "1:45:42", "remaining_time": "0:32:09", "throughput": 5167.27, "total_tokens": 32772416} {"current_steps": 32530, "total_steps": 42420, "loss": 0.3251, "lr": 0.0001566697484796316, "epoch": 15.337105139085336, "percentage": 76.69, "elapsed_time": "1:45:43", "remaining_time": "0:32:08", "throughput": 5167.38, "total_tokens": 32778560} {"current_steps": 32535, "total_steps": 42420, "loss": 0.3518, "lr": 0.00015652022351527207, "epoch": 15.339462517680339, "percentage": 76.7, "elapsed_time": "1:45:44", "remaining_time": "0:32:07", "throughput": 5167.37, "total_tokens": 32782720} {"current_steps": 32540, "total_steps": 42420, "loss": 0.2692, "lr": 0.00015637075669620677, "epoch": 15.341819896275341, "percentage": 76.71, "elapsed_time": "1:45:45", "remaining_time": "0:32:06", "throughput": 5167.36, "total_tokens": 32786976} {"current_steps": 32545, "total_steps": 42420, "loss": 0.3755, "lr": 0.00015622134804773796, "epoch": 15.344177274870344, "percentage": 76.72, "elapsed_time": "1:45:45", "remaining_time": "0:32:05", "throughput": 5167.38, "total_tokens": 32791552} {"current_steps": 32550, "total_steps": 42420, "loss": 0.3029, "lr": 0.00015607199759515777, "epoch": 15.346534653465346, "percentage": 76.73, "elapsed_time": "1:45:46", "remaining_time": "0:32:04", "throughput": 5167.43, "total_tokens": 32796640} {"current_steps": 32555, "total_steps": 42420, "loss": 0.34, "lr": 0.00015592270536374887, "epoch": 15.348892032060348, "percentage": 76.74, "elapsed_time": "1:45:47", "remaining_time": "0:32:03", "throughput": 5167.46, "total_tokens": 32801568} {"current_steps": 32560, "total_steps": 42420, "loss": 0.2743, "lr": 0.00015577347137878377, "epoch": 15.35124941065535, "percentage": 76.76, "elapsed_time": "1:45:48", "remaining_time": "0:32:02", "throughput": 5167.52, "total_tokens": 32807008} {"current_steps": 32565, "total_steps": 42420, "loss": 0.3271, "lr": 0.0001556242956655254, "epoch": 15.353606789250353, "percentage": 76.77, "elapsed_time": "1:45:49", "remaining_time": "0:32:01", "throughput": 5167.52, "total_tokens": 32811392} {"current_steps": 32570, "total_steps": 42420, "loss": 0.2904, "lr": 0.00015547517824922657, "epoch": 15.355964167845356, "percentage": 76.78, "elapsed_time": "1:45:50", "remaining_time": "0:32:00", "throughput": 5167.54, "total_tokens": 32816064} {"current_steps": 32575, "total_steps": 42420, "loss": 0.3746, "lr": 0.00015532611915513035, "epoch": 15.358321546440358, "percentage": 76.79, "elapsed_time": "1:45:51", "remaining_time": "0:31:59", "throughput": 5167.6, "total_tokens": 32821408} {"current_steps": 32580, "total_steps": 42420, "loss": 0.3437, "lr": 0.00015517711840846998, "epoch": 15.36067892503536, "percentage": 76.8, "elapsed_time": "1:45:52", "remaining_time": "0:31:58", "throughput": 5167.6, "total_tokens": 32825856} {"current_steps": 32585, "total_steps": 42420, "loss": 0.3878, "lr": 0.00015502817603446883, "epoch": 15.363036303630363, "percentage": 76.82, "elapsed_time": "1:45:53", "remaining_time": "0:31:57", "throughput": 5167.64, "total_tokens": 32831072} {"current_steps": 32590, "total_steps": 42420, "loss": 0.3359, "lr": 0.00015487929205834002, "epoch": 15.365393682225365, "percentage": 76.83, "elapsed_time": "1:45:54", "remaining_time": "0:31:56", "throughput": 5167.78, "total_tokens": 32837696} {"current_steps": 32595, "total_steps": 42420, "loss": 0.3131, "lr": 0.00015473046650528745, "epoch": 15.367751060820368, "percentage": 76.84, "elapsed_time": "1:45:55", "remaining_time": "0:31:55", "throughput": 5167.84, "total_tokens": 32843072} {"current_steps": 32600, "total_steps": 42420, "loss": 0.3655, "lr": 0.00015458169940050438, "epoch": 15.37010843941537, "percentage": 76.85, "elapsed_time": "1:45:56", "remaining_time": "0:31:54", "throughput": 5167.89, "total_tokens": 32848416} {"current_steps": 32605, "total_steps": 42420, "loss": 0.375, "lr": 0.00015443299076917466, "epoch": 15.372465818010372, "percentage": 76.86, "elapsed_time": "1:45:57", "remaining_time": "0:31:53", "throughput": 5167.9, "total_tokens": 32852896} {"current_steps": 32610, "total_steps": 42420, "loss": 0.3056, "lr": 0.0001542843406364723, "epoch": 15.374823196605375, "percentage": 76.87, "elapsed_time": "1:45:58", "remaining_time": "0:31:52", "throughput": 5167.94, "total_tokens": 32857792} {"current_steps": 32615, "total_steps": 42420, "loss": 0.3246, "lr": 0.0001541357490275611, "epoch": 15.377180575200377, "percentage": 76.89, "elapsed_time": "1:45:59", "remaining_time": "0:31:51", "throughput": 5168.01, "total_tokens": 32863744} {"current_steps": 32620, "total_steps": 42420, "loss": 0.3279, "lr": 0.0001539872159675952, "epoch": 15.37953795379538, "percentage": 76.9, "elapsed_time": "1:46:00", "remaining_time": "0:31:50", "throughput": 5168.07, "total_tokens": 32869024} {"current_steps": 32625, "total_steps": 42420, "loss": 0.3335, "lr": 0.00015383874148171866, "epoch": 15.381895332390382, "percentage": 76.91, "elapsed_time": "1:46:00", "remaining_time": "0:31:49", "throughput": 5168.09, "total_tokens": 32873696} {"current_steps": 32630, "total_steps": 42420, "loss": 0.314, "lr": 0.0001536903255950657, "epoch": 15.384252710985384, "percentage": 76.92, "elapsed_time": "1:46:01", "remaining_time": "0:31:48", "throughput": 5168.13, "total_tokens": 32878944} {"current_steps": 32635, "total_steps": 42420, "loss": 0.4114, "lr": 0.00015354196833276075, "epoch": 15.386610089580387, "percentage": 76.93, "elapsed_time": "1:46:02", "remaining_time": "0:31:47", "throughput": 5168.16, "total_tokens": 32883712} {"current_steps": 32640, "total_steps": 42420, "loss": 0.2982, "lr": 0.00015339366971991786, "epoch": 15.38896746817539, "percentage": 76.94, "elapsed_time": "1:46:03", "remaining_time": "0:31:46", "throughput": 5168.28, "total_tokens": 32890272} {"current_steps": 32645, "total_steps": 42420, "loss": 0.3265, "lr": 0.00015324542978164169, "epoch": 15.391324846770392, "percentage": 76.96, "elapsed_time": "1:46:04", "remaining_time": "0:31:45", "throughput": 5168.31, "total_tokens": 32895136} {"current_steps": 32650, "total_steps": 42420, "loss": 0.3148, "lr": 0.0001530972485430267, "epoch": 15.393682225365394, "percentage": 76.97, "elapsed_time": "1:46:05", "remaining_time": "0:31:44", "throughput": 5168.39, "total_tokens": 32900800} {"current_steps": 32655, "total_steps": 42420, "loss": 0.3165, "lr": 0.00015294912602915745, "epoch": 15.396039603960396, "percentage": 76.98, "elapsed_time": "1:46:06", "remaining_time": "0:31:43", "throughput": 5168.42, "total_tokens": 32905408} {"current_steps": 32660, "total_steps": 42420, "loss": 0.2941, "lr": 0.00015280106226510853, "epoch": 15.398396982555399, "percentage": 76.99, "elapsed_time": "1:46:07", "remaining_time": "0:31:42", "throughput": 5168.44, "total_tokens": 32910336} {"current_steps": 32665, "total_steps": 42420, "loss": 0.3539, "lr": 0.00015265305727594475, "epoch": 15.400754361150401, "percentage": 77.0, "elapsed_time": "1:46:08", "remaining_time": "0:31:41", "throughput": 5168.43, "total_tokens": 32914752} {"current_steps": 32670, "total_steps": 42420, "loss": 0.4007, "lr": 0.0001525051110867207, "epoch": 15.403111739745404, "percentage": 77.02, "elapsed_time": "1:46:09", "remaining_time": "0:31:40", "throughput": 5168.46, "total_tokens": 32919552} {"current_steps": 32675, "total_steps": 42420, "loss": 0.3, "lr": 0.0001523572237224814, "epoch": 15.405469118340406, "percentage": 77.03, "elapsed_time": "1:46:10", "remaining_time": "0:31:39", "throughput": 5168.53, "total_tokens": 32925280} {"current_steps": 32680, "total_steps": 42420, "loss": 0.3119, "lr": 0.00015220939520826132, "epoch": 15.407826496935408, "percentage": 77.04, "elapsed_time": "1:46:11", "remaining_time": "0:31:38", "throughput": 5168.59, "total_tokens": 32930592} {"current_steps": 32685, "total_steps": 42420, "loss": 0.3459, "lr": 0.00015206162556908549, "epoch": 15.41018387553041, "percentage": 77.05, "elapsed_time": "1:46:12", "remaining_time": "0:31:37", "throughput": 5168.66, "total_tokens": 32936224} {"current_steps": 32690, "total_steps": 42420, "loss": 0.355, "lr": 0.00015191391482996874, "epoch": 15.412541254125413, "percentage": 77.06, "elapsed_time": "1:46:13", "remaining_time": "0:31:36", "throughput": 5168.69, "total_tokens": 32941216} {"current_steps": 32695, "total_steps": 42420, "loss": 0.3257, "lr": 0.0001517662630159161, "epoch": 15.414898632720416, "percentage": 77.07, "elapsed_time": "1:46:14", "remaining_time": "0:31:35", "throughput": 5168.71, "total_tokens": 32945760} {"current_steps": 32700, "total_steps": 42420, "loss": 0.3606, "lr": 0.0001516186701519225, "epoch": 15.417256011315418, "percentage": 77.09, "elapsed_time": "1:46:15", "remaining_time": "0:31:34", "throughput": 5168.78, "total_tokens": 32951296} {"current_steps": 32705, "total_steps": 42420, "loss": 0.3184, "lr": 0.00015147113626297278, "epoch": 15.41961338991042, "percentage": 77.1, "elapsed_time": "1:46:15", "remaining_time": "0:31:33", "throughput": 5168.78, "total_tokens": 32955488} {"current_steps": 32710, "total_steps": 42420, "loss": 0.2891, "lr": 0.00015132366137404201, "epoch": 15.421970768505423, "percentage": 77.11, "elapsed_time": "1:46:16", "remaining_time": "0:31:32", "throughput": 5168.8, "total_tokens": 32960064} {"current_steps": 32715, "total_steps": 42420, "loss": 0.379, "lr": 0.0001511762455100954, "epoch": 15.424328147100425, "percentage": 77.12, "elapsed_time": "1:46:17", "remaining_time": "0:31:31", "throughput": 5168.82, "total_tokens": 32964864} {"current_steps": 32720, "total_steps": 42420, "loss": 0.2982, "lr": 0.00015102888869608755, "epoch": 15.426685525695426, "percentage": 77.13, "elapsed_time": "1:46:18", "remaining_time": "0:31:30", "throughput": 5168.84, "total_tokens": 32969536} {"current_steps": 32725, "total_steps": 42420, "loss": 0.3429, "lr": 0.00015088159095696363, "epoch": 15.429042904290428, "percentage": 77.15, "elapsed_time": "1:46:19", "remaining_time": "0:31:29", "throughput": 5168.87, "total_tokens": 32974592} {"current_steps": 32730, "total_steps": 42420, "loss": 0.3338, "lr": 0.0001507343523176587, "epoch": 15.43140028288543, "percentage": 77.16, "elapsed_time": "1:46:20", "remaining_time": "0:31:28", "throughput": 5168.83, "total_tokens": 32978560} {"current_steps": 32735, "total_steps": 42420, "loss": 0.2817, "lr": 0.0001505871728030977, "epoch": 15.433757661480433, "percentage": 77.17, "elapsed_time": "1:46:21", "remaining_time": "0:31:27", "throughput": 5168.84, "total_tokens": 32983136} {"current_steps": 32740, "total_steps": 42420, "loss": 0.3334, "lr": 0.0001504400524381957, "epoch": 15.436115040075435, "percentage": 77.18, "elapsed_time": "1:46:22", "remaining_time": "0:31:26", "throughput": 5168.88, "total_tokens": 32988160} {"current_steps": 32745, "total_steps": 42420, "loss": 0.2868, "lr": 0.00015029299124785756, "epoch": 15.438472418670438, "percentage": 77.19, "elapsed_time": "1:46:22", "remaining_time": "0:31:25", "throughput": 5168.86, "total_tokens": 32992480} {"current_steps": 32750, "total_steps": 42420, "loss": 0.3433, "lr": 0.00015014598925697836, "epoch": 15.44082979726544, "percentage": 77.2, "elapsed_time": "1:46:23", "remaining_time": "0:31:24", "throughput": 5168.96, "total_tokens": 32998496} {"current_steps": 32755, "total_steps": 42420, "loss": 0.3324, "lr": 0.00014999904649044317, "epoch": 15.443187175860443, "percentage": 77.22, "elapsed_time": "1:46:24", "remaining_time": "0:31:24", "throughput": 5169.04, "total_tokens": 33004160} {"current_steps": 32760, "total_steps": 42420, "loss": 0.312, "lr": 0.00014985216297312655, "epoch": 15.445544554455445, "percentage": 77.23, "elapsed_time": "1:46:25", "remaining_time": "0:31:23", "throughput": 5169.06, "total_tokens": 33008704} {"current_steps": 32765, "total_steps": 42420, "loss": 0.3435, "lr": 0.00014970533872989356, "epoch": 15.447901933050447, "percentage": 77.24, "elapsed_time": "1:46:26", "remaining_time": "0:31:22", "throughput": 5169.11, "total_tokens": 33013920} {"current_steps": 32770, "total_steps": 42420, "loss": 0.3783, "lr": 0.00014955857378559905, "epoch": 15.45025931164545, "percentage": 77.25, "elapsed_time": "1:46:27", "remaining_time": "0:31:21", "throughput": 5169.18, "total_tokens": 33019552} {"current_steps": 32775, "total_steps": 42420, "loss": 0.3984, "lr": 0.0001494118681650879, "epoch": 15.452616690240452, "percentage": 77.26, "elapsed_time": "1:46:28", "remaining_time": "0:31:20", "throughput": 5169.24, "total_tokens": 33024864} {"current_steps": 32780, "total_steps": 42420, "loss": 0.3458, "lr": 0.00014926522189319475, "epoch": 15.454974068835455, "percentage": 77.27, "elapsed_time": "1:46:29", "remaining_time": "0:31:19", "throughput": 5169.27, "total_tokens": 33029632} {"current_steps": 32785, "total_steps": 42420, "loss": 0.3127, "lr": 0.00014911863499474443, "epoch": 15.457331447430457, "percentage": 77.29, "elapsed_time": "1:46:30", "remaining_time": "0:31:18", "throughput": 5169.33, "total_tokens": 33035200} {"current_steps": 32790, "total_steps": 42420, "loss": 0.336, "lr": 0.0001489721074945516, "epoch": 15.45968882602546, "percentage": 77.3, "elapsed_time": "1:46:31", "remaining_time": "0:31:17", "throughput": 5169.35, "total_tokens": 33040000} {"current_steps": 32795, "total_steps": 42420, "loss": 0.3623, "lr": 0.00014882563941742095, "epoch": 15.462046204620462, "percentage": 77.31, "elapsed_time": "1:46:32", "remaining_time": "0:31:16", "throughput": 5169.38, "total_tokens": 33044704} {"current_steps": 32800, "total_steps": 42420, "loss": 0.3128, "lr": 0.00014867923078814682, "epoch": 15.464403583215464, "percentage": 77.32, "elapsed_time": "1:46:33", "remaining_time": "0:31:15", "throughput": 5169.39, "total_tokens": 33049312} {"current_steps": 32805, "total_steps": 42420, "loss": 0.3417, "lr": 0.00014853288163151385, "epoch": 15.466760961810467, "percentage": 77.33, "elapsed_time": "1:46:34", "remaining_time": "0:31:14", "throughput": 5169.42, "total_tokens": 33054176} {"current_steps": 32810, "total_steps": 42420, "loss": 0.3186, "lr": 0.0001483865919722966, "epoch": 15.469118340405469, "percentage": 77.35, "elapsed_time": "1:46:35", "remaining_time": "0:31:13", "throughput": 5169.47, "total_tokens": 33059584} {"current_steps": 32815, "total_steps": 42420, "loss": 0.3009, "lr": 0.00014824036183525913, "epoch": 15.471475719000471, "percentage": 77.36, "elapsed_time": "1:46:36", "remaining_time": "0:31:12", "throughput": 5169.47, "total_tokens": 33064032} {"current_steps": 32820, "total_steps": 42420, "loss": 0.3262, "lr": 0.00014809419124515584, "epoch": 15.473833097595474, "percentage": 77.37, "elapsed_time": "1:46:36", "remaining_time": "0:31:11", "throughput": 5169.5, "total_tokens": 33068992} {"current_steps": 32825, "total_steps": 42420, "loss": 0.3329, "lr": 0.00014794808022673095, "epoch": 15.476190476190476, "percentage": 77.38, "elapsed_time": "1:46:37", "remaining_time": "0:31:10", "throughput": 5169.52, "total_tokens": 33073504} {"current_steps": 32830, "total_steps": 42420, "loss": 0.3874, "lr": 0.00014780202880471856, "epoch": 15.478547854785479, "percentage": 77.39, "elapsed_time": "1:46:38", "remaining_time": "0:31:09", "throughput": 5169.55, "total_tokens": 33078112} {"current_steps": 32835, "total_steps": 42420, "loss": 0.3934, "lr": 0.00014765603700384272, "epoch": 15.480905233380481, "percentage": 77.4, "elapsed_time": "1:46:39", "remaining_time": "0:31:08", "throughput": 5169.56, "total_tokens": 33082720} {"current_steps": 32840, "total_steps": 42420, "loss": 0.3096, "lr": 0.00014751010484881737, "epoch": 15.483262611975483, "percentage": 77.42, "elapsed_time": "1:46:40", "remaining_time": "0:31:07", "throughput": 5169.65, "total_tokens": 33088896} {"current_steps": 32845, "total_steps": 42420, "loss": 0.3492, "lr": 0.0001473642323643465, "epoch": 15.485619990570486, "percentage": 77.43, "elapsed_time": "1:46:41", "remaining_time": "0:31:06", "throughput": 5169.66, "total_tokens": 33093760} {"current_steps": 32850, "total_steps": 42420, "loss": 0.3658, "lr": 0.0001472184195751235, "epoch": 15.487977369165488, "percentage": 77.44, "elapsed_time": "1:46:42", "remaining_time": "0:31:05", "throughput": 5169.64, "total_tokens": 33097760} {"current_steps": 32855, "total_steps": 42420, "loss": 0.3054, "lr": 0.00014707266650583218, "epoch": 15.49033474776049, "percentage": 77.45, "elapsed_time": "1:46:43", "remaining_time": "0:31:04", "throughput": 5169.67, "total_tokens": 33102816} {"current_steps": 32860, "total_steps": 42420, "loss": 0.3476, "lr": 0.00014692697318114607, "epoch": 15.492692126355493, "percentage": 77.46, "elapsed_time": "1:46:44", "remaining_time": "0:31:03", "throughput": 5169.86, "total_tokens": 33111168} {"current_steps": 32865, "total_steps": 42420, "loss": 0.3346, "lr": 0.00014678133962572854, "epoch": 15.495049504950495, "percentage": 77.48, "elapsed_time": "1:46:45", "remaining_time": "0:31:02", "throughput": 5169.84, "total_tokens": 33115392} {"current_steps": 32870, "total_steps": 42420, "loss": 0.3406, "lr": 0.0001466357658642329, "epoch": 15.497406883545498, "percentage": 77.49, "elapsed_time": "1:46:46", "remaining_time": "0:31:01", "throughput": 5169.82, "total_tokens": 33119584} {"current_steps": 32875, "total_steps": 42420, "loss": 0.3187, "lr": 0.00014649025192130233, "epoch": 15.4997642621405, "percentage": 77.5, "elapsed_time": "1:46:47", "remaining_time": "0:31:00", "throughput": 5169.81, "total_tokens": 33124096} {"current_steps": 32880, "total_steps": 42420, "loss": 0.3417, "lr": 0.00014634479782156984, "epoch": 15.502121640735503, "percentage": 77.51, "elapsed_time": "1:46:48", "remaining_time": "0:30:59", "throughput": 5169.84, "total_tokens": 33129120} {"current_steps": 32885, "total_steps": 42420, "loss": 0.3163, "lr": 0.0001461994035896585, "epoch": 15.504479019330505, "percentage": 77.52, "elapsed_time": "1:46:49", "remaining_time": "0:30:58", "throughput": 5169.87, "total_tokens": 33133888} {"current_steps": 32890, "total_steps": 42420, "loss": 0.3495, "lr": 0.00014605406925018082, "epoch": 15.506836397925507, "percentage": 77.53, "elapsed_time": "1:46:50", "remaining_time": "0:30:57", "throughput": 5169.96, "total_tokens": 33139808} {"current_steps": 32895, "total_steps": 42420, "loss": 0.3664, "lr": 0.00014590879482773955, "epoch": 15.50919377652051, "percentage": 77.55, "elapsed_time": "1:46:51", "remaining_time": "0:30:56", "throughput": 5170.03, "total_tokens": 33145184} {"current_steps": 32900, "total_steps": 42420, "loss": 0.349, "lr": 0.00014576358034692715, "epoch": 15.511551155115512, "percentage": 77.56, "elapsed_time": "1:46:51", "remaining_time": "0:30:55", "throughput": 5170.06, "total_tokens": 33149856} {"current_steps": 32905, "total_steps": 42420, "loss": 0.3221, "lr": 0.00014561842583232605, "epoch": 15.513908533710515, "percentage": 77.57, "elapsed_time": "1:46:52", "remaining_time": "0:30:54", "throughput": 5170.05, "total_tokens": 33154048} {"current_steps": 32910, "total_steps": 42420, "loss": 0.3438, "lr": 0.00014547333130850837, "epoch": 15.516265912305517, "percentage": 77.58, "elapsed_time": "1:46:53", "remaining_time": "0:30:53", "throughput": 5170.07, "total_tokens": 33158528} {"current_steps": 32915, "total_steps": 42420, "loss": 0.2979, "lr": 0.00014532829680003622, "epoch": 15.518623290900518, "percentage": 77.59, "elapsed_time": "1:46:54", "remaining_time": "0:30:52", "throughput": 5170.1, "total_tokens": 33163552} {"current_steps": 32920, "total_steps": 42420, "loss": 0.3454, "lr": 0.00014518332233146142, "epoch": 15.520980669495522, "percentage": 77.6, "elapsed_time": "1:46:55", "remaining_time": "0:30:51", "throughput": 5170.12, "total_tokens": 33168416} {"current_steps": 32925, "total_steps": 42420, "loss": 0.3399, "lr": 0.00014503840792732587, "epoch": 15.523338048090523, "percentage": 77.62, "elapsed_time": "1:46:56", "remaining_time": "0:30:50", "throughput": 5170.14, "total_tokens": 33173024} {"current_steps": 32930, "total_steps": 42420, "loss": 0.3942, "lr": 0.00014489355361216078, "epoch": 15.525695426685525, "percentage": 77.63, "elapsed_time": "1:46:57", "remaining_time": "0:30:49", "throughput": 5170.16, "total_tokens": 33177920} {"current_steps": 32935, "total_steps": 42420, "loss": 0.3645, "lr": 0.00014474875941048772, "epoch": 15.528052805280527, "percentage": 77.64, "elapsed_time": "1:46:58", "remaining_time": "0:30:48", "throughput": 5170.21, "total_tokens": 33183296} {"current_steps": 32940, "total_steps": 42420, "loss": 0.3302, "lr": 0.00014460402534681782, "epoch": 15.53041018387553, "percentage": 77.65, "elapsed_time": "1:46:59", "remaining_time": "0:30:47", "throughput": 5170.26, "total_tokens": 33188800} {"current_steps": 32945, "total_steps": 42420, "loss": 0.3476, "lr": 0.0001444593514456522, "epoch": 15.532767562470532, "percentage": 77.66, "elapsed_time": "1:47:00", "remaining_time": "0:30:46", "throughput": 5170.29, "total_tokens": 33193536} {"current_steps": 32950, "total_steps": 42420, "loss": 0.3188, "lr": 0.00014431473773148158, "epoch": 15.535124941065535, "percentage": 77.68, "elapsed_time": "1:47:01", "remaining_time": "0:30:45", "throughput": 5170.33, "total_tokens": 33198816} {"current_steps": 32955, "total_steps": 42420, "loss": 0.3282, "lr": 0.00014417018422878674, "epoch": 15.537482319660537, "percentage": 77.69, "elapsed_time": "1:47:01", "remaining_time": "0:30:44", "throughput": 5170.34, "total_tokens": 33203488} {"current_steps": 32960, "total_steps": 42420, "loss": 0.3367, "lr": 0.00014402569096203804, "epoch": 15.53983969825554, "percentage": 77.7, "elapsed_time": "1:47:02", "remaining_time": "0:30:43", "throughput": 5170.3, "total_tokens": 33207072} {"current_steps": 32965, "total_steps": 42420, "loss": 0.3367, "lr": 0.0001438812579556959, "epoch": 15.542197076850542, "percentage": 77.71, "elapsed_time": "1:47:03", "remaining_time": "0:30:42", "throughput": 5170.33, "total_tokens": 33212128} {"current_steps": 32970, "total_steps": 42420, "loss": 0.3394, "lr": 0.00014373688523421007, "epoch": 15.544554455445544, "percentage": 77.72, "elapsed_time": "1:47:04", "remaining_time": "0:30:41", "throughput": 5170.36, "total_tokens": 33216960} {"current_steps": 32975, "total_steps": 42420, "loss": 0.2864, "lr": 0.0001435925728220206, "epoch": 15.546911834040547, "percentage": 77.73, "elapsed_time": "1:47:05", "remaining_time": "0:30:40", "throughput": 5170.4, "total_tokens": 33222048} {"current_steps": 32980, "total_steps": 42420, "loss": 0.3147, "lr": 0.00014344832074355708, "epoch": 15.549269212635549, "percentage": 77.75, "elapsed_time": "1:47:06", "remaining_time": "0:30:39", "throughput": 5170.48, "total_tokens": 33227552} {"current_steps": 32985, "total_steps": 42420, "loss": 0.3144, "lr": 0.00014330412902323897, "epoch": 15.551626591230551, "percentage": 77.76, "elapsed_time": "1:47:07", "remaining_time": "0:30:38", "throughput": 5170.49, "total_tokens": 33232160} {"current_steps": 32990, "total_steps": 42420, "loss": 0.3085, "lr": 0.00014315999768547543, "epoch": 15.553983969825554, "percentage": 77.77, "elapsed_time": "1:47:08", "remaining_time": "0:30:37", "throughput": 5170.51, "total_tokens": 33236704} {"current_steps": 32995, "total_steps": 42420, "loss": 0.396, "lr": 0.00014301592675466552, "epoch": 15.556341348420556, "percentage": 77.78, "elapsed_time": "1:47:09", "remaining_time": "0:30:36", "throughput": 5170.52, "total_tokens": 33241440} {"current_steps": 33000, "total_steps": 42420, "loss": 0.3777, "lr": 0.00014287191625519797, "epoch": 15.558698727015559, "percentage": 77.79, "elapsed_time": "1:47:10", "remaining_time": "0:30:35", "throughput": 5170.58, "total_tokens": 33247104} {"current_steps": 33005, "total_steps": 42420, "loss": 0.3258, "lr": 0.00014272796621145144, "epoch": 15.561056105610561, "percentage": 77.81, "elapsed_time": "1:47:10", "remaining_time": "0:30:34", "throughput": 5170.6, "total_tokens": 33251744} {"current_steps": 33010, "total_steps": 42420, "loss": 0.3312, "lr": 0.000142584076647794, "epoch": 15.563413484205563, "percentage": 77.82, "elapsed_time": "1:47:11", "remaining_time": "0:30:33", "throughput": 5170.63, "total_tokens": 33256416} {"current_steps": 33015, "total_steps": 42420, "loss": 0.3741, "lr": 0.0001424402475885837, "epoch": 15.565770862800566, "percentage": 77.83, "elapsed_time": "1:47:12", "remaining_time": "0:30:32", "throughput": 5170.67, "total_tokens": 33261728} {"current_steps": 33020, "total_steps": 42420, "loss": 0.2975, "lr": 0.00014229647905816855, "epoch": 15.568128241395568, "percentage": 77.84, "elapsed_time": "1:47:13", "remaining_time": "0:30:31", "throughput": 5170.69, "total_tokens": 33266176} {"current_steps": 33025, "total_steps": 42420, "loss": 0.3567, "lr": 0.00014215277108088604, "epoch": 15.57048561999057, "percentage": 77.85, "elapsed_time": "1:47:14", "remaining_time": "0:30:30", "throughput": 5170.74, "total_tokens": 33271488} {"current_steps": 33030, "total_steps": 42420, "loss": 0.3406, "lr": 0.00014200912368106345, "epoch": 15.572842998585573, "percentage": 77.86, "elapsed_time": "1:47:15", "remaining_time": "0:30:29", "throughput": 5170.74, "total_tokens": 33275872} {"current_steps": 33035, "total_steps": 42420, "loss": 0.3426, "lr": 0.00014186553688301802, "epoch": 15.575200377180575, "percentage": 77.88, "elapsed_time": "1:47:16", "remaining_time": "0:30:28", "throughput": 5170.8, "total_tokens": 33280864} {"current_steps": 33040, "total_steps": 42420, "loss": 0.3404, "lr": 0.00014172201071105628, "epoch": 15.577557755775578, "percentage": 77.89, "elapsed_time": "1:47:17", "remaining_time": "0:30:27", "throughput": 5170.89, "total_tokens": 33286720} {"current_steps": 33045, "total_steps": 42420, "loss": 0.3376, "lr": 0.0001415785451894749, "epoch": 15.57991513437058, "percentage": 77.9, "elapsed_time": "1:47:18", "remaining_time": "0:30:26", "throughput": 5170.96, "total_tokens": 33292256} {"current_steps": 33050, "total_steps": 42420, "loss": 0.3307, "lr": 0.00014143514034256022, "epoch": 15.582272512965583, "percentage": 77.91, "elapsed_time": "1:47:19", "remaining_time": "0:30:25", "throughput": 5171.05, "total_tokens": 33298368} {"current_steps": 33055, "total_steps": 42420, "loss": 0.3498, "lr": 0.00014129179619458833, "epoch": 15.584629891560585, "percentage": 77.92, "elapsed_time": "1:47:20", "remaining_time": "0:30:24", "throughput": 5171.13, "total_tokens": 33304160} {"current_steps": 33060, "total_steps": 42420, "loss": 0.3448, "lr": 0.0001411485127698246, "epoch": 15.586987270155587, "percentage": 77.93, "elapsed_time": "1:47:21", "remaining_time": "0:30:23", "throughput": 5171.15, "total_tokens": 33308832} {"current_steps": 33065, "total_steps": 42420, "loss": 0.351, "lr": 0.00014100529009252477, "epoch": 15.58934464875059, "percentage": 77.95, "elapsed_time": "1:47:22", "remaining_time": "0:30:22", "throughput": 5171.14, "total_tokens": 33312928} {"current_steps": 33070, "total_steps": 42420, "loss": 0.3415, "lr": 0.00014086212818693395, "epoch": 15.591702027345592, "percentage": 77.96, "elapsed_time": "1:47:23", "remaining_time": "0:30:21", "throughput": 5171.18, "total_tokens": 33317984} {"current_steps": 33075, "total_steps": 42420, "loss": 0.3355, "lr": 0.000140719027077287, "epoch": 15.594059405940595, "percentage": 77.97, "elapsed_time": "1:47:23", "remaining_time": "0:30:20", "throughput": 5171.2, "total_tokens": 33322720} {"current_steps": 33080, "total_steps": 42420, "loss": 0.2847, "lr": 0.00014057598678780852, "epoch": 15.596416784535597, "percentage": 77.98, "elapsed_time": "1:47:24", "remaining_time": "0:30:19", "throughput": 5171.24, "total_tokens": 33328000} {"current_steps": 33085, "total_steps": 42420, "loss": 0.3506, "lr": 0.0001404330073427128, "epoch": 15.5987741631306, "percentage": 77.99, "elapsed_time": "1:47:25", "remaining_time": "0:30:18", "throughput": 5171.27, "total_tokens": 33332800} {"current_steps": 33090, "total_steps": 42420, "loss": 0.3207, "lr": 0.00014029008876620386, "epoch": 15.601131541725602, "percentage": 78.01, "elapsed_time": "1:47:26", "remaining_time": "0:30:17", "throughput": 5171.29, "total_tokens": 33337728} {"current_steps": 33095, "total_steps": 42420, "loss": 0.3624, "lr": 0.0001401472310824755, "epoch": 15.603488920320604, "percentage": 78.02, "elapsed_time": "1:47:27", "remaining_time": "0:30:16", "throughput": 5171.34, "total_tokens": 33343008} {"current_steps": 33100, "total_steps": 42420, "loss": 0.3459, "lr": 0.0001400044343157108, "epoch": 15.605846298915607, "percentage": 78.03, "elapsed_time": "1:47:29", "remaining_time": "0:30:15", "throughput": 5171.56, "total_tokens": 33351936} {"current_steps": 33105, "total_steps": 42420, "loss": 0.3043, "lr": 0.00013986169849008301, "epoch": 15.608203677510609, "percentage": 78.04, "elapsed_time": "1:47:29", "remaining_time": "0:30:14", "throughput": 5171.58, "total_tokens": 33356576} {"current_steps": 33110, "total_steps": 42420, "loss": 0.3547, "lr": 0.00013971902362975487, "epoch": 15.61056105610561, "percentage": 78.05, "elapsed_time": "1:47:30", "remaining_time": "0:30:13", "throughput": 5171.6, "total_tokens": 33361120} {"current_steps": 33115, "total_steps": 42420, "loss": 0.3288, "lr": 0.0001395764097588788, "epoch": 15.612918434700614, "percentage": 78.06, "elapsed_time": "1:47:31", "remaining_time": "0:30:12", "throughput": 5171.64, "total_tokens": 33366496} {"current_steps": 33120, "total_steps": 42420, "loss": 0.3441, "lr": 0.00013943385690159694, "epoch": 15.615275813295614, "percentage": 78.08, "elapsed_time": "1:47:32", "remaining_time": "0:30:11", "throughput": 5171.65, "total_tokens": 33371328} {"current_steps": 33125, "total_steps": 42420, "loss": 0.3312, "lr": 0.00013929136508204098, "epoch": 15.617633191890617, "percentage": 78.09, "elapsed_time": "1:47:33", "remaining_time": "0:30:10", "throughput": 5171.7, "total_tokens": 33376640} {"current_steps": 33130, "total_steps": 42420, "loss": 0.3001, "lr": 0.00013914893432433252, "epoch": 15.61999057048562, "percentage": 78.1, "elapsed_time": "1:47:34", "remaining_time": "0:30:09", "throughput": 5171.77, "total_tokens": 33382400} {"current_steps": 33135, "total_steps": 42420, "loss": 0.3482, "lr": 0.00013900656465258267, "epoch": 15.622347949080622, "percentage": 78.11, "elapsed_time": "1:47:35", "remaining_time": "0:30:08", "throughput": 5171.79, "total_tokens": 33387104} {"current_steps": 33140, "total_steps": 42420, "loss": 0.2587, "lr": 0.00013886425609089192, "epoch": 15.624705327675624, "percentage": 78.12, "elapsed_time": "1:47:36", "remaining_time": "0:30:07", "throughput": 5171.84, "total_tokens": 33392288} {"current_steps": 33145, "total_steps": 42420, "loss": 0.3073, "lr": 0.00013872200866335094, "epoch": 15.627062706270626, "percentage": 78.14, "elapsed_time": "1:47:37", "remaining_time": "0:30:07", "throughput": 5171.91, "total_tokens": 33397856} {"current_steps": 33150, "total_steps": 42420, "loss": 0.3667, "lr": 0.00013857982239403972, "epoch": 15.629420084865629, "percentage": 78.15, "elapsed_time": "1:47:38", "remaining_time": "0:30:06", "throughput": 5171.98, "total_tokens": 33403584} {"current_steps": 33155, "total_steps": 42420, "loss": 0.4098, "lr": 0.000138437697307028, "epoch": 15.631777463460631, "percentage": 78.16, "elapsed_time": "1:47:39", "remaining_time": "0:30:05", "throughput": 5172.09, "total_tokens": 33409856} {"current_steps": 33160, "total_steps": 42420, "loss": 0.3289, "lr": 0.00013829563342637513, "epoch": 15.634134842055634, "percentage": 78.17, "elapsed_time": "1:47:40", "remaining_time": "0:30:04", "throughput": 5172.17, "total_tokens": 33415520} {"current_steps": 33165, "total_steps": 42420, "loss": 0.3528, "lr": 0.00013815363077613013, "epoch": 15.636492220650636, "percentage": 78.18, "elapsed_time": "1:47:41", "remaining_time": "0:30:03", "throughput": 5172.22, "total_tokens": 33420640} {"current_steps": 33170, "total_steps": 42420, "loss": 0.277, "lr": 0.00013801168938033166, "epoch": 15.638849599245638, "percentage": 78.19, "elapsed_time": "1:47:42", "remaining_time": "0:30:02", "throughput": 5172.22, "total_tokens": 33425248} {"current_steps": 33175, "total_steps": 42420, "loss": 0.2997, "lr": 0.00013786980926300813, "epoch": 15.64120697784064, "percentage": 78.21, "elapsed_time": "1:47:43", "remaining_time": "0:30:01", "throughput": 5172.24, "total_tokens": 33430272} {"current_steps": 33180, "total_steps": 42420, "loss": 0.3513, "lr": 0.00013772799044817706, "epoch": 15.643564356435643, "percentage": 78.22, "elapsed_time": "1:47:44", "remaining_time": "0:30:00", "throughput": 5172.25, "total_tokens": 33434880} {"current_steps": 33185, "total_steps": 42420, "loss": 0.3154, "lr": 0.00013758623295984624, "epoch": 15.645921735030646, "percentage": 78.23, "elapsed_time": "1:47:45", "remaining_time": "0:29:59", "throughput": 5172.26, "total_tokens": 33439328} {"current_steps": 33190, "total_steps": 42420, "loss": 0.321, "lr": 0.00013744453682201275, "epoch": 15.648279113625648, "percentage": 78.24, "elapsed_time": "1:47:46", "remaining_time": "0:29:58", "throughput": 5172.35, "total_tokens": 33445312} {"current_steps": 33195, "total_steps": 42420, "loss": 0.3309, "lr": 0.00013730290205866331, "epoch": 15.65063649222065, "percentage": 78.25, "elapsed_time": "1:47:47", "remaining_time": "0:29:57", "throughput": 5172.37, "total_tokens": 33450080} {"current_steps": 33200, "total_steps": 42420, "loss": 0.3763, "lr": 0.0001371613286937743, "epoch": 15.652993870815653, "percentage": 78.26, "elapsed_time": "1:47:47", "remaining_time": "0:29:56", "throughput": 5172.36, "total_tokens": 33454464} {"current_steps": 33205, "total_steps": 42420, "loss": 0.3381, "lr": 0.0001370198167513117, "epoch": 15.655351249410655, "percentage": 78.28, "elapsed_time": "1:47:48", "remaining_time": "0:29:55", "throughput": 5172.4, "total_tokens": 33459648} {"current_steps": 33210, "total_steps": 42420, "loss": 0.3472, "lr": 0.00013687836625523109, "epoch": 15.657708628005658, "percentage": 78.29, "elapsed_time": "1:47:49", "remaining_time": "0:29:54", "throughput": 5172.49, "total_tokens": 33465344} {"current_steps": 33215, "total_steps": 42420, "loss": 0.3291, "lr": 0.00013673697722947775, "epoch": 15.66006600660066, "percentage": 78.3, "elapsed_time": "1:47:50", "remaining_time": "0:29:53", "throughput": 5172.46, "total_tokens": 33469248} {"current_steps": 33220, "total_steps": 42420, "loss": 0.3285, "lr": 0.00013659564969798615, "epoch": 15.662423385195662, "percentage": 78.31, "elapsed_time": "1:47:51", "remaining_time": "0:29:52", "throughput": 5172.48, "total_tokens": 33474016} {"current_steps": 33225, "total_steps": 42420, "loss": 0.3467, "lr": 0.00013645438368468078, "epoch": 15.664780763790665, "percentage": 78.32, "elapsed_time": "1:47:52", "remaining_time": "0:29:51", "throughput": 5172.48, "total_tokens": 33478656} {"current_steps": 33230, "total_steps": 42420, "loss": 0.269, "lr": 0.00013631317921347562, "epoch": 15.667138142385667, "percentage": 78.34, "elapsed_time": "1:47:53", "remaining_time": "0:29:50", "throughput": 5172.49, "total_tokens": 33483104} {"current_steps": 33235, "total_steps": 42420, "loss": 0.3119, "lr": 0.00013617203630827417, "epoch": 15.66949552098067, "percentage": 78.35, "elapsed_time": "1:47:54", "remaining_time": "0:29:49", "throughput": 5172.55, "total_tokens": 33488448} {"current_steps": 33240, "total_steps": 42420, "loss": 0.2791, "lr": 0.0001360309549929696, "epoch": 15.671852899575672, "percentage": 78.36, "elapsed_time": "1:47:55", "remaining_time": "0:29:48", "throughput": 5172.59, "total_tokens": 33493504} {"current_steps": 33245, "total_steps": 42420, "loss": 0.3095, "lr": 0.00013588993529144438, "epoch": 15.674210278170674, "percentage": 78.37, "elapsed_time": "1:47:56", "remaining_time": "0:29:47", "throughput": 5172.63, "total_tokens": 33498368} {"current_steps": 33250, "total_steps": 42420, "loss": 0.3864, "lr": 0.00013574897722757096, "epoch": 15.676567656765677, "percentage": 78.38, "elapsed_time": "1:47:57", "remaining_time": "0:29:46", "throughput": 5172.65, "total_tokens": 33503296} {"current_steps": 33255, "total_steps": 42420, "loss": 0.3887, "lr": 0.0001356080808252112, "epoch": 15.67892503536068, "percentage": 78.39, "elapsed_time": "1:47:57", "remaining_time": "0:29:45", "throughput": 5172.69, "total_tokens": 33508480} {"current_steps": 33260, "total_steps": 42420, "loss": 0.3535, "lr": 0.0001354672461082163, "epoch": 15.681282413955682, "percentage": 78.41, "elapsed_time": "1:47:58", "remaining_time": "0:29:44", "throughput": 5172.7, "total_tokens": 33513024} {"current_steps": 33265, "total_steps": 42420, "loss": 0.3028, "lr": 0.00013532647310042712, "epoch": 15.683639792550684, "percentage": 78.42, "elapsed_time": "1:47:59", "remaining_time": "0:29:43", "throughput": 5172.75, "total_tokens": 33518048} {"current_steps": 33270, "total_steps": 42420, "loss": 0.3564, "lr": 0.0001351857618256742, "epoch": 15.685997171145686, "percentage": 78.43, "elapsed_time": "1:48:00", "remaining_time": "0:29:42", "throughput": 5172.81, "total_tokens": 33523424} {"current_steps": 33275, "total_steps": 42420, "loss": 0.3409, "lr": 0.00013504511230777767, "epoch": 15.688354549740689, "percentage": 78.44, "elapsed_time": "1:48:01", "remaining_time": "0:29:41", "throughput": 5172.86, "total_tokens": 33528864} {"current_steps": 33280, "total_steps": 42420, "loss": 0.3312, "lr": 0.00013490452457054697, "epoch": 15.690711928335691, "percentage": 78.45, "elapsed_time": "1:48:02", "remaining_time": "0:29:40", "throughput": 5172.89, "total_tokens": 33533664} {"current_steps": 33285, "total_steps": 42420, "loss": 0.3513, "lr": 0.00013476399863778134, "epoch": 15.693069306930694, "percentage": 78.47, "elapsed_time": "1:48:03", "remaining_time": "0:29:39", "throughput": 5172.89, "total_tokens": 33538144} {"current_steps": 33290, "total_steps": 42420, "loss": 0.3463, "lr": 0.00013462353453326937, "epoch": 15.695426685525696, "percentage": 78.48, "elapsed_time": "1:48:04", "remaining_time": "0:29:38", "throughput": 5172.87, "total_tokens": 33542176} {"current_steps": 33295, "total_steps": 42420, "loss": 0.3379, "lr": 0.00013448313228078923, "epoch": 15.697784064120698, "percentage": 78.49, "elapsed_time": "1:48:05", "remaining_time": "0:29:37", "throughput": 5172.9, "total_tokens": 33546976} {"current_steps": 33300, "total_steps": 42420, "loss": 0.3554, "lr": 0.00013434279190410885, "epoch": 15.700141442715701, "percentage": 78.5, "elapsed_time": "1:48:06", "remaining_time": "0:29:36", "throughput": 5172.99, "total_tokens": 33553056} {"current_steps": 33305, "total_steps": 42420, "loss": 0.348, "lr": 0.00013420251342698519, "epoch": 15.702498821310703, "percentage": 78.51, "elapsed_time": "1:48:07", "remaining_time": "0:29:35", "throughput": 5173.07, "total_tokens": 33558784} {"current_steps": 33310, "total_steps": 42420, "loss": 0.3418, "lr": 0.00013406229687316508, "epoch": 15.704856199905706, "percentage": 78.52, "elapsed_time": "1:48:08", "remaining_time": "0:29:34", "throughput": 5173.12, "total_tokens": 33563936} {"current_steps": 33315, "total_steps": 42420, "loss": 0.3598, "lr": 0.00013392214226638484, "epoch": 15.707213578500706, "percentage": 78.54, "elapsed_time": "1:48:08", "remaining_time": "0:29:33", "throughput": 5173.12, "total_tokens": 33568128} {"current_steps": 33320, "total_steps": 42420, "loss": 0.3363, "lr": 0.00013378204963037028, "epoch": 15.70957095709571, "percentage": 78.55, "elapsed_time": "1:48:09", "remaining_time": "0:29:32", "throughput": 5173.17, "total_tokens": 33573408} {"current_steps": 33325, "total_steps": 42420, "loss": 0.2897, "lr": 0.00013364201898883666, "epoch": 15.711928335690711, "percentage": 78.56, "elapsed_time": "1:48:10", "remaining_time": "0:29:31", "throughput": 5173.22, "total_tokens": 33578944} {"current_steps": 33330, "total_steps": 42420, "loss": 0.3093, "lr": 0.00013350205036548885, "epoch": 15.714285714285714, "percentage": 78.57, "elapsed_time": "1:48:11", "remaining_time": "0:29:30", "throughput": 5173.23, "total_tokens": 33583296} {"current_steps": 33335, "total_steps": 42420, "loss": 0.3291, "lr": 0.0001333621437840211, "epoch": 15.716643092880716, "percentage": 78.58, "elapsed_time": "1:48:12", "remaining_time": "0:29:29", "throughput": 5173.2, "total_tokens": 33587040} {"current_steps": 33340, "total_steps": 42420, "loss": 0.2888, "lr": 0.0001332222992681174, "epoch": 15.719000471475718, "percentage": 78.6, "elapsed_time": "1:48:13", "remaining_time": "0:29:28", "throughput": 5173.3, "total_tokens": 33593504} {"current_steps": 33345, "total_steps": 42420, "loss": 0.3551, "lr": 0.0001330825168414508, "epoch": 15.72135785007072, "percentage": 78.61, "elapsed_time": "1:48:14", "remaining_time": "0:29:27", "throughput": 5173.3, "total_tokens": 33597792} {"current_steps": 33350, "total_steps": 42420, "loss": 0.3408, "lr": 0.00013294279652768416, "epoch": 15.723715228665723, "percentage": 78.62, "elapsed_time": "1:48:15", "remaining_time": "0:29:26", "throughput": 5173.34, "total_tokens": 33602784} {"current_steps": 33355, "total_steps": 42420, "loss": 0.3147, "lr": 0.0001328031383504698, "epoch": 15.726072607260726, "percentage": 78.63, "elapsed_time": "1:48:16", "remaining_time": "0:29:25", "throughput": 5173.34, "total_tokens": 33607040} {"current_steps": 33360, "total_steps": 42420, "loss": 0.3552, "lr": 0.00013266354233344946, "epoch": 15.728429985855728, "percentage": 78.64, "elapsed_time": "1:48:17", "remaining_time": "0:29:24", "throughput": 5173.33, "total_tokens": 33611488} {"current_steps": 33365, "total_steps": 42420, "loss": 0.3311, "lr": 0.00013252400850025442, "epoch": 15.73078736445073, "percentage": 78.65, "elapsed_time": "1:48:17", "remaining_time": "0:29:23", "throughput": 5173.29, "total_tokens": 33615136} {"current_steps": 33370, "total_steps": 42420, "loss": 0.3418, "lr": 0.00013238453687450531, "epoch": 15.733144743045733, "percentage": 78.67, "elapsed_time": "1:48:18", "remaining_time": "0:29:22", "throughput": 5173.3, "total_tokens": 33619648} {"current_steps": 33375, "total_steps": 42420, "loss": 0.3144, "lr": 0.00013224512747981237, "epoch": 15.735502121640735, "percentage": 78.68, "elapsed_time": "1:48:19", "remaining_time": "0:29:21", "throughput": 5173.37, "total_tokens": 33625504} {"current_steps": 33380, "total_steps": 42420, "loss": 0.3666, "lr": 0.00013210578033977523, "epoch": 15.737859500235738, "percentage": 78.69, "elapsed_time": "1:48:20", "remaining_time": "0:29:20", "throughput": 5173.36, "total_tokens": 33629984} {"current_steps": 33385, "total_steps": 42420, "loss": 0.2743, "lr": 0.00013196649547798312, "epoch": 15.74021687883074, "percentage": 78.7, "elapsed_time": "1:48:21", "remaining_time": "0:29:19", "throughput": 5173.44, "total_tokens": 33635616} {"current_steps": 33390, "total_steps": 42420, "loss": 0.3706, "lr": 0.00013182727291801438, "epoch": 15.742574257425742, "percentage": 78.71, "elapsed_time": "1:48:22", "remaining_time": "0:29:18", "throughput": 5173.55, "total_tokens": 33642400} {"current_steps": 33395, "total_steps": 42420, "loss": 0.3192, "lr": 0.00013168811268343705, "epoch": 15.744931636020745, "percentage": 78.72, "elapsed_time": "1:48:23", "remaining_time": "0:29:17", "throughput": 5173.56, "total_tokens": 33647040} {"current_steps": 33400, "total_steps": 42420, "loss": 0.3381, "lr": 0.0001315490147978087, "epoch": 15.747289014615747, "percentage": 78.74, "elapsed_time": "1:48:24", "remaining_time": "0:29:16", "throughput": 5173.63, "total_tokens": 33652352} {"current_steps": 33405, "total_steps": 42420, "loss": 0.3704, "lr": 0.0001314099792846762, "epoch": 15.74964639321075, "percentage": 78.75, "elapsed_time": "1:48:25", "remaining_time": "0:29:15", "throughput": 5173.68, "total_tokens": 33657632} {"current_steps": 33410, "total_steps": 42420, "loss": 0.3245, "lr": 0.00013127100616757586, "epoch": 15.752003771805752, "percentage": 78.76, "elapsed_time": "1:48:26", "remaining_time": "0:29:14", "throughput": 5173.66, "total_tokens": 33661824} {"current_steps": 33415, "total_steps": 42420, "loss": 0.3385, "lr": 0.00013113209547003358, "epoch": 15.754361150400754, "percentage": 78.77, "elapsed_time": "1:48:27", "remaining_time": "0:29:13", "throughput": 5173.7, "total_tokens": 33666912} {"current_steps": 33420, "total_steps": 42420, "loss": 0.3614, "lr": 0.0001309932472155645, "epoch": 15.756718528995757, "percentage": 78.78, "elapsed_time": "1:48:28", "remaining_time": "0:29:12", "throughput": 5173.81, "total_tokens": 33673440} {"current_steps": 33425, "total_steps": 42420, "loss": 0.3256, "lr": 0.00013085446142767343, "epoch": 15.75907590759076, "percentage": 78.8, "elapsed_time": "1:48:29", "remaining_time": "0:29:11", "throughput": 5173.81, "total_tokens": 33677888} {"current_steps": 33430, "total_steps": 42420, "loss": 0.3461, "lr": 0.00013071573812985417, "epoch": 15.761433286185762, "percentage": 78.81, "elapsed_time": "1:48:30", "remaining_time": "0:29:10", "throughput": 5173.82, "total_tokens": 33682432} {"current_steps": 33435, "total_steps": 42420, "loss": 0.359, "lr": 0.00013057707734559032, "epoch": 15.763790664780764, "percentage": 78.82, "elapsed_time": "1:48:31", "remaining_time": "0:29:09", "throughput": 5173.83, "total_tokens": 33686976} {"current_steps": 33440, "total_steps": 42420, "loss": 0.3318, "lr": 0.00013043847909835482, "epoch": 15.766148043375766, "percentage": 78.83, "elapsed_time": "1:48:32", "remaining_time": "0:29:08", "throughput": 5173.93, "total_tokens": 33692768} {"current_steps": 33445, "total_steps": 42420, "loss": 0.3214, "lr": 0.00013029994341161004, "epoch": 15.768505421970769, "percentage": 78.84, "elapsed_time": "1:48:33", "remaining_time": "0:29:07", "throughput": 5174.04, "total_tokens": 33698880} {"current_steps": 33450, "total_steps": 42420, "loss": 0.328, "lr": 0.00013016147030880772, "epoch": 15.770862800565771, "percentage": 78.85, "elapsed_time": "1:48:33", "remaining_time": "0:29:06", "throughput": 5174.1, "total_tokens": 33703968} {"current_steps": 33455, "total_steps": 42420, "loss": 0.3158, "lr": 0.00013002305981338891, "epoch": 15.773220179160774, "percentage": 78.87, "elapsed_time": "1:48:34", "remaining_time": "0:29:05", "throughput": 5174.14, "total_tokens": 33709120} {"current_steps": 33460, "total_steps": 42420, "loss": 0.3231, "lr": 0.00012988471194878427, "epoch": 15.775577557755776, "percentage": 78.88, "elapsed_time": "1:48:35", "remaining_time": "0:29:04", "throughput": 5174.17, "total_tokens": 33713952} {"current_steps": 33465, "total_steps": 42420, "loss": 0.4141, "lr": 0.0001297464267384138, "epoch": 15.777934936350778, "percentage": 78.89, "elapsed_time": "1:48:36", "remaining_time": "0:29:03", "throughput": 5174.17, "total_tokens": 33718432} {"current_steps": 33470, "total_steps": 42420, "loss": 0.3538, "lr": 0.00012960820420568664, "epoch": 15.78029231494578, "percentage": 78.9, "elapsed_time": "1:48:37", "remaining_time": "0:29:02", "throughput": 5174.29, "total_tokens": 33724736} {"current_steps": 33475, "total_steps": 42420, "loss": 0.3849, "lr": 0.00012947004437400162, "epoch": 15.782649693540783, "percentage": 78.91, "elapsed_time": "1:48:38", "remaining_time": "0:29:01", "throughput": 5174.25, "total_tokens": 33728256} {"current_steps": 33480, "total_steps": 42420, "loss": 0.3322, "lr": 0.00012933194726674679, "epoch": 15.785007072135786, "percentage": 78.93, "elapsed_time": "1:48:39", "remaining_time": "0:29:00", "throughput": 5174.31, "total_tokens": 33733536} {"current_steps": 33485, "total_steps": 42420, "loss": 0.3333, "lr": 0.00012919391290729988, "epoch": 15.787364450730788, "percentage": 78.94, "elapsed_time": "1:48:40", "remaining_time": "0:28:59", "throughput": 5174.34, "total_tokens": 33738368} {"current_steps": 33490, "total_steps": 42420, "loss": 0.3349, "lr": 0.0001290559413190275, "epoch": 15.78972182932579, "percentage": 78.95, "elapsed_time": "1:48:41", "remaining_time": "0:28:58", "throughput": 5174.36, "total_tokens": 33742848} {"current_steps": 33495, "total_steps": 42420, "loss": 0.3743, "lr": 0.00012891803252528588, "epoch": 15.792079207920793, "percentage": 78.96, "elapsed_time": "1:48:42", "remaining_time": "0:28:57", "throughput": 5174.51, "total_tokens": 33749600} {"current_steps": 33500, "total_steps": 42420, "loss": 0.3633, "lr": 0.00012878018654942076, "epoch": 15.794436586515795, "percentage": 78.97, "elapsed_time": "1:48:43", "remaining_time": "0:28:56", "throughput": 5174.54, "total_tokens": 33754368} {"current_steps": 33505, "total_steps": 42420, "loss": 0.3219, "lr": 0.00012864240341476713, "epoch": 15.796793965110798, "percentage": 78.98, "elapsed_time": "1:48:44", "remaining_time": "0:28:55", "throughput": 5174.55, "total_tokens": 33759040} {"current_steps": 33510, "total_steps": 42420, "loss": 0.3328, "lr": 0.00012850468314464948, "epoch": 15.799151343705798, "percentage": 79.0, "elapsed_time": "1:48:44", "remaining_time": "0:28:54", "throughput": 5174.57, "total_tokens": 33763776} {"current_steps": 33515, "total_steps": 42420, "loss": 0.3402, "lr": 0.00012836702576238113, "epoch": 15.801508722300802, "percentage": 79.01, "elapsed_time": "1:48:45", "remaining_time": "0:28:53", "throughput": 5174.57, "total_tokens": 33767968} {"current_steps": 33520, "total_steps": 42420, "loss": 0.3358, "lr": 0.00012822943129126542, "epoch": 15.803866100895803, "percentage": 79.02, "elapsed_time": "1:48:46", "remaining_time": "0:28:52", "throughput": 5174.66, "total_tokens": 33773536} {"current_steps": 33525, "total_steps": 42420, "loss": 0.3383, "lr": 0.00012809189975459463, "epoch": 15.806223479490805, "percentage": 79.03, "elapsed_time": "1:48:47", "remaining_time": "0:28:51", "throughput": 5174.73, "total_tokens": 33779104} {"current_steps": 33530, "total_steps": 42420, "loss": 0.3646, "lr": 0.00012795443117565058, "epoch": 15.808580858085808, "percentage": 79.04, "elapsed_time": "1:48:48", "remaining_time": "0:28:51", "throughput": 5174.85, "total_tokens": 33785600} {"current_steps": 33535, "total_steps": 42420, "loss": 0.3551, "lr": 0.0001278170255777044, "epoch": 15.81093823668081, "percentage": 79.05, "elapsed_time": "1:48:49", "remaining_time": "0:28:50", "throughput": 5174.97, "total_tokens": 33792128} {"current_steps": 33540, "total_steps": 42420, "loss": 0.3462, "lr": 0.00012767968298401644, "epoch": 15.813295615275813, "percentage": 79.07, "elapsed_time": "1:48:50", "remaining_time": "0:28:49", "throughput": 5174.99, "total_tokens": 33796736} {"current_steps": 33545, "total_steps": 42420, "loss": 0.3202, "lr": 0.00012754240341783656, "epoch": 15.815652993870815, "percentage": 79.08, "elapsed_time": "1:48:51", "remaining_time": "0:28:48", "throughput": 5175.05, "total_tokens": 33802176} {"current_steps": 33550, "total_steps": 42420, "loss": 0.3411, "lr": 0.00012740518690240393, "epoch": 15.818010372465817, "percentage": 79.09, "elapsed_time": "1:48:52", "remaining_time": "0:28:47", "throughput": 5175.04, "total_tokens": 33806304} {"current_steps": 33555, "total_steps": 42420, "loss": 0.3251, "lr": 0.00012726803346094668, "epoch": 15.82036775106082, "percentage": 79.1, "elapsed_time": "1:48:53", "remaining_time": "0:28:46", "throughput": 5175.06, "total_tokens": 33811072} {"current_steps": 33560, "total_steps": 42420, "loss": 0.3046, "lr": 0.00012713094311668277, "epoch": 15.822725129655822, "percentage": 79.11, "elapsed_time": "1:48:54", "remaining_time": "0:28:45", "throughput": 5175.09, "total_tokens": 33815904} {"current_steps": 33565, "total_steps": 42420, "loss": 0.3212, "lr": 0.0001269939158928192, "epoch": 15.825082508250825, "percentage": 79.13, "elapsed_time": "1:48:55", "remaining_time": "0:28:44", "throughput": 5175.17, "total_tokens": 33821984} {"current_steps": 33570, "total_steps": 42420, "loss": 0.3793, "lr": 0.00012685695181255247, "epoch": 15.827439886845827, "percentage": 79.14, "elapsed_time": "1:48:56", "remaining_time": "0:28:43", "throughput": 5175.19, "total_tokens": 33826688} {"current_steps": 33575, "total_steps": 42420, "loss": 0.2896, "lr": 0.00012672005089906812, "epoch": 15.82979726544083, "percentage": 79.15, "elapsed_time": "1:48:57", "remaining_time": "0:28:42", "throughput": 5175.19, "total_tokens": 33831040} {"current_steps": 33580, "total_steps": 42420, "loss": 0.3035, "lr": 0.00012658321317554123, "epoch": 15.832154644035832, "percentage": 79.16, "elapsed_time": "1:48:58", "remaining_time": "0:28:41", "throughput": 5175.19, "total_tokens": 33835488} {"current_steps": 33585, "total_steps": 42420, "loss": 0.2836, "lr": 0.00012644643866513612, "epoch": 15.834512022630834, "percentage": 79.17, "elapsed_time": "1:48:58", "remaining_time": "0:28:40", "throughput": 5175.25, "total_tokens": 33840768} {"current_steps": 33590, "total_steps": 42420, "loss": 0.359, "lr": 0.00012630972739100644, "epoch": 15.836869401225837, "percentage": 79.18, "elapsed_time": "1:48:59", "remaining_time": "0:28:39", "throughput": 5175.27, "total_tokens": 33845440} {"current_steps": 33595, "total_steps": 42420, "loss": 0.2723, "lr": 0.0001261730793762949, "epoch": 15.839226779820839, "percentage": 79.2, "elapsed_time": "1:49:00", "remaining_time": "0:28:38", "throughput": 5175.31, "total_tokens": 33850368} {"current_steps": 33600, "total_steps": 42420, "loss": 0.3667, "lr": 0.00012603649464413386, "epoch": 15.841584158415841, "percentage": 79.21, "elapsed_time": "1:49:01", "remaining_time": "0:28:37", "throughput": 5175.38, "total_tokens": 33856160} {"current_steps": 33605, "total_steps": 42420, "loss": 0.323, "lr": 0.0001258999732176447, "epoch": 15.843941537010844, "percentage": 79.22, "elapsed_time": "1:49:02", "remaining_time": "0:28:36", "throughput": 5175.51, "total_tokens": 33862880} {"current_steps": 33610, "total_steps": 42420, "loss": 0.313, "lr": 0.00012576351511993822, "epoch": 15.846298915605846, "percentage": 79.23, "elapsed_time": "1:49:03", "remaining_time": "0:28:35", "throughput": 5175.53, "total_tokens": 33867392} {"current_steps": 33615, "total_steps": 42420, "loss": 0.2576, "lr": 0.00012562712037411444, "epoch": 15.848656294200849, "percentage": 79.24, "elapsed_time": "1:49:04", "remaining_time": "0:28:34", "throughput": 5175.6, "total_tokens": 33872928} {"current_steps": 33620, "total_steps": 42420, "loss": 0.3331, "lr": 0.00012549078900326267, "epoch": 15.851013672795851, "percentage": 79.26, "elapsed_time": "1:49:05", "remaining_time": "0:28:33", "throughput": 5175.63, "total_tokens": 33877760} {"current_steps": 33625, "total_steps": 42420, "loss": 0.3724, "lr": 0.00012535452103046158, "epoch": 15.853371051390853, "percentage": 79.27, "elapsed_time": "1:49:06", "remaining_time": "0:28:32", "throughput": 5175.69, "total_tokens": 33882976} {"current_steps": 33630, "total_steps": 42420, "loss": 0.3133, "lr": 0.00012521831647877908, "epoch": 15.855728429985856, "percentage": 79.28, "elapsed_time": "1:49:07", "remaining_time": "0:28:31", "throughput": 5175.71, "total_tokens": 33887616} {"current_steps": 33635, "total_steps": 42420, "loss": 0.2808, "lr": 0.00012508217537127197, "epoch": 15.858085808580858, "percentage": 79.29, "elapsed_time": "1:49:08", "remaining_time": "0:28:30", "throughput": 5175.72, "total_tokens": 33891968} {"current_steps": 33640, "total_steps": 42420, "loss": 0.3581, "lr": 0.00012494609773098692, "epoch": 15.86044318717586, "percentage": 79.3, "elapsed_time": "1:49:09", "remaining_time": "0:28:29", "throughput": 5175.76, "total_tokens": 33897088} {"current_steps": 33645, "total_steps": 42420, "loss": 0.3022, "lr": 0.00012481008358095942, "epoch": 15.862800565770863, "percentage": 79.31, "elapsed_time": "1:49:10", "remaining_time": "0:28:28", "throughput": 5175.8, "total_tokens": 33902112} {"current_steps": 33650, "total_steps": 42420, "loss": 0.2884, "lr": 0.0001246741329442144, "epoch": 15.865157944365865, "percentage": 79.33, "elapsed_time": "1:49:11", "remaining_time": "0:28:27", "throughput": 5175.82, "total_tokens": 33906816} {"current_steps": 33655, "total_steps": 42420, "loss": 0.3151, "lr": 0.000124538245843766, "epoch": 15.867515322960868, "percentage": 79.34, "elapsed_time": "1:49:12", "remaining_time": "0:28:26", "throughput": 5175.89, "total_tokens": 33912480} {"current_steps": 33660, "total_steps": 42420, "loss": 0.311, "lr": 0.00012440242230261762, "epoch": 15.86987270155587, "percentage": 79.35, "elapsed_time": "1:49:13", "remaining_time": "0:28:25", "throughput": 5176.01, "total_tokens": 33918912} {"current_steps": 33665, "total_steps": 42420, "loss": 0.3781, "lr": 0.0001242666623437619, "epoch": 15.872230080150873, "percentage": 79.36, "elapsed_time": "1:49:14", "remaining_time": "0:28:24", "throughput": 5176.09, "total_tokens": 33924608} {"current_steps": 33670, "total_steps": 42420, "loss": 0.3264, "lr": 0.00012413096599018077, "epoch": 15.874587458745875, "percentage": 79.37, "elapsed_time": "1:49:14", "remaining_time": "0:28:23", "throughput": 5176.12, "total_tokens": 33929408} {"current_steps": 33675, "total_steps": 42420, "loss": 0.3337, "lr": 0.00012399533326484504, "epoch": 15.876944837340877, "percentage": 79.38, "elapsed_time": "1:49:16", "remaining_time": "0:28:22", "throughput": 5176.2, "total_tokens": 33935264} {"current_steps": 33680, "total_steps": 42420, "loss": 0.3285, "lr": 0.00012385976419071516, "epoch": 15.87930221593588, "percentage": 79.4, "elapsed_time": "1:49:16", "remaining_time": "0:28:21", "throughput": 5176.2, "total_tokens": 33939648} {"current_steps": 33685, "total_steps": 42420, "loss": 0.3871, "lr": 0.0001237242587907408, "epoch": 15.881659594530882, "percentage": 79.41, "elapsed_time": "1:49:17", "remaining_time": "0:28:20", "throughput": 5176.25, "total_tokens": 33945120} {"current_steps": 33690, "total_steps": 42420, "loss": 0.3173, "lr": 0.00012358881708786053, "epoch": 15.884016973125885, "percentage": 79.42, "elapsed_time": "1:49:18", "remaining_time": "0:28:19", "throughput": 5176.27, "total_tokens": 33949760} {"current_steps": 33695, "total_steps": 42420, "loss": 0.3583, "lr": 0.0001234534391050025, "epoch": 15.886374351720887, "percentage": 79.43, "elapsed_time": "1:49:19", "remaining_time": "0:28:18", "throughput": 5176.32, "total_tokens": 33954784} {"current_steps": 33700, "total_steps": 42420, "loss": 0.3113, "lr": 0.00012331812486508382, "epoch": 15.88873173031589, "percentage": 79.44, "elapsed_time": "1:49:20", "remaining_time": "0:28:17", "throughput": 5176.33, "total_tokens": 33959424} {"current_steps": 33705, "total_steps": 42420, "loss": 0.3187, "lr": 0.00012318287439101082, "epoch": 15.891089108910892, "percentage": 79.46, "elapsed_time": "1:49:21", "remaining_time": "0:28:16", "throughput": 5176.33, "total_tokens": 33963872} {"current_steps": 33710, "total_steps": 42420, "loss": 0.3707, "lr": 0.00012304768770567947, "epoch": 15.893446487505894, "percentage": 79.47, "elapsed_time": "1:49:22", "remaining_time": "0:28:15", "throughput": 5176.37, "total_tokens": 33968896} {"current_steps": 33715, "total_steps": 42420, "loss": 0.3276, "lr": 0.0001229125648319741, "epoch": 15.895803866100895, "percentage": 79.48, "elapsed_time": "1:49:23", "remaining_time": "0:28:14", "throughput": 5176.42, "total_tokens": 33974048} {"current_steps": 33720, "total_steps": 42420, "loss": 0.3701, "lr": 0.00012277750579276907, "epoch": 15.898161244695899, "percentage": 79.49, "elapsed_time": "1:49:24", "remaining_time": "0:28:13", "throughput": 5176.41, "total_tokens": 33978400} {"current_steps": 33725, "total_steps": 42420, "loss": 0.3362, "lr": 0.0001226425106109273, "epoch": 15.9005186232909, "percentage": 79.5, "elapsed_time": "1:49:24", "remaining_time": "0:28:12", "throughput": 5176.41, "total_tokens": 33983008} {"current_steps": 33730, "total_steps": 42420, "loss": 0.3358, "lr": 0.0001225075793093013, "epoch": 15.902876001885902, "percentage": 79.51, "elapsed_time": "1:49:25", "remaining_time": "0:28:11", "throughput": 5176.45, "total_tokens": 33987968} {"current_steps": 33735, "total_steps": 42420, "loss": 0.3223, "lr": 0.0001223727119107327, "epoch": 15.905233380480905, "percentage": 79.53, "elapsed_time": "1:49:26", "remaining_time": "0:28:10", "throughput": 5176.51, "total_tokens": 33993504} {"current_steps": 33740, "total_steps": 42420, "loss": 0.3477, "lr": 0.00012223790843805216, "epoch": 15.907590759075907, "percentage": 79.54, "elapsed_time": "1:49:27", "remaining_time": "0:28:09", "throughput": 5176.48, "total_tokens": 33997504} {"current_steps": 33745, "total_steps": 42420, "loss": 0.309, "lr": 0.00012210316891407975, "epoch": 15.90994813767091, "percentage": 79.55, "elapsed_time": "1:49:28", "remaining_time": "0:28:08", "throughput": 5176.59, "total_tokens": 34003552} {"current_steps": 33750, "total_steps": 42420, "loss": 0.2639, "lr": 0.00012196849336162453, "epoch": 15.912305516265912, "percentage": 79.56, "elapsed_time": "1:49:29", "remaining_time": "0:28:07", "throughput": 5176.67, "total_tokens": 34009248} {"current_steps": 33755, "total_steps": 42420, "loss": 0.3172, "lr": 0.00012183388180348471, "epoch": 15.914662894860914, "percentage": 79.57, "elapsed_time": "1:49:30", "remaining_time": "0:28:06", "throughput": 5176.67, "total_tokens": 34013760} {"current_steps": 33760, "total_steps": 42420, "loss": 0.3298, "lr": 0.000121699334262448, "epoch": 15.917020273455917, "percentage": 79.59, "elapsed_time": "1:49:31", "remaining_time": "0:28:05", "throughput": 5176.66, "total_tokens": 34018016} {"current_steps": 33765, "total_steps": 42420, "loss": 0.4022, "lr": 0.00012156485076129064, "epoch": 15.919377652050919, "percentage": 79.6, "elapsed_time": "1:49:32", "remaining_time": "0:28:04", "throughput": 5176.71, "total_tokens": 34023488} {"current_steps": 33770, "total_steps": 42420, "loss": 0.3701, "lr": 0.0001214304313227786, "epoch": 15.921735030645921, "percentage": 79.61, "elapsed_time": "1:49:33", "remaining_time": "0:28:03", "throughput": 5176.68, "total_tokens": 34027296} {"current_steps": 33775, "total_steps": 42420, "loss": 0.3545, "lr": 0.0001212960759696668, "epoch": 15.924092409240924, "percentage": 79.62, "elapsed_time": "1:49:34", "remaining_time": "0:28:02", "throughput": 5176.67, "total_tokens": 34031520} {"current_steps": 33780, "total_steps": 42420, "loss": 0.3693, "lr": 0.0001211617847246993, "epoch": 15.926449787835926, "percentage": 79.63, "elapsed_time": "1:49:34", "remaining_time": "0:28:01", "throughput": 5176.69, "total_tokens": 34036064} {"current_steps": 33785, "total_steps": 42420, "loss": 0.2844, "lr": 0.00012102755761060936, "epoch": 15.928807166430929, "percentage": 79.64, "elapsed_time": "1:49:35", "remaining_time": "0:28:00", "throughput": 5176.77, "total_tokens": 34041952} {"current_steps": 33790, "total_steps": 42420, "loss": 0.2791, "lr": 0.00012089339465011934, "epoch": 15.931164545025931, "percentage": 79.66, "elapsed_time": "1:49:37", "remaining_time": "0:27:59", "throughput": 5176.86, "total_tokens": 34048288} {"current_steps": 33795, "total_steps": 42420, "loss": 0.3849, "lr": 0.0001207592958659407, "epoch": 15.933521923620933, "percentage": 79.67, "elapsed_time": "1:49:37", "remaining_time": "0:27:58", "throughput": 5176.89, "total_tokens": 34053184} {"current_steps": 33800, "total_steps": 42420, "loss": 0.3398, "lr": 0.00012062526128077428, "epoch": 15.935879302215936, "percentage": 79.68, "elapsed_time": "1:49:38", "remaining_time": "0:27:57", "throughput": 5176.93, "total_tokens": 34058560} {"current_steps": 33805, "total_steps": 42420, "loss": 0.3434, "lr": 0.00012049129091730959, "epoch": 15.938236680810938, "percentage": 79.69, "elapsed_time": "1:49:39", "remaining_time": "0:27:56", "throughput": 5177.02, "total_tokens": 34064544} {"current_steps": 33810, "total_steps": 42420, "loss": 0.3084, "lr": 0.0001203573847982256, "epoch": 15.94059405940594, "percentage": 79.7, "elapsed_time": "1:49:40", "remaining_time": "0:27:55", "throughput": 5177.07, "total_tokens": 34069600} {"current_steps": 33815, "total_steps": 42420, "loss": 0.3044, "lr": 0.00012022354294619048, "epoch": 15.942951438000943, "percentage": 79.71, "elapsed_time": "1:49:41", "remaining_time": "0:27:54", "throughput": 5177.08, "total_tokens": 34074208} {"current_steps": 33820, "total_steps": 42420, "loss": 0.3466, "lr": 0.00012008976538386124, "epoch": 15.945308816595945, "percentage": 79.73, "elapsed_time": "1:49:42", "remaining_time": "0:27:53", "throughput": 5177.1, "total_tokens": 34079040} {"current_steps": 33825, "total_steps": 42420, "loss": 0.3224, "lr": 0.00011995605213388427, "epoch": 15.947666195190948, "percentage": 79.74, "elapsed_time": "1:49:43", "remaining_time": "0:27:52", "throughput": 5177.21, "total_tokens": 34085408} {"current_steps": 33830, "total_steps": 42420, "loss": 0.3823, "lr": 0.00011982240321889487, "epoch": 15.95002357378595, "percentage": 79.75, "elapsed_time": "1:49:44", "remaining_time": "0:27:51", "throughput": 5177.26, "total_tokens": 34090624} {"current_steps": 33835, "total_steps": 42420, "loss": 0.3675, "lr": 0.00011968881866151753, "epoch": 15.952380952380953, "percentage": 79.76, "elapsed_time": "1:49:45", "remaining_time": "0:27:50", "throughput": 5177.31, "total_tokens": 34095904} {"current_steps": 33840, "total_steps": 42420, "loss": 0.3289, "lr": 0.00011955529848436602, "epoch": 15.954738330975955, "percentage": 79.77, "elapsed_time": "1:49:46", "remaining_time": "0:27:49", "throughput": 5177.36, "total_tokens": 34100960} {"current_steps": 33845, "total_steps": 42420, "loss": 0.2979, "lr": 0.0001194218427100428, "epoch": 15.957095709570957, "percentage": 79.79, "elapsed_time": "1:49:47", "remaining_time": "0:27:49", "throughput": 5177.43, "total_tokens": 34106464} {"current_steps": 33850, "total_steps": 42420, "loss": 0.3044, "lr": 0.0001192884513611397, "epoch": 15.95945308816596, "percentage": 79.8, "elapsed_time": "1:49:48", "remaining_time": "0:27:48", "throughput": 5177.49, "total_tokens": 34111680} {"current_steps": 33855, "total_steps": 42420, "loss": 0.3696, "lr": 0.00011915512446023768, "epoch": 15.961810466760962, "percentage": 79.81, "elapsed_time": "1:49:49", "remaining_time": "0:27:47", "throughput": 5177.5, "total_tokens": 34116256} {"current_steps": 33860, "total_steps": 42420, "loss": 0.3151, "lr": 0.00011902186202990672, "epoch": 15.964167845355965, "percentage": 79.82, "elapsed_time": "1:49:50", "remaining_time": "0:27:46", "throughput": 5177.57, "total_tokens": 34121792} {"current_steps": 33865, "total_steps": 42420, "loss": 0.3676, "lr": 0.00011888866409270593, "epoch": 15.966525223950967, "percentage": 79.83, "elapsed_time": "1:49:51", "remaining_time": "0:27:45", "throughput": 5177.59, "total_tokens": 34126464} {"current_steps": 33870, "total_steps": 42420, "loss": 0.329, "lr": 0.00011875553067118333, "epoch": 15.96888260254597, "percentage": 79.84, "elapsed_time": "1:49:51", "remaining_time": "0:27:44", "throughput": 5177.52, "total_tokens": 34129920} {"current_steps": 33875, "total_steps": 42420, "loss": 0.3168, "lr": 0.00011862246178787628, "epoch": 15.971239981140972, "percentage": 79.86, "elapsed_time": "1:49:52", "remaining_time": "0:27:43", "throughput": 5177.54, "total_tokens": 34134464} {"current_steps": 33880, "total_steps": 42420, "loss": 0.3241, "lr": 0.00011848945746531114, "epoch": 15.973597359735974, "percentage": 79.87, "elapsed_time": "1:49:53", "remaining_time": "0:27:42", "throughput": 5177.52, "total_tokens": 34138496} {"current_steps": 33885, "total_steps": 42420, "loss": 0.2961, "lr": 0.00011835651772600304, "epoch": 15.975954738330977, "percentage": 79.88, "elapsed_time": "1:49:54", "remaining_time": "0:27:41", "throughput": 5177.63, "total_tokens": 34144960} {"current_steps": 33890, "total_steps": 42420, "loss": 0.3172, "lr": 0.00011822364259245655, "epoch": 15.978312116925979, "percentage": 79.89, "elapsed_time": "1:49:55", "remaining_time": "0:27:40", "throughput": 5177.67, "total_tokens": 34149792} {"current_steps": 33895, "total_steps": 42420, "loss": 0.2928, "lr": 0.00011809083208716515, "epoch": 15.980669495520981, "percentage": 79.9, "elapsed_time": "1:49:56", "remaining_time": "0:27:39", "throughput": 5177.7, "total_tokens": 34154912} {"current_steps": 33900, "total_steps": 42420, "loss": 0.3204, "lr": 0.00011795808623261145, "epoch": 15.983026874115984, "percentage": 79.92, "elapsed_time": "1:49:57", "remaining_time": "0:27:38", "throughput": 5177.76, "total_tokens": 34160160} {"current_steps": 33905, "total_steps": 42420, "loss": 0.2869, "lr": 0.00011782540505126704, "epoch": 15.985384252710986, "percentage": 79.93, "elapsed_time": "1:49:58", "remaining_time": "0:27:37", "throughput": 5177.78, "total_tokens": 34164800} {"current_steps": 33910, "total_steps": 42420, "loss": 0.3486, "lr": 0.00011769278856559262, "epoch": 15.987741631305987, "percentage": 79.94, "elapsed_time": "1:49:59", "remaining_time": "0:27:36", "throughput": 5177.82, "total_tokens": 34169728} {"current_steps": 33915, "total_steps": 42420, "loss": 0.3681, "lr": 0.0001175602367980379, "epoch": 15.990099009900991, "percentage": 79.95, "elapsed_time": "1:50:00", "remaining_time": "0:27:35", "throughput": 5177.83, "total_tokens": 34174368} {"current_steps": 33920, "total_steps": 42420, "loss": 0.314, "lr": 0.00011742774977104176, "epoch": 15.992456388495992, "percentage": 79.96, "elapsed_time": "1:50:01", "remaining_time": "0:27:34", "throughput": 5177.89, "total_tokens": 34179648} {"current_steps": 33925, "total_steps": 42420, "loss": 0.3359, "lr": 0.0001172953275070317, "epoch": 15.994813767090994, "percentage": 79.97, "elapsed_time": "1:50:02", "remaining_time": "0:27:33", "throughput": 5177.93, "total_tokens": 34184736} {"current_steps": 33930, "total_steps": 42420, "loss": 0.3033, "lr": 0.0001171629700284248, "epoch": 15.997171145685996, "percentage": 79.99, "elapsed_time": "1:50:03", "remaining_time": "0:27:32", "throughput": 5178.03, "total_tokens": 34190880} {"current_steps": 33935, "total_steps": 42420, "loss": 0.3823, "lr": 0.00011703067735762696, "epoch": 15.999528524280999, "percentage": 80.0, "elapsed_time": "1:50:04", "remaining_time": "0:27:31", "throughput": 5178.17, "total_tokens": 34198528} {"current_steps": 33936, "total_steps": 42420, "eval_loss": 0.338424414396286, "epoch": 16.0, "percentage": 80.0, "elapsed_time": "1:50:23", "remaining_time": "0:27:35", "throughput": 5163.26, "total_tokens": 34199392} {"current_steps": 33940, "total_steps": 42420, "loss": 0.3907, "lr": 0.00011689844951703288, "epoch": 16.001885902876, "percentage": 80.01, "elapsed_time": "1:50:26", "remaining_time": "0:27:35", "throughput": 5161.6, "total_tokens": 34203648} {"current_steps": 33945, "total_steps": 42420, "loss": 0.3362, "lr": 0.0001167662865290266, "epoch": 16.004243281471005, "percentage": 80.02, "elapsed_time": "1:50:27", "remaining_time": "0:27:34", "throughput": 5161.57, "total_tokens": 34207392} {"current_steps": 33950, "total_steps": 42420, "loss": 0.3049, "lr": 0.00011663418841598105, "epoch": 16.006600660066006, "percentage": 80.03, "elapsed_time": "1:50:28", "remaining_time": "0:27:33", "throughput": 5161.61, "total_tokens": 34212480} {"current_steps": 33955, "total_steps": 42420, "loss": 0.3507, "lr": 0.00011650215520025825, "epoch": 16.00895803866101, "percentage": 80.04, "elapsed_time": "1:50:29", "remaining_time": "0:27:32", "throughput": 5161.65, "total_tokens": 34217856} {"current_steps": 33960, "total_steps": 42420, "loss": 0.3458, "lr": 0.00011637018690420914, "epoch": 16.01131541725601, "percentage": 80.06, "elapsed_time": "1:50:30", "remaining_time": "0:27:31", "throughput": 5161.63, "total_tokens": 34221728} {"current_steps": 33965, "total_steps": 42420, "loss": 0.346, "lr": 0.00011623828355017369, "epoch": 16.013672795851015, "percentage": 80.07, "elapsed_time": "1:50:30", "remaining_time": "0:27:30", "throughput": 5161.65, "total_tokens": 34226432} {"current_steps": 33970, "total_steps": 42420, "loss": 0.3201, "lr": 0.00011610644516048114, "epoch": 16.016030174446016, "percentage": 80.08, "elapsed_time": "1:50:31", "remaining_time": "0:27:29", "throughput": 5161.71, "total_tokens": 34232160} {"current_steps": 33975, "total_steps": 42420, "loss": 0.3159, "lr": 0.00011597467175744908, "epoch": 16.01838755304102, "percentage": 80.09, "elapsed_time": "1:50:32", "remaining_time": "0:27:28", "throughput": 5161.74, "total_tokens": 34237120} {"current_steps": 33980, "total_steps": 42420, "loss": 0.3643, "lr": 0.00011584296336338473, "epoch": 16.02074493163602, "percentage": 80.1, "elapsed_time": "1:50:33", "remaining_time": "0:27:27", "throughput": 5161.82, "total_tokens": 34242880} {"current_steps": 33985, "total_steps": 42420, "loss": 0.3837, "lr": 0.0001157113200005841, "epoch": 16.023102310231025, "percentage": 80.12, "elapsed_time": "1:50:34", "remaining_time": "0:27:26", "throughput": 5161.89, "total_tokens": 34248640} {"current_steps": 33990, "total_steps": 42420, "loss": 0.324, "lr": 0.00011557974169133212, "epoch": 16.025459688826025, "percentage": 80.13, "elapsed_time": "1:50:35", "remaining_time": "0:27:25", "throughput": 5161.91, "total_tokens": 34253504} {"current_steps": 33995, "total_steps": 42420, "loss": 0.3116, "lr": 0.00011544822845790277, "epoch": 16.02781706742103, "percentage": 80.14, "elapsed_time": "1:50:36", "remaining_time": "0:27:24", "throughput": 5161.88, "total_tokens": 34257408} {"current_steps": 34000, "total_steps": 42420, "loss": 0.352, "lr": 0.00011531678032255904, "epoch": 16.03017444601603, "percentage": 80.15, "elapsed_time": "1:50:37", "remaining_time": "0:27:23", "throughput": 5161.89, "total_tokens": 34261856} {"current_steps": 34005, "total_steps": 42420, "loss": 0.2988, "lr": 0.00011518539730755284, "epoch": 16.032531824611034, "percentage": 80.16, "elapsed_time": "1:50:38", "remaining_time": "0:27:22", "throughput": 5162.01, "total_tokens": 34268864} {"current_steps": 34010, "total_steps": 42420, "loss": 0.3242, "lr": 0.00011505407943512525, "epoch": 16.034889203206035, "percentage": 80.17, "elapsed_time": "1:50:39", "remaining_time": "0:27:21", "throughput": 5162.08, "total_tokens": 34274560} {"current_steps": 34015, "total_steps": 42420, "loss": 0.3717, "lr": 0.00011492282672750587, "epoch": 16.03724658180104, "percentage": 80.19, "elapsed_time": "1:50:40", "remaining_time": "0:27:20", "throughput": 5162.05, "total_tokens": 34278432} {"current_steps": 34020, "total_steps": 42420, "loss": 0.3594, "lr": 0.00011479163920691366, "epoch": 16.03960396039604, "percentage": 80.2, "elapsed_time": "1:50:41", "remaining_time": "0:27:19", "throughput": 5162.1, "total_tokens": 34283552} {"current_steps": 34025, "total_steps": 42420, "loss": 0.3059, "lr": 0.00011466051689555651, "epoch": 16.04196133899104, "percentage": 80.21, "elapsed_time": "1:50:42", "remaining_time": "0:27:18", "throughput": 5162.14, "total_tokens": 34288928} {"current_steps": 34030, "total_steps": 42420, "loss": 0.3561, "lr": 0.00011452945981563111, "epoch": 16.044318717586044, "percentage": 80.22, "elapsed_time": "1:50:43", "remaining_time": "0:27:17", "throughput": 5162.16, "total_tokens": 34293504} {"current_steps": 34035, "total_steps": 42420, "loss": 0.3575, "lr": 0.00011439846798932329, "epoch": 16.046676096181045, "percentage": 80.23, "elapsed_time": "1:50:44", "remaining_time": "0:27:16", "throughput": 5162.18, "total_tokens": 34298272} {"current_steps": 34040, "total_steps": 42420, "loss": 0.3317, "lr": 0.00011426754143880768, "epoch": 16.04903347477605, "percentage": 80.25, "elapsed_time": "1:50:45", "remaining_time": "0:27:15", "throughput": 5162.25, "total_tokens": 34304096} {"current_steps": 34045, "total_steps": 42420, "loss": 0.3439, "lr": 0.00011413668018624795, "epoch": 16.05139085337105, "percentage": 80.26, "elapsed_time": "1:50:46", "remaining_time": "0:27:14", "throughput": 5162.31, "total_tokens": 34309216} {"current_steps": 34050, "total_steps": 42420, "loss": 0.3222, "lr": 0.00011400588425379676, "epoch": 16.053748231966054, "percentage": 80.27, "elapsed_time": "1:50:47", "remaining_time": "0:27:13", "throughput": 5162.36, "total_tokens": 34314336} {"current_steps": 34055, "total_steps": 42420, "loss": 0.3676, "lr": 0.00011387515366359547, "epoch": 16.056105610561055, "percentage": 80.28, "elapsed_time": "1:50:47", "remaining_time": "0:27:12", "throughput": 5162.35, "total_tokens": 34318720} {"current_steps": 34060, "total_steps": 42420, "loss": 0.3499, "lr": 0.00011374448843777463, "epoch": 16.05846298915606, "percentage": 80.29, "elapsed_time": "1:50:48", "remaining_time": "0:27:11", "throughput": 5162.38, "total_tokens": 34323808} {"current_steps": 34065, "total_steps": 42420, "loss": 0.3228, "lr": 0.00011361388859845367, "epoch": 16.06082036775106, "percentage": 80.3, "elapsed_time": "1:50:49", "remaining_time": "0:27:10", "throughput": 5162.45, "total_tokens": 34329632} {"current_steps": 34070, "total_steps": 42420, "loss": 0.3746, "lr": 0.00011348335416774092, "epoch": 16.063177746346064, "percentage": 80.32, "elapsed_time": "1:50:50", "remaining_time": "0:27:09", "throughput": 5162.46, "total_tokens": 34333984} {"current_steps": 34075, "total_steps": 42420, "loss": 0.3165, "lr": 0.00011335288516773357, "epoch": 16.065535124941064, "percentage": 80.33, "elapsed_time": "1:50:51", "remaining_time": "0:27:09", "throughput": 5162.51, "total_tokens": 34339328} {"current_steps": 34080, "total_steps": 42420, "loss": 0.331, "lr": 0.0001132224816205179, "epoch": 16.06789250353607, "percentage": 80.34, "elapsed_time": "1:50:52", "remaining_time": "0:27:08", "throughput": 5162.55, "total_tokens": 34344448} {"current_steps": 34085, "total_steps": 42420, "loss": 0.3243, "lr": 0.00011309214354816899, "epoch": 16.07024988213107, "percentage": 80.35, "elapsed_time": "1:50:53", "remaining_time": "0:27:07", "throughput": 5162.6, "total_tokens": 34349632} {"current_steps": 34090, "total_steps": 42420, "loss": 0.3416, "lr": 0.00011296187097275101, "epoch": 16.072607260726073, "percentage": 80.36, "elapsed_time": "1:50:54", "remaining_time": "0:27:06", "throughput": 5162.62, "total_tokens": 34354336} {"current_steps": 34095, "total_steps": 42420, "loss": 0.3096, "lr": 0.00011283166391631655, "epoch": 16.074964639321074, "percentage": 80.37, "elapsed_time": "1:50:55", "remaining_time": "0:27:05", "throughput": 5162.68, "total_tokens": 34359360} {"current_steps": 34100, "total_steps": 42420, "loss": 0.3443, "lr": 0.00011270152240090769, "epoch": 16.077322017916078, "percentage": 80.39, "elapsed_time": "1:50:56", "remaining_time": "0:27:04", "throughput": 5162.71, "total_tokens": 34364480} {"current_steps": 34105, "total_steps": 42420, "loss": 0.3506, "lr": 0.0001125714464485551, "epoch": 16.07967939651108, "percentage": 80.4, "elapsed_time": "1:50:57", "remaining_time": "0:27:03", "throughput": 5162.74, "total_tokens": 34369440} {"current_steps": 34110, "total_steps": 42420, "loss": 0.3068, "lr": 0.00011244143608127849, "epoch": 16.082036775106083, "percentage": 80.41, "elapsed_time": "1:50:58", "remaining_time": "0:27:02", "throughput": 5162.81, "total_tokens": 34374752} {"current_steps": 34115, "total_steps": 42420, "loss": 0.3627, "lr": 0.00011231149132108637, "epoch": 16.084394153701083, "percentage": 80.42, "elapsed_time": "1:50:59", "remaining_time": "0:27:01", "throughput": 5162.88, "total_tokens": 34380352} {"current_steps": 34120, "total_steps": 42420, "loss": 0.329, "lr": 0.00011218161218997619, "epoch": 16.086751532296088, "percentage": 80.43, "elapsed_time": "1:51:00", "remaining_time": "0:27:00", "throughput": 5162.98, "total_tokens": 34386560} {"current_steps": 34125, "total_steps": 42420, "loss": 0.3498, "lr": 0.00011205179870993432, "epoch": 16.08910891089109, "percentage": 80.45, "elapsed_time": "1:51:01", "remaining_time": "0:26:59", "throughput": 5163.03, "total_tokens": 34391840} {"current_steps": 34130, "total_steps": 42420, "loss": 0.3599, "lr": 0.00011192205090293606, "epoch": 16.091466289486092, "percentage": 80.46, "elapsed_time": "1:51:02", "remaining_time": "0:26:58", "throughput": 5163.07, "total_tokens": 34397152} {"current_steps": 34135, "total_steps": 42420, "loss": 0.3333, "lr": 0.00011179236879094523, "epoch": 16.093823668081093, "percentage": 80.47, "elapsed_time": "1:51:03", "remaining_time": "0:26:57", "throughput": 5163.11, "total_tokens": 34402112} {"current_steps": 34140, "total_steps": 42420, "loss": 0.3147, "lr": 0.000111662752395915, "epoch": 16.096181046676097, "percentage": 80.48, "elapsed_time": "1:51:03", "remaining_time": "0:26:56", "throughput": 5163.15, "total_tokens": 34407168} {"current_steps": 34145, "total_steps": 42420, "loss": 0.3419, "lr": 0.00011153320173978715, "epoch": 16.098538425271098, "percentage": 80.49, "elapsed_time": "1:51:04", "remaining_time": "0:26:55", "throughput": 5163.19, "total_tokens": 34412224} {"current_steps": 34150, "total_steps": 42420, "loss": 0.3247, "lr": 0.00011140371684449251, "epoch": 16.100895803866102, "percentage": 80.5, "elapsed_time": "1:51:05", "remaining_time": "0:26:54", "throughput": 5163.2, "total_tokens": 34416512} {"current_steps": 34155, "total_steps": 42420, "loss": 0.3277, "lr": 0.00011127429773195058, "epoch": 16.103253182461103, "percentage": 80.52, "elapsed_time": "1:51:06", "remaining_time": "0:26:53", "throughput": 5163.27, "total_tokens": 34422176} {"current_steps": 34160, "total_steps": 42420, "loss": 0.3228, "lr": 0.00011114494442407003, "epoch": 16.105610561056107, "percentage": 80.53, "elapsed_time": "1:51:07", "remaining_time": "0:26:52", "throughput": 5163.43, "total_tokens": 34429504} {"current_steps": 34165, "total_steps": 42420, "loss": 0.3582, "lr": 0.00011101565694274785, "epoch": 16.107967939651108, "percentage": 80.54, "elapsed_time": "1:51:08", "remaining_time": "0:26:51", "throughput": 5163.51, "total_tokens": 34435200} {"current_steps": 34170, "total_steps": 42420, "loss": 0.3098, "lr": 0.00011088643530987042, "epoch": 16.11032531824611, "percentage": 80.55, "elapsed_time": "1:51:09", "remaining_time": "0:26:50", "throughput": 5163.55, "total_tokens": 34440192} {"current_steps": 34175, "total_steps": 42420, "loss": 0.3005, "lr": 0.00011075727954731269, "epoch": 16.112682696841112, "percentage": 80.56, "elapsed_time": "1:51:10", "remaining_time": "0:26:49", "throughput": 5163.58, "total_tokens": 34445152} {"current_steps": 34180, "total_steps": 42420, "loss": 0.2902, "lr": 0.00011062818967693866, "epoch": 16.115040075436116, "percentage": 80.58, "elapsed_time": "1:51:11", "remaining_time": "0:26:48", "throughput": 5163.7, "total_tokens": 34451584} {"current_steps": 34185, "total_steps": 42420, "loss": 0.3396, "lr": 0.00011049916572060087, "epoch": 16.117397454031117, "percentage": 80.59, "elapsed_time": "1:51:12", "remaining_time": "0:26:47", "throughput": 5163.74, "total_tokens": 34456704} {"current_steps": 34190, "total_steps": 42420, "loss": 0.3933, "lr": 0.00011037020770014094, "epoch": 16.11975483262612, "percentage": 80.6, "elapsed_time": "1:51:13", "remaining_time": "0:26:46", "throughput": 5163.71, "total_tokens": 34460672} {"current_steps": 34195, "total_steps": 42420, "loss": 0.3904, "lr": 0.00011024131563738932, "epoch": 16.122112211221122, "percentage": 80.61, "elapsed_time": "1:51:14", "remaining_time": "0:26:45", "throughput": 5163.73, "total_tokens": 34465344} {"current_steps": 34200, "total_steps": 42420, "loss": 0.308, "lr": 0.00011011248955416525, "epoch": 16.124469589816126, "percentage": 80.62, "elapsed_time": "1:51:15", "remaining_time": "0:26:44", "throughput": 5163.72, "total_tokens": 34469504} {"current_steps": 34205, "total_steps": 42420, "loss": 0.2957, "lr": 0.00010998372947227681, "epoch": 16.126826968411127, "percentage": 80.63, "elapsed_time": "1:51:16", "remaining_time": "0:26:43", "throughput": 5163.72, "total_tokens": 34473824} {"current_steps": 34210, "total_steps": 42420, "loss": 0.3466, "lr": 0.00010985503541352081, "epoch": 16.12918434700613, "percentage": 80.65, "elapsed_time": "1:51:17", "remaining_time": "0:26:42", "throughput": 5163.77, "total_tokens": 34479040} {"current_steps": 34215, "total_steps": 42420, "loss": 0.3464, "lr": 0.0001097264073996831, "epoch": 16.13154172560113, "percentage": 80.66, "elapsed_time": "1:51:18", "remaining_time": "0:26:41", "throughput": 5163.8, "total_tokens": 34484000} {"current_steps": 34220, "total_steps": 42420, "loss": 0.3407, "lr": 0.00010959784545253826, "epoch": 16.133899104196132, "percentage": 80.67, "elapsed_time": "1:51:19", "remaining_time": "0:26:40", "throughput": 5163.87, "total_tokens": 34489472} {"current_steps": 34225, "total_steps": 42420, "loss": 0.3614, "lr": 0.00010946934959384936, "epoch": 16.136256482791136, "percentage": 80.68, "elapsed_time": "1:51:19", "remaining_time": "0:26:39", "throughput": 5163.88, "total_tokens": 34493984} {"current_steps": 34230, "total_steps": 42420, "loss": 0.3141, "lr": 0.00010934091984536876, "epoch": 16.138613861386137, "percentage": 80.69, "elapsed_time": "1:51:20", "remaining_time": "0:26:38", "throughput": 5163.93, "total_tokens": 34499296} {"current_steps": 34235, "total_steps": 42420, "loss": 0.3172, "lr": 0.00010921255622883741, "epoch": 16.14097123998114, "percentage": 80.7, "elapsed_time": "1:51:21", "remaining_time": "0:26:37", "throughput": 5163.96, "total_tokens": 34504096} {"current_steps": 34240, "total_steps": 42420, "loss": 0.3317, "lr": 0.0001090842587659851, "epoch": 16.14332861857614, "percentage": 80.72, "elapsed_time": "1:51:22", "remaining_time": "0:26:36", "throughput": 5163.98, "total_tokens": 34508960} {"current_steps": 34245, "total_steps": 42420, "loss": 0.288, "lr": 0.00010895602747853046, "epoch": 16.145685997171146, "percentage": 80.73, "elapsed_time": "1:51:23", "remaining_time": "0:26:35", "throughput": 5163.99, "total_tokens": 34513664} {"current_steps": 34250, "total_steps": 42420, "loss": 0.3509, "lr": 0.00010882786238818077, "epoch": 16.148043375766147, "percentage": 80.74, "elapsed_time": "1:51:24", "remaining_time": "0:26:34", "throughput": 5164.05, "total_tokens": 34519040} {"current_steps": 34255, "total_steps": 42420, "loss": 0.2968, "lr": 0.00010869976351663224, "epoch": 16.15040075436115, "percentage": 80.75, "elapsed_time": "1:51:25", "remaining_time": "0:26:33", "throughput": 5164.07, "total_tokens": 34523712} {"current_steps": 34260, "total_steps": 42420, "loss": 0.3734, "lr": 0.00010857173088557004, "epoch": 16.15275813295615, "percentage": 80.76, "elapsed_time": "1:51:26", "remaining_time": "0:26:32", "throughput": 5164.16, "total_tokens": 34529568} {"current_steps": 34265, "total_steps": 42420, "loss": 0.3624, "lr": 0.00010844376451666755, "epoch": 16.155115511551156, "percentage": 80.78, "elapsed_time": "1:51:27", "remaining_time": "0:26:31", "throughput": 5164.19, "total_tokens": 34534464} {"current_steps": 34270, "total_steps": 42420, "loss": 0.3339, "lr": 0.00010831586443158748, "epoch": 16.157472890146156, "percentage": 80.79, "elapsed_time": "1:51:28", "remaining_time": "0:26:30", "throughput": 5164.31, "total_tokens": 34541824} {"current_steps": 34275, "total_steps": 42420, "loss": 0.3615, "lr": 0.00010818803065198113, "epoch": 16.15983026874116, "percentage": 80.8, "elapsed_time": "1:51:29", "remaining_time": "0:26:29", "throughput": 5164.31, "total_tokens": 34546112} {"current_steps": 34280, "total_steps": 42420, "loss": 0.3137, "lr": 0.00010806026319948864, "epoch": 16.16218764733616, "percentage": 80.81, "elapsed_time": "1:51:30", "remaining_time": "0:26:28", "throughput": 5164.33, "total_tokens": 34551104} {"current_steps": 34285, "total_steps": 42420, "loss": 0.3415, "lr": 0.00010793256209573882, "epoch": 16.164545025931165, "percentage": 80.82, "elapsed_time": "1:51:31", "remaining_time": "0:26:27", "throughput": 5164.4, "total_tokens": 34556928} {"current_steps": 34290, "total_steps": 42420, "loss": 0.3484, "lr": 0.00010780492736234932, "epoch": 16.166902404526166, "percentage": 80.83, "elapsed_time": "1:51:32", "remaining_time": "0:26:26", "throughput": 5164.46, "total_tokens": 34562208} {"current_steps": 34295, "total_steps": 42420, "loss": 0.3484, "lr": 0.00010767735902092652, "epoch": 16.16925978312117, "percentage": 80.85, "elapsed_time": "1:51:33", "remaining_time": "0:26:25", "throughput": 5164.45, "total_tokens": 34566464} {"current_steps": 34300, "total_steps": 42420, "loss": 0.3591, "lr": 0.00010754985709306569, "epoch": 16.17161716171617, "percentage": 80.86, "elapsed_time": "1:51:34", "remaining_time": "0:26:24", "throughput": 5164.49, "total_tokens": 34571648} {"current_steps": 34305, "total_steps": 42420, "loss": 0.336, "lr": 0.00010742242160035049, "epoch": 16.173974540311175, "percentage": 80.87, "elapsed_time": "1:51:35", "remaining_time": "0:26:23", "throughput": 5164.57, "total_tokens": 34577216} {"current_steps": 34310, "total_steps": 42420, "loss": 0.3372, "lr": 0.00010729505256435373, "epoch": 16.176331918906175, "percentage": 80.88, "elapsed_time": "1:51:36", "remaining_time": "0:26:22", "throughput": 5164.63, "total_tokens": 34582560} {"current_steps": 34315, "total_steps": 42420, "loss": 0.2942, "lr": 0.00010716775000663686, "epoch": 16.17868929750118, "percentage": 80.89, "elapsed_time": "1:51:36", "remaining_time": "0:26:21", "throughput": 5164.62, "total_tokens": 34586592} {"current_steps": 34320, "total_steps": 42420, "loss": 0.3036, "lr": 0.00010704051394874998, "epoch": 16.18104667609618, "percentage": 80.91, "elapsed_time": "1:51:37", "remaining_time": "0:26:20", "throughput": 5164.64, "total_tokens": 34591008} {"current_steps": 34325, "total_steps": 42420, "loss": 0.3382, "lr": 0.000106913344412232, "epoch": 16.183404054691184, "percentage": 80.92, "elapsed_time": "1:51:38", "remaining_time": "0:26:19", "throughput": 5164.7, "total_tokens": 34596480} {"current_steps": 34330, "total_steps": 42420, "loss": 0.3091, "lr": 0.0001067862414186106, "epoch": 16.185761433286185, "percentage": 80.93, "elapsed_time": "1:51:39", "remaining_time": "0:26:18", "throughput": 5164.73, "total_tokens": 34601152} {"current_steps": 34335, "total_steps": 42420, "loss": 0.2875, "lr": 0.00010665920498940213, "epoch": 16.18811881188119, "percentage": 80.94, "elapsed_time": "1:51:40", "remaining_time": "0:26:17", "throughput": 5164.75, "total_tokens": 34606176} {"current_steps": 34340, "total_steps": 42420, "loss": 0.3086, "lr": 0.00010653223514611183, "epoch": 16.19047619047619, "percentage": 80.95, "elapsed_time": "1:51:41", "remaining_time": "0:26:16", "throughput": 5164.82, "total_tokens": 34611552} {"current_steps": 34345, "total_steps": 42420, "loss": 0.307, "lr": 0.00010640533191023333, "epoch": 16.192833569071194, "percentage": 80.96, "elapsed_time": "1:51:42", "remaining_time": "0:26:15", "throughput": 5164.82, "total_tokens": 34616256} {"current_steps": 34350, "total_steps": 42420, "loss": 0.3608, "lr": 0.00010627849530324924, "epoch": 16.195190947666195, "percentage": 80.98, "elapsed_time": "1:51:43", "remaining_time": "0:26:14", "throughput": 5164.96, "total_tokens": 34622752} {"current_steps": 34355, "total_steps": 42420, "loss": 0.339, "lr": 0.00010615172534663087, "epoch": 16.1975483262612, "percentage": 80.99, "elapsed_time": "1:51:44", "remaining_time": "0:26:13", "throughput": 5164.91, "total_tokens": 34626368} {"current_steps": 34360, "total_steps": 42420, "loss": 0.3153, "lr": 0.00010602502206183828, "epoch": 16.1999057048562, "percentage": 81.0, "elapsed_time": "1:51:44", "remaining_time": "0:26:12", "throughput": 5164.86, "total_tokens": 34629920} {"current_steps": 34365, "total_steps": 42420, "loss": 0.4059, "lr": 0.0001058983854703201, "epoch": 16.202263083451204, "percentage": 81.01, "elapsed_time": "1:51:45", "remaining_time": "0:26:11", "throughput": 5164.88, "total_tokens": 34634400} {"current_steps": 34370, "total_steps": 42420, "loss": 0.3693, "lr": 0.00010577181559351384, "epoch": 16.204620462046204, "percentage": 81.02, "elapsed_time": "1:51:46", "remaining_time": "0:26:10", "throughput": 5164.9, "total_tokens": 34639200} {"current_steps": 34375, "total_steps": 42420, "loss": 0.4528, "lr": 0.00010564531245284553, "epoch": 16.20697784064121, "percentage": 81.03, "elapsed_time": "1:51:47", "remaining_time": "0:26:09", "throughput": 5164.9, "total_tokens": 34643424} {"current_steps": 34380, "total_steps": 42420, "loss": 0.3396, "lr": 0.00010551887606973026, "epoch": 16.20933521923621, "percentage": 81.05, "elapsed_time": "1:51:48", "remaining_time": "0:26:08", "throughput": 5164.94, "total_tokens": 34648384} {"current_steps": 34385, "total_steps": 42420, "loss": 0.3611, "lr": 0.0001053925064655713, "epoch": 16.211692597831213, "percentage": 81.06, "elapsed_time": "1:51:49", "remaining_time": "0:26:07", "throughput": 5165.0, "total_tokens": 34654464} {"current_steps": 34390, "total_steps": 42420, "loss": 0.3336, "lr": 0.00010526620366176087, "epoch": 16.214049976426214, "percentage": 81.07, "elapsed_time": "1:51:50", "remaining_time": "0:26:06", "throughput": 5165.03, "total_tokens": 34659392} {"current_steps": 34395, "total_steps": 42420, "loss": 0.3324, "lr": 0.00010513996767967993, "epoch": 16.216407355021218, "percentage": 81.08, "elapsed_time": "1:51:51", "remaining_time": "0:26:05", "throughput": 5165.05, "total_tokens": 34664224} {"current_steps": 34400, "total_steps": 42420, "loss": 0.3268, "lr": 0.00010501379854069809, "epoch": 16.21876473361622, "percentage": 81.09, "elapsed_time": "1:51:52", "remaining_time": "0:26:04", "throughput": 5165.12, "total_tokens": 34669568} {"current_steps": 34405, "total_steps": 42420, "loss": 0.3478, "lr": 0.00010488769626617373, "epoch": 16.221122112211223, "percentage": 81.11, "elapsed_time": "1:51:53", "remaining_time": "0:26:03", "throughput": 5165.2, "total_tokens": 34675264} {"current_steps": 34410, "total_steps": 42420, "loss": 0.3434, "lr": 0.0001047616608774537, "epoch": 16.223479490806223, "percentage": 81.12, "elapsed_time": "1:51:54", "remaining_time": "0:26:02", "throughput": 5165.2, "total_tokens": 34679648} {"current_steps": 34415, "total_steps": 42420, "loss": 0.3427, "lr": 0.00010463569239587378, "epoch": 16.225836869401228, "percentage": 81.13, "elapsed_time": "1:51:55", "remaining_time": "0:26:01", "throughput": 5165.27, "total_tokens": 34685504} {"current_steps": 34420, "total_steps": 42420, "loss": 0.3204, "lr": 0.0001045097908427582, "epoch": 16.22819424799623, "percentage": 81.14, "elapsed_time": "1:51:56", "remaining_time": "0:26:00", "throughput": 5165.28, "total_tokens": 34690080} {"current_steps": 34425, "total_steps": 42420, "loss": 0.3029, "lr": 0.00010438395623942015, "epoch": 16.23055162659123, "percentage": 81.15, "elapsed_time": "1:51:57", "remaining_time": "0:25:59", "throughput": 5165.37, "total_tokens": 34696000} {"current_steps": 34430, "total_steps": 42420, "loss": 0.4006, "lr": 0.00010425818860716097, "epoch": 16.232909005186233, "percentage": 81.16, "elapsed_time": "1:51:57", "remaining_time": "0:25:58", "throughput": 5165.39, "total_tokens": 34700704} {"current_steps": 34435, "total_steps": 42420, "loss": 0.3146, "lr": 0.00010413248796727115, "epoch": 16.235266383781234, "percentage": 81.18, "elapsed_time": "1:51:58", "remaining_time": "0:25:58", "throughput": 5165.4, "total_tokens": 34705440} {"current_steps": 34440, "total_steps": 42420, "loss": 0.3404, "lr": 0.00010400685434102969, "epoch": 16.237623762376238, "percentage": 81.19, "elapsed_time": "1:51:59", "remaining_time": "0:25:57", "throughput": 5165.43, "total_tokens": 34710400} {"current_steps": 34445, "total_steps": 42420, "loss": 0.3394, "lr": 0.00010388128774970418, "epoch": 16.23998114097124, "percentage": 81.2, "elapsed_time": "1:52:00", "remaining_time": "0:25:56", "throughput": 5165.45, "total_tokens": 34715072} {"current_steps": 34450, "total_steps": 42420, "loss": 0.3033, "lr": 0.00010375578821455095, "epoch": 16.242338519566243, "percentage": 81.21, "elapsed_time": "1:52:01", "remaining_time": "0:25:55", "throughput": 5165.46, "total_tokens": 34719584} {"current_steps": 34455, "total_steps": 42420, "loss": 0.3301, "lr": 0.0001036303557568149, "epoch": 16.244695898161243, "percentage": 81.22, "elapsed_time": "1:52:02", "remaining_time": "0:25:54", "throughput": 5165.45, "total_tokens": 34723680} {"current_steps": 34460, "total_steps": 42420, "loss": 0.3097, "lr": 0.0001035049903977297, "epoch": 16.247053276756247, "percentage": 81.24, "elapsed_time": "1:52:03", "remaining_time": "0:25:53", "throughput": 5165.48, "total_tokens": 34728576} {"current_steps": 34465, "total_steps": 42420, "loss": 0.333, "lr": 0.00010337969215851766, "epoch": 16.249410655351248, "percentage": 81.25, "elapsed_time": "1:52:04", "remaining_time": "0:25:52", "throughput": 5165.53, "total_tokens": 34733952} {"current_steps": 34470, "total_steps": 42420, "loss": 0.3247, "lr": 0.00010325446106038933, "epoch": 16.251768033946252, "percentage": 81.26, "elapsed_time": "1:52:05", "remaining_time": "0:25:51", "throughput": 5165.56, "total_tokens": 34738688} {"current_steps": 34475, "total_steps": 42420, "loss": 0.3539, "lr": 0.00010312929712454445, "epoch": 16.254125412541253, "percentage": 81.27, "elapsed_time": "1:52:06", "remaining_time": "0:25:50", "throughput": 5165.62, "total_tokens": 34744000} {"current_steps": 34480, "total_steps": 42420, "loss": 0.3362, "lr": 0.00010300420037217101, "epoch": 16.256482791136257, "percentage": 81.28, "elapsed_time": "1:52:06", "remaining_time": "0:25:49", "throughput": 5165.66, "total_tokens": 34749088} {"current_steps": 34485, "total_steps": 42420, "loss": 0.3741, "lr": 0.00010287917082444587, "epoch": 16.258840169731258, "percentage": 81.29, "elapsed_time": "1:52:07", "remaining_time": "0:25:48", "throughput": 5165.72, "total_tokens": 34754688} {"current_steps": 34490, "total_steps": 42420, "loss": 0.3201, "lr": 0.00010275420850253442, "epoch": 16.261197548326262, "percentage": 81.31, "elapsed_time": "1:52:08", "remaining_time": "0:25:47", "throughput": 5165.79, "total_tokens": 34760064} {"current_steps": 34495, "total_steps": 42420, "loss": 0.3428, "lr": 0.00010262931342759057, "epoch": 16.263554926921262, "percentage": 81.32, "elapsed_time": "1:52:09", "remaining_time": "0:25:46", "throughput": 5165.81, "total_tokens": 34764832} {"current_steps": 34500, "total_steps": 42420, "loss": 0.3739, "lr": 0.00010250448562075703, "epoch": 16.265912305516267, "percentage": 81.33, "elapsed_time": "1:52:10", "remaining_time": "0:25:45", "throughput": 5165.81, "total_tokens": 34769408} {"current_steps": 34505, "total_steps": 42420, "loss": 0.3391, "lr": 0.00010237972510316507, "epoch": 16.268269684111267, "percentage": 81.34, "elapsed_time": "1:52:11", "remaining_time": "0:25:44", "throughput": 5165.92, "total_tokens": 34776224} {"current_steps": 34510, "total_steps": 42420, "loss": 0.338, "lr": 0.00010225503189593433, "epoch": 16.27062706270627, "percentage": 81.35, "elapsed_time": "1:52:12", "remaining_time": "0:25:43", "throughput": 5166.03, "total_tokens": 34782880} {"current_steps": 34515, "total_steps": 42420, "loss": 0.3427, "lr": 0.00010213040602017337, "epoch": 16.272984441301272, "percentage": 81.36, "elapsed_time": "1:52:13", "remaining_time": "0:25:42", "throughput": 5166.02, "total_tokens": 34786944} {"current_steps": 34520, "total_steps": 42420, "loss": 0.3321, "lr": 0.00010200584749697922, "epoch": 16.275341819896276, "percentage": 81.38, "elapsed_time": "1:52:14", "remaining_time": "0:25:41", "throughput": 5166.06, "total_tokens": 34792320} {"current_steps": 34525, "total_steps": 42420, "loss": 0.2913, "lr": 0.00010188135634743756, "epoch": 16.277699198491277, "percentage": 81.39, "elapsed_time": "1:52:15", "remaining_time": "0:25:40", "throughput": 5166.11, "total_tokens": 34797664} {"current_steps": 34530, "total_steps": 42420, "loss": 0.3287, "lr": 0.00010175693259262259, "epoch": 16.28005657708628, "percentage": 81.4, "elapsed_time": "1:52:16", "remaining_time": "0:25:39", "throughput": 5166.12, "total_tokens": 34802368} {"current_steps": 34535, "total_steps": 42420, "loss": 0.3206, "lr": 0.00010163257625359712, "epoch": 16.28241395568128, "percentage": 81.41, "elapsed_time": "1:52:17", "remaining_time": "0:25:38", "throughput": 5166.2, "total_tokens": 34808288} {"current_steps": 34540, "total_steps": 42420, "loss": 0.3157, "lr": 0.00010150828735141265, "epoch": 16.284771334276286, "percentage": 81.42, "elapsed_time": "1:52:18", "remaining_time": "0:25:37", "throughput": 5166.19, "total_tokens": 34812576} {"current_steps": 34545, "total_steps": 42420, "loss": 0.3418, "lr": 0.00010138406590710924, "epoch": 16.287128712871286, "percentage": 81.44, "elapsed_time": "1:52:19", "remaining_time": "0:25:36", "throughput": 5166.27, "total_tokens": 34818720} {"current_steps": 34550, "total_steps": 42420, "loss": 0.3356, "lr": 0.00010125991194171519, "epoch": 16.28948609146629, "percentage": 81.45, "elapsed_time": "1:52:20", "remaining_time": "0:25:35", "throughput": 5166.27, "total_tokens": 34823168} {"current_steps": 34555, "total_steps": 42420, "loss": 0.3794, "lr": 0.00010113582547624784, "epoch": 16.29184347006129, "percentage": 81.46, "elapsed_time": "1:52:21", "remaining_time": "0:25:34", "throughput": 5166.33, "total_tokens": 34828736} {"current_steps": 34560, "total_steps": 42420, "loss": 0.3574, "lr": 0.00010101180653171293, "epoch": 16.294200848656295, "percentage": 81.47, "elapsed_time": "1:52:22", "remaining_time": "0:25:33", "throughput": 5166.36, "total_tokens": 34833344} {"current_steps": 34565, "total_steps": 42420, "loss": 0.3412, "lr": 0.00010088785512910476, "epoch": 16.296558227251296, "percentage": 81.48, "elapsed_time": "1:52:23", "remaining_time": "0:25:32", "throughput": 5166.42, "total_tokens": 34838496} {"current_steps": 34570, "total_steps": 42420, "loss": 0.3564, "lr": 0.00010076397128940612, "epoch": 16.2989156058463, "percentage": 81.49, "elapsed_time": "1:52:24", "remaining_time": "0:25:31", "throughput": 5166.47, "total_tokens": 34843840} {"current_steps": 34575, "total_steps": 42420, "loss": 0.3415, "lr": 0.00010064015503358853, "epoch": 16.3012729844413, "percentage": 81.51, "elapsed_time": "1:52:25", "remaining_time": "0:25:30", "throughput": 5166.47, "total_tokens": 34847968} {"current_steps": 34580, "total_steps": 42420, "loss": 0.2789, "lr": 0.00010051640638261195, "epoch": 16.303630363036305, "percentage": 81.52, "elapsed_time": "1:52:26", "remaining_time": "0:25:29", "throughput": 5166.58, "total_tokens": 34854336} {"current_steps": 34585, "total_steps": 42420, "loss": 0.279, "lr": 0.00010039272535742505, "epoch": 16.305987741631306, "percentage": 81.53, "elapsed_time": "1:52:26", "remaining_time": "0:25:28", "throughput": 5166.6, "total_tokens": 34858784} {"current_steps": 34590, "total_steps": 42420, "loss": 0.3547, "lr": 0.00010026911197896466, "epoch": 16.30834512022631, "percentage": 81.54, "elapsed_time": "1:52:27", "remaining_time": "0:25:27", "throughput": 5166.65, "total_tokens": 34863936} {"current_steps": 34595, "total_steps": 42420, "loss": 0.2654, "lr": 0.00010014556626815658, "epoch": 16.31070249882131, "percentage": 81.55, "elapsed_time": "1:52:28", "remaining_time": "0:25:26", "throughput": 5166.66, "total_tokens": 34868448} {"current_steps": 34600, "total_steps": 42420, "loss": 0.3553, "lr": 0.00010002208824591502, "epoch": 16.313059877416315, "percentage": 81.57, "elapsed_time": "1:52:29", "remaining_time": "0:25:25", "throughput": 5166.68, "total_tokens": 34872960} {"current_steps": 34605, "total_steps": 42420, "loss": 0.3364, "lr": 9.989867793314262e-05, "epoch": 16.315417256011315, "percentage": 81.58, "elapsed_time": "1:52:30", "remaining_time": "0:25:24", "throughput": 5166.7, "total_tokens": 34877952} {"current_steps": 34610, "total_steps": 42420, "loss": 0.4039, "lr": 9.977533535073091e-05, "epoch": 16.31777463460632, "percentage": 81.59, "elapsed_time": "1:52:31", "remaining_time": "0:25:23", "throughput": 5166.68, "total_tokens": 34882144} {"current_steps": 34615, "total_steps": 42420, "loss": 0.329, "lr": 9.965206051955934e-05, "epoch": 16.32013201320132, "percentage": 81.6, "elapsed_time": "1:52:32", "remaining_time": "0:25:22", "throughput": 5166.74, "total_tokens": 34887616} {"current_steps": 34620, "total_steps": 42420, "loss": 0.3439, "lr": 9.952885346049634e-05, "epoch": 16.32248939179632, "percentage": 81.61, "elapsed_time": "1:52:33", "remaining_time": "0:25:21", "throughput": 5166.78, "total_tokens": 34892576} {"current_steps": 34625, "total_steps": 42420, "loss": 0.3017, "lr": 9.940571419439892e-05, "epoch": 16.324846770391325, "percentage": 81.62, "elapsed_time": "1:52:34", "remaining_time": "0:25:20", "throughput": 5166.83, "total_tokens": 34897792} {"current_steps": 34630, "total_steps": 42420, "loss": 0.3189, "lr": 9.928264274211234e-05, "epoch": 16.327204148986326, "percentage": 81.64, "elapsed_time": "1:52:35", "remaining_time": "0:25:19", "throughput": 5166.87, "total_tokens": 34902720} {"current_steps": 34635, "total_steps": 42420, "loss": 0.4086, "lr": 9.915963912447068e-05, "epoch": 16.32956152758133, "percentage": 81.65, "elapsed_time": "1:52:35", "remaining_time": "0:25:18", "throughput": 5166.83, "total_tokens": 34906624} {"current_steps": 34640, "total_steps": 42420, "loss": 0.3112, "lr": 9.903670336229615e-05, "epoch": 16.33191890617633, "percentage": 81.66, "elapsed_time": "1:52:36", "remaining_time": "0:25:17", "throughput": 5166.82, "total_tokens": 34911072} {"current_steps": 34645, "total_steps": 42420, "loss": 0.3602, "lr": 9.89138354763997e-05, "epoch": 16.334276284771335, "percentage": 81.67, "elapsed_time": "1:52:37", "remaining_time": "0:25:16", "throughput": 5166.85, "total_tokens": 34915584} {"current_steps": 34650, "total_steps": 42420, "loss": 0.3217, "lr": 9.879103548758089e-05, "epoch": 16.336633663366335, "percentage": 81.68, "elapsed_time": "1:52:38", "remaining_time": "0:25:15", "throughput": 5166.84, "total_tokens": 34919872} {"current_steps": 34655, "total_steps": 42420, "loss": 0.3339, "lr": 9.866830341662764e-05, "epoch": 16.33899104196134, "percentage": 81.69, "elapsed_time": "1:52:39", "remaining_time": "0:25:14", "throughput": 5166.87, "total_tokens": 34924864} {"current_steps": 34660, "total_steps": 42420, "loss": 0.3718, "lr": 9.854563928431637e-05, "epoch": 16.34134842055634, "percentage": 81.71, "elapsed_time": "1:52:40", "remaining_time": "0:25:13", "throughput": 5166.93, "total_tokens": 34930144} {"current_steps": 34665, "total_steps": 42420, "loss": 0.346, "lr": 9.84230431114121e-05, "epoch": 16.343705799151344, "percentage": 81.72, "elapsed_time": "1:52:41", "remaining_time": "0:25:12", "throughput": 5167.01, "total_tokens": 34935936} {"current_steps": 34670, "total_steps": 42420, "loss": 0.3459, "lr": 9.83005149186682e-05, "epoch": 16.346063177746345, "percentage": 81.73, "elapsed_time": "1:52:42", "remaining_time": "0:25:11", "throughput": 5167.07, "total_tokens": 34941568} {"current_steps": 34675, "total_steps": 42420, "loss": 0.335, "lr": 9.817805472682684e-05, "epoch": 16.34842055634135, "percentage": 81.74, "elapsed_time": "1:52:43", "remaining_time": "0:25:10", "throughput": 5167.1, "total_tokens": 34946048} {"current_steps": 34680, "total_steps": 42420, "loss": 0.3148, "lr": 9.805566255661807e-05, "epoch": 16.35077793493635, "percentage": 81.75, "elapsed_time": "1:52:44", "remaining_time": "0:25:09", "throughput": 5167.18, "total_tokens": 34951968} {"current_steps": 34685, "total_steps": 42420, "loss": 0.3111, "lr": 9.793333842876106e-05, "epoch": 16.353135313531354, "percentage": 81.77, "elapsed_time": "1:52:45", "remaining_time": "0:25:08", "throughput": 5167.18, "total_tokens": 34956448} {"current_steps": 34690, "total_steps": 42420, "loss": 0.302, "lr": 9.781108236396313e-05, "epoch": 16.355492692126354, "percentage": 81.78, "elapsed_time": "1:52:45", "remaining_time": "0:25:07", "throughput": 5167.2, "total_tokens": 34960960} {"current_steps": 34695, "total_steps": 42420, "loss": 0.2966, "lr": 9.768889438292017e-05, "epoch": 16.35785007072136, "percentage": 81.79, "elapsed_time": "1:52:46", "remaining_time": "0:25:06", "throughput": 5167.23, "total_tokens": 34965824} {"current_steps": 34700, "total_steps": 42420, "loss": 0.3666, "lr": 9.756677450631651e-05, "epoch": 16.36020744931636, "percentage": 81.8, "elapsed_time": "1:52:47", "remaining_time": "0:25:05", "throughput": 5167.26, "total_tokens": 34970816} {"current_steps": 34705, "total_steps": 42420, "loss": 0.3308, "lr": 9.744472275482502e-05, "epoch": 16.362564827911363, "percentage": 81.81, "elapsed_time": "1:52:48", "remaining_time": "0:25:04", "throughput": 5167.3, "total_tokens": 34975840} {"current_steps": 34710, "total_steps": 42420, "loss": 0.3244, "lr": 9.7322739149107e-05, "epoch": 16.364922206506364, "percentage": 81.82, "elapsed_time": "1:52:49", "remaining_time": "0:25:03", "throughput": 5167.36, "total_tokens": 34981312} {"current_steps": 34715, "total_steps": 42420, "loss": 0.3421, "lr": 9.720082370981226e-05, "epoch": 16.367279585101368, "percentage": 81.84, "elapsed_time": "1:52:50", "remaining_time": "0:25:02", "throughput": 5167.35, "total_tokens": 34985568} {"current_steps": 34720, "total_steps": 42420, "loss": 0.3285, "lr": 9.707897645757885e-05, "epoch": 16.36963696369637, "percentage": 81.85, "elapsed_time": "1:52:51", "remaining_time": "0:25:01", "throughput": 5167.44, "total_tokens": 34991392} {"current_steps": 34725, "total_steps": 42420, "loss": 0.319, "lr": 9.695719741303344e-05, "epoch": 16.371994342291373, "percentage": 81.86, "elapsed_time": "1:52:52", "remaining_time": "0:25:00", "throughput": 5167.56, "total_tokens": 34998208} {"current_steps": 34730, "total_steps": 42420, "loss": 0.3656, "lr": 9.683548659679125e-05, "epoch": 16.374351720886374, "percentage": 81.87, "elapsed_time": "1:52:53", "remaining_time": "0:24:59", "throughput": 5167.55, "total_tokens": 35002304} {"current_steps": 34735, "total_steps": 42420, "loss": 0.3262, "lr": 9.671384402945587e-05, "epoch": 16.376709099481378, "percentage": 81.88, "elapsed_time": "1:52:54", "remaining_time": "0:24:58", "throughput": 5167.58, "total_tokens": 35007232} {"current_steps": 34740, "total_steps": 42420, "loss": 0.3184, "lr": 9.659226973161927e-05, "epoch": 16.37906647807638, "percentage": 81.9, "elapsed_time": "1:52:55", "remaining_time": "0:24:57", "throughput": 5167.61, "total_tokens": 35011968} {"current_steps": 34745, "total_steps": 42420, "loss": 0.3366, "lr": 9.647076372386193e-05, "epoch": 16.381423856671383, "percentage": 81.91, "elapsed_time": "1:52:56", "remaining_time": "0:24:56", "throughput": 5167.68, "total_tokens": 35017632} {"current_steps": 34750, "total_steps": 42420, "loss": 0.3572, "lr": 9.634932602675278e-05, "epoch": 16.383781235266383, "percentage": 81.92, "elapsed_time": "1:52:57", "remaining_time": "0:24:55", "throughput": 5167.7, "total_tokens": 35022176} {"current_steps": 34755, "total_steps": 42420, "loss": 0.3248, "lr": 9.622795666084921e-05, "epoch": 16.386138613861387, "percentage": 81.93, "elapsed_time": "1:52:58", "remaining_time": "0:24:54", "throughput": 5167.75, "total_tokens": 35027712} {"current_steps": 34760, "total_steps": 42420, "loss": 0.3391, "lr": 9.610665564669685e-05, "epoch": 16.388495992456388, "percentage": 81.94, "elapsed_time": "1:52:59", "remaining_time": "0:24:53", "throughput": 5167.8, "total_tokens": 35032832} {"current_steps": 34765, "total_steps": 42420, "loss": 0.3094, "lr": 9.598542300482993e-05, "epoch": 16.390853371051392, "percentage": 81.95, "elapsed_time": "1:53:00", "remaining_time": "0:24:52", "throughput": 5167.88, "total_tokens": 35038592} {"current_steps": 34770, "total_steps": 42420, "loss": 0.3676, "lr": 9.586425875577115e-05, "epoch": 16.393210749646393, "percentage": 81.97, "elapsed_time": "1:53:00", "remaining_time": "0:24:51", "throughput": 5167.93, "total_tokens": 35043584} {"current_steps": 34775, "total_steps": 42420, "loss": 0.2995, "lr": 9.574316292003149e-05, "epoch": 16.395568128241397, "percentage": 81.98, "elapsed_time": "1:53:01", "remaining_time": "0:24:50", "throughput": 5167.98, "total_tokens": 35048736} {"current_steps": 34780, "total_steps": 42420, "loss": 0.3514, "lr": 9.56221355181105e-05, "epoch": 16.397925506836398, "percentage": 81.99, "elapsed_time": "1:53:02", "remaining_time": "0:24:49", "throughput": 5167.98, "total_tokens": 35053056} {"current_steps": 34785, "total_steps": 42420, "loss": 0.3468, "lr": 9.550117657049601e-05, "epoch": 16.400282885431402, "percentage": 82.0, "elapsed_time": "1:53:03", "remaining_time": "0:24:48", "throughput": 5167.98, "total_tokens": 35057408} {"current_steps": 34790, "total_steps": 42420, "loss": 0.3134, "lr": 9.53802860976643e-05, "epoch": 16.402640264026402, "percentage": 82.01, "elapsed_time": "1:53:04", "remaining_time": "0:24:47", "throughput": 5168.02, "total_tokens": 35062560} {"current_steps": 34795, "total_steps": 42420, "loss": 0.2515, "lr": 9.525946412008024e-05, "epoch": 16.404997642621407, "percentage": 82.02, "elapsed_time": "1:53:05", "remaining_time": "0:24:46", "throughput": 5168.06, "total_tokens": 35067488} {"current_steps": 34800, "total_steps": 42420, "loss": 0.3348, "lr": 9.513871065819669e-05, "epoch": 16.407355021216407, "percentage": 82.04, "elapsed_time": "1:53:06", "remaining_time": "0:24:45", "throughput": 5168.08, "total_tokens": 35072288} {"current_steps": 34805, "total_steps": 42420, "loss": 0.2943, "lr": 9.501802573245527e-05, "epoch": 16.40971239981141, "percentage": 82.05, "elapsed_time": "1:53:07", "remaining_time": "0:24:44", "throughput": 5168.1, "total_tokens": 35077024} {"current_steps": 34810, "total_steps": 42420, "loss": 0.3678, "lr": 9.489740936328583e-05, "epoch": 16.412069778406412, "percentage": 82.06, "elapsed_time": "1:53:08", "remaining_time": "0:24:43", "throughput": 5168.13, "total_tokens": 35081536} {"current_steps": 34815, "total_steps": 42420, "loss": 0.2745, "lr": 9.477686157110682e-05, "epoch": 16.414427157001413, "percentage": 82.07, "elapsed_time": "1:53:09", "remaining_time": "0:24:43", "throughput": 5168.21, "total_tokens": 35087360} {"current_steps": 34820, "total_steps": 42420, "loss": 0.3924, "lr": 9.465638237632479e-05, "epoch": 16.416784535596417, "percentage": 82.08, "elapsed_time": "1:53:10", "remaining_time": "0:24:42", "throughput": 5168.27, "total_tokens": 35093120} {"current_steps": 34825, "total_steps": 42420, "loss": 0.3627, "lr": 9.453597179933493e-05, "epoch": 16.419141914191417, "percentage": 82.1, "elapsed_time": "1:53:11", "remaining_time": "0:24:41", "throughput": 5168.38, "total_tokens": 35099328} {"current_steps": 34830, "total_steps": 42420, "loss": 0.2719, "lr": 9.44156298605206e-05, "epoch": 16.42149929278642, "percentage": 82.11, "elapsed_time": "1:53:12", "remaining_time": "0:24:40", "throughput": 5168.39, "total_tokens": 35103904} {"current_steps": 34835, "total_steps": 42420, "loss": 0.3423, "lr": 9.42953565802539e-05, "epoch": 16.423856671381422, "percentage": 82.12, "elapsed_time": "1:53:12", "remaining_time": "0:24:39", "throughput": 5168.42, "total_tokens": 35108640} {"current_steps": 34840, "total_steps": 42420, "loss": 0.3006, "lr": 9.41751519788947e-05, "epoch": 16.426214049976426, "percentage": 82.13, "elapsed_time": "1:53:14", "remaining_time": "0:24:38", "throughput": 5168.58, "total_tokens": 35116320} {"current_steps": 34845, "total_steps": 42420, "loss": 0.2984, "lr": 9.405501607679184e-05, "epoch": 16.428571428571427, "percentage": 82.14, "elapsed_time": "1:53:15", "remaining_time": "0:24:37", "throughput": 5168.59, "total_tokens": 35120864} {"current_steps": 34850, "total_steps": 42420, "loss": 0.3583, "lr": 9.393494889428212e-05, "epoch": 16.43092880716643, "percentage": 82.15, "elapsed_time": "1:53:15", "remaining_time": "0:24:36", "throughput": 5168.63, "total_tokens": 35125824} {"current_steps": 34855, "total_steps": 42420, "loss": 0.3149, "lr": 9.381495045169098e-05, "epoch": 16.433286185761432, "percentage": 82.17, "elapsed_time": "1:53:17", "remaining_time": "0:24:35", "throughput": 5168.78, "total_tokens": 35132800} {"current_steps": 34860, "total_steps": 42420, "loss": 0.3452, "lr": 9.369502076933206e-05, "epoch": 16.435643564356436, "percentage": 82.18, "elapsed_time": "1:53:17", "remaining_time": "0:24:34", "throughput": 5168.8, "total_tokens": 35137312} {"current_steps": 34865, "total_steps": 42420, "loss": 0.3057, "lr": 9.357515986750748e-05, "epoch": 16.438000942951437, "percentage": 82.19, "elapsed_time": "1:53:18", "remaining_time": "0:24:33", "throughput": 5168.84, "total_tokens": 35142112} {"current_steps": 34870, "total_steps": 42420, "loss": 0.4007, "lr": 9.345536776650764e-05, "epoch": 16.44035832154644, "percentage": 82.2, "elapsed_time": "1:53:19", "remaining_time": "0:24:32", "throughput": 5168.88, "total_tokens": 35147072} {"current_steps": 34875, "total_steps": 42420, "loss": 0.3538, "lr": 9.333564448661125e-05, "epoch": 16.44271570014144, "percentage": 82.21, "elapsed_time": "1:53:20", "remaining_time": "0:24:31", "throughput": 5168.9, "total_tokens": 35151712} {"current_steps": 34880, "total_steps": 42420, "loss": 0.2928, "lr": 9.321599004808557e-05, "epoch": 16.445073078736446, "percentage": 82.23, "elapsed_time": "1:53:21", "remaining_time": "0:24:30", "throughput": 5168.98, "total_tokens": 35157344} {"current_steps": 34885, "total_steps": 42420, "loss": 0.3473, "lr": 9.3096404471186e-05, "epoch": 16.447430457331446, "percentage": 82.24, "elapsed_time": "1:53:22", "remaining_time": "0:24:29", "throughput": 5168.99, "total_tokens": 35161856} {"current_steps": 34890, "total_steps": 42420, "loss": 0.3629, "lr": 9.297688777615626e-05, "epoch": 16.44978783592645, "percentage": 82.25, "elapsed_time": "1:53:23", "remaining_time": "0:24:28", "throughput": 5168.99, "total_tokens": 35166336} {"current_steps": 34895, "total_steps": 42420, "loss": 0.3337, "lr": 9.285743998322854e-05, "epoch": 16.45214521452145, "percentage": 82.26, "elapsed_time": "1:53:24", "remaining_time": "0:24:27", "throughput": 5169.03, "total_tokens": 35171456} {"current_steps": 34900, "total_steps": 42420, "loss": 0.3153, "lr": 9.273806111262339e-05, "epoch": 16.454502593116455, "percentage": 82.27, "elapsed_time": "1:53:25", "remaining_time": "0:24:26", "throughput": 5169.13, "total_tokens": 35177568} {"current_steps": 34905, "total_steps": 42420, "loss": 0.3532, "lr": 9.261875118454954e-05, "epoch": 16.456859971711456, "percentage": 82.28, "elapsed_time": "1:53:26", "remaining_time": "0:24:25", "throughput": 5169.21, "total_tokens": 35183264} {"current_steps": 34910, "total_steps": 42420, "loss": 0.3049, "lr": 9.249951021920416e-05, "epoch": 16.45921735030646, "percentage": 82.3, "elapsed_time": "1:53:27", "remaining_time": "0:24:24", "throughput": 5169.21, "total_tokens": 35187616} {"current_steps": 34915, "total_steps": 42420, "loss": 0.3118, "lr": 9.238033823677277e-05, "epoch": 16.46157472890146, "percentage": 82.31, "elapsed_time": "1:53:28", "remaining_time": "0:24:23", "throughput": 5169.25, "total_tokens": 35192608} {"current_steps": 34920, "total_steps": 42420, "loss": 0.3307, "lr": 9.226123525742907e-05, "epoch": 16.463932107496465, "percentage": 82.32, "elapsed_time": "1:53:29", "remaining_time": "0:24:22", "throughput": 5169.36, "total_tokens": 35199232} {"current_steps": 34925, "total_steps": 42420, "loss": 0.3113, "lr": 9.214220130133544e-05, "epoch": 16.466289486091465, "percentage": 82.33, "elapsed_time": "1:53:30", "remaining_time": "0:24:21", "throughput": 5169.37, "total_tokens": 35203712} {"current_steps": 34930, "total_steps": 42420, "loss": 0.2457, "lr": 9.20232363886419e-05, "epoch": 16.46864686468647, "percentage": 82.34, "elapsed_time": "1:53:31", "remaining_time": "0:24:20", "throughput": 5169.44, "total_tokens": 35209536} {"current_steps": 34935, "total_steps": 42420, "loss": 0.3386, "lr": 9.190434053948743e-05, "epoch": 16.47100424328147, "percentage": 82.36, "elapsed_time": "1:53:32", "remaining_time": "0:24:19", "throughput": 5169.46, "total_tokens": 35214560} {"current_steps": 34940, "total_steps": 42420, "loss": 0.377, "lr": 9.178551377399901e-05, "epoch": 16.473361621876474, "percentage": 82.37, "elapsed_time": "1:53:33", "remaining_time": "0:24:18", "throughput": 5169.52, "total_tokens": 35219968} {"current_steps": 34945, "total_steps": 42420, "loss": 0.3599, "lr": 9.166675611229203e-05, "epoch": 16.475719000471475, "percentage": 82.38, "elapsed_time": "1:53:33", "remaining_time": "0:24:17", "throughput": 5169.52, "total_tokens": 35224288} {"current_steps": 34950, "total_steps": 42420, "loss": 0.2978, "lr": 9.154806757447015e-05, "epoch": 16.47807637906648, "percentage": 82.39, "elapsed_time": "1:53:34", "remaining_time": "0:24:16", "throughput": 5169.56, "total_tokens": 35229248} {"current_steps": 34955, "total_steps": 42420, "loss": 0.3054, "lr": 9.142944818062532e-05, "epoch": 16.48043375766148, "percentage": 82.4, "elapsed_time": "1:53:35", "remaining_time": "0:24:15", "throughput": 5169.59, "total_tokens": 35234112} {"current_steps": 34960, "total_steps": 42420, "loss": 0.3506, "lr": 9.131089795083774e-05, "epoch": 16.482791136256484, "percentage": 82.41, "elapsed_time": "1:53:36", "remaining_time": "0:24:14", "throughput": 5169.65, "total_tokens": 35239552} {"current_steps": 34965, "total_steps": 42420, "loss": 0.3662, "lr": 9.119241690517616e-05, "epoch": 16.485148514851485, "percentage": 82.43, "elapsed_time": "1:53:37", "remaining_time": "0:24:13", "throughput": 5169.71, "total_tokens": 35245152} {"current_steps": 34970, "total_steps": 42420, "loss": 0.345, "lr": 9.107400506369706e-05, "epoch": 16.48750589344649, "percentage": 82.44, "elapsed_time": "1:53:38", "remaining_time": "0:24:12", "throughput": 5169.74, "total_tokens": 35250048} {"current_steps": 34975, "total_steps": 42420, "loss": 0.3449, "lr": 9.095566244644581e-05, "epoch": 16.48986327204149, "percentage": 82.45, "elapsed_time": "1:53:39", "remaining_time": "0:24:11", "throughput": 5169.74, "total_tokens": 35254592} {"current_steps": 34980, "total_steps": 42420, "loss": 0.2965, "lr": 9.083738907345567e-05, "epoch": 16.492220650636494, "percentage": 82.46, "elapsed_time": "1:53:40", "remaining_time": "0:24:10", "throughput": 5169.75, "total_tokens": 35259296} {"current_steps": 34985, "total_steps": 42420, "loss": 0.3217, "lr": 9.07191849647484e-05, "epoch": 16.494578029231494, "percentage": 82.47, "elapsed_time": "1:53:41", "remaining_time": "0:24:09", "throughput": 5169.76, "total_tokens": 35263968} {"current_steps": 34990, "total_steps": 42420, "loss": 0.2706, "lr": 9.060105014033392e-05, "epoch": 16.4969354078265, "percentage": 82.48, "elapsed_time": "1:53:42", "remaining_time": "0:24:08", "throughput": 5169.78, "total_tokens": 35268608} {"current_steps": 34995, "total_steps": 42420, "loss": 0.2974, "lr": 9.048298462021043e-05, "epoch": 16.4992927864215, "percentage": 82.5, "elapsed_time": "1:53:42", "remaining_time": "0:24:07", "throughput": 5169.77, "total_tokens": 35273152} {"current_steps": 35000, "total_steps": 42420, "loss": 0.2891, "lr": 9.036498842436447e-05, "epoch": 16.501650165016503, "percentage": 82.51, "elapsed_time": "1:53:43", "remaining_time": "0:24:06", "throughput": 5169.78, "total_tokens": 35277536} {"current_steps": 35005, "total_steps": 42420, "loss": 0.3286, "lr": 9.024706157277085e-05, "epoch": 16.504007543611504, "percentage": 82.52, "elapsed_time": "1:53:44", "remaining_time": "0:24:05", "throughput": 5169.78, "total_tokens": 35281984} {"current_steps": 35010, "total_steps": 42420, "loss": 0.2561, "lr": 9.012920408539243e-05, "epoch": 16.506364922206508, "percentage": 82.53, "elapsed_time": "1:53:45", "remaining_time": "0:24:04", "throughput": 5169.88, "total_tokens": 35288192} {"current_steps": 35015, "total_steps": 42420, "loss": 0.2966, "lr": 9.00114159821805e-05, "epoch": 16.50872230080151, "percentage": 82.54, "elapsed_time": "1:53:46", "remaining_time": "0:24:03", "throughput": 5169.9, "total_tokens": 35292768} {"current_steps": 35020, "total_steps": 42420, "loss": 0.3906, "lr": 8.989369728307467e-05, "epoch": 16.51107967939651, "percentage": 82.56, "elapsed_time": "1:53:47", "remaining_time": "0:24:02", "throughput": 5169.89, "total_tokens": 35297120} {"current_steps": 35025, "total_steps": 42420, "loss": 0.2989, "lr": 8.977604800800266e-05, "epoch": 16.513437057991514, "percentage": 82.57, "elapsed_time": "1:53:48", "remaining_time": "0:24:01", "throughput": 5169.95, "total_tokens": 35302432} {"current_steps": 35030, "total_steps": 42420, "loss": 0.3162, "lr": 8.96584681768805e-05, "epoch": 16.515794436586514, "percentage": 82.58, "elapsed_time": "1:53:49", "remaining_time": "0:24:00", "throughput": 5170.0, "total_tokens": 35307648} {"current_steps": 35035, "total_steps": 42420, "loss": 0.4043, "lr": 8.954095780961252e-05, "epoch": 16.51815181518152, "percentage": 82.59, "elapsed_time": "1:53:50", "remaining_time": "0:23:59", "throughput": 5170.04, "total_tokens": 35312768} {"current_steps": 35040, "total_steps": 42420, "loss": 0.254, "lr": 8.942351692609113e-05, "epoch": 16.52050919377652, "percentage": 82.6, "elapsed_time": "1:53:51", "remaining_time": "0:23:58", "throughput": 5170.07, "total_tokens": 35317920} {"current_steps": 35045, "total_steps": 42420, "loss": 0.3742, "lr": 8.930614554619732e-05, "epoch": 16.522866572371523, "percentage": 82.61, "elapsed_time": "1:53:52", "remaining_time": "0:23:57", "throughput": 5170.17, "total_tokens": 35324128} {"current_steps": 35050, "total_steps": 42420, "loss": 0.2985, "lr": 8.918884368979968e-05, "epoch": 16.525223950966524, "percentage": 82.63, "elapsed_time": "1:53:53", "remaining_time": "0:23:56", "throughput": 5170.19, "total_tokens": 35329088} {"current_steps": 35055, "total_steps": 42420, "loss": 0.3302, "lr": 8.907161137675567e-05, "epoch": 16.527581329561528, "percentage": 82.64, "elapsed_time": "1:53:54", "remaining_time": "0:23:55", "throughput": 5170.27, "total_tokens": 35334848} {"current_steps": 35060, "total_steps": 42420, "loss": 0.3961, "lr": 8.895444862691072e-05, "epoch": 16.52993870815653, "percentage": 82.65, "elapsed_time": "1:53:55", "remaining_time": "0:23:54", "throughput": 5170.29, "total_tokens": 35339776} {"current_steps": 35065, "total_steps": 42420, "loss": 0.3548, "lr": 8.883735546009857e-05, "epoch": 16.532296086751533, "percentage": 82.66, "elapsed_time": "1:53:56", "remaining_time": "0:23:53", "throughput": 5170.37, "total_tokens": 35345728} {"current_steps": 35070, "total_steps": 42420, "loss": 0.314, "lr": 8.872033189614081e-05, "epoch": 16.534653465346533, "percentage": 82.67, "elapsed_time": "1:53:57", "remaining_time": "0:23:52", "throughput": 5170.42, "total_tokens": 35350912} {"current_steps": 35075, "total_steps": 42420, "loss": 0.3056, "lr": 8.86033779548478e-05, "epoch": 16.537010843941538, "percentage": 82.69, "elapsed_time": "1:53:58", "remaining_time": "0:23:51", "throughput": 5170.43, "total_tokens": 35355552} {"current_steps": 35080, "total_steps": 42420, "loss": 0.3712, "lr": 8.848649365601774e-05, "epoch": 16.539368222536538, "percentage": 82.7, "elapsed_time": "1:53:58", "remaining_time": "0:23:50", "throughput": 5170.48, "total_tokens": 35360736} {"current_steps": 35085, "total_steps": 42420, "loss": 0.3207, "lr": 8.836967901943727e-05, "epoch": 16.541725601131542, "percentage": 82.71, "elapsed_time": "1:53:59", "remaining_time": "0:23:49", "throughput": 5170.49, "total_tokens": 35365376} {"current_steps": 35090, "total_steps": 42420, "loss": 0.3707, "lr": 8.825293406488099e-05, "epoch": 16.544082979726543, "percentage": 82.72, "elapsed_time": "1:54:00", "remaining_time": "0:23:48", "throughput": 5170.53, "total_tokens": 35370464} {"current_steps": 35095, "total_steps": 42420, "loss": 0.3402, "lr": 8.813625881211207e-05, "epoch": 16.546440358321547, "percentage": 82.73, "elapsed_time": "1:54:01", "remaining_time": "0:23:48", "throughput": 5170.6, "total_tokens": 35376096} {"current_steps": 35100, "total_steps": 42420, "loss": 0.343, "lr": 8.801965328088135e-05, "epoch": 16.548797736916548, "percentage": 82.74, "elapsed_time": "1:54:02", "remaining_time": "0:23:47", "throughput": 5170.58, "total_tokens": 35380160} {"current_steps": 35105, "total_steps": 42420, "loss": 0.3409, "lr": 8.790311749092827e-05, "epoch": 16.551155115511552, "percentage": 82.76, "elapsed_time": "1:54:03", "remaining_time": "0:23:46", "throughput": 5170.62, "total_tokens": 35385376} {"current_steps": 35110, "total_steps": 42420, "loss": 0.3209, "lr": 8.778665146198039e-05, "epoch": 16.553512494106553, "percentage": 82.77, "elapsed_time": "1:54:04", "remaining_time": "0:23:45", "throughput": 5170.59, "total_tokens": 35389088} {"current_steps": 35115, "total_steps": 42420, "loss": 0.3164, "lr": 8.76702552137535e-05, "epoch": 16.555869872701557, "percentage": 82.78, "elapsed_time": "1:54:05", "remaining_time": "0:23:44", "throughput": 5170.65, "total_tokens": 35394432} {"current_steps": 35120, "total_steps": 42420, "loss": 0.3434, "lr": 8.755392876595136e-05, "epoch": 16.558227251296557, "percentage": 82.79, "elapsed_time": "1:54:06", "remaining_time": "0:23:43", "throughput": 5170.67, "total_tokens": 35399168} {"current_steps": 35125, "total_steps": 42420, "loss": 0.3162, "lr": 8.74376721382662e-05, "epoch": 16.56058462989156, "percentage": 82.8, "elapsed_time": "1:54:07", "remaining_time": "0:23:42", "throughput": 5170.68, "total_tokens": 35403744} {"current_steps": 35130, "total_steps": 42420, "loss": 0.3217, "lr": 8.73214853503782e-05, "epoch": 16.562942008486562, "percentage": 82.81, "elapsed_time": "1:54:08", "remaining_time": "0:23:41", "throughput": 5170.74, "total_tokens": 35409408} {"current_steps": 35135, "total_steps": 42420, "loss": 0.3434, "lr": 8.720536842195603e-05, "epoch": 16.565299387081566, "percentage": 82.83, "elapsed_time": "1:54:08", "remaining_time": "0:23:40", "throughput": 5170.75, "total_tokens": 35413696} {"current_steps": 35140, "total_steps": 42420, "loss": 0.3268, "lr": 8.708932137265596e-05, "epoch": 16.567656765676567, "percentage": 82.84, "elapsed_time": "1:54:09", "remaining_time": "0:23:39", "throughput": 5170.82, "total_tokens": 35419328} {"current_steps": 35145, "total_steps": 42420, "loss": 0.3228, "lr": 8.697334422212294e-05, "epoch": 16.57001414427157, "percentage": 82.85, "elapsed_time": "1:54:10", "remaining_time": "0:23:38", "throughput": 5170.83, "total_tokens": 35423840} {"current_steps": 35150, "total_steps": 42420, "loss": 0.3303, "lr": 8.685743698999e-05, "epoch": 16.572371522866572, "percentage": 82.86, "elapsed_time": "1:54:11", "remaining_time": "0:23:37", "throughput": 5170.81, "total_tokens": 35427936} {"current_steps": 35155, "total_steps": 42420, "loss": 0.3356, "lr": 8.674159969587819e-05, "epoch": 16.574728901461576, "percentage": 82.87, "elapsed_time": "1:54:12", "remaining_time": "0:23:36", "throughput": 5170.78, "total_tokens": 35431680} {"current_steps": 35160, "total_steps": 42420, "loss": 0.3627, "lr": 8.662583235939687e-05, "epoch": 16.577086280056577, "percentage": 82.89, "elapsed_time": "1:54:13", "remaining_time": "0:23:35", "throughput": 5170.79, "total_tokens": 35436128} {"current_steps": 35165, "total_steps": 42420, "loss": 0.3709, "lr": 8.651013500014337e-05, "epoch": 16.57944365865158, "percentage": 82.9, "elapsed_time": "1:54:14", "remaining_time": "0:23:34", "throughput": 5170.83, "total_tokens": 35440928} {"current_steps": 35170, "total_steps": 42420, "loss": 0.3298, "lr": 8.63945076377034e-05, "epoch": 16.58180103724658, "percentage": 82.91, "elapsed_time": "1:54:14", "remaining_time": "0:23:33", "throughput": 5170.85, "total_tokens": 35445536} {"current_steps": 35175, "total_steps": 42420, "loss": 0.3555, "lr": 8.627895029165072e-05, "epoch": 16.584158415841586, "percentage": 82.92, "elapsed_time": "1:54:15", "remaining_time": "0:23:32", "throughput": 5170.82, "total_tokens": 35449408} {"current_steps": 35180, "total_steps": 42420, "loss": 0.2927, "lr": 8.616346298154715e-05, "epoch": 16.586515794436586, "percentage": 82.93, "elapsed_time": "1:54:16", "remaining_time": "0:23:31", "throughput": 5170.83, "total_tokens": 35453792} {"current_steps": 35185, "total_steps": 42420, "loss": 0.3417, "lr": 8.604804572694269e-05, "epoch": 16.58887317303159, "percentage": 82.94, "elapsed_time": "1:54:17", "remaining_time": "0:23:30", "throughput": 5170.8, "total_tokens": 35457728} {"current_steps": 35190, "total_steps": 42420, "loss": 0.3217, "lr": 8.593269854737556e-05, "epoch": 16.59123055162659, "percentage": 82.96, "elapsed_time": "1:54:18", "remaining_time": "0:23:29", "throughput": 5170.78, "total_tokens": 35461824} {"current_steps": 35195, "total_steps": 42420, "loss": 0.3707, "lr": 8.581742146237215e-05, "epoch": 16.593587930221595, "percentage": 82.97, "elapsed_time": "1:54:19", "remaining_time": "0:23:28", "throughput": 5170.82, "total_tokens": 35466848} {"current_steps": 35200, "total_steps": 42420, "loss": 0.3541, "lr": 8.570221449144689e-05, "epoch": 16.595945308816596, "percentage": 82.98, "elapsed_time": "1:54:19", "remaining_time": "0:23:27", "throughput": 5170.86, "total_tokens": 35471872} {"current_steps": 35205, "total_steps": 42420, "loss": 0.3631, "lr": 8.558707765410228e-05, "epoch": 16.5983026874116, "percentage": 82.99, "elapsed_time": "1:54:20", "remaining_time": "0:23:26", "throughput": 5170.92, "total_tokens": 35477312} {"current_steps": 35210, "total_steps": 42420, "loss": 0.348, "lr": 8.547201096982915e-05, "epoch": 16.6006600660066, "percentage": 83.0, "elapsed_time": "1:54:21", "remaining_time": "0:23:25", "throughput": 5170.98, "total_tokens": 35482560} {"current_steps": 35215, "total_steps": 42420, "loss": 0.3351, "lr": 8.535701445810635e-05, "epoch": 16.603017444601605, "percentage": 83.02, "elapsed_time": "1:54:22", "remaining_time": "0:23:24", "throughput": 5171.0, "total_tokens": 35487232} {"current_steps": 35220, "total_steps": 42420, "loss": 0.2865, "lr": 8.524208813840068e-05, "epoch": 16.605374823196605, "percentage": 83.03, "elapsed_time": "1:54:23", "remaining_time": "0:23:23", "throughput": 5171.03, "total_tokens": 35492032} {"current_steps": 35225, "total_steps": 42420, "loss": 0.3175, "lr": 8.512723203016726e-05, "epoch": 16.607732201791606, "percentage": 83.04, "elapsed_time": "1:54:24", "remaining_time": "0:23:22", "throughput": 5171.1, "total_tokens": 35497632} {"current_steps": 35230, "total_steps": 42420, "loss": 0.3366, "lr": 8.501244615284937e-05, "epoch": 16.61008958038661, "percentage": 83.05, "elapsed_time": "1:54:25", "remaining_time": "0:23:21", "throughput": 5171.11, "total_tokens": 35502144} {"current_steps": 35235, "total_steps": 42420, "loss": 0.3186, "lr": 8.489773052587818e-05, "epoch": 16.61244695898161, "percentage": 83.06, "elapsed_time": "1:54:26", "remaining_time": "0:23:20", "throughput": 5171.16, "total_tokens": 35507136} {"current_steps": 35240, "total_steps": 42420, "loss": 0.3484, "lr": 8.478308516867323e-05, "epoch": 16.614804337576615, "percentage": 83.07, "elapsed_time": "1:54:27", "remaining_time": "0:23:19", "throughput": 5171.17, "total_tokens": 35511584} {"current_steps": 35245, "total_steps": 42420, "loss": 0.3449, "lr": 8.466851010064197e-05, "epoch": 16.617161716171616, "percentage": 83.09, "elapsed_time": "1:54:28", "remaining_time": "0:23:18", "throughput": 5171.22, "total_tokens": 35516800} {"current_steps": 35250, "total_steps": 42420, "loss": 0.2915, "lr": 8.455400534118007e-05, "epoch": 16.61951909476662, "percentage": 83.1, "elapsed_time": "1:54:29", "remaining_time": "0:23:17", "throughput": 5171.21, "total_tokens": 35521216} {"current_steps": 35255, "total_steps": 42420, "loss": 0.3175, "lr": 8.443957090967125e-05, "epoch": 16.62187647336162, "percentage": 83.11, "elapsed_time": "1:54:29", "remaining_time": "0:23:16", "throughput": 5171.23, "total_tokens": 35525824} {"current_steps": 35260, "total_steps": 42420, "loss": 0.3068, "lr": 8.432520682548716e-05, "epoch": 16.624233851956625, "percentage": 83.12, "elapsed_time": "1:54:30", "remaining_time": "0:23:15", "throughput": 5171.2, "total_tokens": 35529728} {"current_steps": 35265, "total_steps": 42420, "loss": 0.3411, "lr": 8.421091310798779e-05, "epoch": 16.626591230551625, "percentage": 83.13, "elapsed_time": "1:54:31", "remaining_time": "0:23:14", "throughput": 5171.23, "total_tokens": 35534496} {"current_steps": 35270, "total_steps": 42420, "loss": 0.3695, "lr": 8.409668977652118e-05, "epoch": 16.62894860914663, "percentage": 83.14, "elapsed_time": "1:54:32", "remaining_time": "0:23:13", "throughput": 5171.3, "total_tokens": 35540160} {"current_steps": 35275, "total_steps": 42420, "loss": 0.3724, "lr": 8.398253685042328e-05, "epoch": 16.63130598774163, "percentage": 83.16, "elapsed_time": "1:54:33", "remaining_time": "0:23:12", "throughput": 5171.4, "total_tokens": 35546368} {"current_steps": 35280, "total_steps": 42420, "loss": 0.3462, "lr": 8.386845434901841e-05, "epoch": 16.633663366336634, "percentage": 83.17, "elapsed_time": "1:54:34", "remaining_time": "0:23:11", "throughput": 5171.41, "total_tokens": 35550752} {"current_steps": 35285, "total_steps": 42420, "loss": 0.3265, "lr": 8.375444229161861e-05, "epoch": 16.636020744931635, "percentage": 83.18, "elapsed_time": "1:54:35", "remaining_time": "0:23:10", "throughput": 5171.47, "total_tokens": 35556256} {"current_steps": 35290, "total_steps": 42420, "loss": 0.3486, "lr": 8.364050069752449e-05, "epoch": 16.63837812352664, "percentage": 83.19, "elapsed_time": "1:54:36", "remaining_time": "0:23:09", "throughput": 5171.53, "total_tokens": 35561856} {"current_steps": 35295, "total_steps": 42420, "loss": 0.3092, "lr": 8.352662958602403e-05, "epoch": 16.64073550212164, "percentage": 83.2, "elapsed_time": "1:54:37", "remaining_time": "0:23:08", "throughput": 5171.54, "total_tokens": 35566368} {"current_steps": 35300, "total_steps": 42420, "loss": 0.3239, "lr": 8.341282897639402e-05, "epoch": 16.643092880716644, "percentage": 83.22, "elapsed_time": "1:54:38", "remaining_time": "0:23:07", "throughput": 5171.52, "total_tokens": 35570368} {"current_steps": 35305, "total_steps": 42420, "loss": 0.3589, "lr": 8.329909888789866e-05, "epoch": 16.645450259311644, "percentage": 83.23, "elapsed_time": "1:54:39", "remaining_time": "0:23:06", "throughput": 5171.53, "total_tokens": 35575104} {"current_steps": 35310, "total_steps": 42420, "loss": 0.3316, "lr": 8.318543933979073e-05, "epoch": 16.64780763790665, "percentage": 83.24, "elapsed_time": "1:54:40", "remaining_time": "0:23:05", "throughput": 5171.63, "total_tokens": 35581248} {"current_steps": 35315, "total_steps": 42420, "loss": 0.3583, "lr": 8.307185035131081e-05, "epoch": 16.65016501650165, "percentage": 83.25, "elapsed_time": "1:54:41", "remaining_time": "0:23:04", "throughput": 5171.7, "total_tokens": 35586912} {"current_steps": 35320, "total_steps": 42420, "loss": 0.3689, "lr": 8.295833194168756e-05, "epoch": 16.652522395096653, "percentage": 83.26, "elapsed_time": "1:54:41", "remaining_time": "0:23:03", "throughput": 5171.68, "total_tokens": 35591104} {"current_steps": 35325, "total_steps": 42420, "loss": 0.386, "lr": 8.284488413013775e-05, "epoch": 16.654879773691654, "percentage": 83.27, "elapsed_time": "1:54:42", "remaining_time": "0:23:02", "throughput": 5171.75, "total_tokens": 35596448} {"current_steps": 35330, "total_steps": 42420, "loss": 0.368, "lr": 8.273150693586617e-05, "epoch": 16.65723715228666, "percentage": 83.29, "elapsed_time": "1:54:43", "remaining_time": "0:23:01", "throughput": 5171.72, "total_tokens": 35600288} {"current_steps": 35335, "total_steps": 42420, "loss": 0.334, "lr": 8.261820037806566e-05, "epoch": 16.65959453088166, "percentage": 83.3, "elapsed_time": "1:54:44", "remaining_time": "0:23:00", "throughput": 5171.76, "total_tokens": 35605536} {"current_steps": 35340, "total_steps": 42420, "loss": 0.3413, "lr": 8.250496447591722e-05, "epoch": 16.661951909476663, "percentage": 83.31, "elapsed_time": "1:54:45", "remaining_time": "0:22:59", "throughput": 5171.75, "total_tokens": 35609760} {"current_steps": 35345, "total_steps": 42420, "loss": 0.3343, "lr": 8.239179924858948e-05, "epoch": 16.664309288071664, "percentage": 83.32, "elapsed_time": "1:54:46", "remaining_time": "0:22:58", "throughput": 5171.81, "total_tokens": 35615296} {"current_steps": 35350, "total_steps": 42420, "loss": 0.3544, "lr": 8.227870471523957e-05, "epoch": 16.666666666666668, "percentage": 83.33, "elapsed_time": "1:54:47", "remaining_time": "0:22:57", "throughput": 5171.88, "total_tokens": 35620832} {"current_steps": 35355, "total_steps": 42420, "loss": 0.3237, "lr": 8.216568089501242e-05, "epoch": 16.66902404526167, "percentage": 83.35, "elapsed_time": "1:54:48", "remaining_time": "0:22:56", "throughput": 5171.87, "total_tokens": 35625152} {"current_steps": 35360, "total_steps": 42420, "loss": 0.3618, "lr": 8.20527278070411e-05, "epoch": 16.671381423856673, "percentage": 83.36, "elapsed_time": "1:54:49", "remaining_time": "0:22:55", "throughput": 5171.87, "total_tokens": 35629472} {"current_steps": 35365, "total_steps": 42420, "loss": 0.3, "lr": 8.193984547044659e-05, "epoch": 16.673738802451673, "percentage": 83.37, "elapsed_time": "1:54:50", "remaining_time": "0:22:54", "throughput": 5171.94, "total_tokens": 35635232} {"current_steps": 35370, "total_steps": 42420, "loss": 0.2954, "lr": 8.182703390433798e-05, "epoch": 16.676096181046677, "percentage": 83.38, "elapsed_time": "1:54:51", "remaining_time": "0:22:53", "throughput": 5172.0, "total_tokens": 35640704} {"current_steps": 35375, "total_steps": 42420, "loss": 0.3329, "lr": 8.171429312781237e-05, "epoch": 16.678453559641678, "percentage": 83.39, "elapsed_time": "1:54:52", "remaining_time": "0:22:52", "throughput": 5172.05, "total_tokens": 35645856} {"current_steps": 35380, "total_steps": 42420, "loss": 0.3668, "lr": 8.160162315995501e-05, "epoch": 16.680810938236682, "percentage": 83.4, "elapsed_time": "1:54:52", "remaining_time": "0:22:51", "throughput": 5172.1, "total_tokens": 35650880} {"current_steps": 35385, "total_steps": 42420, "loss": 0.2859, "lr": 8.148902401983865e-05, "epoch": 16.683168316831683, "percentage": 83.42, "elapsed_time": "1:54:53", "remaining_time": "0:22:50", "throughput": 5172.18, "total_tokens": 35656416} {"current_steps": 35390, "total_steps": 42420, "loss": 0.337, "lr": 8.137649572652466e-05, "epoch": 16.685525695426687, "percentage": 83.43, "elapsed_time": "1:54:54", "remaining_time": "0:22:49", "throughput": 5172.23, "total_tokens": 35661728} {"current_steps": 35395, "total_steps": 42420, "loss": 0.2978, "lr": 8.126403829906215e-05, "epoch": 16.687883074021688, "percentage": 83.44, "elapsed_time": "1:54:55", "remaining_time": "0:22:48", "throughput": 5172.28, "total_tokens": 35666656} {"current_steps": 35400, "total_steps": 42420, "loss": 0.393, "lr": 8.11516517564882e-05, "epoch": 16.690240452616692, "percentage": 83.45, "elapsed_time": "1:54:56", "remaining_time": "0:22:47", "throughput": 5172.33, "total_tokens": 35671904} {"current_steps": 35405, "total_steps": 42420, "loss": 0.3493, "lr": 8.103933611782793e-05, "epoch": 16.692597831211692, "percentage": 83.46, "elapsed_time": "1:54:57", "remaining_time": "0:22:46", "throughput": 5172.39, "total_tokens": 35677216} {"current_steps": 35410, "total_steps": 42420, "loss": 0.2966, "lr": 8.092709140209453e-05, "epoch": 16.694955209806697, "percentage": 83.47, "elapsed_time": "1:54:58", "remaining_time": "0:22:45", "throughput": 5172.42, "total_tokens": 35682048} {"current_steps": 35415, "total_steps": 42420, "loss": 0.3472, "lr": 8.081491762828913e-05, "epoch": 16.697312588401697, "percentage": 83.49, "elapsed_time": "1:54:59", "remaining_time": "0:22:44", "throughput": 5172.44, "total_tokens": 35686656} {"current_steps": 35420, "total_steps": 42420, "loss": 0.3362, "lr": 8.070281481540093e-05, "epoch": 16.6996699669967, "percentage": 83.5, "elapsed_time": "1:55:00", "remaining_time": "0:22:43", "throughput": 5172.45, "total_tokens": 35691072} {"current_steps": 35425, "total_steps": 42420, "loss": 0.3237, "lr": 8.059078298240675e-05, "epoch": 16.702027345591702, "percentage": 83.51, "elapsed_time": "1:55:01", "remaining_time": "0:22:42", "throughput": 5172.57, "total_tokens": 35697664} {"current_steps": 35430, "total_steps": 42420, "loss": 0.338, "lr": 8.04788221482719e-05, "epoch": 16.704384724186703, "percentage": 83.52, "elapsed_time": "1:55:02", "remaining_time": "0:22:41", "throughput": 5172.61, "total_tokens": 35702784} {"current_steps": 35435, "total_steps": 42420, "loss": 0.3561, "lr": 8.036693233194936e-05, "epoch": 16.706742102781707, "percentage": 83.53, "elapsed_time": "1:55:03", "remaining_time": "0:22:40", "throughput": 5172.62, "total_tokens": 35707264} {"current_steps": 35440, "total_steps": 42420, "loss": 0.3456, "lr": 8.025511355238024e-05, "epoch": 16.709099481376708, "percentage": 83.55, "elapsed_time": "1:55:04", "remaining_time": "0:22:39", "throughput": 5172.69, "total_tokens": 35712832} {"current_steps": 35445, "total_steps": 42420, "loss": 0.3304, "lr": 8.014336582849352e-05, "epoch": 16.71145685997171, "percentage": 83.56, "elapsed_time": "1:55:05", "remaining_time": "0:22:38", "throughput": 5172.72, "total_tokens": 35718080} {"current_steps": 35450, "total_steps": 42420, "loss": 0.3435, "lr": 8.003168917920622e-05, "epoch": 16.713814238566712, "percentage": 83.57, "elapsed_time": "1:55:05", "remaining_time": "0:22:37", "throughput": 5172.73, "total_tokens": 35722464} {"current_steps": 35455, "total_steps": 42420, "loss": 0.3408, "lr": 7.992008362342328e-05, "epoch": 16.716171617161717, "percentage": 83.58, "elapsed_time": "1:55:06", "remaining_time": "0:22:36", "throughput": 5172.72, "total_tokens": 35726912} {"current_steps": 35460, "total_steps": 42420, "loss": 0.3285, "lr": 7.98085491800376e-05, "epoch": 16.718528995756717, "percentage": 83.59, "elapsed_time": "1:55:07", "remaining_time": "0:22:35", "throughput": 5172.77, "total_tokens": 35732064} {"current_steps": 35465, "total_steps": 42420, "loss": 0.3636, "lr": 7.969708586793029e-05, "epoch": 16.72088637435172, "percentage": 83.6, "elapsed_time": "1:55:08", "remaining_time": "0:22:34", "throughput": 5172.79, "total_tokens": 35736768} {"current_steps": 35470, "total_steps": 42420, "loss": 0.3035, "lr": 7.958569370596986e-05, "epoch": 16.723243752946722, "percentage": 83.62, "elapsed_time": "1:55:09", "remaining_time": "0:22:33", "throughput": 5172.88, "total_tokens": 35742560} {"current_steps": 35475, "total_steps": 42420, "loss": 0.3601, "lr": 7.947437271301322e-05, "epoch": 16.725601131541726, "percentage": 83.63, "elapsed_time": "1:55:10", "remaining_time": "0:22:32", "throughput": 5172.98, "total_tokens": 35749024} {"current_steps": 35480, "total_steps": 42420, "loss": 0.3409, "lr": 7.936312290790515e-05, "epoch": 16.727958510136727, "percentage": 83.64, "elapsed_time": "1:55:11", "remaining_time": "0:22:31", "throughput": 5172.99, "total_tokens": 35753568} {"current_steps": 35485, "total_steps": 42420, "loss": 0.2987, "lr": 7.925194430947841e-05, "epoch": 16.73031588873173, "percentage": 83.65, "elapsed_time": "1:55:12", "remaining_time": "0:22:30", "throughput": 5173.02, "total_tokens": 35758304} {"current_steps": 35490, "total_steps": 42420, "loss": 0.3625, "lr": 7.914083693655361e-05, "epoch": 16.73267326732673, "percentage": 83.66, "elapsed_time": "1:55:13", "remaining_time": "0:22:29", "throughput": 5173.16, "total_tokens": 35765184} {"current_steps": 35495, "total_steps": 42420, "loss": 0.3036, "lr": 7.902980080793931e-05, "epoch": 16.735030645921736, "percentage": 83.68, "elapsed_time": "1:55:14", "remaining_time": "0:22:28", "throughput": 5173.15, "total_tokens": 35769568} {"current_steps": 35500, "total_steps": 42420, "loss": 0.3531, "lr": 7.891883594243205e-05, "epoch": 16.737388024516736, "percentage": 83.69, "elapsed_time": "1:55:15", "remaining_time": "0:22:28", "throughput": 5173.18, "total_tokens": 35774656} {"current_steps": 35505, "total_steps": 42420, "loss": 0.3295, "lr": 7.880794235881645e-05, "epoch": 16.73974540311174, "percentage": 83.7, "elapsed_time": "1:55:16", "remaining_time": "0:22:27", "throughput": 5173.31, "total_tokens": 35781472} {"current_steps": 35510, "total_steps": 42420, "loss": 0.33, "lr": 7.869712007586465e-05, "epoch": 16.74210278170674, "percentage": 83.71, "elapsed_time": "1:55:17", "remaining_time": "0:22:26", "throughput": 5173.33, "total_tokens": 35786144} {"current_steps": 35515, "total_steps": 42420, "loss": 0.3594, "lr": 7.858636911233724e-05, "epoch": 16.744460160301745, "percentage": 83.72, "elapsed_time": "1:55:18", "remaining_time": "0:22:25", "throughput": 5173.37, "total_tokens": 35791296} {"current_steps": 35520, "total_steps": 42420, "loss": 0.3849, "lr": 7.847568948698225e-05, "epoch": 16.746817538896746, "percentage": 83.73, "elapsed_time": "1:55:19", "remaining_time": "0:22:24", "throughput": 5173.42, "total_tokens": 35796640} {"current_steps": 35525, "total_steps": 42420, "loss": 0.327, "lr": 7.836508121853591e-05, "epoch": 16.74917491749175, "percentage": 83.75, "elapsed_time": "1:55:20", "remaining_time": "0:22:23", "throughput": 5173.45, "total_tokens": 35801280} {"current_steps": 35530, "total_steps": 42420, "loss": 0.2938, "lr": 7.825454432572238e-05, "epoch": 16.75153229608675, "percentage": 83.76, "elapsed_time": "1:55:21", "remaining_time": "0:22:22", "throughput": 5173.43, "total_tokens": 35805440} {"current_steps": 35535, "total_steps": 42420, "loss": 0.3068, "lr": 7.814407882725366e-05, "epoch": 16.753889674681755, "percentage": 83.77, "elapsed_time": "1:55:21", "remaining_time": "0:22:21", "throughput": 5173.46, "total_tokens": 35810176} {"current_steps": 35540, "total_steps": 42420, "loss": 0.331, "lr": 7.80336847418297e-05, "epoch": 16.756247053276756, "percentage": 83.78, "elapsed_time": "1:55:23", "remaining_time": "0:22:20", "throughput": 5173.61, "total_tokens": 35817376} {"current_steps": 35545, "total_steps": 42420, "loss": 0.3128, "lr": 7.792336208813833e-05, "epoch": 16.75860443187176, "percentage": 83.79, "elapsed_time": "1:55:24", "remaining_time": "0:22:19", "throughput": 5173.69, "total_tokens": 35823424} {"current_steps": 35550, "total_steps": 42420, "loss": 0.3341, "lr": 7.781311088485537e-05, "epoch": 16.76096181046676, "percentage": 83.8, "elapsed_time": "1:55:25", "remaining_time": "0:22:18", "throughput": 5173.77, "total_tokens": 35829600} {"current_steps": 35555, "total_steps": 42420, "loss": 0.3248, "lr": 7.770293115064436e-05, "epoch": 16.763319189061765, "percentage": 83.82, "elapsed_time": "1:55:26", "remaining_time": "0:22:17", "throughput": 5173.81, "total_tokens": 35834688} {"current_steps": 35560, "total_steps": 42420, "loss": 0.3125, "lr": 7.759282290415692e-05, "epoch": 16.765676567656765, "percentage": 83.83, "elapsed_time": "1:55:27", "remaining_time": "0:22:16", "throughput": 5173.89, "total_tokens": 35840480} {"current_steps": 35565, "total_steps": 42420, "loss": 0.3159, "lr": 7.748278616403242e-05, "epoch": 16.76803394625177, "percentage": 83.84, "elapsed_time": "1:55:28", "remaining_time": "0:22:15", "throughput": 5173.9, "total_tokens": 35844832} {"current_steps": 35570, "total_steps": 42420, "loss": 0.3245, "lr": 7.737282094889836e-05, "epoch": 16.77039132484677, "percentage": 83.85, "elapsed_time": "1:55:28", "remaining_time": "0:22:14", "throughput": 5173.94, "total_tokens": 35849984} {"current_steps": 35575, "total_steps": 42420, "loss": 0.4058, "lr": 7.726292727736989e-05, "epoch": 16.772748703441774, "percentage": 83.86, "elapsed_time": "1:55:29", "remaining_time": "0:22:13", "throughput": 5173.95, "total_tokens": 35854464} {"current_steps": 35580, "total_steps": 42420, "loss": 0.3426, "lr": 7.715310516805018e-05, "epoch": 16.775106082036775, "percentage": 83.88, "elapsed_time": "1:55:30", "remaining_time": "0:22:12", "throughput": 5174.01, "total_tokens": 35859776} {"current_steps": 35585, "total_steps": 42420, "loss": 0.2962, "lr": 7.704335463953021e-05, "epoch": 16.77746346063178, "percentage": 83.89, "elapsed_time": "1:55:31", "remaining_time": "0:22:11", "throughput": 5174.06, "total_tokens": 35865216} {"current_steps": 35590, "total_steps": 42420, "loss": 0.3642, "lr": 7.693367571038901e-05, "epoch": 16.77982083922678, "percentage": 83.9, "elapsed_time": "1:55:32", "remaining_time": "0:22:10", "throughput": 5174.12, "total_tokens": 35871104} {"current_steps": 35595, "total_steps": 42420, "loss": 0.3473, "lr": 7.682406839919315e-05, "epoch": 16.782178217821784, "percentage": 83.91, "elapsed_time": "1:55:33", "remaining_time": "0:22:09", "throughput": 5174.24, "total_tokens": 35877920} {"current_steps": 35600, "total_steps": 42420, "loss": 0.3214, "lr": 7.67145327244974e-05, "epoch": 16.784535596416784, "percentage": 83.92, "elapsed_time": "1:55:34", "remaining_time": "0:22:08", "throughput": 5174.22, "total_tokens": 35881760} {"current_steps": 35605, "total_steps": 42420, "loss": 0.2965, "lr": 7.660506870484429e-05, "epoch": 16.78689297501179, "percentage": 83.93, "elapsed_time": "1:55:35", "remaining_time": "0:22:07", "throughput": 5174.25, "total_tokens": 35886400} {"current_steps": 35610, "total_steps": 42420, "loss": 0.3144, "lr": 7.64956763587642e-05, "epoch": 16.78925035360679, "percentage": 83.95, "elapsed_time": "1:55:36", "remaining_time": "0:22:06", "throughput": 5174.23, "total_tokens": 35890560} {"current_steps": 35615, "total_steps": 42420, "loss": 0.2942, "lr": 7.638635570477542e-05, "epoch": 16.79160773220179, "percentage": 83.96, "elapsed_time": "1:55:37", "remaining_time": "0:22:05", "throughput": 5174.23, "total_tokens": 35894816} {"current_steps": 35620, "total_steps": 42420, "loss": 0.3279, "lr": 7.627710676138405e-05, "epoch": 16.793965110796794, "percentage": 83.97, "elapsed_time": "1:55:38", "remaining_time": "0:22:04", "throughput": 5174.24, "total_tokens": 35899360} {"current_steps": 35625, "total_steps": 42420, "loss": 0.3048, "lr": 7.616792954708413e-05, "epoch": 16.796322489391795, "percentage": 83.98, "elapsed_time": "1:55:39", "remaining_time": "0:22:03", "throughput": 5174.37, "total_tokens": 35905920} {"current_steps": 35630, "total_steps": 42420, "loss": 0.2731, "lr": 7.605882408035758e-05, "epoch": 16.7986798679868, "percentage": 83.99, "elapsed_time": "1:55:40", "remaining_time": "0:22:02", "throughput": 5174.41, "total_tokens": 35910880} {"current_steps": 35635, "total_steps": 42420, "loss": 0.3665, "lr": 7.594979037967386e-05, "epoch": 16.8010372465818, "percentage": 84.01, "elapsed_time": "1:55:40", "remaining_time": "0:22:01", "throughput": 5174.42, "total_tokens": 35915232} {"current_steps": 35640, "total_steps": 42420, "loss": 0.3625, "lr": 7.584082846349071e-05, "epoch": 16.803394625176804, "percentage": 84.02, "elapsed_time": "1:55:41", "remaining_time": "0:22:00", "throughput": 5174.41, "total_tokens": 35919488} {"current_steps": 35645, "total_steps": 42420, "loss": 0.3684, "lr": 7.573193835025344e-05, "epoch": 16.805752003771804, "percentage": 84.03, "elapsed_time": "1:55:42", "remaining_time": "0:21:59", "throughput": 5174.42, "total_tokens": 35923904} {"current_steps": 35650, "total_steps": 42420, "loss": 0.3648, "lr": 7.562312005839539e-05, "epoch": 16.80810938236681, "percentage": 84.04, "elapsed_time": "1:55:43", "remaining_time": "0:21:58", "throughput": 5174.49, "total_tokens": 35929600} {"current_steps": 35655, "total_steps": 42420, "loss": 0.3556, "lr": 7.551437360633757e-05, "epoch": 16.81046676096181, "percentage": 84.05, "elapsed_time": "1:55:44", "remaining_time": "0:21:57", "throughput": 5174.51, "total_tokens": 35934400} {"current_steps": 35660, "total_steps": 42420, "loss": 0.3361, "lr": 7.540569901248895e-05, "epoch": 16.812824139556813, "percentage": 84.06, "elapsed_time": "1:55:45", "remaining_time": "0:21:56", "throughput": 5174.56, "total_tokens": 35939584} {"current_steps": 35665, "total_steps": 42420, "loss": 0.31, "lr": 7.529709629524628e-05, "epoch": 16.815181518151814, "percentage": 84.08, "elapsed_time": "1:55:46", "remaining_time": "0:21:55", "throughput": 5174.59, "total_tokens": 35944384} {"current_steps": 35670, "total_steps": 42420, "loss": 0.3477, "lr": 7.51885654729943e-05, "epoch": 16.817538896746818, "percentage": 84.09, "elapsed_time": "1:55:47", "remaining_time": "0:21:54", "throughput": 5174.69, "total_tokens": 35950496} {"current_steps": 35675, "total_steps": 42420, "loss": 0.328, "lr": 7.508010656410519e-05, "epoch": 16.81989627534182, "percentage": 84.1, "elapsed_time": "1:55:48", "remaining_time": "0:21:53", "throughput": 5174.72, "total_tokens": 35955360} {"current_steps": 35680, "total_steps": 42420, "loss": 0.3269, "lr": 7.497171958693927e-05, "epoch": 16.822253653936823, "percentage": 84.11, "elapsed_time": "1:55:49", "remaining_time": "0:21:52", "throughput": 5174.78, "total_tokens": 35960960} {"current_steps": 35685, "total_steps": 42420, "loss": 0.3612, "lr": 7.48634045598447e-05, "epoch": 16.824611032531823, "percentage": 84.12, "elapsed_time": "1:55:50", "remaining_time": "0:21:51", "throughput": 5174.81, "total_tokens": 35966144} {"current_steps": 35690, "total_steps": 42420, "loss": 0.3156, "lr": 7.47551615011573e-05, "epoch": 16.826968411126828, "percentage": 84.13, "elapsed_time": "1:55:51", "remaining_time": "0:21:50", "throughput": 5174.87, "total_tokens": 35971360} {"current_steps": 35695, "total_steps": 42420, "loss": 0.3552, "lr": 7.464699042920081e-05, "epoch": 16.82932578972183, "percentage": 84.15, "elapsed_time": "1:55:52", "remaining_time": "0:21:49", "throughput": 5174.94, "total_tokens": 35977184} {"current_steps": 35700, "total_steps": 42420, "loss": 0.3251, "lr": 7.453889136228681e-05, "epoch": 16.831683168316832, "percentage": 84.16, "elapsed_time": "1:55:53", "remaining_time": "0:21:48", "throughput": 5174.95, "total_tokens": 35981792} {"current_steps": 35705, "total_steps": 42420, "loss": 0.3324, "lr": 7.443086431871454e-05, "epoch": 16.834040546911833, "percentage": 84.17, "elapsed_time": "1:55:53", "remaining_time": "0:21:47", "throughput": 5174.95, "total_tokens": 35986048} {"current_steps": 35710, "total_steps": 42420, "loss": 0.3364, "lr": 7.432290931677126e-05, "epoch": 16.836397925506837, "percentage": 84.18, "elapsed_time": "1:55:54", "remaining_time": "0:21:46", "throughput": 5174.93, "total_tokens": 35990080} {"current_steps": 35715, "total_steps": 42420, "loss": 0.3255, "lr": 7.421502637473182e-05, "epoch": 16.838755304101838, "percentage": 84.19, "elapsed_time": "1:55:55", "remaining_time": "0:21:45", "throughput": 5174.93, "total_tokens": 35994272} {"current_steps": 35720, "total_steps": 42420, "loss": 0.3131, "lr": 7.410721551085891e-05, "epoch": 16.841112682696842, "percentage": 84.21, "elapsed_time": "1:55:56", "remaining_time": "0:21:44", "throughput": 5174.98, "total_tokens": 35999392} {"current_steps": 35725, "total_steps": 42420, "loss": 0.3245, "lr": 7.399947674340324e-05, "epoch": 16.843470061291843, "percentage": 84.22, "elapsed_time": "1:55:57", "remaining_time": "0:21:43", "throughput": 5175.04, "total_tokens": 36004768} {"current_steps": 35730, "total_steps": 42420, "loss": 0.2582, "lr": 7.389181009060303e-05, "epoch": 16.845827439886847, "percentage": 84.23, "elapsed_time": "1:55:58", "remaining_time": "0:21:42", "throughput": 5175.09, "total_tokens": 36010080} {"current_steps": 35735, "total_steps": 42420, "loss": 0.323, "lr": 7.378421557068448e-05, "epoch": 16.848184818481847, "percentage": 84.24, "elapsed_time": "1:55:59", "remaining_time": "0:21:41", "throughput": 5175.13, "total_tokens": 36015136} {"current_steps": 35740, "total_steps": 42420, "loss": 0.3276, "lr": 7.367669320186165e-05, "epoch": 16.85054219707685, "percentage": 84.25, "elapsed_time": "1:56:00", "remaining_time": "0:21:40", "throughput": 5175.15, "total_tokens": 36019904} {"current_steps": 35745, "total_steps": 42420, "loss": 0.346, "lr": 7.356924300233597e-05, "epoch": 16.852899575671852, "percentage": 84.26, "elapsed_time": "1:56:01", "remaining_time": "0:21:39", "throughput": 5175.15, "total_tokens": 36024480} {"current_steps": 35750, "total_steps": 42420, "loss": 0.3697, "lr": 7.346186499029706e-05, "epoch": 16.855256954266856, "percentage": 84.28, "elapsed_time": "1:56:01", "remaining_time": "0:21:38", "throughput": 5175.19, "total_tokens": 36029568} {"current_steps": 35755, "total_steps": 42420, "loss": 0.3571, "lr": 7.335455918392219e-05, "epoch": 16.857614332861857, "percentage": 84.29, "elapsed_time": "1:56:02", "remaining_time": "0:21:37", "throughput": 5175.19, "total_tokens": 36034240} {"current_steps": 35760, "total_steps": 42420, "loss": 0.3252, "lr": 7.324732560137653e-05, "epoch": 16.85997171145686, "percentage": 84.3, "elapsed_time": "1:56:03", "remaining_time": "0:21:36", "throughput": 5175.16, "total_tokens": 36038144} {"current_steps": 35765, "total_steps": 42420, "loss": 0.3056, "lr": 7.314016426081271e-05, "epoch": 16.862329090051862, "percentage": 84.31, "elapsed_time": "1:56:04", "remaining_time": "0:21:35", "throughput": 5175.23, "total_tokens": 36043712} {"current_steps": 35770, "total_steps": 42420, "loss": 0.3347, "lr": 7.303307518037139e-05, "epoch": 16.864686468646866, "percentage": 84.32, "elapsed_time": "1:56:05", "remaining_time": "0:21:34", "throughput": 5175.27, "total_tokens": 36049120} {"current_steps": 35775, "total_steps": 42420, "loss": 0.2868, "lr": 7.2926058378181e-05, "epoch": 16.867043847241867, "percentage": 84.34, "elapsed_time": "1:56:06", "remaining_time": "0:21:33", "throughput": 5175.29, "total_tokens": 36053760} {"current_steps": 35780, "total_steps": 42420, "loss": 0.3239, "lr": 7.281911387235757e-05, "epoch": 16.86940122583687, "percentage": 84.35, "elapsed_time": "1:56:07", "remaining_time": "0:21:33", "throughput": 5175.41, "total_tokens": 36060416} {"current_steps": 35785, "total_steps": 42420, "loss": 0.3685, "lr": 7.27122416810051e-05, "epoch": 16.87175860443187, "percentage": 84.36, "elapsed_time": "1:56:08", "remaining_time": "0:21:32", "throughput": 5175.41, "total_tokens": 36064928} {"current_steps": 35790, "total_steps": 42420, "loss": 0.3167, "lr": 7.260544182221513e-05, "epoch": 16.874115983026876, "percentage": 84.37, "elapsed_time": "1:56:09", "remaining_time": "0:21:31", "throughput": 5175.43, "total_tokens": 36069600} {"current_steps": 35795, "total_steps": 42420, "loss": 0.3601, "lr": 7.249871431406712e-05, "epoch": 16.876473361621876, "percentage": 84.38, "elapsed_time": "1:56:10", "remaining_time": "0:21:30", "throughput": 5175.54, "total_tokens": 36075968} {"current_steps": 35800, "total_steps": 42420, "loss": 0.2933, "lr": 7.239205917462838e-05, "epoch": 16.87883074021688, "percentage": 84.39, "elapsed_time": "1:56:11", "remaining_time": "0:21:29", "throughput": 5175.58, "total_tokens": 36080928} {"current_steps": 35805, "total_steps": 42420, "loss": 0.3419, "lr": 7.228547642195349e-05, "epoch": 16.88118811881188, "percentage": 84.41, "elapsed_time": "1:56:12", "remaining_time": "0:21:28", "throughput": 5175.66, "total_tokens": 36086624} {"current_steps": 35810, "total_steps": 42420, "loss": 0.3551, "lr": 7.217896607408526e-05, "epoch": 16.883545497406885, "percentage": 84.42, "elapsed_time": "1:56:13", "remaining_time": "0:21:27", "throughput": 5175.75, "total_tokens": 36092896} {"current_steps": 35815, "total_steps": 42420, "loss": 0.348, "lr": 7.207252814905407e-05, "epoch": 16.885902876001886, "percentage": 84.43, "elapsed_time": "1:56:14", "remaining_time": "0:21:26", "throughput": 5175.93, "total_tokens": 36100928} {"current_steps": 35820, "total_steps": 42420, "loss": 0.3051, "lr": 7.196616266487816e-05, "epoch": 16.888260254596887, "percentage": 84.44, "elapsed_time": "1:56:15", "remaining_time": "0:21:25", "throughput": 5175.99, "total_tokens": 36106240} {"current_steps": 35825, "total_steps": 42420, "loss": 0.2848, "lr": 7.18598696395632e-05, "epoch": 16.89061763319189, "percentage": 84.45, "elapsed_time": "1:56:16", "remaining_time": "0:21:24", "throughput": 5176.04, "total_tokens": 36111680} {"current_steps": 35830, "total_steps": 42420, "loss": 0.3176, "lr": 7.175364909110299e-05, "epoch": 16.89297501178689, "percentage": 84.46, "elapsed_time": "1:56:17", "remaining_time": "0:21:23", "throughput": 5176.07, "total_tokens": 36116448} {"current_steps": 35835, "total_steps": 42420, "loss": 0.3453, "lr": 7.164750103747875e-05, "epoch": 16.895332390381895, "percentage": 84.48, "elapsed_time": "1:56:18", "remaining_time": "0:21:22", "throughput": 5176.15, "total_tokens": 36122112} {"current_steps": 35840, "total_steps": 42420, "loss": 0.3576, "lr": 7.154142549665965e-05, "epoch": 16.897689768976896, "percentage": 84.49, "elapsed_time": "1:56:19", "remaining_time": "0:21:21", "throughput": 5176.16, "total_tokens": 36126784} {"current_steps": 35845, "total_steps": 42420, "loss": 0.37, "lr": 7.143542248660234e-05, "epoch": 16.9000471475719, "percentage": 84.5, "elapsed_time": "1:56:20", "remaining_time": "0:21:20", "throughput": 5176.17, "total_tokens": 36131136} {"current_steps": 35850, "total_steps": 42420, "loss": 0.3128, "lr": 7.132949202525135e-05, "epoch": 16.9024045261669, "percentage": 84.51, "elapsed_time": "1:56:21", "remaining_time": "0:21:19", "throughput": 5176.21, "total_tokens": 36136448} {"current_steps": 35855, "total_steps": 42420, "loss": 0.2993, "lr": 7.122363413053895e-05, "epoch": 16.904761904761905, "percentage": 84.52, "elapsed_time": "1:56:22", "remaining_time": "0:21:18", "throughput": 5176.25, "total_tokens": 36141504} {"current_steps": 35860, "total_steps": 42420, "loss": 0.3196, "lr": 7.111784882038513e-05, "epoch": 16.907119283356906, "percentage": 84.54, "elapsed_time": "1:56:23", "remaining_time": "0:21:17", "throughput": 5176.33, "total_tokens": 36147232} {"current_steps": 35865, "total_steps": 42420, "loss": 0.3184, "lr": 7.101213611269747e-05, "epoch": 16.90947666195191, "percentage": 84.55, "elapsed_time": "1:56:24", "remaining_time": "0:21:16", "throughput": 5176.37, "total_tokens": 36152224} {"current_steps": 35870, "total_steps": 42420, "loss": 0.3605, "lr": 7.090649602537136e-05, "epoch": 16.91183404054691, "percentage": 84.56, "elapsed_time": "1:56:25", "remaining_time": "0:21:15", "throughput": 5176.43, "total_tokens": 36157504} {"current_steps": 35875, "total_steps": 42420, "loss": 0.3189, "lr": 7.080092857628989e-05, "epoch": 16.914191419141915, "percentage": 84.57, "elapsed_time": "1:56:25", "remaining_time": "0:21:14", "throughput": 5176.48, "total_tokens": 36162752} {"current_steps": 35880, "total_steps": 42420, "loss": 0.3208, "lr": 7.069543378332393e-05, "epoch": 16.916548797736915, "percentage": 84.58, "elapsed_time": "1:56:26", "remaining_time": "0:21:13", "throughput": 5176.49, "total_tokens": 36167136} {"current_steps": 35885, "total_steps": 42420, "loss": 0.2979, "lr": 7.059001166433176e-05, "epoch": 16.91890617633192, "percentage": 84.59, "elapsed_time": "1:56:27", "remaining_time": "0:21:12", "throughput": 5176.52, "total_tokens": 36172096} {"current_steps": 35890, "total_steps": 42420, "loss": 0.3774, "lr": 7.04846622371596e-05, "epoch": 16.92126355492692, "percentage": 84.61, "elapsed_time": "1:56:28", "remaining_time": "0:21:11", "throughput": 5176.56, "total_tokens": 36177312} {"current_steps": 35895, "total_steps": 42420, "loss": 0.3182, "lr": 7.037938551964141e-05, "epoch": 16.923620933521924, "percentage": 84.62, "elapsed_time": "1:56:29", "remaining_time": "0:21:10", "throughput": 5176.57, "total_tokens": 36181888} {"current_steps": 35900, "total_steps": 42420, "loss": 0.3203, "lr": 7.027418152959875e-05, "epoch": 16.925978312116925, "percentage": 84.63, "elapsed_time": "1:56:30", "remaining_time": "0:21:09", "throughput": 5176.63, "total_tokens": 36187072} {"current_steps": 35905, "total_steps": 42420, "loss": 0.3395, "lr": 7.016905028484077e-05, "epoch": 16.92833569071193, "percentage": 84.64, "elapsed_time": "1:56:31", "remaining_time": "0:21:08", "throughput": 5176.63, "total_tokens": 36191168} {"current_steps": 35910, "total_steps": 42420, "loss": 0.3949, "lr": 7.006399180316447e-05, "epoch": 16.93069306930693, "percentage": 84.65, "elapsed_time": "1:56:32", "remaining_time": "0:21:07", "throughput": 5176.68, "total_tokens": 36196576} {"current_steps": 35915, "total_steps": 42420, "loss": 0.2698, "lr": 6.995900610235445e-05, "epoch": 16.933050447901934, "percentage": 84.67, "elapsed_time": "1:56:33", "remaining_time": "0:21:06", "throughput": 5176.67, "total_tokens": 36200672} {"current_steps": 35920, "total_steps": 42420, "loss": 0.3103, "lr": 6.985409320018315e-05, "epoch": 16.935407826496935, "percentage": 84.68, "elapsed_time": "1:56:33", "remaining_time": "0:21:05", "throughput": 5176.7, "total_tokens": 36205440} {"current_steps": 35925, "total_steps": 42420, "loss": 0.2444, "lr": 6.974925311441028e-05, "epoch": 16.93776520509194, "percentage": 84.69, "elapsed_time": "1:56:34", "remaining_time": "0:21:04", "throughput": 5176.73, "total_tokens": 36210432} {"current_steps": 35930, "total_steps": 42420, "loss": 0.3914, "lr": 6.964448586278365e-05, "epoch": 16.94012258368694, "percentage": 84.7, "elapsed_time": "1:56:36", "remaining_time": "0:21:03", "throughput": 5176.85, "total_tokens": 36217248} {"current_steps": 35935, "total_steps": 42420, "loss": 0.3081, "lr": 6.953979146303851e-05, "epoch": 16.942479962281944, "percentage": 84.71, "elapsed_time": "1:56:36", "remaining_time": "0:21:02", "throughput": 5176.88, "total_tokens": 36222208} {"current_steps": 35940, "total_steps": 42420, "loss": 0.3746, "lr": 6.943516993289789e-05, "epoch": 16.944837340876944, "percentage": 84.72, "elapsed_time": "1:56:37", "remaining_time": "0:21:01", "throughput": 5176.87, "total_tokens": 36226432} {"current_steps": 35945, "total_steps": 42420, "loss": 0.3538, "lr": 6.93306212900724e-05, "epoch": 16.94719471947195, "percentage": 84.74, "elapsed_time": "1:56:38", "remaining_time": "0:21:00", "throughput": 5176.88, "total_tokens": 36230784} {"current_steps": 35950, "total_steps": 42420, "loss": 0.357, "lr": 6.922614555226042e-05, "epoch": 16.94955209806695, "percentage": 84.75, "elapsed_time": "1:56:39", "remaining_time": "0:20:59", "throughput": 5176.85, "total_tokens": 36234880} {"current_steps": 35955, "total_steps": 42420, "loss": 0.3662, "lr": 6.912174273714783e-05, "epoch": 16.951909476661953, "percentage": 84.76, "elapsed_time": "1:56:40", "remaining_time": "0:20:58", "throughput": 5176.91, "total_tokens": 36240288} {"current_steps": 35960, "total_steps": 42420, "loss": 0.3288, "lr": 6.901741286240843e-05, "epoch": 16.954266855256954, "percentage": 84.77, "elapsed_time": "1:56:41", "remaining_time": "0:20:57", "throughput": 5176.93, "total_tokens": 36244928} {"current_steps": 35965, "total_steps": 42420, "loss": 0.2875, "lr": 6.891315594570325e-05, "epoch": 16.956624233851958, "percentage": 84.78, "elapsed_time": "1:56:42", "remaining_time": "0:20:56", "throughput": 5176.96, "total_tokens": 36249696} {"current_steps": 35970, "total_steps": 42420, "loss": 0.3863, "lr": 6.880897200468145e-05, "epoch": 16.95898161244696, "percentage": 84.79, "elapsed_time": "1:56:43", "remaining_time": "0:20:55", "throughput": 5177.0, "total_tokens": 36254624} {"current_steps": 35975, "total_steps": 42420, "loss": 0.3372, "lr": 6.870486105697938e-05, "epoch": 16.961338991041963, "percentage": 84.81, "elapsed_time": "1:56:43", "remaining_time": "0:20:54", "throughput": 5176.96, "total_tokens": 36258368} {"current_steps": 35980, "total_steps": 42420, "loss": 0.3259, "lr": 6.860082312022132e-05, "epoch": 16.963696369636963, "percentage": 84.82, "elapsed_time": "1:56:44", "remaining_time": "0:20:53", "throughput": 5177.02, "total_tokens": 36263808} {"current_steps": 35985, "total_steps": 42420, "loss": 0.3897, "lr": 6.849685821201923e-05, "epoch": 16.966053748231968, "percentage": 84.83, "elapsed_time": "1:56:45", "remaining_time": "0:20:52", "throughput": 5177.07, "total_tokens": 36268832} {"current_steps": 35990, "total_steps": 42420, "loss": 0.3259, "lr": 6.83929663499725e-05, "epoch": 16.968411126826968, "percentage": 84.84, "elapsed_time": "1:56:46", "remaining_time": "0:20:51", "throughput": 5177.06, "total_tokens": 36273056} {"current_steps": 35995, "total_steps": 42420, "loss": 0.3482, "lr": 6.828914755166827e-05, "epoch": 16.970768505421972, "percentage": 84.85, "elapsed_time": "1:56:47", "remaining_time": "0:20:50", "throughput": 5177.15, "total_tokens": 36279200} {"current_steps": 36000, "total_steps": 42420, "loss": 0.3254, "lr": 6.818540183468136e-05, "epoch": 16.973125884016973, "percentage": 84.87, "elapsed_time": "1:56:48", "remaining_time": "0:20:49", "throughput": 5177.23, "total_tokens": 36285056} {"current_steps": 36005, "total_steps": 42420, "loss": 0.3657, "lr": 6.808172921657413e-05, "epoch": 16.975483262611977, "percentage": 84.88, "elapsed_time": "1:56:49", "remaining_time": "0:20:48", "throughput": 5177.26, "total_tokens": 36290016} {"current_steps": 36010, "total_steps": 42420, "loss": 0.3623, "lr": 6.79781297148967e-05, "epoch": 16.977840641206978, "percentage": 84.89, "elapsed_time": "1:56:50", "remaining_time": "0:20:47", "throughput": 5177.34, "total_tokens": 36295584} {"current_steps": 36015, "total_steps": 42420, "loss": 0.3053, "lr": 6.787460334718643e-05, "epoch": 16.980198019801982, "percentage": 84.9, "elapsed_time": "1:56:51", "remaining_time": "0:20:46", "throughput": 5177.42, "total_tokens": 36301600} {"current_steps": 36020, "total_steps": 42420, "loss": 0.3352, "lr": 6.77711501309688e-05, "epoch": 16.982555398396983, "percentage": 84.91, "elapsed_time": "1:56:52", "remaining_time": "0:20:45", "throughput": 5177.44, "total_tokens": 36306208} {"current_steps": 36025, "total_steps": 42420, "loss": 0.347, "lr": 6.766777008375657e-05, "epoch": 16.984912776991983, "percentage": 84.92, "elapsed_time": "1:56:53", "remaining_time": "0:20:44", "throughput": 5177.4, "total_tokens": 36309888} {"current_steps": 36030, "total_steps": 42420, "loss": 0.3277, "lr": 6.756446322305038e-05, "epoch": 16.987270155586987, "percentage": 84.94, "elapsed_time": "1:56:54", "remaining_time": "0:20:43", "throughput": 5177.47, "total_tokens": 36315520} {"current_steps": 36035, "total_steps": 42420, "loss": 0.3392, "lr": 6.746122956633816e-05, "epoch": 16.989627534181988, "percentage": 84.95, "elapsed_time": "1:56:55", "remaining_time": "0:20:43", "throughput": 5177.56, "total_tokens": 36321376} {"current_steps": 36040, "total_steps": 42420, "loss": 0.339, "lr": 6.735806913109565e-05, "epoch": 16.991984912776992, "percentage": 84.96, "elapsed_time": "1:56:56", "remaining_time": "0:20:42", "throughput": 5177.58, "total_tokens": 36326272} {"current_steps": 36045, "total_steps": 42420, "loss": 0.3083, "lr": 6.725498193478624e-05, "epoch": 16.994342291371993, "percentage": 84.97, "elapsed_time": "1:56:57", "remaining_time": "0:20:41", "throughput": 5177.65, "total_tokens": 36331616} {"current_steps": 36050, "total_steps": 42420, "loss": 0.3285, "lr": 6.71519679948609e-05, "epoch": 16.996699669966997, "percentage": 84.98, "elapsed_time": "1:56:57", "remaining_time": "0:20:40", "throughput": 5177.7, "total_tokens": 36336992} {"current_steps": 36055, "total_steps": 42420, "loss": 0.3576, "lr": 6.704902732875795e-05, "epoch": 16.999057048561998, "percentage": 85.0, "elapsed_time": "1:56:59", "remaining_time": "0:20:39", "throughput": 5177.77, "total_tokens": 36343040} {"current_steps": 36057, "total_steps": 42420, "eval_loss": 0.33738505840301514, "epoch": 17.0, "percentage": 85.0, "elapsed_time": "1:57:18", "remaining_time": "0:20:42", "throughput": 5163.75, "total_tokens": 36344896} {"current_steps": 36060, "total_steps": 42420, "loss": 0.3016, "lr": 6.694615995390358e-05, "epoch": 17.001414427157002, "percentage": 85.01, "elapsed_time": "1:57:21", "remaining_time": "0:20:41", "throughput": 5162.08, "total_tokens": 36347872} {"current_steps": 36065, "total_steps": 42420, "loss": 0.3192, "lr": 6.684336588771146e-05, "epoch": 17.003771805752002, "percentage": 85.02, "elapsed_time": "1:57:22", "remaining_time": "0:20:40", "throughput": 5162.13, "total_tokens": 36352928} {"current_steps": 36070, "total_steps": 42420, "loss": 0.3177, "lr": 6.674064514758294e-05, "epoch": 17.006129184347007, "percentage": 85.03, "elapsed_time": "1:57:23", "remaining_time": "0:20:39", "throughput": 5162.14, "total_tokens": 36357408} {"current_steps": 36075, "total_steps": 42420, "loss": 0.345, "lr": 6.663799775090684e-05, "epoch": 17.008486562942007, "percentage": 85.04, "elapsed_time": "1:57:24", "remaining_time": "0:20:38", "throughput": 5162.19, "total_tokens": 36362752} {"current_steps": 36080, "total_steps": 42420, "loss": 0.2955, "lr": 6.653542371505971e-05, "epoch": 17.01084394153701, "percentage": 85.05, "elapsed_time": "1:57:24", "remaining_time": "0:20:37", "throughput": 5162.19, "total_tokens": 36367072} {"current_steps": 36085, "total_steps": 42420, "loss": 0.3029, "lr": 6.64329230574055e-05, "epoch": 17.013201320132012, "percentage": 85.07, "elapsed_time": "1:57:25", "remaining_time": "0:20:36", "throughput": 5162.21, "total_tokens": 36371776} {"current_steps": 36090, "total_steps": 42420, "loss": 0.4061, "lr": 6.633049579529598e-05, "epoch": 17.015558698727016, "percentage": 85.08, "elapsed_time": "1:57:26", "remaining_time": "0:20:35", "throughput": 5162.23, "total_tokens": 36376448} {"current_steps": 36095, "total_steps": 42420, "loss": 0.3017, "lr": 6.622814194607014e-05, "epoch": 17.017916077322017, "percentage": 85.09, "elapsed_time": "1:57:27", "remaining_time": "0:20:34", "throughput": 5162.25, "total_tokens": 36381344} {"current_steps": 36100, "total_steps": 42420, "loss": 0.3535, "lr": 6.61258615270548e-05, "epoch": 17.02027345591702, "percentage": 85.1, "elapsed_time": "1:57:28", "remaining_time": "0:20:33", "throughput": 5162.26, "total_tokens": 36385824} {"current_steps": 36105, "total_steps": 42420, "loss": 0.312, "lr": 6.602365455556436e-05, "epoch": 17.02263083451202, "percentage": 85.11, "elapsed_time": "1:57:29", "remaining_time": "0:20:32", "throughput": 5162.28, "total_tokens": 36390720} {"current_steps": 36110, "total_steps": 42420, "loss": 0.3233, "lr": 6.59215210489007e-05, "epoch": 17.024988213107026, "percentage": 85.12, "elapsed_time": "1:57:30", "remaining_time": "0:20:31", "throughput": 5162.3, "total_tokens": 36395360} {"current_steps": 36115, "total_steps": 42420, "loss": 0.3405, "lr": 6.581946102435332e-05, "epoch": 17.027345591702026, "percentage": 85.14, "elapsed_time": "1:57:31", "remaining_time": "0:20:30", "throughput": 5162.32, "total_tokens": 36400064} {"current_steps": 36120, "total_steps": 42420, "loss": 0.385, "lr": 6.571747449919918e-05, "epoch": 17.02970297029703, "percentage": 85.15, "elapsed_time": "1:57:32", "remaining_time": "0:20:30", "throughput": 5162.38, "total_tokens": 36405472} {"current_steps": 36125, "total_steps": 42420, "loss": 0.3248, "lr": 6.561556149070292e-05, "epoch": 17.03206034889203, "percentage": 85.16, "elapsed_time": "1:57:32", "remaining_time": "0:20:29", "throughput": 5162.41, "total_tokens": 36410368} {"current_steps": 36130, "total_steps": 42420, "loss": 0.3602, "lr": 6.551372201611682e-05, "epoch": 17.034417727487035, "percentage": 85.17, "elapsed_time": "1:57:33", "remaining_time": "0:20:28", "throughput": 5162.43, "total_tokens": 36415072} {"current_steps": 36135, "total_steps": 42420, "loss": 0.3008, "lr": 6.54119560926803e-05, "epoch": 17.036775106082036, "percentage": 85.18, "elapsed_time": "1:57:34", "remaining_time": "0:20:27", "throughput": 5162.43, "total_tokens": 36419648} {"current_steps": 36140, "total_steps": 42420, "loss": 0.3179, "lr": 6.531026373762073e-05, "epoch": 17.03913248467704, "percentage": 85.2, "elapsed_time": "1:57:35", "remaining_time": "0:20:26", "throughput": 5162.5, "total_tokens": 36425184} {"current_steps": 36145, "total_steps": 42420, "loss": 0.3845, "lr": 6.520864496815288e-05, "epoch": 17.04148986327204, "percentage": 85.21, "elapsed_time": "1:57:36", "remaining_time": "0:20:25", "throughput": 5162.49, "total_tokens": 36429312} {"current_steps": 36150, "total_steps": 42420, "loss": 0.3635, "lr": 6.51070998014791e-05, "epoch": 17.043847241867045, "percentage": 85.22, "elapsed_time": "1:57:37", "remaining_time": "0:20:24", "throughput": 5162.53, "total_tokens": 36434528} {"current_steps": 36155, "total_steps": 42420, "loss": 0.3823, "lr": 6.500562825478929e-05, "epoch": 17.046204620462046, "percentage": 85.23, "elapsed_time": "1:57:38", "remaining_time": "0:20:23", "throughput": 5162.61, "total_tokens": 36440704} {"current_steps": 36160, "total_steps": 42420, "loss": 0.2896, "lr": 6.490423034526077e-05, "epoch": 17.04856199905705, "percentage": 85.24, "elapsed_time": "1:57:39", "remaining_time": "0:20:22", "throughput": 5162.72, "total_tokens": 36447168} {"current_steps": 36165, "total_steps": 42420, "loss": 0.3469, "lr": 6.480290609005862e-05, "epoch": 17.05091937765205, "percentage": 85.25, "elapsed_time": "1:57:40", "remaining_time": "0:20:21", "throughput": 5162.7, "total_tokens": 36451232} {"current_steps": 36170, "total_steps": 42420, "loss": 0.3122, "lr": 6.470165550633528e-05, "epoch": 17.053276756247055, "percentage": 85.27, "elapsed_time": "1:57:41", "remaining_time": "0:20:20", "throughput": 5162.78, "total_tokens": 36456992} {"current_steps": 36175, "total_steps": 42420, "loss": 0.3237, "lr": 6.46004786112307e-05, "epoch": 17.055634134842055, "percentage": 85.28, "elapsed_time": "1:57:42", "remaining_time": "0:20:19", "throughput": 5162.8, "total_tokens": 36461760} {"current_steps": 36180, "total_steps": 42420, "loss": 0.3368, "lr": 6.449937542187235e-05, "epoch": 17.05799151343706, "percentage": 85.29, "elapsed_time": "1:57:43", "remaining_time": "0:20:18", "throughput": 5162.89, "total_tokens": 36467968} {"current_steps": 36185, "total_steps": 42420, "loss": 0.3491, "lr": 6.439834595537542e-05, "epoch": 17.06034889203206, "percentage": 85.3, "elapsed_time": "1:57:44", "remaining_time": "0:20:17", "throughput": 5162.93, "total_tokens": 36472800} {"current_steps": 36190, "total_steps": 42420, "loss": 0.3117, "lr": 6.429739022884246e-05, "epoch": 17.062706270627064, "percentage": 85.31, "elapsed_time": "1:57:45", "remaining_time": "0:20:16", "throughput": 5162.96, "total_tokens": 36477792} {"current_steps": 36195, "total_steps": 42420, "loss": 0.269, "lr": 6.419650825936346e-05, "epoch": 17.065063649222065, "percentage": 85.33, "elapsed_time": "1:57:46", "remaining_time": "0:20:15", "throughput": 5162.99, "total_tokens": 36482720} {"current_steps": 36200, "total_steps": 42420, "loss": 0.2619, "lr": 6.40957000640161e-05, "epoch": 17.06742102781707, "percentage": 85.34, "elapsed_time": "1:57:47", "remaining_time": "0:20:14", "throughput": 5163.07, "total_tokens": 36488672} {"current_steps": 36205, "total_steps": 42420, "loss": 0.3741, "lr": 6.399496565986545e-05, "epoch": 17.06977840641207, "percentage": 85.35, "elapsed_time": "1:57:48", "remaining_time": "0:20:13", "throughput": 5163.08, "total_tokens": 36493376} {"current_steps": 36210, "total_steps": 42420, "loss": 0.2368, "lr": 6.389430506396421e-05, "epoch": 17.072135785007074, "percentage": 85.36, "elapsed_time": "1:57:49", "remaining_time": "0:20:12", "throughput": 5163.13, "total_tokens": 36498528} {"current_steps": 36215, "total_steps": 42420, "loss": 0.3646, "lr": 6.379371829335245e-05, "epoch": 17.074493163602074, "percentage": 85.37, "elapsed_time": "1:57:49", "remaining_time": "0:20:11", "throughput": 5163.16, "total_tokens": 36503328} {"current_steps": 36220, "total_steps": 42420, "loss": 0.3252, "lr": 6.369320536505796e-05, "epoch": 17.076850542197075, "percentage": 85.38, "elapsed_time": "1:57:50", "remaining_time": "0:20:10", "throughput": 5163.2, "total_tokens": 36508352} {"current_steps": 36225, "total_steps": 42420, "loss": 0.3422, "lr": 6.359276629609556e-05, "epoch": 17.07920792079208, "percentage": 85.4, "elapsed_time": "1:57:52", "remaining_time": "0:20:09", "throughput": 5163.33, "total_tokens": 36515296} {"current_steps": 36230, "total_steps": 42420, "loss": 0.3869, "lr": 6.34924011034681e-05, "epoch": 17.08156529938708, "percentage": 85.41, "elapsed_time": "1:57:52", "remaining_time": "0:20:08", "throughput": 5163.35, "total_tokens": 36519872} {"current_steps": 36235, "total_steps": 42420, "loss": 0.3326, "lr": 6.339210980416566e-05, "epoch": 17.083922677982084, "percentage": 85.42, "elapsed_time": "1:57:54", "remaining_time": "0:20:07", "throughput": 5163.44, "total_tokens": 36526272} {"current_steps": 36240, "total_steps": 42420, "loss": 0.3069, "lr": 6.329189241516587e-05, "epoch": 17.086280056577085, "percentage": 85.43, "elapsed_time": "1:57:54", "remaining_time": "0:20:06", "throughput": 5163.43, "total_tokens": 36530624} {"current_steps": 36245, "total_steps": 42420, "loss": 0.3404, "lr": 6.319174895343378e-05, "epoch": 17.08863743517209, "percentage": 85.44, "elapsed_time": "1:57:55", "remaining_time": "0:20:05", "throughput": 5163.39, "total_tokens": 36534336} {"current_steps": 36250, "total_steps": 42420, "loss": 0.3396, "lr": 6.309167943592198e-05, "epoch": 17.09099481376709, "percentage": 85.45, "elapsed_time": "1:57:56", "remaining_time": "0:20:04", "throughput": 5163.38, "total_tokens": 36538400} {"current_steps": 36255, "total_steps": 42420, "loss": 0.3359, "lr": 6.29916838795706e-05, "epoch": 17.093352192362094, "percentage": 85.47, "elapsed_time": "1:57:57", "remaining_time": "0:20:03", "throughput": 5163.42, "total_tokens": 36543808} {"current_steps": 36260, "total_steps": 42420, "loss": 0.2962, "lr": 6.289176230130727e-05, "epoch": 17.095709570957094, "percentage": 85.48, "elapsed_time": "1:57:58", "remaining_time": "0:20:02", "throughput": 5163.37, "total_tokens": 36547648} {"current_steps": 36265, "total_steps": 42420, "loss": 0.3233, "lr": 6.279191471804679e-05, "epoch": 17.0980669495521, "percentage": 85.49, "elapsed_time": "1:57:59", "remaining_time": "0:20:01", "throughput": 5163.41, "total_tokens": 36552672} {"current_steps": 36270, "total_steps": 42420, "loss": 0.3505, "lr": 6.26921411466918e-05, "epoch": 17.1004243281471, "percentage": 85.5, "elapsed_time": "1:57:59", "remaining_time": "0:20:00", "throughput": 5163.39, "total_tokens": 36556800} {"current_steps": 36275, "total_steps": 42420, "loss": 0.3699, "lr": 6.259244160413225e-05, "epoch": 17.102781706742103, "percentage": 85.51, "elapsed_time": "1:58:01", "remaining_time": "0:19:59", "throughput": 5163.48, "total_tokens": 36563264} {"current_steps": 36280, "total_steps": 42420, "loss": 0.2952, "lr": 6.249281610724555e-05, "epoch": 17.105139085337104, "percentage": 85.53, "elapsed_time": "1:58:02", "remaining_time": "0:19:58", "throughput": 5163.54, "total_tokens": 36568704} {"current_steps": 36285, "total_steps": 42420, "loss": 0.3423, "lr": 6.239326467289663e-05, "epoch": 17.107496463932108, "percentage": 85.54, "elapsed_time": "1:58:03", "remaining_time": "0:19:57", "throughput": 5163.6, "total_tokens": 36574176} {"current_steps": 36290, "total_steps": 42420, "loss": 0.337, "lr": 6.229378731793789e-05, "epoch": 17.10985384252711, "percentage": 85.55, "elapsed_time": "1:58:03", "remaining_time": "0:19:56", "throughput": 5163.59, "total_tokens": 36578208} {"current_steps": 36295, "total_steps": 42420, "loss": 0.3122, "lr": 6.219438405920913e-05, "epoch": 17.112211221122113, "percentage": 85.56, "elapsed_time": "1:58:04", "remaining_time": "0:19:55", "throughput": 5163.67, "total_tokens": 36584160} {"current_steps": 36300, "total_steps": 42420, "loss": 0.3662, "lr": 6.209505491353773e-05, "epoch": 17.114568599717114, "percentage": 85.57, "elapsed_time": "1:58:05", "remaining_time": "0:19:54", "throughput": 5163.72, "total_tokens": 36589376} {"current_steps": 36305, "total_steps": 42420, "loss": 0.3837, "lr": 6.199579989773818e-05, "epoch": 17.116925978312118, "percentage": 85.58, "elapsed_time": "1:58:06", "remaining_time": "0:19:53", "throughput": 5163.75, "total_tokens": 36594240} {"current_steps": 36310, "total_steps": 42420, "loss": 0.3309, "lr": 6.189661902861288e-05, "epoch": 17.11928335690712, "percentage": 85.6, "elapsed_time": "1:58:07", "remaining_time": "0:19:52", "throughput": 5163.82, "total_tokens": 36599968} {"current_steps": 36315, "total_steps": 42420, "loss": 0.3458, "lr": 6.179751232295139e-05, "epoch": 17.121640735502123, "percentage": 85.61, "elapsed_time": "1:58:08", "remaining_time": "0:19:51", "throughput": 5163.81, "total_tokens": 36604160} {"current_steps": 36320, "total_steps": 42420, "loss": 0.3055, "lr": 6.16984797975308e-05, "epoch": 17.123998114097123, "percentage": 85.62, "elapsed_time": "1:58:09", "remaining_time": "0:19:50", "throughput": 5163.87, "total_tokens": 36609504} {"current_steps": 36325, "total_steps": 42420, "loss": 0.3112, "lr": 6.159952146911568e-05, "epoch": 17.126355492692127, "percentage": 85.63, "elapsed_time": "1:58:10", "remaining_time": "0:19:49", "throughput": 5163.93, "total_tokens": 36615008} {"current_steps": 36330, "total_steps": 42420, "loss": 0.3785, "lr": 6.150063735445794e-05, "epoch": 17.128712871287128, "percentage": 85.64, "elapsed_time": "1:58:11", "remaining_time": "0:19:48", "throughput": 5163.99, "total_tokens": 36620448} {"current_steps": 36335, "total_steps": 42420, "loss": 0.3739, "lr": 6.140182747029693e-05, "epoch": 17.131070249882132, "percentage": 85.66, "elapsed_time": "1:58:12", "remaining_time": "0:19:47", "throughput": 5164.0, "total_tokens": 36624992} {"current_steps": 36340, "total_steps": 42420, "loss": 0.3467, "lr": 6.130309183335974e-05, "epoch": 17.133427628477133, "percentage": 85.67, "elapsed_time": "1:58:13", "remaining_time": "0:19:46", "throughput": 5164.13, "total_tokens": 36631936} {"current_steps": 36345, "total_steps": 42420, "loss": 0.3353, "lr": 6.120443046036034e-05, "epoch": 17.135785007072137, "percentage": 85.68, "elapsed_time": "1:58:14", "remaining_time": "0:19:45", "throughput": 5164.14, "total_tokens": 36636480} {"current_steps": 36350, "total_steps": 42420, "loss": 0.3509, "lr": 6.110584336800051e-05, "epoch": 17.138142385667138, "percentage": 85.69, "elapsed_time": "1:58:15", "remaining_time": "0:19:44", "throughput": 5164.16, "total_tokens": 36641120} {"current_steps": 36355, "total_steps": 42420, "loss": 0.3267, "lr": 6.100733057296942e-05, "epoch": 17.14049976426214, "percentage": 85.7, "elapsed_time": "1:58:16", "remaining_time": "0:19:43", "throughput": 5164.21, "total_tokens": 36646240} {"current_steps": 36360, "total_steps": 42420, "loss": 0.3788, "lr": 6.090889209194361e-05, "epoch": 17.142857142857142, "percentage": 85.71, "elapsed_time": "1:58:17", "remaining_time": "0:19:42", "throughput": 5164.21, "total_tokens": 36650432} {"current_steps": 36365, "total_steps": 42420, "loss": 0.3186, "lr": 6.081052794158698e-05, "epoch": 17.145214521452147, "percentage": 85.73, "elapsed_time": "1:58:17", "remaining_time": "0:19:41", "throughput": 5164.19, "total_tokens": 36654560} {"current_steps": 36370, "total_steps": 42420, "loss": 0.2866, "lr": 6.071223813855098e-05, "epoch": 17.147571900047147, "percentage": 85.74, "elapsed_time": "1:58:18", "remaining_time": "0:19:40", "throughput": 5164.21, "total_tokens": 36659296} {"current_steps": 36375, "total_steps": 42420, "loss": 0.2966, "lr": 6.0614022699474413e-05, "epoch": 17.14992927864215, "percentage": 85.75, "elapsed_time": "1:58:19", "remaining_time": "0:19:39", "throughput": 5164.34, "total_tokens": 36665984} {"current_steps": 36380, "total_steps": 42420, "loss": 0.3363, "lr": 6.051588164098354e-05, "epoch": 17.152286657237152, "percentage": 85.76, "elapsed_time": "1:58:20", "remaining_time": "0:19:38", "throughput": 5164.39, "total_tokens": 36671328} {"current_steps": 36385, "total_steps": 42420, "loss": 0.3843, "lr": 6.041781497969173e-05, "epoch": 17.154644035832156, "percentage": 85.77, "elapsed_time": "1:58:21", "remaining_time": "0:19:37", "throughput": 5164.46, "total_tokens": 36676992} {"current_steps": 36390, "total_steps": 42420, "loss": 0.3736, "lr": 6.0319822732200214e-05, "epoch": 17.157001414427157, "percentage": 85.79, "elapsed_time": "1:58:22", "remaining_time": "0:19:36", "throughput": 5164.51, "total_tokens": 36682624} {"current_steps": 36395, "total_steps": 42420, "loss": 0.3539, "lr": 6.022190491509738e-05, "epoch": 17.15935879302216, "percentage": 85.8, "elapsed_time": "1:58:23", "remaining_time": "0:19:35", "throughput": 5164.54, "total_tokens": 36687520} {"current_steps": 36400, "total_steps": 42420, "loss": 0.3088, "lr": 6.012406154495903e-05, "epoch": 17.16171617161716, "percentage": 85.81, "elapsed_time": "1:58:24", "remaining_time": "0:19:35", "throughput": 5164.61, "total_tokens": 36693248} {"current_steps": 36405, "total_steps": 42420, "loss": 0.3307, "lr": 6.002629263834836e-05, "epoch": 17.164073550212166, "percentage": 85.82, "elapsed_time": "1:58:25", "remaining_time": "0:19:34", "throughput": 5164.68, "total_tokens": 36698656} {"current_steps": 36410, "total_steps": 42420, "loss": 0.3448, "lr": 5.992859821181607e-05, "epoch": 17.166430928807166, "percentage": 85.83, "elapsed_time": "1:58:26", "remaining_time": "0:19:33", "throughput": 5164.71, "total_tokens": 36703584} {"current_steps": 36415, "total_steps": 42420, "loss": 0.3193, "lr": 5.983097828190015e-05, "epoch": 17.16878830740217, "percentage": 85.84, "elapsed_time": "1:58:27", "remaining_time": "0:19:32", "throughput": 5164.72, "total_tokens": 36707936} {"current_steps": 36420, "total_steps": 42420, "loss": 0.2873, "lr": 5.973343286512589e-05, "epoch": 17.17114568599717, "percentage": 85.86, "elapsed_time": "1:58:28", "remaining_time": "0:19:31", "throughput": 5164.78, "total_tokens": 36713728} {"current_steps": 36425, "total_steps": 42420, "loss": 0.319, "lr": 5.9635961978006246e-05, "epoch": 17.173503064592172, "percentage": 85.87, "elapsed_time": "1:58:29", "remaining_time": "0:19:30", "throughput": 5164.86, "total_tokens": 36719456} {"current_steps": 36430, "total_steps": 42420, "loss": 0.3563, "lr": 5.9538565637041185e-05, "epoch": 17.175860443187176, "percentage": 85.88, "elapsed_time": "1:58:30", "remaining_time": "0:19:29", "throughput": 5164.92, "total_tokens": 36724832} {"current_steps": 36435, "total_steps": 42420, "loss": 0.3143, "lr": 5.944124385871835e-05, "epoch": 17.178217821782177, "percentage": 85.89, "elapsed_time": "1:58:31", "remaining_time": "0:19:28", "throughput": 5164.99, "total_tokens": 36730496} {"current_steps": 36440, "total_steps": 42420, "loss": 0.3046, "lr": 5.9343996659512676e-05, "epoch": 17.18057520037718, "percentage": 85.9, "elapsed_time": "1:58:32", "remaining_time": "0:19:27", "throughput": 5164.98, "total_tokens": 36734560} {"current_steps": 36445, "total_steps": 42420, "loss": 0.256, "lr": 5.9246824055886437e-05, "epoch": 17.18293257897218, "percentage": 85.91, "elapsed_time": "1:58:33", "remaining_time": "0:19:26", "throughput": 5165.02, "total_tokens": 36739552} {"current_steps": 36450, "total_steps": 42420, "loss": 0.3351, "lr": 5.9149726064289365e-05, "epoch": 17.185289957567186, "percentage": 85.93, "elapsed_time": "1:58:34", "remaining_time": "0:19:25", "throughput": 5165.01, "total_tokens": 36744032} {"current_steps": 36455, "total_steps": 42420, "loss": 0.2734, "lr": 5.905270270115842e-05, "epoch": 17.187647336162186, "percentage": 85.94, "elapsed_time": "1:58:35", "remaining_time": "0:19:24", "throughput": 5165.07, "total_tokens": 36749472} {"current_steps": 36460, "total_steps": 42420, "loss": 0.3271, "lr": 5.895575398291808e-05, "epoch": 17.19000471475719, "percentage": 85.95, "elapsed_time": "1:58:35", "remaining_time": "0:19:23", "throughput": 5165.05, "total_tokens": 36753600} {"current_steps": 36465, "total_steps": 42420, "loss": 0.4674, "lr": 5.885887992598016e-05, "epoch": 17.19236209335219, "percentage": 85.96, "elapsed_time": "1:58:36", "remaining_time": "0:19:22", "throughput": 5165.08, "total_tokens": 36758400} {"current_steps": 36470, "total_steps": 42420, "loss": 0.3514, "lr": 5.876208054674359e-05, "epoch": 17.194719471947195, "percentage": 85.97, "elapsed_time": "1:58:37", "remaining_time": "0:19:21", "throughput": 5165.11, "total_tokens": 36763360} {"current_steps": 36475, "total_steps": 42420, "loss": 0.2974, "lr": 5.866535586159505e-05, "epoch": 17.197076850542196, "percentage": 85.99, "elapsed_time": "1:58:38", "remaining_time": "0:19:20", "throughput": 5165.23, "total_tokens": 36770784} {"current_steps": 36480, "total_steps": 42420, "loss": 0.3192, "lr": 5.856870588690827e-05, "epoch": 17.1994342291372, "percentage": 86.0, "elapsed_time": "1:58:39", "remaining_time": "0:19:19", "throughput": 5165.24, "total_tokens": 36775424} {"current_steps": 36485, "total_steps": 42420, "loss": 0.2782, "lr": 5.847213063904455e-05, "epoch": 17.2017916077322, "percentage": 86.01, "elapsed_time": "1:58:40", "remaining_time": "0:19:18", "throughput": 5165.27, "total_tokens": 36780352} {"current_steps": 36490, "total_steps": 42420, "loss": 0.3615, "lr": 5.8375630134352374e-05, "epoch": 17.204148986327205, "percentage": 86.02, "elapsed_time": "1:58:41", "remaining_time": "0:19:17", "throughput": 5165.27, "total_tokens": 36784768} {"current_steps": 36495, "total_steps": 42420, "loss": 0.3647, "lr": 5.8279204389167675e-05, "epoch": 17.206506364922205, "percentage": 86.03, "elapsed_time": "1:58:42", "remaining_time": "0:19:16", "throughput": 5165.29, "total_tokens": 36789824} {"current_steps": 36500, "total_steps": 42420, "loss": 0.3644, "lr": 5.8182853419813665e-05, "epoch": 17.20886374351721, "percentage": 86.04, "elapsed_time": "1:58:43", "remaining_time": "0:19:15", "throughput": 5165.31, "total_tokens": 36794400} {"current_steps": 36505, "total_steps": 42420, "loss": 0.3553, "lr": 5.808657724260108e-05, "epoch": 17.21122112211221, "percentage": 86.06, "elapsed_time": "1:58:44", "remaining_time": "0:19:14", "throughput": 5165.35, "total_tokens": 36799456} {"current_steps": 36510, "total_steps": 42420, "loss": 0.3606, "lr": 5.799037587382755e-05, "epoch": 17.213578500707214, "percentage": 86.07, "elapsed_time": "1:58:45", "remaining_time": "0:19:13", "throughput": 5165.4, "total_tokens": 36804608} {"current_steps": 36515, "total_steps": 42420, "loss": 0.3201, "lr": 5.789424932977855e-05, "epoch": 17.215935879302215, "percentage": 86.08, "elapsed_time": "1:58:46", "remaining_time": "0:19:12", "throughput": 5165.41, "total_tokens": 36809088} {"current_steps": 36520, "total_steps": 42420, "loss": 0.3282, "lr": 5.779819762672656e-05, "epoch": 17.21829325789722, "percentage": 86.09, "elapsed_time": "1:58:46", "remaining_time": "0:19:11", "throughput": 5165.41, "total_tokens": 36813504} {"current_steps": 36525, "total_steps": 42420, "loss": 0.3224, "lr": 5.7702220780931566e-05, "epoch": 17.22065063649222, "percentage": 86.1, "elapsed_time": "1:58:48", "remaining_time": "0:19:10", "throughput": 5165.55, "total_tokens": 36820864} {"current_steps": 36530, "total_steps": 42420, "loss": 0.3299, "lr": 5.760631880864081e-05, "epoch": 17.223008015087224, "percentage": 86.12, "elapsed_time": "1:58:49", "remaining_time": "0:19:09", "throughput": 5165.59, "total_tokens": 36825792} {"current_steps": 36535, "total_steps": 42420, "loss": 0.3212, "lr": 5.751049172608885e-05, "epoch": 17.225365393682225, "percentage": 86.13, "elapsed_time": "1:58:50", "remaining_time": "0:19:08", "throughput": 5165.64, "total_tokens": 36831296} {"current_steps": 36540, "total_steps": 42420, "loss": 0.314, "lr": 5.7414739549497606e-05, "epoch": 17.22772277227723, "percentage": 86.14, "elapsed_time": "1:58:50", "remaining_time": "0:19:07", "throughput": 5165.68, "total_tokens": 36836416} {"current_steps": 36545, "total_steps": 42420, "loss": 0.3209, "lr": 5.731906229507638e-05, "epoch": 17.23008015087223, "percentage": 86.15, "elapsed_time": "1:58:51", "remaining_time": "0:19:06", "throughput": 5165.73, "total_tokens": 36841984} {"current_steps": 36550, "total_steps": 42420, "loss": 0.3738, "lr": 5.7223459979021565e-05, "epoch": 17.232437529467234, "percentage": 86.16, "elapsed_time": "1:58:52", "remaining_time": "0:19:05", "throughput": 5165.8, "total_tokens": 36847584} {"current_steps": 36555, "total_steps": 42420, "loss": 0.3687, "lr": 5.7127932617516976e-05, "epoch": 17.234794908062234, "percentage": 86.17, "elapsed_time": "1:58:53", "remaining_time": "0:19:04", "throughput": 5165.83, "total_tokens": 36852608} {"current_steps": 36560, "total_steps": 42420, "loss": 0.3154, "lr": 5.7032480226733907e-05, "epoch": 17.23715228665724, "percentage": 86.19, "elapsed_time": "1:58:54", "remaining_time": "0:19:03", "throughput": 5165.87, "total_tokens": 36857664} {"current_steps": 36565, "total_steps": 42420, "loss": 0.3092, "lr": 5.693710282283077e-05, "epoch": 17.23950966525224, "percentage": 86.2, "elapsed_time": "1:58:55", "remaining_time": "0:19:02", "throughput": 5165.85, "total_tokens": 36861760} {"current_steps": 36570, "total_steps": 42420, "loss": 0.2993, "lr": 5.684180042195336e-05, "epoch": 17.241867043847243, "percentage": 86.21, "elapsed_time": "1:58:56", "remaining_time": "0:19:01", "throughput": 5165.86, "total_tokens": 36866496} {"current_steps": 36575, "total_steps": 42420, "loss": 0.3182, "lr": 5.674657304023467e-05, "epoch": 17.244224422442244, "percentage": 86.22, "elapsed_time": "1:58:57", "remaining_time": "0:19:00", "throughput": 5165.9, "total_tokens": 36871552} {"current_steps": 36580, "total_steps": 42420, "loss": 0.3201, "lr": 5.66514206937952e-05, "epoch": 17.246581801037248, "percentage": 86.23, "elapsed_time": "1:58:58", "remaining_time": "0:18:59", "throughput": 5165.96, "total_tokens": 36876928} {"current_steps": 36585, "total_steps": 42420, "loss": 0.3032, "lr": 5.6556343398742715e-05, "epoch": 17.24893917963225, "percentage": 86.24, "elapsed_time": "1:58:59", "remaining_time": "0:18:58", "throughput": 5166.02, "total_tokens": 36882432} {"current_steps": 36590, "total_steps": 42420, "loss": 0.335, "lr": 5.646134117117185e-05, "epoch": 17.251296558227253, "percentage": 86.26, "elapsed_time": "1:59:00", "remaining_time": "0:18:57", "throughput": 5166.04, "total_tokens": 36887360} {"current_steps": 36595, "total_steps": 42420, "loss": 0.2981, "lr": 5.6366414027165134e-05, "epoch": 17.253653936822253, "percentage": 86.27, "elapsed_time": "1:59:01", "remaining_time": "0:18:56", "throughput": 5166.09, "total_tokens": 36892384} {"current_steps": 36600, "total_steps": 42420, "loss": 0.2942, "lr": 5.627156198279204e-05, "epoch": 17.256011315417258, "percentage": 86.28, "elapsed_time": "1:59:02", "remaining_time": "0:18:55", "throughput": 5166.11, "total_tokens": 36897248} {"current_steps": 36605, "total_steps": 42420, "loss": 0.3894, "lr": 5.617678505410945e-05, "epoch": 17.25836869401226, "percentage": 86.29, "elapsed_time": "1:59:03", "remaining_time": "0:18:54", "throughput": 5166.15, "total_tokens": 36902144} {"current_steps": 36610, "total_steps": 42420, "loss": 0.3318, "lr": 5.608208325716141e-05, "epoch": 17.260726072607262, "percentage": 86.3, "elapsed_time": "1:59:03", "remaining_time": "0:18:53", "throughput": 5166.14, "total_tokens": 36906496} {"current_steps": 36615, "total_steps": 42420, "loss": 0.3444, "lr": 5.598745660797944e-05, "epoch": 17.263083451202263, "percentage": 86.32, "elapsed_time": "1:59:05", "remaining_time": "0:18:52", "throughput": 5166.23, "total_tokens": 36912864} {"current_steps": 36620, "total_steps": 42420, "loss": 0.3252, "lr": 5.58929051225821e-05, "epoch": 17.265440829797264, "percentage": 86.33, "elapsed_time": "1:59:05", "remaining_time": "0:18:51", "throughput": 5166.24, "total_tokens": 36917408} {"current_steps": 36625, "total_steps": 42420, "loss": 0.2981, "lr": 5.579842881697556e-05, "epoch": 17.267798208392268, "percentage": 86.34, "elapsed_time": "1:59:06", "remaining_time": "0:18:50", "throughput": 5166.27, "total_tokens": 36922336} {"current_steps": 36630, "total_steps": 42420, "loss": 0.3414, "lr": 5.570402770715277e-05, "epoch": 17.27015558698727, "percentage": 86.35, "elapsed_time": "1:59:07", "remaining_time": "0:18:49", "throughput": 5166.33, "total_tokens": 36928096} {"current_steps": 36635, "total_steps": 42420, "loss": 0.3054, "lr": 5.5609701809094416e-05, "epoch": 17.272512965582273, "percentage": 86.36, "elapsed_time": "1:59:08", "remaining_time": "0:18:48", "throughput": 5166.46, "total_tokens": 36934912} {"current_steps": 36640, "total_steps": 42420, "loss": 0.3521, "lr": 5.551545113876827e-05, "epoch": 17.274870344177273, "percentage": 86.37, "elapsed_time": "1:59:10", "remaining_time": "0:18:47", "throughput": 5166.55, "total_tokens": 36941152} {"current_steps": 36645, "total_steps": 42420, "loss": 0.3636, "lr": 5.542127571212929e-05, "epoch": 17.277227722772277, "percentage": 86.39, "elapsed_time": "1:59:10", "remaining_time": "0:18:46", "throughput": 5166.56, "total_tokens": 36945856} {"current_steps": 36650, "total_steps": 42420, "loss": 0.3519, "lr": 5.532717554511979e-05, "epoch": 17.279585101367278, "percentage": 86.4, "elapsed_time": "1:59:11", "remaining_time": "0:18:45", "throughput": 5166.57, "total_tokens": 36950272} {"current_steps": 36655, "total_steps": 42420, "loss": 0.388, "lr": 5.5233150653669354e-05, "epoch": 17.281942479962282, "percentage": 86.41, "elapsed_time": "1:59:12", "remaining_time": "0:18:44", "throughput": 5166.61, "total_tokens": 36955360} {"current_steps": 36660, "total_steps": 42420, "loss": 0.3969, "lr": 5.5139201053694745e-05, "epoch": 17.284299858557283, "percentage": 86.42, "elapsed_time": "1:59:13", "remaining_time": "0:18:43", "throughput": 5166.64, "total_tokens": 36960192} {"current_steps": 36665, "total_steps": 42420, "loss": 0.3826, "lr": 5.504532676110019e-05, "epoch": 17.286657237152287, "percentage": 86.43, "elapsed_time": "1:59:14", "remaining_time": "0:18:42", "throughput": 5166.63, "total_tokens": 36964384} {"current_steps": 36670, "total_steps": 42420, "loss": 0.3604, "lr": 5.4951527791776866e-05, "epoch": 17.289014615747288, "percentage": 86.45, "elapsed_time": "1:59:15", "remaining_time": "0:18:41", "throughput": 5166.64, "total_tokens": 36968864} {"current_steps": 36675, "total_steps": 42420, "loss": 0.3459, "lr": 5.4857804161603464e-05, "epoch": 17.291371994342292, "percentage": 86.46, "elapsed_time": "1:59:16", "remaining_time": "0:18:41", "throughput": 5166.72, "total_tokens": 36975008} {"current_steps": 36680, "total_steps": 42420, "loss": 0.3506, "lr": 5.476415588644568e-05, "epoch": 17.293729372937293, "percentage": 86.47, "elapsed_time": "1:59:17", "remaining_time": "0:18:40", "throughput": 5166.74, "total_tokens": 36979712} {"current_steps": 36685, "total_steps": 42420, "loss": 0.3404, "lr": 5.4670582982156605e-05, "epoch": 17.296086751532297, "percentage": 86.48, "elapsed_time": "1:59:18", "remaining_time": "0:18:39", "throughput": 5166.77, "total_tokens": 36984960} {"current_steps": 36690, "total_steps": 42420, "loss": 0.3408, "lr": 5.457708546457657e-05, "epoch": 17.298444130127297, "percentage": 86.49, "elapsed_time": "1:59:19", "remaining_time": "0:18:38", "throughput": 5166.75, "total_tokens": 36989120} {"current_steps": 36695, "total_steps": 42420, "loss": 0.3415, "lr": 5.448366334953314e-05, "epoch": 17.3008015087223, "percentage": 86.5, "elapsed_time": "1:59:19", "remaining_time": "0:18:37", "throughput": 5166.76, "total_tokens": 36993792} {"current_steps": 36700, "total_steps": 42420, "loss": 0.3268, "lr": 5.439031665284105e-05, "epoch": 17.303158887317302, "percentage": 86.52, "elapsed_time": "1:59:20", "remaining_time": "0:18:36", "throughput": 5166.75, "total_tokens": 36997824} {"current_steps": 36705, "total_steps": 42420, "loss": 0.3236, "lr": 5.4297045390302335e-05, "epoch": 17.305516265912306, "percentage": 86.53, "elapsed_time": "1:59:21", "remaining_time": "0:18:35", "throughput": 5166.8, "total_tokens": 37002912} {"current_steps": 36710, "total_steps": 42420, "loss": 0.3372, "lr": 5.420384957770624e-05, "epoch": 17.307873644507307, "percentage": 86.54, "elapsed_time": "1:59:22", "remaining_time": "0:18:34", "throughput": 5166.8, "total_tokens": 37007296} {"current_steps": 36715, "total_steps": 42420, "loss": 0.3327, "lr": 5.411072923082927e-05, "epoch": 17.31023102310231, "percentage": 86.55, "elapsed_time": "1:59:23", "remaining_time": "0:18:33", "throughput": 5166.86, "total_tokens": 37012736} {"current_steps": 36720, "total_steps": 42420, "loss": 0.351, "lr": 5.4017684365435024e-05, "epoch": 17.31258840169731, "percentage": 86.56, "elapsed_time": "1:59:24", "remaining_time": "0:18:32", "throughput": 5166.87, "total_tokens": 37017280} {"current_steps": 36725, "total_steps": 42420, "loss": 0.3182, "lr": 5.392471499727447e-05, "epoch": 17.314945780292316, "percentage": 86.57, "elapsed_time": "1:59:25", "remaining_time": "0:18:31", "throughput": 5166.85, "total_tokens": 37021344} {"current_steps": 36730, "total_steps": 42420, "loss": 0.3776, "lr": 5.383182114208574e-05, "epoch": 17.317303158887317, "percentage": 86.59, "elapsed_time": "1:59:26", "remaining_time": "0:18:30", "throughput": 5166.85, "total_tokens": 37025728} {"current_steps": 36735, "total_steps": 42420, "loss": 0.2974, "lr": 5.373900281559413e-05, "epoch": 17.31966053748232, "percentage": 86.6, "elapsed_time": "1:59:26", "remaining_time": "0:18:29", "throughput": 5166.87, "total_tokens": 37030624} {"current_steps": 36740, "total_steps": 42420, "loss": 0.3101, "lr": 5.364626003351231e-05, "epoch": 17.32201791607732, "percentage": 86.61, "elapsed_time": "1:59:27", "remaining_time": "0:18:28", "throughput": 5166.9, "total_tokens": 37035680} {"current_steps": 36745, "total_steps": 42420, "loss": 0.3174, "lr": 5.355359281153998e-05, "epoch": 17.324375294672326, "percentage": 86.62, "elapsed_time": "1:59:28", "remaining_time": "0:18:27", "throughput": 5166.91, "total_tokens": 37040288} {"current_steps": 36750, "total_steps": 42420, "loss": 0.3602, "lr": 5.3461001165364234e-05, "epoch": 17.326732673267326, "percentage": 86.63, "elapsed_time": "1:59:29", "remaining_time": "0:18:26", "throughput": 5166.95, "total_tokens": 37045344} {"current_steps": 36755, "total_steps": 42420, "loss": 0.3173, "lr": 5.3368485110659225e-05, "epoch": 17.32909005186233, "percentage": 86.65, "elapsed_time": "1:59:30", "remaining_time": "0:18:25", "throughput": 5167.01, "total_tokens": 37050784} {"current_steps": 36760, "total_steps": 42420, "loss": 0.3352, "lr": 5.327604466308622e-05, "epoch": 17.33144743045733, "percentage": 86.66, "elapsed_time": "1:59:31", "remaining_time": "0:18:24", "throughput": 5167.03, "total_tokens": 37055456} {"current_steps": 36765, "total_steps": 42420, "loss": 0.2982, "lr": 5.318367983829392e-05, "epoch": 17.333804809052335, "percentage": 86.67, "elapsed_time": "1:59:32", "remaining_time": "0:18:23", "throughput": 5167.07, "total_tokens": 37060576} {"current_steps": 36770, "total_steps": 42420, "loss": 0.3542, "lr": 5.309139065191809e-05, "epoch": 17.336162187647336, "percentage": 86.68, "elapsed_time": "1:59:33", "remaining_time": "0:18:22", "throughput": 5167.08, "total_tokens": 37065024} {"current_steps": 36775, "total_steps": 42420, "loss": 0.3049, "lr": 5.29991771195818e-05, "epoch": 17.33851956624234, "percentage": 86.69, "elapsed_time": "1:59:34", "remaining_time": "0:18:21", "throughput": 5167.12, "total_tokens": 37070304} {"current_steps": 36780, "total_steps": 42420, "loss": 0.3047, "lr": 5.290703925689516e-05, "epoch": 17.34087694483734, "percentage": 86.7, "elapsed_time": "1:59:35", "remaining_time": "0:18:20", "throughput": 5167.14, "total_tokens": 37074912} {"current_steps": 36785, "total_steps": 42420, "loss": 0.2736, "lr": 5.281497707945554e-05, "epoch": 17.343234323432345, "percentage": 86.72, "elapsed_time": "1:59:36", "remaining_time": "0:18:19", "throughput": 5167.19, "total_tokens": 37080160} {"current_steps": 36790, "total_steps": 42420, "loss": 0.2998, "lr": 5.2722990602847574e-05, "epoch": 17.345591702027345, "percentage": 86.73, "elapsed_time": "1:59:37", "remaining_time": "0:18:18", "throughput": 5167.28, "total_tokens": 37085952} {"current_steps": 36795, "total_steps": 42420, "loss": 0.276, "lr": 5.2631079842643034e-05, "epoch": 17.34794908062235, "percentage": 86.74, "elapsed_time": "1:59:37", "remaining_time": "0:18:17", "throughput": 5167.27, "total_tokens": 37090400} {"current_steps": 36800, "total_steps": 42420, "loss": 0.2461, "lr": 5.25392448144007e-05, "epoch": 17.35030645921735, "percentage": 86.75, "elapsed_time": "1:59:38", "remaining_time": "0:18:16", "throughput": 5167.24, "total_tokens": 37094368} {"current_steps": 36805, "total_steps": 42420, "loss": 0.3885, "lr": 5.244748553366674e-05, "epoch": 17.352663837812354, "percentage": 86.76, "elapsed_time": "1:59:39", "remaining_time": "0:18:15", "throughput": 5167.33, "total_tokens": 37100320} {"current_steps": 36810, "total_steps": 42420, "loss": 0.334, "lr": 5.235580201597451e-05, "epoch": 17.355021216407355, "percentage": 86.78, "elapsed_time": "1:59:40", "remaining_time": "0:18:14", "throughput": 5167.31, "total_tokens": 37104384} {"current_steps": 36815, "total_steps": 42420, "loss": 0.2475, "lr": 5.226419427684442e-05, "epoch": 17.35737859500236, "percentage": 86.79, "elapsed_time": "1:59:41", "remaining_time": "0:18:13", "throughput": 5167.31, "total_tokens": 37108800} {"current_steps": 36820, "total_steps": 42420, "loss": 0.3857, "lr": 5.2172662331784085e-05, "epoch": 17.35973597359736, "percentage": 86.8, "elapsed_time": "1:59:42", "remaining_time": "0:18:12", "throughput": 5167.36, "total_tokens": 37114144} {"current_steps": 36825, "total_steps": 42420, "loss": 0.3435, "lr": 5.208120619628842e-05, "epoch": 17.36209335219236, "percentage": 86.81, "elapsed_time": "1:59:43", "remaining_time": "0:18:11", "throughput": 5167.36, "total_tokens": 37118400} {"current_steps": 36830, "total_steps": 42420, "loss": 0.375, "lr": 5.198982588583922e-05, "epoch": 17.364450730787365, "percentage": 86.82, "elapsed_time": "1:59:44", "remaining_time": "0:18:10", "throughput": 5167.42, "total_tokens": 37124064} {"current_steps": 36835, "total_steps": 42420, "loss": 0.3811, "lr": 5.18985214159059e-05, "epoch": 17.366808109382365, "percentage": 86.83, "elapsed_time": "1:59:45", "remaining_time": "0:18:09", "throughput": 5167.43, "total_tokens": 37128512} {"current_steps": 36840, "total_steps": 42420, "loss": 0.3056, "lr": 5.180729280194441e-05, "epoch": 17.36916548797737, "percentage": 86.85, "elapsed_time": "1:59:46", "remaining_time": "0:18:08", "throughput": 5167.46, "total_tokens": 37133408} {"current_steps": 36845, "total_steps": 42420, "loss": 0.341, "lr": 5.1716140059398396e-05, "epoch": 17.37152286657237, "percentage": 86.86, "elapsed_time": "1:59:47", "remaining_time": "0:18:07", "throughput": 5167.53, "total_tokens": 37139424} {"current_steps": 36850, "total_steps": 42420, "loss": 0.3192, "lr": 5.1625063203698475e-05, "epoch": 17.373880245167374, "percentage": 86.87, "elapsed_time": "1:59:47", "remaining_time": "0:18:06", "throughput": 5167.55, "total_tokens": 37144320} {"current_steps": 36855, "total_steps": 42420, "loss": 0.3215, "lr": 5.153406225026236e-05, "epoch": 17.376237623762375, "percentage": 86.88, "elapsed_time": "1:59:48", "remaining_time": "0:18:05", "throughput": 5167.58, "total_tokens": 37149120} {"current_steps": 36860, "total_steps": 42420, "loss": 0.3548, "lr": 5.144313721449495e-05, "epoch": 17.37859500235738, "percentage": 86.89, "elapsed_time": "1:59:49", "remaining_time": "0:18:04", "throughput": 5167.61, "total_tokens": 37154176} {"current_steps": 36865, "total_steps": 42420, "loss": 0.3359, "lr": 5.1352288111788495e-05, "epoch": 17.38095238095238, "percentage": 86.9, "elapsed_time": "1:59:50", "remaining_time": "0:18:03", "throughput": 5167.67, "total_tokens": 37159872} {"current_steps": 36870, "total_steps": 42420, "loss": 0.3256, "lr": 5.126151495752196e-05, "epoch": 17.383309759547384, "percentage": 86.92, "elapsed_time": "1:59:51", "remaining_time": "0:18:02", "throughput": 5167.72, "total_tokens": 37165024} {"current_steps": 36875, "total_steps": 42420, "loss": 0.4153, "lr": 5.117081776706178e-05, "epoch": 17.385667138142384, "percentage": 86.93, "elapsed_time": "1:59:52", "remaining_time": "0:18:01", "throughput": 5167.74, "total_tokens": 37169664} {"current_steps": 36880, "total_steps": 42420, "loss": 0.3353, "lr": 5.108019655576146e-05, "epoch": 17.38802451673739, "percentage": 86.94, "elapsed_time": "1:59:53", "remaining_time": "0:18:00", "throughput": 5167.79, "total_tokens": 37174880} {"current_steps": 36885, "total_steps": 42420, "loss": 0.2894, "lr": 5.098965133896172e-05, "epoch": 17.39038189533239, "percentage": 86.95, "elapsed_time": "1:59:54", "remaining_time": "0:17:59", "throughput": 5167.85, "total_tokens": 37180896} {"current_steps": 36890, "total_steps": 42420, "loss": 0.3165, "lr": 5.089918213199018e-05, "epoch": 17.392739273927393, "percentage": 86.96, "elapsed_time": "1:59:55", "remaining_time": "0:17:58", "throughput": 5167.93, "total_tokens": 37186496} {"current_steps": 36895, "total_steps": 42420, "loss": 0.3473, "lr": 5.080878895016178e-05, "epoch": 17.395096652522394, "percentage": 86.98, "elapsed_time": "1:59:56", "remaining_time": "0:17:57", "throughput": 5167.98, "total_tokens": 37191808} {"current_steps": 36900, "total_steps": 42420, "loss": 0.3379, "lr": 5.071847180877864e-05, "epoch": 17.397454031117398, "percentage": 86.99, "elapsed_time": "1:59:57", "remaining_time": "0:17:56", "throughput": 5167.97, "total_tokens": 37195936} {"current_steps": 36905, "total_steps": 42420, "loss": 0.3511, "lr": 5.0628230723129774e-05, "epoch": 17.3998114097124, "percentage": 87.0, "elapsed_time": "1:59:58", "remaining_time": "0:17:55", "throughput": 5167.97, "total_tokens": 37200704} {"current_steps": 36910, "total_steps": 42420, "loss": 0.3341, "lr": 5.053806570849162e-05, "epoch": 17.402168788307403, "percentage": 87.01, "elapsed_time": "1:59:59", "remaining_time": "0:17:54", "throughput": 5168.04, "total_tokens": 37206144} {"current_steps": 36915, "total_steps": 42420, "loss": 0.3401, "lr": 5.044797678012747e-05, "epoch": 17.404526166902404, "percentage": 87.02, "elapsed_time": "2:00:00", "remaining_time": "0:17:53", "throughput": 5168.07, "total_tokens": 37210880} {"current_steps": 36920, "total_steps": 42420, "loss": 0.3728, "lr": 5.03579639532879e-05, "epoch": 17.406883545497408, "percentage": 87.03, "elapsed_time": "2:00:00", "remaining_time": "0:17:52", "throughput": 5168.05, "total_tokens": 37215104} {"current_steps": 36925, "total_steps": 42420, "loss": 0.3182, "lr": 5.026802724321061e-05, "epoch": 17.40924092409241, "percentage": 87.05, "elapsed_time": "2:00:01", "remaining_time": "0:17:51", "throughput": 5168.1, "total_tokens": 37220352} {"current_steps": 36930, "total_steps": 42420, "loss": 0.2975, "lr": 5.017816666512026e-05, "epoch": 17.411598302687413, "percentage": 87.06, "elapsed_time": "2:00:02", "remaining_time": "0:17:50", "throughput": 5168.09, "total_tokens": 37224448} {"current_steps": 36935, "total_steps": 42420, "loss": 0.3399, "lr": 5.008838223422874e-05, "epoch": 17.413955681282413, "percentage": 87.07, "elapsed_time": "2:00:03", "remaining_time": "0:17:49", "throughput": 5168.09, "total_tokens": 37229088} {"current_steps": 36940, "total_steps": 42420, "loss": 0.3319, "lr": 4.9998673965734985e-05, "epoch": 17.416313059877417, "percentage": 87.08, "elapsed_time": "2:00:04", "remaining_time": "0:17:48", "throughput": 5168.13, "total_tokens": 37233792} {"current_steps": 36945, "total_steps": 42420, "loss": 0.3566, "lr": 4.99090418748252e-05, "epoch": 17.418670438472418, "percentage": 87.09, "elapsed_time": "2:00:05", "remaining_time": "0:17:47", "throughput": 5168.17, "total_tokens": 37238784} {"current_steps": 36950, "total_steps": 42420, "loss": 0.4189, "lr": 4.981948597667252e-05, "epoch": 17.421027817067422, "percentage": 87.11, "elapsed_time": "2:00:06", "remaining_time": "0:17:46", "throughput": 5168.19, "total_tokens": 37243552} {"current_steps": 36955, "total_steps": 42420, "loss": 0.3273, "lr": 4.9730006286437205e-05, "epoch": 17.423385195662423, "percentage": 87.12, "elapsed_time": "2:00:07", "remaining_time": "0:17:45", "throughput": 5168.25, "total_tokens": 37249056} {"current_steps": 36960, "total_steps": 42420, "loss": 0.3414, "lr": 4.964060281926669e-05, "epoch": 17.425742574257427, "percentage": 87.13, "elapsed_time": "2:00:08", "remaining_time": "0:17:44", "throughput": 5168.29, "total_tokens": 37254272} {"current_steps": 36965, "total_steps": 42420, "loss": 0.3569, "lr": 4.955127559029554e-05, "epoch": 17.428099952852428, "percentage": 87.14, "elapsed_time": "2:00:09", "remaining_time": "0:17:43", "throughput": 5168.29, "total_tokens": 37258656} {"current_steps": 36970, "total_steps": 42420, "loss": 0.3131, "lr": 4.946202461464516e-05, "epoch": 17.430457331447432, "percentage": 87.15, "elapsed_time": "2:00:10", "remaining_time": "0:17:42", "throughput": 5168.3, "total_tokens": 37263520} {"current_steps": 36975, "total_steps": 42420, "loss": 0.3449, "lr": 4.93728499074243e-05, "epoch": 17.432814710042432, "percentage": 87.16, "elapsed_time": "2:00:10", "remaining_time": "0:17:41", "throughput": 5168.36, "total_tokens": 37269024} {"current_steps": 36980, "total_steps": 42420, "loss": 0.381, "lr": 4.928375148372871e-05, "epoch": 17.435172088637437, "percentage": 87.18, "elapsed_time": "2:00:11", "remaining_time": "0:17:40", "throughput": 5168.37, "total_tokens": 37273376} {"current_steps": 36985, "total_steps": 42420, "loss": 0.3647, "lr": 4.919472935864133e-05, "epoch": 17.437529467232437, "percentage": 87.19, "elapsed_time": "2:00:12", "remaining_time": "0:17:39", "throughput": 5168.48, "total_tokens": 37279776} {"current_steps": 36990, "total_steps": 42420, "loss": 0.3025, "lr": 4.9105783547232e-05, "epoch": 17.43988684582744, "percentage": 87.2, "elapsed_time": "2:00:13", "remaining_time": "0:17:38", "throughput": 5168.48, "total_tokens": 37283968} {"current_steps": 36995, "total_steps": 42420, "loss": 0.3168, "lr": 4.9016914064557775e-05, "epoch": 17.442244224422442, "percentage": 87.21, "elapsed_time": "2:00:14", "remaining_time": "0:17:37", "throughput": 5168.56, "total_tokens": 37289792} {"current_steps": 37000, "total_steps": 42420, "loss": 0.3099, "lr": 4.8928120925662736e-05, "epoch": 17.444601603017446, "percentage": 87.22, "elapsed_time": "2:00:15", "remaining_time": "0:17:36", "throughput": 5168.58, "total_tokens": 37294624} {"current_steps": 37005, "total_steps": 42420, "loss": 0.3165, "lr": 4.883940414557814e-05, "epoch": 17.446958981612447, "percentage": 87.23, "elapsed_time": "2:00:16", "remaining_time": "0:17:36", "throughput": 5168.65, "total_tokens": 37300320} {"current_steps": 37010, "total_steps": 42420, "loss": 0.3659, "lr": 4.875076373932208e-05, "epoch": 17.44931636020745, "percentage": 87.25, "elapsed_time": "2:00:17", "remaining_time": "0:17:35", "throughput": 5168.63, "total_tokens": 37304160} {"current_steps": 37015, "total_steps": 42420, "loss": 0.3408, "lr": 4.8662199721899944e-05, "epoch": 17.45167373880245, "percentage": 87.26, "elapsed_time": "2:00:18", "remaining_time": "0:17:34", "throughput": 5168.68, "total_tokens": 37309632} {"current_steps": 37020, "total_steps": 42420, "loss": 0.3277, "lr": 4.857371210830408e-05, "epoch": 17.454031117397456, "percentage": 87.27, "elapsed_time": "2:00:19", "remaining_time": "0:17:33", "throughput": 5168.7, "total_tokens": 37314336} {"current_steps": 37025, "total_steps": 42420, "loss": 0.3035, "lr": 4.848530091351405e-05, "epoch": 17.456388495992456, "percentage": 87.28, "elapsed_time": "2:00:20", "remaining_time": "0:17:32", "throughput": 5168.8, "total_tokens": 37320704} {"current_steps": 37030, "total_steps": 42420, "loss": 0.36, "lr": 4.8396966152496225e-05, "epoch": 17.458745874587457, "percentage": 87.29, "elapsed_time": "2:00:21", "remaining_time": "0:17:31", "throughput": 5168.82, "total_tokens": 37325888} {"current_steps": 37035, "total_steps": 42420, "loss": 0.3874, "lr": 4.830870784020425e-05, "epoch": 17.46110325318246, "percentage": 87.31, "elapsed_time": "2:00:22", "remaining_time": "0:17:30", "throughput": 5168.89, "total_tokens": 37331616} {"current_steps": 37040, "total_steps": 42420, "loss": 0.3704, "lr": 4.822052599157878e-05, "epoch": 17.463460631777462, "percentage": 87.32, "elapsed_time": "2:00:23", "remaining_time": "0:17:29", "throughput": 5168.93, "total_tokens": 37336672} {"current_steps": 37045, "total_steps": 42420, "loss": 0.347, "lr": 4.813242062154755e-05, "epoch": 17.465818010372466, "percentage": 87.33, "elapsed_time": "2:00:24", "remaining_time": "0:17:28", "throughput": 5168.96, "total_tokens": 37341312} {"current_steps": 37050, "total_steps": 42420, "loss": 0.338, "lr": 4.8044391745025175e-05, "epoch": 17.468175388967467, "percentage": 87.34, "elapsed_time": "2:00:25", "remaining_time": "0:17:27", "throughput": 5169.06, "total_tokens": 37347488} {"current_steps": 37055, "total_steps": 42420, "loss": 0.311, "lr": 4.795643937691352e-05, "epoch": 17.47053276756247, "percentage": 87.35, "elapsed_time": "2:00:26", "remaining_time": "0:17:26", "throughput": 5169.06, "total_tokens": 37351904} {"current_steps": 37060, "total_steps": 42420, "loss": 0.3127, "lr": 4.786856353210134e-05, "epoch": 17.47289014615747, "percentage": 87.36, "elapsed_time": "2:00:26", "remaining_time": "0:17:25", "throughput": 5169.07, "total_tokens": 37356544} {"current_steps": 37065, "total_steps": 42420, "loss": 0.2942, "lr": 4.7780764225464676e-05, "epoch": 17.475247524752476, "percentage": 87.38, "elapsed_time": "2:00:27", "remaining_time": "0:17:24", "throughput": 5169.11, "total_tokens": 37361728} {"current_steps": 37070, "total_steps": 42420, "loss": 0.3094, "lr": 4.769304147186632e-05, "epoch": 17.477604903347476, "percentage": 87.39, "elapsed_time": "2:00:28", "remaining_time": "0:17:23", "throughput": 5169.14, "total_tokens": 37366592} {"current_steps": 37075, "total_steps": 42420, "loss": 0.3645, "lr": 4.760539528615637e-05, "epoch": 17.47996228194248, "percentage": 87.4, "elapsed_time": "2:00:29", "remaining_time": "0:17:22", "throughput": 5169.15, "total_tokens": 37371456} {"current_steps": 37080, "total_steps": 42420, "loss": 0.3764, "lr": 4.7517825683171756e-05, "epoch": 17.48231966053748, "percentage": 87.41, "elapsed_time": "2:00:30", "remaining_time": "0:17:21", "throughput": 5169.19, "total_tokens": 37376640} {"current_steps": 37085, "total_steps": 42420, "loss": 0.3481, "lr": 4.743033267773661e-05, "epoch": 17.484677039132485, "percentage": 87.42, "elapsed_time": "2:00:31", "remaining_time": "0:17:20", "throughput": 5169.24, "total_tokens": 37382112} {"current_steps": 37090, "total_steps": 42420, "loss": 0.3831, "lr": 4.734291628466186e-05, "epoch": 17.487034417727486, "percentage": 87.44, "elapsed_time": "2:00:32", "remaining_time": "0:17:19", "throughput": 5169.24, "total_tokens": 37386336} {"current_steps": 37095, "total_steps": 42420, "loss": 0.3221, "lr": 4.725557651874579e-05, "epoch": 17.48939179632249, "percentage": 87.45, "elapsed_time": "2:00:33", "remaining_time": "0:17:18", "throughput": 5169.32, "total_tokens": 37392320} {"current_steps": 37100, "total_steps": 42420, "loss": 0.3462, "lr": 4.7168313394773396e-05, "epoch": 17.49174917491749, "percentage": 87.46, "elapsed_time": "2:00:34", "remaining_time": "0:17:17", "throughput": 5169.29, "total_tokens": 37396256} {"current_steps": 37105, "total_steps": 42420, "loss": 0.3694, "lr": 4.708112692751687e-05, "epoch": 17.494106553512495, "percentage": 87.47, "elapsed_time": "2:00:35", "remaining_time": "0:17:16", "throughput": 5169.27, "total_tokens": 37400512} {"current_steps": 37110, "total_steps": 42420, "loss": 0.3042, "lr": 4.6994017131735454e-05, "epoch": 17.496463932107496, "percentage": 87.48, "elapsed_time": "2:00:36", "remaining_time": "0:17:15", "throughput": 5169.29, "total_tokens": 37405216} {"current_steps": 37115, "total_steps": 42420, "loss": 0.3356, "lr": 4.69069840221753e-05, "epoch": 17.4988213107025, "percentage": 87.49, "elapsed_time": "2:00:36", "remaining_time": "0:17:14", "throughput": 5169.34, "total_tokens": 37410400} {"current_steps": 37120, "total_steps": 42420, "loss": 0.278, "lr": 4.6820027613569684e-05, "epoch": 17.5011786892975, "percentage": 87.51, "elapsed_time": "2:00:37", "remaining_time": "0:17:13", "throughput": 5169.34, "total_tokens": 37415040} {"current_steps": 37125, "total_steps": 42420, "loss": 0.3385, "lr": 4.6733147920638874e-05, "epoch": 17.503536067892504, "percentage": 87.52, "elapsed_time": "2:00:38", "remaining_time": "0:17:12", "throughput": 5169.37, "total_tokens": 37419840} {"current_steps": 37130, "total_steps": 42420, "loss": 0.3535, "lr": 4.664634495809006e-05, "epoch": 17.505893446487505, "percentage": 87.53, "elapsed_time": "2:00:39", "remaining_time": "0:17:11", "throughput": 5169.37, "total_tokens": 37424352} {"current_steps": 37135, "total_steps": 42420, "loss": 0.3795, "lr": 4.655961874061759e-05, "epoch": 17.50825082508251, "percentage": 87.54, "elapsed_time": "2:00:40", "remaining_time": "0:17:10", "throughput": 5169.43, "total_tokens": 37430144} {"current_steps": 37140, "total_steps": 42420, "loss": 0.3238, "lr": 4.6472969282902623e-05, "epoch": 17.51060820367751, "percentage": 87.55, "elapsed_time": "2:00:41", "remaining_time": "0:17:09", "throughput": 5169.39, "total_tokens": 37433760} {"current_steps": 37145, "total_steps": 42420, "loss": 0.3557, "lr": 4.638639659961352e-05, "epoch": 17.512965582272514, "percentage": 87.56, "elapsed_time": "2:00:42", "remaining_time": "0:17:08", "throughput": 5169.39, "total_tokens": 37438144} {"current_steps": 37150, "total_steps": 42420, "loss": 0.3316, "lr": 4.6299900705405576e-05, "epoch": 17.515322960867515, "percentage": 87.58, "elapsed_time": "2:00:43", "remaining_time": "0:17:07", "throughput": 5169.41, "total_tokens": 37442944} {"current_steps": 37155, "total_steps": 42420, "loss": 0.3738, "lr": 4.6213481614921005e-05, "epoch": 17.51768033946252, "percentage": 87.59, "elapsed_time": "2:00:44", "remaining_time": "0:17:06", "throughput": 5169.45, "total_tokens": 37448096} {"current_steps": 37160, "total_steps": 42420, "loss": 0.3122, "lr": 4.612713934278917e-05, "epoch": 17.52003771805752, "percentage": 87.6, "elapsed_time": "2:00:45", "remaining_time": "0:17:05", "throughput": 5169.48, "total_tokens": 37453152} {"current_steps": 37165, "total_steps": 42420, "loss": 0.3342, "lr": 4.604087390362627e-05, "epoch": 17.522395096652524, "percentage": 87.61, "elapsed_time": "2:00:46", "remaining_time": "0:17:04", "throughput": 5169.53, "total_tokens": 37458784} {"current_steps": 37170, "total_steps": 42420, "loss": 0.3058, "lr": 4.595468531203567e-05, "epoch": 17.524752475247524, "percentage": 87.62, "elapsed_time": "2:00:47", "remaining_time": "0:17:03", "throughput": 5169.64, "total_tokens": 37465568} {"current_steps": 37175, "total_steps": 42420, "loss": 0.3126, "lr": 4.586857358260771e-05, "epoch": 17.52710985384253, "percentage": 87.64, "elapsed_time": "2:00:48", "remaining_time": "0:17:02", "throughput": 5169.64, "total_tokens": 37470048} {"current_steps": 37180, "total_steps": 42420, "loss": 0.3129, "lr": 4.578253872991939e-05, "epoch": 17.52946723243753, "percentage": 87.65, "elapsed_time": "2:00:48", "remaining_time": "0:17:01", "throughput": 5169.69, "total_tokens": 37475104} {"current_steps": 37185, "total_steps": 42420, "loss": 0.3366, "lr": 4.5696580768535115e-05, "epoch": 17.531824611032533, "percentage": 87.66, "elapsed_time": "2:00:50", "remaining_time": "0:17:00", "throughput": 5169.78, "total_tokens": 37481216} {"current_steps": 37190, "total_steps": 42420, "loss": 0.2912, "lr": 4.5610699713006074e-05, "epoch": 17.534181989627534, "percentage": 87.67, "elapsed_time": "2:00:50", "remaining_time": "0:16:59", "throughput": 5169.79, "total_tokens": 37485536} {"current_steps": 37195, "total_steps": 42420, "loss": 0.3488, "lr": 4.552489557787048e-05, "epoch": 17.536539368222538, "percentage": 87.68, "elapsed_time": "2:00:51", "remaining_time": "0:16:58", "throughput": 5169.82, "total_tokens": 37490656} {"current_steps": 37200, "total_steps": 42420, "loss": 0.3682, "lr": 4.543916837765355e-05, "epoch": 17.53889674681754, "percentage": 87.69, "elapsed_time": "2:00:52", "remaining_time": "0:16:57", "throughput": 5169.84, "total_tokens": 37495200} {"current_steps": 37205, "total_steps": 42420, "loss": 0.3397, "lr": 4.535351812686739e-05, "epoch": 17.541254125412543, "percentage": 87.71, "elapsed_time": "2:00:53", "remaining_time": "0:16:56", "throughput": 5169.86, "total_tokens": 37499776} {"current_steps": 37210, "total_steps": 42420, "loss": 0.3044, "lr": 4.526794484001112e-05, "epoch": 17.543611504007544, "percentage": 87.72, "elapsed_time": "2:00:54", "remaining_time": "0:16:55", "throughput": 5169.87, "total_tokens": 37504256} {"current_steps": 37215, "total_steps": 42420, "loss": 0.3252, "lr": 4.518244853157105e-05, "epoch": 17.545968882602544, "percentage": 87.73, "elapsed_time": "2:00:55", "remaining_time": "0:16:54", "throughput": 5169.88, "total_tokens": 37508768} {"current_steps": 37220, "total_steps": 42420, "loss": 0.3094, "lr": 4.5097029216019935e-05, "epoch": 17.54832626119755, "percentage": 87.74, "elapsed_time": "2:00:56", "remaining_time": "0:16:53", "throughput": 5169.93, "total_tokens": 37513760} {"current_steps": 37225, "total_steps": 42420, "loss": 0.334, "lr": 4.5011686907817925e-05, "epoch": 17.55068363979255, "percentage": 87.75, "elapsed_time": "2:00:57", "remaining_time": "0:16:52", "throughput": 5169.94, "total_tokens": 37518464} {"current_steps": 37230, "total_steps": 42420, "loss": 0.3319, "lr": 4.492642162141214e-05, "epoch": 17.553041018387553, "percentage": 87.77, "elapsed_time": "2:00:57", "remaining_time": "0:16:51", "throughput": 5169.97, "total_tokens": 37523328} {"current_steps": 37235, "total_steps": 42420, "loss": 0.3415, "lr": 4.484123337123641e-05, "epoch": 17.555398396982554, "percentage": 87.78, "elapsed_time": "2:00:58", "remaining_time": "0:16:50", "throughput": 5170.01, "total_tokens": 37528416} {"current_steps": 37240, "total_steps": 42420, "loss": 0.3127, "lr": 4.475612217171171e-05, "epoch": 17.557755775577558, "percentage": 87.79, "elapsed_time": "2:00:59", "remaining_time": "0:16:49", "throughput": 5170.06, "total_tokens": 37533696} {"current_steps": 37245, "total_steps": 42420, "loss": 0.3072, "lr": 4.467108803724596e-05, "epoch": 17.56011315417256, "percentage": 87.8, "elapsed_time": "2:01:00", "remaining_time": "0:16:48", "throughput": 5170.15, "total_tokens": 37539616} {"current_steps": 37250, "total_steps": 42420, "loss": 0.304, "lr": 4.4586130982233905e-05, "epoch": 17.562470532767563, "percentage": 87.81, "elapsed_time": "2:01:01", "remaining_time": "0:16:47", "throughput": 5170.13, "total_tokens": 37543776} {"current_steps": 37255, "total_steps": 42420, "loss": 0.3599, "lr": 4.4501251021057564e-05, "epoch": 17.564827911362563, "percentage": 87.82, "elapsed_time": "2:01:02", "remaining_time": "0:16:46", "throughput": 5170.16, "total_tokens": 37548672} {"current_steps": 37260, "total_steps": 42420, "loss": 0.3702, "lr": 4.441644816808532e-05, "epoch": 17.567185289957568, "percentage": 87.84, "elapsed_time": "2:01:03", "remaining_time": "0:16:45", "throughput": 5170.21, "total_tokens": 37553920} {"current_steps": 37265, "total_steps": 42420, "loss": 0.374, "lr": 4.4331722437673086e-05, "epoch": 17.569542668552568, "percentage": 87.85, "elapsed_time": "2:01:04", "remaining_time": "0:16:44", "throughput": 5170.22, "total_tokens": 37558912} {"current_steps": 37270, "total_steps": 42420, "loss": 0.3686, "lr": 4.4247073844163435e-05, "epoch": 17.571900047147572, "percentage": 87.86, "elapsed_time": "2:01:05", "remaining_time": "0:16:43", "throughput": 5170.22, "total_tokens": 37563200} {"current_steps": 37275, "total_steps": 42420, "loss": 0.3206, "lr": 4.416250240188591e-05, "epoch": 17.574257425742573, "percentage": 87.87, "elapsed_time": "2:01:06", "remaining_time": "0:16:42", "throughput": 5170.23, "total_tokens": 37567648} {"current_steps": 37280, "total_steps": 42420, "loss": 0.2731, "lr": 4.4078008125157055e-05, "epoch": 17.576614804337577, "percentage": 87.88, "elapsed_time": "2:01:07", "remaining_time": "0:16:41", "throughput": 5170.27, "total_tokens": 37572608} {"current_steps": 37285, "total_steps": 42420, "loss": 0.3232, "lr": 4.399359102828032e-05, "epoch": 17.578972182932578, "percentage": 87.89, "elapsed_time": "2:01:07", "remaining_time": "0:16:40", "throughput": 5170.31, "total_tokens": 37577760} {"current_steps": 37290, "total_steps": 42420, "loss": 0.3248, "lr": 4.3909251125546054e-05, "epoch": 17.581329561527582, "percentage": 87.91, "elapsed_time": "2:01:08", "remaining_time": "0:16:39", "throughput": 5170.35, "total_tokens": 37582624} {"current_steps": 37295, "total_steps": 42420, "loss": 0.317, "lr": 4.3824988431231726e-05, "epoch": 17.583686940122583, "percentage": 87.92, "elapsed_time": "2:01:09", "remaining_time": "0:16:39", "throughput": 5170.41, "total_tokens": 37588096} {"current_steps": 37300, "total_steps": 42420, "loss": 0.4173, "lr": 4.3740802959601314e-05, "epoch": 17.586044318717587, "percentage": 87.93, "elapsed_time": "2:01:10", "remaining_time": "0:16:38", "throughput": 5170.42, "total_tokens": 37592768} {"current_steps": 37305, "total_steps": 42420, "loss": 0.329, "lr": 4.365669472490619e-05, "epoch": 17.588401697312587, "percentage": 87.94, "elapsed_time": "2:01:11", "remaining_time": "0:16:37", "throughput": 5170.45, "total_tokens": 37597888} {"current_steps": 37310, "total_steps": 42420, "loss": 0.327, "lr": 4.35726637413843e-05, "epoch": 17.59075907590759, "percentage": 87.95, "elapsed_time": "2:01:12", "remaining_time": "0:16:36", "throughput": 5170.5, "total_tokens": 37603424} {"current_steps": 37315, "total_steps": 42420, "loss": 0.3422, "lr": 4.348871002326088e-05, "epoch": 17.593116454502592, "percentage": 87.97, "elapsed_time": "2:01:13", "remaining_time": "0:16:35", "throughput": 5170.53, "total_tokens": 37608224} {"current_steps": 37320, "total_steps": 42420, "loss": 0.3142, "lr": 4.3404833584747615e-05, "epoch": 17.595473833097596, "percentage": 87.98, "elapsed_time": "2:01:14", "remaining_time": "0:16:34", "throughput": 5170.54, "total_tokens": 37612704} {"current_steps": 37325, "total_steps": 42420, "loss": 0.3429, "lr": 4.332103444004348e-05, "epoch": 17.597831211692597, "percentage": 87.99, "elapsed_time": "2:01:15", "remaining_time": "0:16:33", "throughput": 5170.59, "total_tokens": 37617760} {"current_steps": 37330, "total_steps": 42420, "loss": 0.3144, "lr": 4.323731260333424e-05, "epoch": 17.6001885902876, "percentage": 88.0, "elapsed_time": "2:01:16", "remaining_time": "0:16:32", "throughput": 5170.61, "total_tokens": 37622592} {"current_steps": 37335, "total_steps": 42420, "loss": 0.2955, "lr": 4.315366808879262e-05, "epoch": 17.602545968882602, "percentage": 88.01, "elapsed_time": "2:01:17", "remaining_time": "0:16:31", "throughput": 5170.67, "total_tokens": 37628256} {"current_steps": 37340, "total_steps": 42420, "loss": 0.3245, "lr": 4.307010091057828e-05, "epoch": 17.604903347477606, "percentage": 88.02, "elapsed_time": "2:01:18", "remaining_time": "0:16:30", "throughput": 5170.76, "total_tokens": 37634272} {"current_steps": 37345, "total_steps": 42420, "loss": 0.3518, "lr": 4.298661108283747e-05, "epoch": 17.607260726072607, "percentage": 88.04, "elapsed_time": "2:01:19", "remaining_time": "0:16:29", "throughput": 5170.75, "total_tokens": 37638432} {"current_steps": 37350, "total_steps": 42420, "loss": 0.3448, "lr": 4.290319861970382e-05, "epoch": 17.60961810466761, "percentage": 88.05, "elapsed_time": "2:01:20", "remaining_time": "0:16:28", "throughput": 5170.81, "total_tokens": 37643968} {"current_steps": 37355, "total_steps": 42420, "loss": 0.2893, "lr": 4.28198635352976e-05, "epoch": 17.61197548326261, "percentage": 88.06, "elapsed_time": "2:01:21", "remaining_time": "0:16:27", "throughput": 5170.83, "total_tokens": 37648928} {"current_steps": 37360, "total_steps": 42420, "loss": 0.3379, "lr": 4.273660584372596e-05, "epoch": 17.614332861857616, "percentage": 88.07, "elapsed_time": "2:01:21", "remaining_time": "0:16:26", "throughput": 5170.83, "total_tokens": 37653344} {"current_steps": 37365, "total_steps": 42420, "loss": 0.3305, "lr": 4.265342555908308e-05, "epoch": 17.616690240452616, "percentage": 88.08, "elapsed_time": "2:01:22", "remaining_time": "0:16:25", "throughput": 5170.87, "total_tokens": 37658816} {"current_steps": 37370, "total_steps": 42420, "loss": 0.3395, "lr": 4.257032269544991e-05, "epoch": 17.61904761904762, "percentage": 88.1, "elapsed_time": "2:01:23", "remaining_time": "0:16:24", "throughput": 5170.87, "total_tokens": 37663072} {"current_steps": 37375, "total_steps": 42420, "loss": 0.3137, "lr": 4.248729726689443e-05, "epoch": 17.62140499764262, "percentage": 88.11, "elapsed_time": "2:01:24", "remaining_time": "0:16:23", "throughput": 5170.9, "total_tokens": 37668160} {"current_steps": 37380, "total_steps": 42420, "loss": 0.3225, "lr": 4.24043492874715e-05, "epoch": 17.623762376237625, "percentage": 88.12, "elapsed_time": "2:01:25", "remaining_time": "0:16:22", "throughput": 5170.93, "total_tokens": 37673216} {"current_steps": 37385, "total_steps": 42420, "loss": 0.3439, "lr": 4.232147877122261e-05, "epoch": 17.626119754832626, "percentage": 88.13, "elapsed_time": "2:01:26", "remaining_time": "0:16:21", "throughput": 5170.96, "total_tokens": 37678176} {"current_steps": 37390, "total_steps": 42420, "loss": 0.3077, "lr": 4.2238685732176376e-05, "epoch": 17.62847713342763, "percentage": 88.14, "elapsed_time": "2:01:27", "remaining_time": "0:16:20", "throughput": 5170.93, "total_tokens": 37682080} {"current_steps": 37395, "total_steps": 42420, "loss": 0.3029, "lr": 4.215597018434836e-05, "epoch": 17.63083451202263, "percentage": 88.15, "elapsed_time": "2:01:28", "remaining_time": "0:16:19", "throughput": 5170.93, "total_tokens": 37686560} {"current_steps": 37400, "total_steps": 42420, "loss": 0.3425, "lr": 4.20733321417408e-05, "epoch": 17.633191890617635, "percentage": 88.17, "elapsed_time": "2:01:29", "remaining_time": "0:16:18", "throughput": 5171.0, "total_tokens": 37692480} {"current_steps": 37405, "total_steps": 42420, "loss": 0.3782, "lr": 4.1990771618342974e-05, "epoch": 17.635549269212635, "percentage": 88.18, "elapsed_time": "2:01:30", "remaining_time": "0:16:17", "throughput": 5171.15, "total_tokens": 37700832} {"current_steps": 37410, "total_steps": 42420, "loss": 0.2595, "lr": 4.190828862813101e-05, "epoch": 17.63790664780764, "percentage": 88.19, "elapsed_time": "2:01:31", "remaining_time": "0:16:16", "throughput": 5171.12, "total_tokens": 37704768} {"current_steps": 37415, "total_steps": 42420, "loss": 0.295, "lr": 4.182588318506775e-05, "epoch": 17.64026402640264, "percentage": 88.2, "elapsed_time": "2:01:32", "remaining_time": "0:16:15", "throughput": 5171.17, "total_tokens": 37710272} {"current_steps": 37420, "total_steps": 42420, "loss": 0.3684, "lr": 4.1743555303103254e-05, "epoch": 17.64262140499764, "percentage": 88.21, "elapsed_time": "2:01:33", "remaining_time": "0:16:14", "throughput": 5171.2, "total_tokens": 37714976} {"current_steps": 37425, "total_steps": 42420, "loss": 0.3372, "lr": 4.166130499617399e-05, "epoch": 17.644978783592645, "percentage": 88.22, "elapsed_time": "2:01:34", "remaining_time": "0:16:13", "throughput": 5171.21, "total_tokens": 37719808} {"current_steps": 37430, "total_steps": 42420, "loss": 0.3096, "lr": 4.157913227820359e-05, "epoch": 17.647336162187646, "percentage": 88.24, "elapsed_time": "2:01:35", "remaining_time": "0:16:12", "throughput": 5171.23, "total_tokens": 37724448} {"current_steps": 37435, "total_steps": 42420, "loss": 0.3496, "lr": 4.1497037163102605e-05, "epoch": 17.64969354078265, "percentage": 88.25, "elapsed_time": "2:01:36", "remaining_time": "0:16:11", "throughput": 5171.3, "total_tokens": 37730176} {"current_steps": 37440, "total_steps": 42420, "loss": 0.3341, "lr": 4.14150196647683e-05, "epoch": 17.65205091937765, "percentage": 88.26, "elapsed_time": "2:01:37", "remaining_time": "0:16:10", "throughput": 5171.39, "total_tokens": 37736320} {"current_steps": 37445, "total_steps": 42420, "loss": 0.3438, "lr": 4.133307979708484e-05, "epoch": 17.654408297972655, "percentage": 88.27, "elapsed_time": "2:01:37", "remaining_time": "0:16:09", "throughput": 5171.4, "total_tokens": 37740800} {"current_steps": 37450, "total_steps": 42420, "loss": 0.3175, "lr": 4.1251217573923195e-05, "epoch": 17.656765676567655, "percentage": 88.28, "elapsed_time": "2:01:38", "remaining_time": "0:16:08", "throughput": 5171.4, "total_tokens": 37745088} {"current_steps": 37455, "total_steps": 42420, "loss": 0.3354, "lr": 4.116943300914133e-05, "epoch": 17.65912305516266, "percentage": 88.3, "elapsed_time": "2:01:39", "remaining_time": "0:16:07", "throughput": 5171.44, "total_tokens": 37750048} {"current_steps": 37460, "total_steps": 42420, "loss": 0.3491, "lr": 4.108772611658407e-05, "epoch": 17.66148043375766, "percentage": 88.31, "elapsed_time": "2:01:40", "remaining_time": "0:16:06", "throughput": 5171.54, "total_tokens": 37756608} {"current_steps": 37465, "total_steps": 42420, "loss": 0.3501, "lr": 4.100609691008278e-05, "epoch": 17.663837812352664, "percentage": 88.32, "elapsed_time": "2:01:41", "remaining_time": "0:16:05", "throughput": 5171.54, "total_tokens": 37761024} {"current_steps": 37470, "total_steps": 42420, "loss": 0.3234, "lr": 4.0924545403456024e-05, "epoch": 17.666195190947665, "percentage": 88.33, "elapsed_time": "2:01:42", "remaining_time": "0:16:04", "throughput": 5171.59, "total_tokens": 37766368} {"current_steps": 37475, "total_steps": 42420, "loss": 0.3425, "lr": 4.0843071610509096e-05, "epoch": 17.66855256954267, "percentage": 88.34, "elapsed_time": "2:01:43", "remaining_time": "0:16:03", "throughput": 5171.6, "total_tokens": 37771040} {"current_steps": 37480, "total_steps": 42420, "loss": 0.343, "lr": 4.0761675545034075e-05, "epoch": 17.67090994813767, "percentage": 88.35, "elapsed_time": "2:01:44", "remaining_time": "0:16:02", "throughput": 5171.66, "total_tokens": 37776672} {"current_steps": 37485, "total_steps": 42420, "loss": 0.3163, "lr": 4.068035722080993e-05, "epoch": 17.673267326732674, "percentage": 88.37, "elapsed_time": "2:01:45", "remaining_time": "0:16:01", "throughput": 5171.68, "total_tokens": 37781312} {"current_steps": 37490, "total_steps": 42420, "loss": 0.3583, "lr": 4.0599116651602544e-05, "epoch": 17.675624705327674, "percentage": 88.38, "elapsed_time": "2:01:46", "remaining_time": "0:16:00", "throughput": 5171.68, "total_tokens": 37785824} {"current_steps": 37495, "total_steps": 42420, "loss": 0.3385, "lr": 4.051795385116452e-05, "epoch": 17.67798208392268, "percentage": 88.39, "elapsed_time": "2:01:47", "remaining_time": "0:15:59", "throughput": 5171.63, "total_tokens": 37789248} {"current_steps": 37500, "total_steps": 42420, "loss": 0.323, "lr": 4.043686883323544e-05, "epoch": 17.68033946251768, "percentage": 88.4, "elapsed_time": "2:01:47", "remaining_time": "0:15:58", "throughput": 5171.63, "total_tokens": 37793440} {"current_steps": 37505, "total_steps": 42420, "loss": 0.3119, "lr": 4.035586161154142e-05, "epoch": 17.682696841112683, "percentage": 88.41, "elapsed_time": "2:01:48", "remaining_time": "0:15:57", "throughput": 5171.64, "total_tokens": 37798240} {"current_steps": 37510, "total_steps": 42420, "loss": 0.3201, "lr": 4.027493219979572e-05, "epoch": 17.685054219707684, "percentage": 88.43, "elapsed_time": "2:01:49", "remaining_time": "0:15:56", "throughput": 5171.67, "total_tokens": 37803200} {"current_steps": 37515, "total_steps": 42420, "loss": 0.3203, "lr": 4.019408061169833e-05, "epoch": 17.68741159830269, "percentage": 88.44, "elapsed_time": "2:01:50", "remaining_time": "0:15:55", "throughput": 5171.67, "total_tokens": 37807680} {"current_steps": 37520, "total_steps": 42420, "loss": 0.3461, "lr": 4.011330686093601e-05, "epoch": 17.68976897689769, "percentage": 88.45, "elapsed_time": "2:01:51", "remaining_time": "0:15:54", "throughput": 5171.69, "total_tokens": 37812640} {"current_steps": 37525, "total_steps": 42420, "loss": 0.3581, "lr": 4.0032610961182435e-05, "epoch": 17.692126355492693, "percentage": 88.46, "elapsed_time": "2:01:52", "remaining_time": "0:15:53", "throughput": 5171.73, "total_tokens": 37818048} {"current_steps": 37530, "total_steps": 42420, "loss": 0.2789, "lr": 3.995199292609802e-05, "epoch": 17.694483734087694, "percentage": 88.47, "elapsed_time": "2:01:53", "remaining_time": "0:15:52", "throughput": 5171.81, "total_tokens": 37824192} {"current_steps": 37535, "total_steps": 42420, "loss": 0.3117, "lr": 3.987145276933002e-05, "epoch": 17.696841112682698, "percentage": 88.48, "elapsed_time": "2:01:54", "remaining_time": "0:15:51", "throughput": 5171.85, "total_tokens": 37829568} {"current_steps": 37540, "total_steps": 42420, "loss": 0.3113, "lr": 3.979099050451257e-05, "epoch": 17.6991984912777, "percentage": 88.5, "elapsed_time": "2:01:55", "remaining_time": "0:15:50", "throughput": 5171.89, "total_tokens": 37834976} {"current_steps": 37545, "total_steps": 42420, "loss": 0.3466, "lr": 3.9710606145266505e-05, "epoch": 17.701555869872703, "percentage": 88.51, "elapsed_time": "2:01:56", "remaining_time": "0:15:49", "throughput": 5171.9, "total_tokens": 37839680} {"current_steps": 37550, "total_steps": 42420, "loss": 0.3704, "lr": 3.96302997051996e-05, "epoch": 17.703913248467703, "percentage": 88.52, "elapsed_time": "2:01:57", "remaining_time": "0:15:49", "throughput": 5171.96, "total_tokens": 37845504} {"current_steps": 37555, "total_steps": 42420, "loss": 0.3036, "lr": 3.9550071197906267e-05, "epoch": 17.706270627062707, "percentage": 88.53, "elapsed_time": "2:01:58", "remaining_time": "0:15:48", "throughput": 5171.97, "total_tokens": 37849920} {"current_steps": 37560, "total_steps": 42420, "loss": 0.3059, "lr": 3.946992063696786e-05, "epoch": 17.708628005657708, "percentage": 88.54, "elapsed_time": "2:01:59", "remaining_time": "0:15:47", "throughput": 5172.04, "total_tokens": 37856064} {"current_steps": 37565, "total_steps": 42420, "loss": 0.2688, "lr": 3.938984803595252e-05, "epoch": 17.710985384252712, "percentage": 88.55, "elapsed_time": "2:02:00", "remaining_time": "0:15:46", "throughput": 5172.08, "total_tokens": 37861088} {"current_steps": 37570, "total_steps": 42420, "loss": 0.3679, "lr": 3.930985340841525e-05, "epoch": 17.713342762847713, "percentage": 88.57, "elapsed_time": "2:02:01", "remaining_time": "0:15:45", "throughput": 5172.09, "total_tokens": 37865664} {"current_steps": 37575, "total_steps": 42420, "loss": 0.264, "lr": 3.9229936767897655e-05, "epoch": 17.715700141442717, "percentage": 88.58, "elapsed_time": "2:02:02", "remaining_time": "0:15:44", "throughput": 5172.16, "total_tokens": 37871456} {"current_steps": 37580, "total_steps": 42420, "loss": 0.3127, "lr": 3.9150098127928414e-05, "epoch": 17.718057520037718, "percentage": 88.59, "elapsed_time": "2:02:03", "remaining_time": "0:15:43", "throughput": 5172.16, "total_tokens": 37876000} {"current_steps": 37585, "total_steps": 42420, "loss": 0.3019, "lr": 3.907033750202271e-05, "epoch": 17.720414898632722, "percentage": 88.6, "elapsed_time": "2:02:03", "remaining_time": "0:15:42", "throughput": 5172.21, "total_tokens": 37881120} {"current_steps": 37590, "total_steps": 42420, "loss": 0.2922, "lr": 3.899065490368281e-05, "epoch": 17.722772277227723, "percentage": 88.61, "elapsed_time": "2:02:04", "remaining_time": "0:15:41", "throughput": 5172.23, "total_tokens": 37885824} {"current_steps": 37595, "total_steps": 42420, "loss": 0.3443, "lr": 3.891105034639747e-05, "epoch": 17.725129655822727, "percentage": 88.63, "elapsed_time": "2:02:05", "remaining_time": "0:15:40", "throughput": 5172.26, "total_tokens": 37890688} {"current_steps": 37600, "total_steps": 42420, "loss": 0.3124, "lr": 3.883152384364247e-05, "epoch": 17.727487034417727, "percentage": 88.64, "elapsed_time": "2:02:06", "remaining_time": "0:15:39", "throughput": 5172.27, "total_tokens": 37895264} {"current_steps": 37605, "total_steps": 42420, "loss": 0.3714, "lr": 3.8752075408880316e-05, "epoch": 17.72984441301273, "percentage": 88.65, "elapsed_time": "2:02:07", "remaining_time": "0:15:38", "throughput": 5172.33, "total_tokens": 37900672} {"current_steps": 37610, "total_steps": 42420, "loss": 0.3428, "lr": 3.86727050555602e-05, "epoch": 17.732201791607732, "percentage": 88.66, "elapsed_time": "2:02:08", "remaining_time": "0:15:37", "throughput": 5172.34, "total_tokens": 37905056} {"current_steps": 37615, "total_steps": 42420, "loss": 0.3788, "lr": 3.859341279711825e-05, "epoch": 17.734559170202736, "percentage": 88.67, "elapsed_time": "2:02:09", "remaining_time": "0:15:36", "throughput": 5172.38, "total_tokens": 37910400} {"current_steps": 37620, "total_steps": 42420, "loss": 0.3316, "lr": 3.851419864697725e-05, "epoch": 17.736916548797737, "percentage": 88.68, "elapsed_time": "2:02:10", "remaining_time": "0:15:35", "throughput": 5172.48, "total_tokens": 37916800} {"current_steps": 37625, "total_steps": 42420, "loss": 0.3516, "lr": 3.8435062618546904e-05, "epoch": 17.739273927392738, "percentage": 88.7, "elapsed_time": "2:02:11", "remaining_time": "0:15:34", "throughput": 5172.51, "total_tokens": 37921696} {"current_steps": 37630, "total_steps": 42420, "loss": 0.2691, "lr": 3.8356004725223546e-05, "epoch": 17.74163130598774, "percentage": 88.71, "elapsed_time": "2:02:12", "remaining_time": "0:15:33", "throughput": 5172.56, "total_tokens": 37927104} {"current_steps": 37635, "total_steps": 42420, "loss": 0.2954, "lr": 3.827702498039026e-05, "epoch": 17.743988684582742, "percentage": 88.72, "elapsed_time": "2:02:13", "remaining_time": "0:15:32", "throughput": 5172.55, "total_tokens": 37931296} {"current_steps": 37640, "total_steps": 42420, "loss": 0.278, "lr": 3.8198123397417016e-05, "epoch": 17.746346063177747, "percentage": 88.73, "elapsed_time": "2:02:14", "remaining_time": "0:15:31", "throughput": 5172.59, "total_tokens": 37936416} {"current_steps": 37645, "total_steps": 42420, "loss": 0.4005, "lr": 3.811929998966057e-05, "epoch": 17.748703441772747, "percentage": 88.74, "elapsed_time": "2:02:15", "remaining_time": "0:15:30", "throughput": 5172.66, "total_tokens": 37942112} {"current_steps": 37650, "total_steps": 42420, "loss": 0.3372, "lr": 3.804055477046431e-05, "epoch": 17.75106082036775, "percentage": 88.76, "elapsed_time": "2:02:15", "remaining_time": "0:15:29", "throughput": 5172.64, "total_tokens": 37945920} {"current_steps": 37655, "total_steps": 42420, "loss": 0.3745, "lr": 3.7961887753158506e-05, "epoch": 17.753418198962752, "percentage": 88.77, "elapsed_time": "2:02:16", "remaining_time": "0:15:28", "throughput": 5172.64, "total_tokens": 37950528} {"current_steps": 37660, "total_steps": 42420, "loss": 0.3516, "lr": 3.7883298951060186e-05, "epoch": 17.755775577557756, "percentage": 88.78, "elapsed_time": "2:02:17", "remaining_time": "0:15:27", "throughput": 5172.71, "total_tokens": 37956256} {"current_steps": 37665, "total_steps": 42420, "loss": 0.312, "lr": 3.780478837747303e-05, "epoch": 17.758132956152757, "percentage": 88.79, "elapsed_time": "2:02:18", "remaining_time": "0:15:26", "throughput": 5172.8, "total_tokens": 37962272} {"current_steps": 37670, "total_steps": 42420, "loss": 0.319, "lr": 3.772635604568764e-05, "epoch": 17.76049033474776, "percentage": 88.8, "elapsed_time": "2:02:19", "remaining_time": "0:15:25", "throughput": 5172.82, "total_tokens": 37967392} {"current_steps": 37675, "total_steps": 42420, "loss": 0.3612, "lr": 3.764800196898116e-05, "epoch": 17.76284771334276, "percentage": 88.81, "elapsed_time": "2:02:20", "remaining_time": "0:15:24", "throughput": 5172.82, "total_tokens": 37971744} {"current_steps": 37680, "total_steps": 42420, "loss": 0.3719, "lr": 3.756972616061766e-05, "epoch": 17.765205091937766, "percentage": 88.83, "elapsed_time": "2:02:21", "remaining_time": "0:15:23", "throughput": 5172.8, "total_tokens": 37975680} {"current_steps": 37685, "total_steps": 42420, "loss": 0.2873, "lr": 3.749152863384786e-05, "epoch": 17.767562470532766, "percentage": 88.84, "elapsed_time": "2:02:22", "remaining_time": "0:15:22", "throughput": 5172.87, "total_tokens": 37981504} {"current_steps": 37690, "total_steps": 42420, "loss": 0.3667, "lr": 3.741340940190935e-05, "epoch": 17.76991984912777, "percentage": 88.85, "elapsed_time": "2:02:23", "remaining_time": "0:15:21", "throughput": 5172.9, "total_tokens": 37986432} {"current_steps": 37695, "total_steps": 42420, "loss": 0.3943, "lr": 3.733536847802638e-05, "epoch": 17.77227722772277, "percentage": 88.86, "elapsed_time": "2:02:24", "remaining_time": "0:15:20", "throughput": 5172.86, "total_tokens": 37990112} {"current_steps": 37700, "total_steps": 42420, "loss": 0.3787, "lr": 3.725740587540988e-05, "epoch": 17.774634606317775, "percentage": 88.87, "elapsed_time": "2:02:24", "remaining_time": "0:15:19", "throughput": 5172.89, "total_tokens": 37994880} {"current_steps": 37705, "total_steps": 42420, "loss": 0.3173, "lr": 3.717952160725763e-05, "epoch": 17.776991984912776, "percentage": 88.88, "elapsed_time": "2:02:25", "remaining_time": "0:15:18", "throughput": 5172.91, "total_tokens": 37999424} {"current_steps": 37710, "total_steps": 42420, "loss": 0.3122, "lr": 3.710171568675425e-05, "epoch": 17.77934936350778, "percentage": 88.9, "elapsed_time": "2:02:26", "remaining_time": "0:15:17", "throughput": 5172.93, "total_tokens": 38004192} {"current_steps": 37715, "total_steps": 42420, "loss": 0.2918, "lr": 3.702398812707075e-05, "epoch": 17.78170674210278, "percentage": 88.91, "elapsed_time": "2:02:27", "remaining_time": "0:15:16", "throughput": 5172.99, "total_tokens": 38009984} {"current_steps": 37720, "total_steps": 42420, "loss": 0.3517, "lr": 3.6946338941365177e-05, "epoch": 17.784064120697785, "percentage": 88.92, "elapsed_time": "2:02:28", "remaining_time": "0:15:15", "throughput": 5172.99, "total_tokens": 38014176} {"current_steps": 37725, "total_steps": 42420, "loss": 0.3613, "lr": 3.686876814278217e-05, "epoch": 17.786421499292786, "percentage": 88.93, "elapsed_time": "2:02:29", "remaining_time": "0:15:14", "throughput": 5173.0, "total_tokens": 38018624} {"current_steps": 37730, "total_steps": 42420, "loss": 0.3417, "lr": 3.679127574445323e-05, "epoch": 17.78877887788779, "percentage": 88.94, "elapsed_time": "2:02:30", "remaining_time": "0:15:13", "throughput": 5173.02, "total_tokens": 38023328} {"current_steps": 37735, "total_steps": 42420, "loss": 0.3002, "lr": 3.6713861759496484e-05, "epoch": 17.79113625648279, "percentage": 88.96, "elapsed_time": "2:02:31", "remaining_time": "0:15:12", "throughput": 5173.1, "total_tokens": 38029184} {"current_steps": 37740, "total_steps": 42420, "loss": 0.3731, "lr": 3.663652620101682e-05, "epoch": 17.793493635077795, "percentage": 88.97, "elapsed_time": "2:02:32", "remaining_time": "0:15:11", "throughput": 5173.14, "total_tokens": 38034304} {"current_steps": 37745, "total_steps": 42420, "loss": 0.3166, "lr": 3.655926908210577e-05, "epoch": 17.795851013672795, "percentage": 88.98, "elapsed_time": "2:02:33", "remaining_time": "0:15:10", "throughput": 5173.1, "total_tokens": 38037888} {"current_steps": 37750, "total_steps": 42420, "loss": 0.339, "lr": 3.648209041584177e-05, "epoch": 17.7982083922678, "percentage": 88.99, "elapsed_time": "2:02:33", "remaining_time": "0:15:09", "throughput": 5173.15, "total_tokens": 38043232} {"current_steps": 37755, "total_steps": 42420, "loss": 0.339, "lr": 3.6404990215289646e-05, "epoch": 17.8005657708628, "percentage": 89.0, "elapsed_time": "2:02:34", "remaining_time": "0:15:08", "throughput": 5173.18, "total_tokens": 38048320} {"current_steps": 37760, "total_steps": 42420, "loss": 0.3121, "lr": 3.632796849350134e-05, "epoch": 17.802923149457804, "percentage": 89.01, "elapsed_time": "2:02:35", "remaining_time": "0:15:07", "throughput": 5173.17, "total_tokens": 38052448} {"current_steps": 37765, "total_steps": 42420, "loss": 0.3272, "lr": 3.625102526351537e-05, "epoch": 17.805280528052805, "percentage": 89.03, "elapsed_time": "2:02:36", "remaining_time": "0:15:06", "throughput": 5173.19, "total_tokens": 38057376} {"current_steps": 37770, "total_steps": 42420, "loss": 0.3355, "lr": 3.6174160538356707e-05, "epoch": 17.80763790664781, "percentage": 89.04, "elapsed_time": "2:02:37", "remaining_time": "0:15:05", "throughput": 5173.22, "total_tokens": 38062624} {"current_steps": 37775, "total_steps": 42420, "loss": 0.2997, "lr": 3.609737433103732e-05, "epoch": 17.80999528524281, "percentage": 89.05, "elapsed_time": "2:02:38", "remaining_time": "0:15:04", "throughput": 5173.27, "total_tokens": 38067712} {"current_steps": 37780, "total_steps": 42420, "loss": 0.3381, "lr": 3.602066665455589e-05, "epoch": 17.812352663837814, "percentage": 89.06, "elapsed_time": "2:02:39", "remaining_time": "0:15:03", "throughput": 5173.39, "total_tokens": 38075168} {"current_steps": 37785, "total_steps": 42420, "loss": 0.372, "lr": 3.5944037521897624e-05, "epoch": 17.814710042432814, "percentage": 89.07, "elapsed_time": "2:02:40", "remaining_time": "0:15:02", "throughput": 5173.42, "total_tokens": 38080000} {"current_steps": 37790, "total_steps": 42420, "loss": 0.3208, "lr": 3.586748694603459e-05, "epoch": 17.81706742102782, "percentage": 89.09, "elapsed_time": "2:02:41", "remaining_time": "0:15:01", "throughput": 5173.44, "total_tokens": 38084832} {"current_steps": 37795, "total_steps": 42420, "loss": 0.3549, "lr": 3.5791014939925534e-05, "epoch": 17.81942479962282, "percentage": 89.1, "elapsed_time": "2:02:42", "remaining_time": "0:15:00", "throughput": 5173.48, "total_tokens": 38090112} {"current_steps": 37800, "total_steps": 42420, "loss": 0.3236, "lr": 3.571462151651589e-05, "epoch": 17.821782178217823, "percentage": 89.11, "elapsed_time": "2:02:43", "remaining_time": "0:14:59", "throughput": 5173.56, "total_tokens": 38096000} {"current_steps": 37805, "total_steps": 42420, "loss": 0.3451, "lr": 3.563830668873763e-05, "epoch": 17.824139556812824, "percentage": 89.12, "elapsed_time": "2:02:44", "remaining_time": "0:14:59", "throughput": 5173.58, "total_tokens": 38100768} {"current_steps": 37810, "total_steps": 42420, "loss": 0.3314, "lr": 3.55620704695096e-05, "epoch": 17.826496935407828, "percentage": 89.13, "elapsed_time": "2:02:45", "remaining_time": "0:14:58", "throughput": 5173.64, "total_tokens": 38106528} {"current_steps": 37815, "total_steps": 42420, "loss": 0.3529, "lr": 3.548591287173736e-05, "epoch": 17.82885431400283, "percentage": 89.14, "elapsed_time": "2:02:46", "remaining_time": "0:14:57", "throughput": 5173.76, "total_tokens": 38113216} {"current_steps": 37820, "total_steps": 42420, "loss": 0.3555, "lr": 3.5409833908313096e-05, "epoch": 17.831211692597833, "percentage": 89.16, "elapsed_time": "2:02:47", "remaining_time": "0:14:56", "throughput": 5173.76, "total_tokens": 38117920} {"current_steps": 37825, "total_steps": 42420, "loss": 0.3101, "lr": 3.533383359211562e-05, "epoch": 17.833569071192834, "percentage": 89.17, "elapsed_time": "2:02:48", "remaining_time": "0:14:55", "throughput": 5173.79, "total_tokens": 38122912} {"current_steps": 37830, "total_steps": 42420, "loss": 0.3626, "lr": 3.525791193601058e-05, "epoch": 17.835926449787834, "percentage": 89.18, "elapsed_time": "2:02:49", "remaining_time": "0:14:54", "throughput": 5173.87, "total_tokens": 38128864} {"current_steps": 37835, "total_steps": 42420, "loss": 0.3597, "lr": 3.5182068952850146e-05, "epoch": 17.83828382838284, "percentage": 89.19, "elapsed_time": "2:02:50", "remaining_time": "0:14:53", "throughput": 5173.87, "total_tokens": 38133184} {"current_steps": 37840, "total_steps": 42420, "loss": 0.3528, "lr": 3.510630465547332e-05, "epoch": 17.84064120697784, "percentage": 89.2, "elapsed_time": "2:02:51", "remaining_time": "0:14:52", "throughput": 5173.92, "total_tokens": 38138336} {"current_steps": 37845, "total_steps": 42420, "loss": 0.2954, "lr": 3.5030619056705625e-05, "epoch": 17.842998585572843, "percentage": 89.21, "elapsed_time": "2:02:52", "remaining_time": "0:14:51", "throughput": 5173.98, "total_tokens": 38143872} {"current_steps": 37850, "total_steps": 42420, "loss": 0.3204, "lr": 3.4955012169359426e-05, "epoch": 17.845355964167844, "percentage": 89.23, "elapsed_time": "2:02:53", "remaining_time": "0:14:50", "throughput": 5174.05, "total_tokens": 38149824} {"current_steps": 37855, "total_steps": 42420, "loss": 0.333, "lr": 3.487948400623364e-05, "epoch": 17.847713342762848, "percentage": 89.24, "elapsed_time": "2:02:54", "remaining_time": "0:14:49", "throughput": 5174.09, "total_tokens": 38155168} {"current_steps": 37860, "total_steps": 42420, "loss": 0.3371, "lr": 3.4804034580113886e-05, "epoch": 17.85007072135785, "percentage": 89.25, "elapsed_time": "2:02:55", "remaining_time": "0:14:48", "throughput": 5174.08, "total_tokens": 38159584} {"current_steps": 37865, "total_steps": 42420, "loss": 0.3098, "lr": 3.47286639037725e-05, "epoch": 17.852428099952853, "percentage": 89.26, "elapsed_time": "2:02:56", "remaining_time": "0:14:47", "throughput": 5174.11, "total_tokens": 38164288} {"current_steps": 37870, "total_steps": 42420, "loss": 0.356, "lr": 3.4653371989968487e-05, "epoch": 17.854785478547853, "percentage": 89.27, "elapsed_time": "2:02:56", "remaining_time": "0:14:46", "throughput": 5174.09, "total_tokens": 38168352} {"current_steps": 37875, "total_steps": 42420, "loss": 0.3032, "lr": 3.457815885144744e-05, "epoch": 17.857142857142858, "percentage": 89.29, "elapsed_time": "2:02:57", "remaining_time": "0:14:45", "throughput": 5174.1, "total_tokens": 38172672} {"current_steps": 37880, "total_steps": 42420, "loss": 0.3294, "lr": 3.450302450094178e-05, "epoch": 17.85950023573786, "percentage": 89.3, "elapsed_time": "2:02:58", "remaining_time": "0:14:44", "throughput": 5174.14, "total_tokens": 38177824} {"current_steps": 37885, "total_steps": 42420, "loss": 0.3795, "lr": 3.4427968951170284e-05, "epoch": 17.861857614332862, "percentage": 89.31, "elapsed_time": "2:02:59", "remaining_time": "0:14:43", "throughput": 5174.16, "total_tokens": 38182624} {"current_steps": 37890, "total_steps": 42420, "loss": 0.3262, "lr": 3.435299221483867e-05, "epoch": 17.864214992927863, "percentage": 89.32, "elapsed_time": "2:03:00", "remaining_time": "0:14:42", "throughput": 5174.17, "total_tokens": 38187360} {"current_steps": 37895, "total_steps": 42420, "loss": 0.3114, "lr": 3.427809430463924e-05, "epoch": 17.866572371522867, "percentage": 89.33, "elapsed_time": "2:03:01", "remaining_time": "0:14:41", "throughput": 5174.23, "total_tokens": 38192576} {"current_steps": 37900, "total_steps": 42420, "loss": 0.3226, "lr": 3.420327523325101e-05, "epoch": 17.868929750117868, "percentage": 89.34, "elapsed_time": "2:03:02", "remaining_time": "0:14:40", "throughput": 5174.22, "total_tokens": 38196736} {"current_steps": 37905, "total_steps": 42420, "loss": 0.3016, "lr": 3.41285350133394e-05, "epoch": 17.871287128712872, "percentage": 89.36, "elapsed_time": "2:03:03", "remaining_time": "0:14:39", "throughput": 5174.23, "total_tokens": 38201440} {"current_steps": 37910, "total_steps": 42420, "loss": 0.3678, "lr": 3.4053873657556856e-05, "epoch": 17.873644507307873, "percentage": 89.37, "elapsed_time": "2:03:03", "remaining_time": "0:14:38", "throughput": 5174.27, "total_tokens": 38206368} {"current_steps": 37915, "total_steps": 42420, "loss": 0.3232, "lr": 3.397929117854215e-05, "epoch": 17.876001885902877, "percentage": 89.38, "elapsed_time": "2:03:04", "remaining_time": "0:14:37", "throughput": 5174.36, "total_tokens": 38212576} {"current_steps": 37920, "total_steps": 42420, "loss": 0.2798, "lr": 3.3904787588920906e-05, "epoch": 17.878359264497877, "percentage": 89.39, "elapsed_time": "2:03:05", "remaining_time": "0:14:36", "throughput": 5174.38, "total_tokens": 38217664} {"current_steps": 37925, "total_steps": 42420, "loss": 0.3221, "lr": 3.383036290130526e-05, "epoch": 17.88071664309288, "percentage": 89.4, "elapsed_time": "2:03:06", "remaining_time": "0:14:35", "throughput": 5174.47, "total_tokens": 38223808} {"current_steps": 37930, "total_steps": 42420, "loss": 0.402, "lr": 3.3756017128294016e-05, "epoch": 17.883074021687882, "percentage": 89.42, "elapsed_time": "2:03:07", "remaining_time": "0:14:34", "throughput": 5174.52, "total_tokens": 38229312} {"current_steps": 37935, "total_steps": 42420, "loss": 0.3695, "lr": 3.3681750282472725e-05, "epoch": 17.885431400282886, "percentage": 89.43, "elapsed_time": "2:03:08", "remaining_time": "0:14:33", "throughput": 5174.54, "total_tokens": 38234016} {"current_steps": 37940, "total_steps": 42420, "loss": 0.3275, "lr": 3.360756237641349e-05, "epoch": 17.887788778877887, "percentage": 89.44, "elapsed_time": "2:03:09", "remaining_time": "0:14:32", "throughput": 5174.56, "total_tokens": 38238592} {"current_steps": 37945, "total_steps": 42420, "loss": 0.353, "lr": 3.3533453422675056e-05, "epoch": 17.89014615747289, "percentage": 89.45, "elapsed_time": "2:03:10", "remaining_time": "0:14:31", "throughput": 5174.6, "total_tokens": 38243616} {"current_steps": 37950, "total_steps": 42420, "loss": 0.3574, "lr": 3.3459423433802764e-05, "epoch": 17.892503536067892, "percentage": 89.46, "elapsed_time": "2:03:11", "remaining_time": "0:14:30", "throughput": 5174.62, "total_tokens": 38248800} {"current_steps": 37955, "total_steps": 42420, "loss": 0.3047, "lr": 3.33854724223287e-05, "epoch": 17.894860914662896, "percentage": 89.47, "elapsed_time": "2:03:12", "remaining_time": "0:14:29", "throughput": 5174.7, "total_tokens": 38254784} {"current_steps": 37960, "total_steps": 42420, "loss": 0.3323, "lr": 3.331160040077158e-05, "epoch": 17.897218293257897, "percentage": 89.49, "elapsed_time": "2:03:13", "remaining_time": "0:14:28", "throughput": 5174.66, "total_tokens": 38258304} {"current_steps": 37965, "total_steps": 42420, "loss": 0.3428, "lr": 3.323780738163651e-05, "epoch": 17.8995756718529, "percentage": 89.5, "elapsed_time": "2:03:14", "remaining_time": "0:14:27", "throughput": 5174.72, "total_tokens": 38263840} {"current_steps": 37970, "total_steps": 42420, "loss": 0.3204, "lr": 3.3164093377415436e-05, "epoch": 17.9019330504479, "percentage": 89.51, "elapsed_time": "2:03:15", "remaining_time": "0:14:26", "throughput": 5174.75, "total_tokens": 38268672} {"current_steps": 37975, "total_steps": 42420, "loss": 0.3514, "lr": 3.3090458400586996e-05, "epoch": 17.904290429042906, "percentage": 89.52, "elapsed_time": "2:03:16", "remaining_time": "0:14:25", "throughput": 5174.81, "total_tokens": 38273952} {"current_steps": 37980, "total_steps": 42420, "loss": 0.3191, "lr": 3.301690246361622e-05, "epoch": 17.906647807637906, "percentage": 89.53, "elapsed_time": "2:03:17", "remaining_time": "0:14:24", "throughput": 5174.8, "total_tokens": 38278112} {"current_steps": 37985, "total_steps": 42420, "loss": 0.3341, "lr": 3.294342557895496e-05, "epoch": 17.90900518623291, "percentage": 89.55, "elapsed_time": "2:03:17", "remaining_time": "0:14:23", "throughput": 5174.85, "total_tokens": 38283488} {"current_steps": 37990, "total_steps": 42420, "loss": 0.3562, "lr": 3.287002775904169e-05, "epoch": 17.91136256482791, "percentage": 89.56, "elapsed_time": "2:03:18", "remaining_time": "0:14:22", "throughput": 5174.88, "total_tokens": 38288224} {"current_steps": 37995, "total_steps": 42420, "loss": 0.3497, "lr": 3.279670901630122e-05, "epoch": 17.913719943422915, "percentage": 89.57, "elapsed_time": "2:03:19", "remaining_time": "0:14:21", "throughput": 5174.94, "total_tokens": 38294080} {"current_steps": 38000, "total_steps": 42420, "loss": 0.3064, "lr": 3.2723469363145184e-05, "epoch": 17.916077322017916, "percentage": 89.58, "elapsed_time": "2:03:20", "remaining_time": "0:14:20", "throughput": 5174.99, "total_tokens": 38299520} {"current_steps": 38005, "total_steps": 42420, "loss": 0.3363, "lr": 3.265030881197195e-05, "epoch": 17.91843470061292, "percentage": 89.59, "elapsed_time": "2:03:21", "remaining_time": "0:14:19", "throughput": 5175.03, "total_tokens": 38304320} {"current_steps": 38010, "total_steps": 42420, "loss": 0.3327, "lr": 3.2577227375166326e-05, "epoch": 17.92079207920792, "percentage": 89.6, "elapsed_time": "2:03:22", "remaining_time": "0:14:18", "throughput": 5175.12, "total_tokens": 38310496} {"current_steps": 38015, "total_steps": 42420, "loss": 0.3674, "lr": 3.25042250650997e-05, "epoch": 17.92314945780292, "percentage": 89.62, "elapsed_time": "2:03:23", "remaining_time": "0:14:17", "throughput": 5175.16, "total_tokens": 38315776} {"current_steps": 38020, "total_steps": 42420, "loss": 0.3465, "lr": 3.243130189413007e-05, "epoch": 17.925506836397926, "percentage": 89.63, "elapsed_time": "2:03:24", "remaining_time": "0:14:16", "throughput": 5175.19, "total_tokens": 38320704} {"current_steps": 38025, "total_steps": 42420, "loss": 0.3655, "lr": 3.235845787460223e-05, "epoch": 17.927864214992926, "percentage": 89.64, "elapsed_time": "2:03:25", "remaining_time": "0:14:15", "throughput": 5175.23, "total_tokens": 38325792} {"current_steps": 38030, "total_steps": 42420, "loss": 0.293, "lr": 3.228569301884732e-05, "epoch": 17.93022159358793, "percentage": 89.65, "elapsed_time": "2:03:26", "remaining_time": "0:14:14", "throughput": 5175.25, "total_tokens": 38330336} {"current_steps": 38035, "total_steps": 42420, "loss": 0.2941, "lr": 3.221300733918331e-05, "epoch": 17.93257897218293, "percentage": 89.66, "elapsed_time": "2:03:27", "remaining_time": "0:14:13", "throughput": 5175.25, "total_tokens": 38334752} {"current_steps": 38040, "total_steps": 42420, "loss": 0.3102, "lr": 3.214040084791453e-05, "epoch": 17.934936350777935, "percentage": 89.67, "elapsed_time": "2:03:28", "remaining_time": "0:14:12", "throughput": 5175.25, "total_tokens": 38339136} {"current_steps": 38045, "total_steps": 42420, "loss": 0.3515, "lr": 3.206787355733209e-05, "epoch": 17.937293729372936, "percentage": 89.69, "elapsed_time": "2:03:29", "remaining_time": "0:14:12", "throughput": 5175.28, "total_tokens": 38344096} {"current_steps": 38050, "total_steps": 42420, "loss": 0.3077, "lr": 3.199542547971368e-05, "epoch": 17.93965110796794, "percentage": 89.7, "elapsed_time": "2:03:30", "remaining_time": "0:14:11", "throughput": 5175.33, "total_tokens": 38349792} {"current_steps": 38055, "total_steps": 42420, "loss": 0.3434, "lr": 3.192305662732342e-05, "epoch": 17.94200848656294, "percentage": 89.71, "elapsed_time": "2:03:31", "remaining_time": "0:14:10", "throughput": 5175.36, "total_tokens": 38354656} {"current_steps": 38060, "total_steps": 42420, "loss": 0.2944, "lr": 3.1850767012412183e-05, "epoch": 17.944365865157945, "percentage": 89.72, "elapsed_time": "2:03:31", "remaining_time": "0:14:09", "throughput": 5175.37, "total_tokens": 38359136} {"current_steps": 38065, "total_steps": 42420, "loss": 0.3758, "lr": 3.177855664721741e-05, "epoch": 17.946723243752945, "percentage": 89.73, "elapsed_time": "2:03:32", "remaining_time": "0:14:08", "throughput": 5175.46, "total_tokens": 38365152} {"current_steps": 38070, "total_steps": 42420, "loss": 0.3854, "lr": 3.1706425543962967e-05, "epoch": 17.94908062234795, "percentage": 89.75, "elapsed_time": "2:03:33", "remaining_time": "0:14:07", "throughput": 5175.53, "total_tokens": 38370560} {"current_steps": 38075, "total_steps": 42420, "loss": 0.3316, "lr": 3.163437371485956e-05, "epoch": 17.95143800094295, "percentage": 89.76, "elapsed_time": "2:03:34", "remaining_time": "0:14:06", "throughput": 5175.54, "total_tokens": 38375008} {"current_steps": 38080, "total_steps": 42420, "loss": 0.2744, "lr": 3.156240117210424e-05, "epoch": 17.953795379537954, "percentage": 89.77, "elapsed_time": "2:03:35", "remaining_time": "0:14:05", "throughput": 5175.6, "total_tokens": 38380480} {"current_steps": 38085, "total_steps": 42420, "loss": 0.3416, "lr": 3.149050792788083e-05, "epoch": 17.956152758132955, "percentage": 89.78, "elapsed_time": "2:03:36", "remaining_time": "0:14:04", "throughput": 5175.58, "total_tokens": 38384416} {"current_steps": 38090, "total_steps": 42420, "loss": 0.3633, "lr": 3.141869399435965e-05, "epoch": 17.95851013672796, "percentage": 89.79, "elapsed_time": "2:03:37", "remaining_time": "0:14:03", "throughput": 5175.63, "total_tokens": 38389696} {"current_steps": 38095, "total_steps": 42420, "loss": 0.3433, "lr": 3.134695938369747e-05, "epoch": 17.96086751532296, "percentage": 89.8, "elapsed_time": "2:03:38", "remaining_time": "0:14:02", "throughput": 5175.66, "total_tokens": 38394592} {"current_steps": 38100, "total_steps": 42420, "loss": 0.3906, "lr": 3.127530410803775e-05, "epoch": 17.963224893917964, "percentage": 89.82, "elapsed_time": "2:03:39", "remaining_time": "0:14:01", "throughput": 5175.76, "total_tokens": 38400832} {"current_steps": 38105, "total_steps": 42420, "loss": 0.356, "lr": 3.120372817951056e-05, "epoch": 17.965582272512965, "percentage": 89.83, "elapsed_time": "2:03:40", "remaining_time": "0:14:00", "throughput": 5175.81, "total_tokens": 38406624} {"current_steps": 38110, "total_steps": 42420, "loss": 0.3247, "lr": 3.1132231610232484e-05, "epoch": 17.96793965110797, "percentage": 89.84, "elapsed_time": "2:03:41", "remaining_time": "0:13:59", "throughput": 5175.92, "total_tokens": 38412992} {"current_steps": 38115, "total_steps": 42420, "loss": 0.3215, "lr": 3.106081441230668e-05, "epoch": 17.97029702970297, "percentage": 89.85, "elapsed_time": "2:03:42", "remaining_time": "0:13:58", "throughput": 5176.07, "total_tokens": 38420480} {"current_steps": 38120, "total_steps": 42420, "loss": 0.283, "lr": 3.0989476597822864e-05, "epoch": 17.972654408297974, "percentage": 89.86, "elapsed_time": "2:03:43", "remaining_time": "0:13:57", "throughput": 5176.08, "total_tokens": 38424864} {"current_steps": 38125, "total_steps": 42420, "loss": 0.3046, "lr": 3.0918218178857316e-05, "epoch": 17.975011786892974, "percentage": 89.88, "elapsed_time": "2:03:44", "remaining_time": "0:13:56", "throughput": 5176.1, "total_tokens": 38429664} {"current_steps": 38130, "total_steps": 42420, "loss": 0.3633, "lr": 3.084703916747295e-05, "epoch": 17.97736916548798, "percentage": 89.89, "elapsed_time": "2:03:45", "remaining_time": "0:13:55", "throughput": 5176.14, "total_tokens": 38434496} {"current_steps": 38135, "total_steps": 42420, "loss": 0.3013, "lr": 3.0775939575719e-05, "epoch": 17.97972654408298, "percentage": 89.9, "elapsed_time": "2:03:46", "remaining_time": "0:13:54", "throughput": 5176.25, "total_tokens": 38441376} {"current_steps": 38140, "total_steps": 42420, "loss": 0.3409, "lr": 3.070491941563153e-05, "epoch": 17.982083922677983, "percentage": 89.91, "elapsed_time": "2:03:47", "remaining_time": "0:13:53", "throughput": 5176.26, "total_tokens": 38445856} {"current_steps": 38145, "total_steps": 42420, "loss": 0.324, "lr": 3.0633978699233015e-05, "epoch": 17.984441301272984, "percentage": 89.92, "elapsed_time": "2:03:48", "remaining_time": "0:13:52", "throughput": 5176.3, "total_tokens": 38450976} {"current_steps": 38150, "total_steps": 42420, "loss": 0.338, "lr": 3.0563117438532517e-05, "epoch": 17.986798679867988, "percentage": 89.93, "elapsed_time": "2:03:49", "remaining_time": "0:13:51", "throughput": 5176.35, "total_tokens": 38456480} {"current_steps": 38155, "total_steps": 42420, "loss": 0.2813, "lr": 3.0492335645525613e-05, "epoch": 17.98915605846299, "percentage": 89.95, "elapsed_time": "2:03:50", "remaining_time": "0:13:50", "throughput": 5176.34, "total_tokens": 38460544} {"current_steps": 38160, "total_steps": 42420, "loss": 0.312, "lr": 3.0421633332194552e-05, "epoch": 17.991513437057993, "percentage": 89.96, "elapsed_time": "2:03:51", "remaining_time": "0:13:49", "throughput": 5176.39, "total_tokens": 38465920} {"current_steps": 38165, "total_steps": 42420, "loss": 0.2874, "lr": 3.0351010510507982e-05, "epoch": 17.993870815652993, "percentage": 89.97, "elapsed_time": "2:03:51", "remaining_time": "0:13:48", "throughput": 5176.43, "total_tokens": 38470848} {"current_steps": 38170, "total_steps": 42420, "loss": 0.3349, "lr": 3.0280467192421222e-05, "epoch": 17.996228194247998, "percentage": 89.98, "elapsed_time": "2:03:52", "remaining_time": "0:13:47", "throughput": 5176.4, "total_tokens": 38474880} {"current_steps": 38175, "total_steps": 42420, "loss": 0.3536, "lr": 3.0210003389875895e-05, "epoch": 17.998585572842998, "percentage": 89.99, "elapsed_time": "2:03:53", "remaining_time": "0:13:46", "throughput": 5176.4, "total_tokens": 38479104} {"current_steps": 38178, "total_steps": 42420, "eval_loss": 0.3389852046966553, "epoch": 18.0, "percentage": 90.0, "elapsed_time": "2:04:13", "remaining_time": "0:13:48", "throughput": 5163.13, "total_tokens": 38482112} {"current_steps": 38180, "total_steps": 42420, "loss": 0.326, "lr": 3.01396191148004e-05, "epoch": 18.000942951438002, "percentage": 90.0, "elapsed_time": "2:04:15", "remaining_time": "0:13:48", "throughput": 5161.57, "total_tokens": 38484192} {"current_steps": 38185, "total_steps": 42420, "loss": 0.3416, "lr": 3.006931437910965e-05, "epoch": 18.003300330033003, "percentage": 90.02, "elapsed_time": "2:04:16", "remaining_time": "0:13:47", "throughput": 5161.59, "total_tokens": 38488960} {"current_steps": 38190, "total_steps": 42420, "loss": 0.3482, "lr": 2.999908919470501e-05, "epoch": 18.005657708628007, "percentage": 90.03, "elapsed_time": "2:04:17", "remaining_time": "0:13:46", "throughput": 5161.58, "total_tokens": 38493248} {"current_steps": 38195, "total_steps": 42420, "loss": 0.3362, "lr": 2.9928943573474365e-05, "epoch": 18.008015087223008, "percentage": 90.04, "elapsed_time": "2:04:18", "remaining_time": "0:13:45", "throughput": 5161.56, "total_tokens": 38497312} {"current_steps": 38200, "total_steps": 42420, "loss": 0.3136, "lr": 2.9858877527292216e-05, "epoch": 18.010372465818012, "percentage": 90.05, "elapsed_time": "2:04:19", "remaining_time": "0:13:44", "throughput": 5161.63, "total_tokens": 38503584} {"current_steps": 38205, "total_steps": 42420, "loss": 0.3271, "lr": 2.978889106801963e-05, "epoch": 18.012729844413013, "percentage": 90.06, "elapsed_time": "2:04:20", "remaining_time": "0:13:43", "throughput": 5161.66, "total_tokens": 38508256} {"current_steps": 38210, "total_steps": 42420, "loss": 0.3344, "lr": 2.9718984207504076e-05, "epoch": 18.015087223008017, "percentage": 90.08, "elapsed_time": "2:04:21", "remaining_time": "0:13:42", "throughput": 5161.7, "total_tokens": 38513600} {"current_steps": 38215, "total_steps": 42420, "loss": 0.3421, "lr": 2.9649156957579526e-05, "epoch": 18.017444601603017, "percentage": 90.09, "elapsed_time": "2:04:22", "remaining_time": "0:13:41", "throughput": 5161.7, "total_tokens": 38518016} {"current_steps": 38220, "total_steps": 42420, "loss": 0.2799, "lr": 2.9579409330066644e-05, "epoch": 18.019801980198018, "percentage": 90.1, "elapsed_time": "2:04:23", "remaining_time": "0:13:40", "throughput": 5161.77, "total_tokens": 38523904} {"current_steps": 38225, "total_steps": 42420, "loss": 0.3037, "lr": 2.950974133677242e-05, "epoch": 18.022159358793022, "percentage": 90.11, "elapsed_time": "2:04:24", "remaining_time": "0:13:39", "throughput": 5161.82, "total_tokens": 38529312} {"current_steps": 38230, "total_steps": 42420, "loss": 0.3718, "lr": 2.9440152989490487e-05, "epoch": 18.024516737388023, "percentage": 90.12, "elapsed_time": "2:04:25", "remaining_time": "0:13:38", "throughput": 5161.86, "total_tokens": 38534432} {"current_steps": 38235, "total_steps": 42420, "loss": 0.3414, "lr": 2.9370644300001082e-05, "epoch": 18.026874115983027, "percentage": 90.13, "elapsed_time": "2:04:26", "remaining_time": "0:13:37", "throughput": 5161.9, "total_tokens": 38539264} {"current_steps": 38240, "total_steps": 42420, "loss": 0.2543, "lr": 2.9301215280070683e-05, "epoch": 18.029231494578028, "percentage": 90.15, "elapsed_time": "2:04:27", "remaining_time": "0:13:36", "throughput": 5161.94, "total_tokens": 38544320} {"current_steps": 38245, "total_steps": 42420, "loss": 0.3243, "lr": 2.923186594145255e-05, "epoch": 18.031588873173032, "percentage": 90.16, "elapsed_time": "2:04:27", "remaining_time": "0:13:35", "throughput": 5161.94, "total_tokens": 38548448} {"current_steps": 38250, "total_steps": 42420, "loss": 0.302, "lr": 2.9162596295886356e-05, "epoch": 18.033946251768032, "percentage": 90.17, "elapsed_time": "2:04:28", "remaining_time": "0:13:34", "throughput": 5162.0, "total_tokens": 38553696} {"current_steps": 38255, "total_steps": 42420, "loss": 0.3617, "lr": 2.909340635509822e-05, "epoch": 18.036303630363037, "percentage": 90.18, "elapsed_time": "2:04:29", "remaining_time": "0:13:33", "throughput": 5162.04, "total_tokens": 38559104} {"current_steps": 38260, "total_steps": 42420, "loss": 0.2832, "lr": 2.902429613080093e-05, "epoch": 18.038661008958037, "percentage": 90.19, "elapsed_time": "2:04:30", "remaining_time": "0:13:32", "throughput": 5162.05, "total_tokens": 38563648} {"current_steps": 38265, "total_steps": 42420, "loss": 0.3199, "lr": 2.895526563469353e-05, "epoch": 18.04101838755304, "percentage": 90.21, "elapsed_time": "2:04:31", "remaining_time": "0:13:31", "throughput": 5162.07, "total_tokens": 38568512} {"current_steps": 38270, "total_steps": 42420, "loss": 0.3372, "lr": 2.888631487846172e-05, "epoch": 18.043375766148042, "percentage": 90.22, "elapsed_time": "2:04:32", "remaining_time": "0:13:30", "throughput": 5162.07, "total_tokens": 38572640} {"current_steps": 38275, "total_steps": 42420, "loss": 0.3446, "lr": 2.8817443873777827e-05, "epoch": 18.045733144743046, "percentage": 90.23, "elapsed_time": "2:04:33", "remaining_time": "0:13:29", "throughput": 5162.1, "total_tokens": 38577600} {"current_steps": 38280, "total_steps": 42420, "loss": 0.3204, "lr": 2.8748652632300364e-05, "epoch": 18.048090523338047, "percentage": 90.24, "elapsed_time": "2:04:34", "remaining_time": "0:13:28", "throughput": 5162.19, "total_tokens": 38583584} {"current_steps": 38285, "total_steps": 42420, "loss": 0.3045, "lr": 2.8679941165674683e-05, "epoch": 18.05044790193305, "percentage": 90.25, "elapsed_time": "2:04:35", "remaining_time": "0:13:27", "throughput": 5162.22, "total_tokens": 38588256} {"current_steps": 38290, "total_steps": 42420, "loss": 0.3202, "lr": 2.861130948553242e-05, "epoch": 18.05280528052805, "percentage": 90.26, "elapsed_time": "2:04:36", "remaining_time": "0:13:26", "throughput": 5162.3, "total_tokens": 38594144} {"current_steps": 38295, "total_steps": 42420, "loss": 0.3191, "lr": 2.854275760349173e-05, "epoch": 18.055162659123056, "percentage": 90.28, "elapsed_time": "2:04:37", "remaining_time": "0:13:25", "throughput": 5162.31, "total_tokens": 38598720} {"current_steps": 38300, "total_steps": 42420, "loss": 0.3259, "lr": 2.8474285531157327e-05, "epoch": 18.057520037718056, "percentage": 90.29, "elapsed_time": "2:04:37", "remaining_time": "0:13:24", "throughput": 5162.3, "total_tokens": 38603232} {"current_steps": 38305, "total_steps": 42420, "loss": 0.2811, "lr": 2.8405893280120277e-05, "epoch": 18.05987741631306, "percentage": 90.3, "elapsed_time": "2:04:38", "remaining_time": "0:13:23", "throughput": 5162.38, "total_tokens": 38609024} {"current_steps": 38310, "total_steps": 42420, "loss": 0.3562, "lr": 2.8337580861958257e-05, "epoch": 18.06223479490806, "percentage": 90.31, "elapsed_time": "2:04:39", "remaining_time": "0:13:22", "throughput": 5162.47, "total_tokens": 38615104} {"current_steps": 38315, "total_steps": 42420, "loss": 0.3796, "lr": 2.8269348288235465e-05, "epoch": 18.064592173503065, "percentage": 90.32, "elapsed_time": "2:04:40", "remaining_time": "0:13:21", "throughput": 5162.49, "total_tokens": 38619904} {"current_steps": 38320, "total_steps": 42420, "loss": 0.3922, "lr": 2.8201195570502436e-05, "epoch": 18.066949552098066, "percentage": 90.33, "elapsed_time": "2:04:41", "remaining_time": "0:13:20", "throughput": 5162.52, "total_tokens": 38624960} {"current_steps": 38325, "total_steps": 42420, "loss": 0.3563, "lr": 2.8133122720296334e-05, "epoch": 18.06930693069307, "percentage": 90.35, "elapsed_time": "2:04:42", "remaining_time": "0:13:19", "throughput": 5162.56, "total_tokens": 38630016} {"current_steps": 38330, "total_steps": 42420, "loss": 0.3092, "lr": 2.8065129749140715e-05, "epoch": 18.07166430928807, "percentage": 90.36, "elapsed_time": "2:04:43", "remaining_time": "0:13:18", "throughput": 5162.56, "total_tokens": 38634432} {"current_steps": 38335, "total_steps": 42420, "loss": 0.3337, "lr": 2.7997216668545654e-05, "epoch": 18.074021687883075, "percentage": 90.37, "elapsed_time": "2:04:44", "remaining_time": "0:13:17", "throughput": 5162.63, "total_tokens": 38639936} {"current_steps": 38340, "total_steps": 42420, "loss": 0.3347, "lr": 2.792938349000773e-05, "epoch": 18.076379066478076, "percentage": 90.38, "elapsed_time": "2:04:45", "remaining_time": "0:13:16", "throughput": 5162.65, "total_tokens": 38645216} {"current_steps": 38345, "total_steps": 42420, "loss": 0.3295, "lr": 2.786163022500976e-05, "epoch": 18.07873644507308, "percentage": 90.39, "elapsed_time": "2:04:46", "remaining_time": "0:13:15", "throughput": 5162.68, "total_tokens": 38650112} {"current_steps": 38350, "total_steps": 42420, "loss": 0.3227, "lr": 2.7793956885021343e-05, "epoch": 18.08109382366808, "percentage": 90.41, "elapsed_time": "2:04:47", "remaining_time": "0:13:14", "throughput": 5162.72, "total_tokens": 38655040} {"current_steps": 38355, "total_steps": 42420, "loss": 0.3428, "lr": 2.7726363481498486e-05, "epoch": 18.083451202263085, "percentage": 90.42, "elapsed_time": "2:04:48", "remaining_time": "0:13:13", "throughput": 5162.8, "total_tokens": 38660640} {"current_steps": 38360, "total_steps": 42420, "loss": 0.3945, "lr": 2.7658850025883474e-05, "epoch": 18.085808580858085, "percentage": 90.43, "elapsed_time": "2:04:49", "remaining_time": "0:13:12", "throughput": 5162.83, "total_tokens": 38665760} {"current_steps": 38365, "total_steps": 42420, "loss": 0.3165, "lr": 2.7591416529605285e-05, "epoch": 18.08816595945309, "percentage": 90.44, "elapsed_time": "2:04:50", "remaining_time": "0:13:11", "throughput": 5162.86, "total_tokens": 38670624} {"current_steps": 38370, "total_steps": 42420, "loss": 0.3123, "lr": 2.7524063004079224e-05, "epoch": 18.09052333804809, "percentage": 90.45, "elapsed_time": "2:04:50", "remaining_time": "0:13:10", "throughput": 5162.85, "total_tokens": 38674752} {"current_steps": 38375, "total_steps": 42420, "loss": 0.3161, "lr": 2.745678946070712e-05, "epoch": 18.092880716643094, "percentage": 90.46, "elapsed_time": "2:04:51", "remaining_time": "0:13:09", "throughput": 5162.86, "total_tokens": 38679232} {"current_steps": 38380, "total_steps": 42420, "loss": 0.3066, "lr": 2.7389595910877306e-05, "epoch": 18.095238095238095, "percentage": 90.48, "elapsed_time": "2:04:52", "remaining_time": "0:13:08", "throughput": 5162.91, "total_tokens": 38684704} {"current_steps": 38385, "total_steps": 42420, "loss": 0.356, "lr": 2.732248236596435e-05, "epoch": 18.0975954738331, "percentage": 90.49, "elapsed_time": "2:04:53", "remaining_time": "0:13:07", "throughput": 5162.93, "total_tokens": 38689344} {"current_steps": 38390, "total_steps": 42420, "loss": 0.3174, "lr": 2.7255448837329556e-05, "epoch": 18.0999528524281, "percentage": 90.5, "elapsed_time": "2:04:54", "remaining_time": "0:13:06", "throughput": 5162.98, "total_tokens": 38694688} {"current_steps": 38395, "total_steps": 42420, "loss": 0.3078, "lr": 2.7188495336320505e-05, "epoch": 18.102310231023104, "percentage": 90.51, "elapsed_time": "2:04:55", "remaining_time": "0:13:05", "throughput": 5163.04, "total_tokens": 38700096} {"current_steps": 38400, "total_steps": 42420, "loss": 0.3388, "lr": 2.7121621874271306e-05, "epoch": 18.104667609618105, "percentage": 90.52, "elapsed_time": "2:04:56", "remaining_time": "0:13:04", "throughput": 5163.12, "total_tokens": 38706176} {"current_steps": 38405, "total_steps": 42420, "loss": 0.3648, "lr": 2.705482846250251e-05, "epoch": 18.10702498821311, "percentage": 90.54, "elapsed_time": "2:04:57", "remaining_time": "0:13:03", "throughput": 5163.25, "total_tokens": 38713408} {"current_steps": 38410, "total_steps": 42420, "loss": 0.3013, "lr": 2.6988115112321134e-05, "epoch": 18.10938236680811, "percentage": 90.55, "elapsed_time": "2:04:58", "remaining_time": "0:13:02", "throughput": 5163.32, "total_tokens": 38719008} {"current_steps": 38415, "total_steps": 42420, "loss": 0.2637, "lr": 2.6921481835020578e-05, "epoch": 18.111739745403113, "percentage": 90.56, "elapsed_time": "2:04:59", "remaining_time": "0:13:01", "throughput": 5163.31, "total_tokens": 38723040} {"current_steps": 38420, "total_steps": 42420, "loss": 0.2724, "lr": 2.685492864188077e-05, "epoch": 18.114097123998114, "percentage": 90.57, "elapsed_time": "2:05:00", "remaining_time": "0:13:00", "throughput": 5163.32, "total_tokens": 38727424} {"current_steps": 38425, "total_steps": 42420, "loss": 0.3326, "lr": 2.678845554416792e-05, "epoch": 18.116454502593115, "percentage": 90.58, "elapsed_time": "2:05:01", "remaining_time": "0:12:59", "throughput": 5163.38, "total_tokens": 38732800} {"current_steps": 38430, "total_steps": 42420, "loss": 0.2963, "lr": 2.672206255313492e-05, "epoch": 18.11881188118812, "percentage": 90.59, "elapsed_time": "2:05:02", "remaining_time": "0:12:58", "throughput": 5163.43, "total_tokens": 38737952} {"current_steps": 38435, "total_steps": 42420, "loss": 0.2677, "lr": 2.665574968002088e-05, "epoch": 18.12116925978312, "percentage": 90.61, "elapsed_time": "2:05:03", "remaining_time": "0:12:57", "throughput": 5163.45, "total_tokens": 38742912} {"current_steps": 38440, "total_steps": 42420, "loss": 0.2911, "lr": 2.6589516936051604e-05, "epoch": 18.123526638378124, "percentage": 90.62, "elapsed_time": "2:05:04", "remaining_time": "0:12:56", "throughput": 5163.44, "total_tokens": 38747008} {"current_steps": 38445, "total_steps": 42420, "loss": 0.2893, "lr": 2.652336433243896e-05, "epoch": 18.125884016973124, "percentage": 90.63, "elapsed_time": "2:05:04", "remaining_time": "0:12:55", "throughput": 5163.44, "total_tokens": 38751552} {"current_steps": 38450, "total_steps": 42420, "loss": 0.3239, "lr": 2.645729188038154e-05, "epoch": 18.12824139556813, "percentage": 90.64, "elapsed_time": "2:05:05", "remaining_time": "0:12:54", "throughput": 5163.45, "total_tokens": 38756096} {"current_steps": 38455, "total_steps": 42420, "loss": 0.3509, "lr": 2.6391299591064344e-05, "epoch": 18.13059877416313, "percentage": 90.65, "elapsed_time": "2:05:06", "remaining_time": "0:12:54", "throughput": 5163.51, "total_tokens": 38761440} {"current_steps": 38460, "total_steps": 42420, "loss": 0.2679, "lr": 2.6325387475658712e-05, "epoch": 18.132956152758133, "percentage": 90.66, "elapsed_time": "2:05:07", "remaining_time": "0:12:53", "throughput": 5163.64, "total_tokens": 38768064} {"current_steps": 38465, "total_steps": 42420, "loss": 0.2828, "lr": 2.62595555453225e-05, "epoch": 18.135313531353134, "percentage": 90.68, "elapsed_time": "2:05:09", "remaining_time": "0:12:52", "throughput": 5163.76, "total_tokens": 38774944} {"current_steps": 38470, "total_steps": 42420, "loss": 0.3102, "lr": 2.619380381119979e-05, "epoch": 18.137670909948138, "percentage": 90.69, "elapsed_time": "2:05:09", "remaining_time": "0:12:51", "throughput": 5163.81, "total_tokens": 38780192} {"current_steps": 38475, "total_steps": 42420, "loss": 0.3898, "lr": 2.6128132284421347e-05, "epoch": 18.14002828854314, "percentage": 90.7, "elapsed_time": "2:05:10", "remaining_time": "0:12:50", "throughput": 5163.82, "total_tokens": 38784512} {"current_steps": 38480, "total_steps": 42420, "loss": 0.3252, "lr": 2.606254097610422e-05, "epoch": 18.142385667138143, "percentage": 90.71, "elapsed_time": "2:05:11", "remaining_time": "0:12:49", "throughput": 5163.82, "total_tokens": 38788768} {"current_steps": 38485, "total_steps": 42420, "loss": 0.3269, "lr": 2.5997029897351866e-05, "epoch": 18.144743045733144, "percentage": 90.72, "elapsed_time": "2:05:12", "remaining_time": "0:12:48", "throughput": 5163.85, "total_tokens": 38793888} {"current_steps": 38490, "total_steps": 42420, "loss": 0.317, "lr": 2.5931599059254296e-05, "epoch": 18.147100424328148, "percentage": 90.74, "elapsed_time": "2:05:13", "remaining_time": "0:12:47", "throughput": 5163.86, "total_tokens": 38798272} {"current_steps": 38495, "total_steps": 42420, "loss": 0.3009, "lr": 2.5866248472887765e-05, "epoch": 18.14945780292315, "percentage": 90.75, "elapsed_time": "2:05:14", "remaining_time": "0:12:46", "throughput": 5163.84, "total_tokens": 38802304} {"current_steps": 38500, "total_steps": 42420, "loss": 0.3398, "lr": 2.5800978149315035e-05, "epoch": 18.151815181518153, "percentage": 90.76, "elapsed_time": "2:05:15", "remaining_time": "0:12:45", "throughput": 5163.83, "total_tokens": 38806496} {"current_steps": 38505, "total_steps": 42420, "loss": 0.3599, "lr": 2.5735788099585324e-05, "epoch": 18.154172560113153, "percentage": 90.77, "elapsed_time": "2:05:15", "remaining_time": "0:12:44", "throughput": 5163.84, "total_tokens": 38810880} {"current_steps": 38510, "total_steps": 42420, "loss": 0.33, "lr": 2.5670678334734076e-05, "epoch": 18.156529938708157, "percentage": 90.78, "elapsed_time": "2:05:16", "remaining_time": "0:12:43", "throughput": 5163.85, "total_tokens": 38815648} {"current_steps": 38515, "total_steps": 42420, "loss": 0.3357, "lr": 2.5605648865783315e-05, "epoch": 18.158887317303158, "percentage": 90.79, "elapsed_time": "2:05:17", "remaining_time": "0:12:42", "throughput": 5163.87, "total_tokens": 38820448} {"current_steps": 38520, "total_steps": 42420, "loss": 0.3196, "lr": 2.5540699703741456e-05, "epoch": 18.161244695898162, "percentage": 90.81, "elapsed_time": "2:05:18", "remaining_time": "0:12:41", "throughput": 5163.87, "total_tokens": 38824576} {"current_steps": 38525, "total_steps": 42420, "loss": 0.3574, "lr": 2.5475830859603255e-05, "epoch": 18.163602074493163, "percentage": 90.82, "elapsed_time": "2:05:19", "remaining_time": "0:12:40", "throughput": 5163.87, "total_tokens": 38828896} {"current_steps": 38530, "total_steps": 42420, "loss": 0.3496, "lr": 2.541104234434999e-05, "epoch": 18.165959453088167, "percentage": 90.83, "elapsed_time": "2:05:20", "remaining_time": "0:12:39", "throughput": 5163.92, "total_tokens": 38834144} {"current_steps": 38535, "total_steps": 42420, "loss": 0.3468, "lr": 2.534633416894916e-05, "epoch": 18.168316831683168, "percentage": 90.84, "elapsed_time": "2:05:21", "remaining_time": "0:12:38", "throughput": 5163.94, "total_tokens": 38838688} {"current_steps": 38540, "total_steps": 42420, "loss": 0.2843, "lr": 2.5281706344354782e-05, "epoch": 18.17067421027817, "percentage": 90.85, "elapsed_time": "2:05:21", "remaining_time": "0:12:37", "throughput": 5163.96, "total_tokens": 38843168} {"current_steps": 38545, "total_steps": 42420, "loss": 0.3389, "lr": 2.521715888150733e-05, "epoch": 18.173031588873172, "percentage": 90.87, "elapsed_time": "2:05:22", "remaining_time": "0:12:36", "throughput": 5164.02, "total_tokens": 38848512} {"current_steps": 38550, "total_steps": 42420, "loss": 0.4137, "lr": 2.515269179133345e-05, "epoch": 18.175388967468177, "percentage": 90.88, "elapsed_time": "2:05:23", "remaining_time": "0:12:35", "throughput": 5164.1, "total_tokens": 38854176} {"current_steps": 38555, "total_steps": 42420, "loss": 0.2812, "lr": 2.508830508474641e-05, "epoch": 18.177746346063177, "percentage": 90.89, "elapsed_time": "2:05:24", "remaining_time": "0:12:34", "throughput": 5164.16, "total_tokens": 38860096} {"current_steps": 38560, "total_steps": 42420, "loss": 0.3171, "lr": 2.5023998772645827e-05, "epoch": 18.18010372465818, "percentage": 90.9, "elapsed_time": "2:05:25", "remaining_time": "0:12:33", "throughput": 5164.17, "total_tokens": 38864640} {"current_steps": 38565, "total_steps": 42420, "loss": 0.3123, "lr": 2.495977286591755e-05, "epoch": 18.182461103253182, "percentage": 90.91, "elapsed_time": "2:05:26", "remaining_time": "0:12:32", "throughput": 5164.23, "total_tokens": 38870112} {"current_steps": 38570, "total_steps": 42420, "loss": 0.3543, "lr": 2.489562737543405e-05, "epoch": 18.184818481848186, "percentage": 90.92, "elapsed_time": "2:05:27", "remaining_time": "0:12:31", "throughput": 5164.27, "total_tokens": 38875040} {"current_steps": 38575, "total_steps": 42420, "loss": 0.3679, "lr": 2.483156231205397e-05, "epoch": 18.187175860443187, "percentage": 90.94, "elapsed_time": "2:05:28", "remaining_time": "0:12:30", "throughput": 5164.38, "total_tokens": 38881664} {"current_steps": 38580, "total_steps": 42420, "loss": 0.3809, "lr": 2.4767577686622522e-05, "epoch": 18.18953323903819, "percentage": 90.95, "elapsed_time": "2:05:29", "remaining_time": "0:12:29", "throughput": 5164.45, "total_tokens": 38887168} {"current_steps": 38585, "total_steps": 42420, "loss": 0.3164, "lr": 2.4703673509971268e-05, "epoch": 18.19189061763319, "percentage": 90.96, "elapsed_time": "2:05:30", "remaining_time": "0:12:28", "throughput": 5164.49, "total_tokens": 38892192} {"current_steps": 38590, "total_steps": 42420, "loss": 0.345, "lr": 2.4639849792917945e-05, "epoch": 18.194247996228196, "percentage": 90.97, "elapsed_time": "2:05:31", "remaining_time": "0:12:27", "throughput": 5164.55, "total_tokens": 38897792} {"current_steps": 38595, "total_steps": 42420, "loss": 0.3282, "lr": 2.457610654626685e-05, "epoch": 18.196605374823196, "percentage": 90.98, "elapsed_time": "2:05:32", "remaining_time": "0:12:26", "throughput": 5164.56, "total_tokens": 38902176} {"current_steps": 38600, "total_steps": 42420, "loss": 0.2831, "lr": 2.4512443780808747e-05, "epoch": 18.1989627534182, "percentage": 90.99, "elapsed_time": "2:05:33", "remaining_time": "0:12:25", "throughput": 5164.57, "total_tokens": 38906720} {"current_steps": 38605, "total_steps": 42420, "loss": 0.3333, "lr": 2.4448861507320564e-05, "epoch": 18.2013201320132, "percentage": 91.01, "elapsed_time": "2:05:34", "remaining_time": "0:12:24", "throughput": 5164.61, "total_tokens": 38911744} {"current_steps": 38610, "total_steps": 42420, "loss": 0.3067, "lr": 2.4385359736565692e-05, "epoch": 18.203677510608205, "percentage": 91.02, "elapsed_time": "2:05:35", "remaining_time": "0:12:23", "throughput": 5164.64, "total_tokens": 38916512} {"current_steps": 38615, "total_steps": 42420, "loss": 0.307, "lr": 2.4321938479293982e-05, "epoch": 18.206034889203206, "percentage": 91.03, "elapsed_time": "2:05:36", "remaining_time": "0:12:22", "throughput": 5164.68, "total_tokens": 38921664} {"current_steps": 38620, "total_steps": 42420, "loss": 0.2922, "lr": 2.4258597746241505e-05, "epoch": 18.208392267798207, "percentage": 91.04, "elapsed_time": "2:05:36", "remaining_time": "0:12:21", "throughput": 5164.65, "total_tokens": 38925376} {"current_steps": 38625, "total_steps": 42420, "loss": 0.2846, "lr": 2.4195337548130914e-05, "epoch": 18.21074964639321, "percentage": 91.05, "elapsed_time": "2:05:38", "remaining_time": "0:12:20", "throughput": 5164.75, "total_tokens": 38932192} {"current_steps": 38630, "total_steps": 42420, "loss": 0.3226, "lr": 2.4132157895670858e-05, "epoch": 18.21310702498821, "percentage": 91.07, "elapsed_time": "2:05:39", "remaining_time": "0:12:19", "throughput": 5164.84, "total_tokens": 38938048} {"current_steps": 38635, "total_steps": 42420, "loss": 0.3636, "lr": 2.4069058799556732e-05, "epoch": 18.215464403583216, "percentage": 91.08, "elapsed_time": "2:05:39", "remaining_time": "0:12:18", "throughput": 5164.85, "total_tokens": 38942528} {"current_steps": 38640, "total_steps": 42420, "loss": 0.3063, "lr": 2.400604027047004e-05, "epoch": 18.217821782178216, "percentage": 91.09, "elapsed_time": "2:05:40", "remaining_time": "0:12:17", "throughput": 5164.93, "total_tokens": 38948736} {"current_steps": 38645, "total_steps": 42420, "loss": 0.3679, "lr": 2.394310231907887e-05, "epoch": 18.22017916077322, "percentage": 91.1, "elapsed_time": "2:05:42", "remaining_time": "0:12:16", "throughput": 5165.0, "total_tokens": 38954688} {"current_steps": 38650, "total_steps": 42420, "loss": 0.3204, "lr": 2.3880244956037467e-05, "epoch": 18.22253653936822, "percentage": 91.11, "elapsed_time": "2:05:42", "remaining_time": "0:12:15", "throughput": 5165.01, "total_tokens": 38959264} {"current_steps": 38655, "total_steps": 42420, "loss": 0.3157, "lr": 2.3817468191986546e-05, "epoch": 18.224893917963225, "percentage": 91.12, "elapsed_time": "2:05:43", "remaining_time": "0:12:14", "throughput": 5165.06, "total_tokens": 38964576} {"current_steps": 38660, "total_steps": 42420, "loss": 0.3178, "lr": 2.3754772037553163e-05, "epoch": 18.227251296558226, "percentage": 91.14, "elapsed_time": "2:05:44", "remaining_time": "0:12:13", "throughput": 5165.15, "total_tokens": 38970848} {"current_steps": 38665, "total_steps": 42420, "loss": 0.3424, "lr": 2.3692156503350772e-05, "epoch": 18.22960867515323, "percentage": 91.15, "elapsed_time": "2:05:45", "remaining_time": "0:12:12", "throughput": 5165.14, "total_tokens": 38975040} {"current_steps": 38670, "total_steps": 42420, "loss": 0.3151, "lr": 2.3629621599978945e-05, "epoch": 18.23196605374823, "percentage": 91.16, "elapsed_time": "2:05:46", "remaining_time": "0:12:11", "throughput": 5165.11, "total_tokens": 38979072} {"current_steps": 38675, "total_steps": 42420, "loss": 0.2945, "lr": 2.3567167338023887e-05, "epoch": 18.234323432343235, "percentage": 91.17, "elapsed_time": "2:05:47", "remaining_time": "0:12:10", "throughput": 5165.16, "total_tokens": 38984352} {"current_steps": 38680, "total_steps": 42420, "loss": 0.3569, "lr": 2.350479372805803e-05, "epoch": 18.236680810938235, "percentage": 91.18, "elapsed_time": "2:05:48", "remaining_time": "0:12:09", "throughput": 5165.3, "total_tokens": 38991808} {"current_steps": 38685, "total_steps": 42420, "loss": 0.384, "lr": 2.344250078064014e-05, "epoch": 18.23903818953324, "percentage": 91.2, "elapsed_time": "2:05:49", "remaining_time": "0:12:08", "throughput": 5165.32, "total_tokens": 38996608} {"current_steps": 38690, "total_steps": 42420, "loss": 0.3701, "lr": 2.3380288506315405e-05, "epoch": 18.24139556812824, "percentage": 91.21, "elapsed_time": "2:05:50", "remaining_time": "0:12:07", "throughput": 5165.39, "total_tokens": 39002368} {"current_steps": 38695, "total_steps": 42420, "loss": 0.2985, "lr": 2.3318156915615287e-05, "epoch": 18.243752946723244, "percentage": 91.22, "elapsed_time": "2:05:51", "remaining_time": "0:12:06", "throughput": 5165.42, "total_tokens": 39007392} {"current_steps": 38700, "total_steps": 42420, "loss": 0.3377, "lr": 2.3256106019057532e-05, "epoch": 18.246110325318245, "percentage": 91.23, "elapsed_time": "2:05:52", "remaining_time": "0:12:05", "throughput": 5165.41, "total_tokens": 39011808} {"current_steps": 38705, "total_steps": 42420, "loss": 0.3056, "lr": 2.3194135827146468e-05, "epoch": 18.24846770391325, "percentage": 91.24, "elapsed_time": "2:05:53", "remaining_time": "0:12:05", "throughput": 5165.48, "total_tokens": 39017856} {"current_steps": 38710, "total_steps": 42420, "loss": 0.2697, "lr": 2.3132246350372476e-05, "epoch": 18.25082508250825, "percentage": 91.25, "elapsed_time": "2:05:54", "remaining_time": "0:12:04", "throughput": 5165.5, "total_tokens": 39022688} {"current_steps": 38715, "total_steps": 42420, "loss": 0.3531, "lr": 2.307043759921246e-05, "epoch": 18.253182461103254, "percentage": 91.27, "elapsed_time": "2:05:55", "remaining_time": "0:12:03", "throughput": 5165.51, "total_tokens": 39027200} {"current_steps": 38720, "total_steps": 42420, "loss": 0.3066, "lr": 2.300870958412954e-05, "epoch": 18.255539839698255, "percentage": 91.28, "elapsed_time": "2:05:56", "remaining_time": "0:12:02", "throughput": 5165.56, "total_tokens": 39032608} {"current_steps": 38725, "total_steps": 42420, "loss": 0.3324, "lr": 2.29470623155732e-05, "epoch": 18.25789721829326, "percentage": 91.29, "elapsed_time": "2:05:57", "remaining_time": "0:12:01", "throughput": 5165.55, "total_tokens": 39036704} {"current_steps": 38730, "total_steps": 42420, "loss": 0.348, "lr": 2.2885495803979305e-05, "epoch": 18.26025459688826, "percentage": 91.3, "elapsed_time": "2:05:58", "remaining_time": "0:12:00", "throughput": 5165.57, "total_tokens": 39041440} {"current_steps": 38735, "total_steps": 42420, "loss": 0.3824, "lr": 2.2824010059770027e-05, "epoch": 18.262611975483264, "percentage": 91.31, "elapsed_time": "2:05:58", "remaining_time": "0:11:59", "throughput": 5165.61, "total_tokens": 39046272} {"current_steps": 38740, "total_steps": 42420, "loss": 0.3443, "lr": 2.276260509335387e-05, "epoch": 18.264969354078264, "percentage": 91.32, "elapsed_time": "2:05:59", "remaining_time": "0:11:58", "throughput": 5165.62, "total_tokens": 39051296} {"current_steps": 38745, "total_steps": 42420, "loss": 0.352, "lr": 2.270128091512569e-05, "epoch": 18.26732673267327, "percentage": 91.34, "elapsed_time": "2:06:00", "remaining_time": "0:11:57", "throughput": 5165.61, "total_tokens": 39055648} {"current_steps": 38750, "total_steps": 42420, "loss": 0.3652, "lr": 2.2640037535466505e-05, "epoch": 18.26968411126827, "percentage": 91.35, "elapsed_time": "2:06:01", "remaining_time": "0:11:56", "throughput": 5165.61, "total_tokens": 39059872} {"current_steps": 38755, "total_steps": 42420, "loss": 0.3176, "lr": 2.2578874964743978e-05, "epoch": 18.272041489863273, "percentage": 91.36, "elapsed_time": "2:06:02", "remaining_time": "0:11:55", "throughput": 5165.63, "total_tokens": 39064640} {"current_steps": 38760, "total_steps": 42420, "loss": 0.3467, "lr": 2.2517793213311655e-05, "epoch": 18.274398868458274, "percentage": 91.37, "elapsed_time": "2:06:03", "remaining_time": "0:11:54", "throughput": 5165.68, "total_tokens": 39069760} {"current_steps": 38765, "total_steps": 42420, "loss": 0.315, "lr": 2.2456792291509763e-05, "epoch": 18.276756247053278, "percentage": 91.38, "elapsed_time": "2:06:04", "remaining_time": "0:11:53", "throughput": 5165.7, "total_tokens": 39074592} {"current_steps": 38770, "total_steps": 42420, "loss": 0.3006, "lr": 2.2395872209664768e-05, "epoch": 18.27911362564828, "percentage": 91.4, "elapsed_time": "2:06:05", "remaining_time": "0:11:52", "throughput": 5165.78, "total_tokens": 39080352} {"current_steps": 38775, "total_steps": 42420, "loss": 0.2989, "lr": 2.233503297808931e-05, "epoch": 18.281471004243283, "percentage": 91.41, "elapsed_time": "2:06:06", "remaining_time": "0:11:51", "throughput": 5165.84, "total_tokens": 39085856} {"current_steps": 38780, "total_steps": 42420, "loss": 0.341, "lr": 2.227427460708248e-05, "epoch": 18.283828382838283, "percentage": 91.42, "elapsed_time": "2:06:07", "remaining_time": "0:11:50", "throughput": 5165.88, "total_tokens": 39091168} {"current_steps": 38785, "total_steps": 42420, "loss": 0.2621, "lr": 2.2213597106929607e-05, "epoch": 18.286185761433288, "percentage": 91.43, "elapsed_time": "2:06:08", "remaining_time": "0:11:49", "throughput": 5165.87, "total_tokens": 39095712} {"current_steps": 38790, "total_steps": 42420, "loss": 0.3336, "lr": 2.215300048790242e-05, "epoch": 18.28854314002829, "percentage": 91.44, "elapsed_time": "2:06:08", "remaining_time": "0:11:48", "throughput": 5165.9, "total_tokens": 39100416} {"current_steps": 38795, "total_steps": 42420, "loss": 0.3367, "lr": 2.209248476025899e-05, "epoch": 18.290900518623292, "percentage": 91.45, "elapsed_time": "2:06:09", "remaining_time": "0:11:47", "throughput": 5165.93, "total_tokens": 39105344} {"current_steps": 38800, "total_steps": 42420, "loss": 0.3334, "lr": 2.2032049934243347e-05, "epoch": 18.293257897218293, "percentage": 91.47, "elapsed_time": "2:06:11", "remaining_time": "0:11:46", "throughput": 5166.06, "total_tokens": 39112448} {"current_steps": 38805, "total_steps": 42420, "loss": 0.3543, "lr": 2.1971696020086252e-05, "epoch": 18.295615275813297, "percentage": 91.48, "elapsed_time": "2:06:12", "remaining_time": "0:11:45", "throughput": 5166.13, "total_tokens": 39118336} {"current_steps": 38810, "total_steps": 42420, "loss": 0.3565, "lr": 2.191142302800453e-05, "epoch": 18.297972654408298, "percentage": 91.49, "elapsed_time": "2:06:12", "remaining_time": "0:11:44", "throughput": 5166.16, "total_tokens": 39123008} {"current_steps": 38815, "total_steps": 42420, "loss": 0.341, "lr": 2.1851230968201407e-05, "epoch": 18.300330033003302, "percentage": 91.5, "elapsed_time": "2:06:13", "remaining_time": "0:11:43", "throughput": 5166.21, "total_tokens": 39128128} {"current_steps": 38820, "total_steps": 42420, "loss": 0.3189, "lr": 2.1791119850866402e-05, "epoch": 18.302687411598303, "percentage": 91.51, "elapsed_time": "2:06:14", "remaining_time": "0:11:42", "throughput": 5166.21, "total_tokens": 39132448} {"current_steps": 38825, "total_steps": 42420, "loss": 0.336, "lr": 2.1731089686175264e-05, "epoch": 18.305044790193303, "percentage": 91.53, "elapsed_time": "2:06:15", "remaining_time": "0:11:41", "throughput": 5166.22, "total_tokens": 39136928} {"current_steps": 38830, "total_steps": 42420, "loss": 0.3611, "lr": 2.167114048429014e-05, "epoch": 18.307402168788308, "percentage": 91.54, "elapsed_time": "2:06:16", "remaining_time": "0:11:40", "throughput": 5166.25, "total_tokens": 39141920} {"current_steps": 38835, "total_steps": 42420, "loss": 0.3429, "lr": 2.1611272255359415e-05, "epoch": 18.309759547383308, "percentage": 91.55, "elapsed_time": "2:06:17", "remaining_time": "0:11:39", "throughput": 5166.25, "total_tokens": 39146240} {"current_steps": 38840, "total_steps": 42420, "loss": 0.3572, "lr": 2.1551485009517592e-05, "epoch": 18.312116925978312, "percentage": 91.56, "elapsed_time": "2:06:18", "remaining_time": "0:11:38", "throughput": 5166.31, "total_tokens": 39151616} {"current_steps": 38845, "total_steps": 42420, "loss": 0.3193, "lr": 2.149177875688585e-05, "epoch": 18.314474304573313, "percentage": 91.57, "elapsed_time": "2:06:19", "remaining_time": "0:11:37", "throughput": 5166.35, "total_tokens": 39156704} {"current_steps": 38850, "total_steps": 42420, "loss": 0.3982, "lr": 2.1432153507571327e-05, "epoch": 18.316831683168317, "percentage": 91.58, "elapsed_time": "2:06:20", "remaining_time": "0:11:36", "throughput": 5166.41, "total_tokens": 39162400} {"current_steps": 38855, "total_steps": 42420, "loss": 0.305, "lr": 2.1372609271667563e-05, "epoch": 18.319189061763318, "percentage": 91.6, "elapsed_time": "2:06:21", "remaining_time": "0:11:35", "throughput": 5166.45, "total_tokens": 39167776} {"current_steps": 38860, "total_steps": 42420, "loss": 0.3469, "lr": 2.131314605925444e-05, "epoch": 18.321546440358322, "percentage": 91.61, "elapsed_time": "2:06:22", "remaining_time": "0:11:34", "throughput": 5166.47, "total_tokens": 39172256} {"current_steps": 38865, "total_steps": 42420, "loss": 0.2808, "lr": 2.1253763880398068e-05, "epoch": 18.323903818953323, "percentage": 91.62, "elapsed_time": "2:06:23", "remaining_time": "0:11:33", "throughput": 5166.53, "total_tokens": 39177792} {"current_steps": 38870, "total_steps": 42420, "loss": 0.2908, "lr": 2.11944627451508e-05, "epoch": 18.326261197548327, "percentage": 91.63, "elapsed_time": "2:06:24", "remaining_time": "0:11:32", "throughput": 5166.63, "total_tokens": 39184352} {"current_steps": 38875, "total_steps": 42420, "loss": 0.3358, "lr": 2.1135242663551434e-05, "epoch": 18.328618576143327, "percentage": 91.64, "elapsed_time": "2:06:25", "remaining_time": "0:11:31", "throughput": 5166.72, "total_tokens": 39190784} {"current_steps": 38880, "total_steps": 42420, "loss": 0.364, "lr": 2.1076103645624733e-05, "epoch": 18.33097595473833, "percentage": 91.65, "elapsed_time": "2:06:26", "remaining_time": "0:11:30", "throughput": 5166.73, "total_tokens": 39195648} {"current_steps": 38885, "total_steps": 42420, "loss": 0.3469, "lr": 2.101704570138202e-05, "epoch": 18.333333333333332, "percentage": 91.67, "elapsed_time": "2:06:27", "remaining_time": "0:11:29", "throughput": 5166.71, "total_tokens": 39199904} {"current_steps": 38890, "total_steps": 42420, "loss": 0.3259, "lr": 2.0958068840820853e-05, "epoch": 18.335690711928336, "percentage": 91.68, "elapsed_time": "2:06:28", "remaining_time": "0:11:28", "throughput": 5166.78, "total_tokens": 39205696} {"current_steps": 38895, "total_steps": 42420, "loss": 0.3361, "lr": 2.0899173073924917e-05, "epoch": 18.338048090523337, "percentage": 91.69, "elapsed_time": "2:06:28", "remaining_time": "0:11:27", "throughput": 5166.78, "total_tokens": 39210112} {"current_steps": 38900, "total_steps": 42420, "loss": 0.3092, "lr": 2.0840358410664283e-05, "epoch": 18.34040546911834, "percentage": 91.7, "elapsed_time": "2:06:29", "remaining_time": "0:11:26", "throughput": 5166.8, "total_tokens": 39214656} {"current_steps": 38905, "total_steps": 42420, "loss": 0.2964, "lr": 2.078162486099533e-05, "epoch": 18.34276284771334, "percentage": 91.71, "elapsed_time": "2:06:30", "remaining_time": "0:11:25", "throughput": 5166.86, "total_tokens": 39220576} {"current_steps": 38910, "total_steps": 42420, "loss": 0.3337, "lr": 2.0722972434860655e-05, "epoch": 18.345120226308346, "percentage": 91.73, "elapsed_time": "2:06:31", "remaining_time": "0:11:24", "throughput": 5166.92, "total_tokens": 39226016} {"current_steps": 38915, "total_steps": 42420, "loss": 0.3203, "lr": 2.0664401142189037e-05, "epoch": 18.347477604903347, "percentage": 91.74, "elapsed_time": "2:06:32", "remaining_time": "0:11:23", "throughput": 5166.94, "total_tokens": 39230688} {"current_steps": 38920, "total_steps": 42420, "loss": 0.378, "lr": 2.060591099289566e-05, "epoch": 18.34983498349835, "percentage": 91.75, "elapsed_time": "2:06:33", "remaining_time": "0:11:22", "throughput": 5167.0, "total_tokens": 39236256} {"current_steps": 38925, "total_steps": 42420, "loss": 0.3647, "lr": 2.054750199688199e-05, "epoch": 18.35219236209335, "percentage": 91.76, "elapsed_time": "2:06:34", "remaining_time": "0:11:21", "throughput": 5167.03, "total_tokens": 39241376} {"current_steps": 38930, "total_steps": 42420, "loss": 0.3392, "lr": 2.0489174164035505e-05, "epoch": 18.354549740688356, "percentage": 91.77, "elapsed_time": "2:06:35", "remaining_time": "0:11:20", "throughput": 5167.03, "total_tokens": 39245728} {"current_steps": 38935, "total_steps": 42420, "loss": 0.3274, "lr": 2.04309275042302e-05, "epoch": 18.356907119283356, "percentage": 91.78, "elapsed_time": "2:06:36", "remaining_time": "0:11:19", "throughput": 5167.04, "total_tokens": 39250464} {"current_steps": 38940, "total_steps": 42420, "loss": 0.3099, "lr": 2.037276202732624e-05, "epoch": 18.35926449787836, "percentage": 91.8, "elapsed_time": "2:06:37", "remaining_time": "0:11:18", "throughput": 5167.07, "total_tokens": 39255136} {"current_steps": 38945, "total_steps": 42420, "loss": 0.333, "lr": 2.0314677743170083e-05, "epoch": 18.36162187647336, "percentage": 91.81, "elapsed_time": "2:06:38", "remaining_time": "0:11:17", "throughput": 5167.09, "total_tokens": 39259936} {"current_steps": 38950, "total_steps": 42420, "loss": 0.3452, "lr": 2.0256674661594366e-05, "epoch": 18.363979255068365, "percentage": 91.82, "elapsed_time": "2:06:39", "remaining_time": "0:11:16", "throughput": 5167.16, "total_tokens": 39265600} {"current_steps": 38955, "total_steps": 42420, "loss": 0.2983, "lr": 2.019875279241806e-05, "epoch": 18.366336633663366, "percentage": 91.83, "elapsed_time": "2:06:39", "remaining_time": "0:11:16", "throughput": 5167.15, "total_tokens": 39269856} {"current_steps": 38960, "total_steps": 42420, "loss": 0.3117, "lr": 2.0140912145446387e-05, "epoch": 18.36869401225837, "percentage": 91.84, "elapsed_time": "2:06:40", "remaining_time": "0:11:15", "throughput": 5167.22, "total_tokens": 39275488} {"current_steps": 38965, "total_steps": 42420, "loss": 0.3701, "lr": 2.0083152730470788e-05, "epoch": 18.37105139085337, "percentage": 91.86, "elapsed_time": "2:06:41", "remaining_time": "0:11:14", "throughput": 5167.29, "total_tokens": 39281248} {"current_steps": 38970, "total_steps": 42420, "loss": 0.296, "lr": 2.002547455726883e-05, "epoch": 18.373408769448375, "percentage": 91.87, "elapsed_time": "2:06:42", "remaining_time": "0:11:13", "throughput": 5167.38, "total_tokens": 39287584} {"current_steps": 38975, "total_steps": 42420, "loss": 0.3388, "lr": 1.9967877635604538e-05, "epoch": 18.375766148043375, "percentage": 91.88, "elapsed_time": "2:06:43", "remaining_time": "0:11:12", "throughput": 5167.37, "total_tokens": 39291552} {"current_steps": 38980, "total_steps": 42420, "loss": 0.3222, "lr": 1.991036197522805e-05, "epoch": 18.37812352663838, "percentage": 91.89, "elapsed_time": "2:06:44", "remaining_time": "0:11:11", "throughput": 5167.38, "total_tokens": 39296064} {"current_steps": 38985, "total_steps": 42420, "loss": 0.3484, "lr": 1.985292758587581e-05, "epoch": 18.38048090523338, "percentage": 91.9, "elapsed_time": "2:06:45", "remaining_time": "0:11:10", "throughput": 5167.47, "total_tokens": 39302304} {"current_steps": 38990, "total_steps": 42420, "loss": 0.3176, "lr": 1.979557447727054e-05, "epoch": 18.382838283828384, "percentage": 91.91, "elapsed_time": "2:06:46", "remaining_time": "0:11:09", "throughput": 5167.51, "total_tokens": 39307200} {"current_steps": 38995, "total_steps": 42420, "loss": 0.3551, "lr": 1.9738302659121133e-05, "epoch": 18.385195662423385, "percentage": 91.93, "elapsed_time": "2:06:47", "remaining_time": "0:11:08", "throughput": 5167.53, "total_tokens": 39312192} {"current_steps": 39000, "total_steps": 42420, "loss": 0.3394, "lr": 1.9681112141122672e-05, "epoch": 18.38755304101839, "percentage": 91.94, "elapsed_time": "2:06:48", "remaining_time": "0:11:07", "throughput": 5167.57, "total_tokens": 39317440} {"current_steps": 39005, "total_steps": 42420, "loss": 0.3897, "lr": 1.962400293295663e-05, "epoch": 18.38991041961339, "percentage": 91.95, "elapsed_time": "2:06:49", "remaining_time": "0:11:06", "throughput": 5167.56, "total_tokens": 39321632} {"current_steps": 39010, "total_steps": 42420, "loss": 0.3368, "lr": 1.95669750442905e-05, "epoch": 18.392267798208394, "percentage": 91.96, "elapsed_time": "2:06:50", "remaining_time": "0:11:05", "throughput": 5167.62, "total_tokens": 39327264} {"current_steps": 39015, "total_steps": 42420, "loss": 0.3578, "lr": 1.951002848477823e-05, "epoch": 18.394625176803395, "percentage": 91.97, "elapsed_time": "2:06:51", "remaining_time": "0:11:04", "throughput": 5167.68, "total_tokens": 39332640} {"current_steps": 39020, "total_steps": 42420, "loss": 0.3153, "lr": 1.9453163264059825e-05, "epoch": 18.396982555398395, "percentage": 91.98, "elapsed_time": "2:06:52", "remaining_time": "0:11:03", "throughput": 5167.75, "total_tokens": 39338432} {"current_steps": 39025, "total_steps": 42420, "loss": 0.3521, "lr": 1.93963793917617e-05, "epoch": 18.3993399339934, "percentage": 92.0, "elapsed_time": "2:06:53", "remaining_time": "0:11:02", "throughput": 5167.83, "total_tokens": 39344384} {"current_steps": 39030, "total_steps": 42420, "loss": 0.3758, "lr": 1.933967687749638e-05, "epoch": 18.4016973125884, "percentage": 92.01, "elapsed_time": "2:06:54", "remaining_time": "0:11:01", "throughput": 5167.79, "total_tokens": 39348064} {"current_steps": 39035, "total_steps": 42420, "loss": 0.3553, "lr": 1.9283055730862587e-05, "epoch": 18.404054691183404, "percentage": 92.02, "elapsed_time": "2:06:54", "remaining_time": "0:11:00", "throughput": 5167.76, "total_tokens": 39351712} {"current_steps": 39040, "total_steps": 42420, "loss": 0.3328, "lr": 1.9226515961445368e-05, "epoch": 18.406412069778405, "percentage": 92.03, "elapsed_time": "2:06:55", "remaining_time": "0:10:59", "throughput": 5167.8, "total_tokens": 39356992} {"current_steps": 39045, "total_steps": 42420, "loss": 0.3493, "lr": 1.9170057578816014e-05, "epoch": 18.40876944837341, "percentage": 92.04, "elapsed_time": "2:06:56", "remaining_time": "0:10:58", "throughput": 5167.82, "total_tokens": 39361728} {"current_steps": 39050, "total_steps": 42420, "loss": 0.3757, "lr": 1.911368059253177e-05, "epoch": 18.41112682696841, "percentage": 92.06, "elapsed_time": "2:06:57", "remaining_time": "0:10:57", "throughput": 5167.83, "total_tokens": 39366112} {"current_steps": 39055, "total_steps": 42420, "loss": 0.3002, "lr": 1.905738501213644e-05, "epoch": 18.413484205563414, "percentage": 92.07, "elapsed_time": "2:06:58", "remaining_time": "0:10:56", "throughput": 5167.87, "total_tokens": 39370944} {"current_steps": 39060, "total_steps": 42420, "loss": 0.3344, "lr": 1.900117084715991e-05, "epoch": 18.415841584158414, "percentage": 92.08, "elapsed_time": "2:06:59", "remaining_time": "0:10:55", "throughput": 5167.88, "total_tokens": 39375552} {"current_steps": 39065, "total_steps": 42420, "loss": 0.3434, "lr": 1.894503810711823e-05, "epoch": 18.41819896275342, "percentage": 92.09, "elapsed_time": "2:07:00", "remaining_time": "0:10:54", "throughput": 5167.95, "total_tokens": 39381120} {"current_steps": 39070, "total_steps": 42420, "loss": 0.3305, "lr": 1.8888986801513742e-05, "epoch": 18.42055634134842, "percentage": 92.1, "elapsed_time": "2:07:01", "remaining_time": "0:10:53", "throughput": 5168.0, "total_tokens": 39386816} {"current_steps": 39075, "total_steps": 42420, "loss": 0.3023, "lr": 1.8833016939835023e-05, "epoch": 18.422913719943423, "percentage": 92.11, "elapsed_time": "2:07:02", "remaining_time": "0:10:52", "throughput": 5168.06, "total_tokens": 39392416} {"current_steps": 39080, "total_steps": 42420, "loss": 0.3044, "lr": 1.877712853155672e-05, "epoch": 18.425271098538424, "percentage": 92.13, "elapsed_time": "2:07:03", "remaining_time": "0:10:51", "throughput": 5168.12, "total_tokens": 39398048} {"current_steps": 39085, "total_steps": 42420, "loss": 0.3684, "lr": 1.8721321586139983e-05, "epoch": 18.427628477133428, "percentage": 92.14, "elapsed_time": "2:07:04", "remaining_time": "0:10:50", "throughput": 5168.16, "total_tokens": 39403136} {"current_steps": 39090, "total_steps": 42420, "loss": 0.3628, "lr": 1.8665596113031758e-05, "epoch": 18.42998585572843, "percentage": 92.15, "elapsed_time": "2:07:05", "remaining_time": "0:10:49", "throughput": 5168.17, "total_tokens": 39407712} {"current_steps": 39095, "total_steps": 42420, "loss": 0.3405, "lr": 1.860995212166555e-05, "epoch": 18.432343234323433, "percentage": 92.16, "elapsed_time": "2:07:06", "remaining_time": "0:10:48", "throughput": 5168.21, "total_tokens": 39412960} {"current_steps": 39100, "total_steps": 42420, "loss": 0.3638, "lr": 1.8554389621460887e-05, "epoch": 18.434700612918434, "percentage": 92.17, "elapsed_time": "2:07:06", "remaining_time": "0:10:47", "throughput": 5168.24, "total_tokens": 39417856} {"current_steps": 39105, "total_steps": 42420, "loss": 0.3037, "lr": 1.8498908621823575e-05, "epoch": 18.437057991513438, "percentage": 92.19, "elapsed_time": "2:07:07", "remaining_time": "0:10:46", "throughput": 5168.27, "total_tokens": 39422720} {"current_steps": 39110, "total_steps": 42420, "loss": 0.3291, "lr": 1.8443509132145653e-05, "epoch": 18.43941537010844, "percentage": 92.2, "elapsed_time": "2:07:08", "remaining_time": "0:10:45", "throughput": 5168.29, "total_tokens": 39427552} {"current_steps": 39115, "total_steps": 42420, "loss": 0.3617, "lr": 1.838819116180529e-05, "epoch": 18.441772748703443, "percentage": 92.21, "elapsed_time": "2:07:09", "remaining_time": "0:10:44", "throughput": 5168.31, "total_tokens": 39432096} {"current_steps": 39120, "total_steps": 42420, "loss": 0.3023, "lr": 1.833295472016683e-05, "epoch": 18.444130127298443, "percentage": 92.22, "elapsed_time": "2:07:10", "remaining_time": "0:10:43", "throughput": 5168.33, "total_tokens": 39436672} {"current_steps": 39125, "total_steps": 42420, "loss": 0.3614, "lr": 1.8277799816580843e-05, "epoch": 18.446487505893447, "percentage": 92.23, "elapsed_time": "2:07:11", "remaining_time": "0:10:42", "throughput": 5168.39, "total_tokens": 39442496} {"current_steps": 39130, "total_steps": 42420, "loss": 0.306, "lr": 1.8222726460384198e-05, "epoch": 18.448844884488448, "percentage": 92.24, "elapsed_time": "2:07:12", "remaining_time": "0:10:41", "throughput": 5168.42, "total_tokens": 39447680} {"current_steps": 39135, "total_steps": 42420, "loss": 0.3444, "lr": 1.8167734660899936e-05, "epoch": 18.451202263083452, "percentage": 92.26, "elapsed_time": "2:07:13", "remaining_time": "0:10:40", "throughput": 5168.54, "total_tokens": 39454976} {"current_steps": 39140, "total_steps": 42420, "loss": 0.3351, "lr": 1.8112824427437057e-05, "epoch": 18.453559641678453, "percentage": 92.27, "elapsed_time": "2:07:14", "remaining_time": "0:10:39", "throughput": 5168.53, "total_tokens": 39459264} {"current_steps": 39145, "total_steps": 42420, "loss": 0.284, "lr": 1.805799576929107e-05, "epoch": 18.455917020273457, "percentage": 92.28, "elapsed_time": "2:07:15", "remaining_time": "0:10:38", "throughput": 5168.6, "total_tokens": 39465088} {"current_steps": 39150, "total_steps": 42420, "loss": 0.3998, "lr": 1.8003248695743436e-05, "epoch": 18.458274398868458, "percentage": 92.29, "elapsed_time": "2:07:16", "remaining_time": "0:10:37", "throughput": 5168.6, "total_tokens": 39469472} {"current_steps": 39155, "total_steps": 42420, "loss": 0.3165, "lr": 1.7948583216061964e-05, "epoch": 18.460631777463462, "percentage": 92.3, "elapsed_time": "2:07:17", "remaining_time": "0:10:36", "throughput": 5168.64, "total_tokens": 39474624} {"current_steps": 39160, "total_steps": 42420, "loss": 0.3686, "lr": 1.789399933950059e-05, "epoch": 18.462989156058462, "percentage": 92.31, "elapsed_time": "2:07:18", "remaining_time": "0:10:35", "throughput": 5168.64, "total_tokens": 39479072} {"current_steps": 39165, "total_steps": 42420, "loss": 0.373, "lr": 1.783949707529947e-05, "epoch": 18.465346534653467, "percentage": 92.33, "elapsed_time": "2:07:18", "remaining_time": "0:10:34", "throughput": 5168.63, "total_tokens": 39483104} {"current_steps": 39170, "total_steps": 42420, "loss": 0.3112, "lr": 1.7785076432684844e-05, "epoch": 18.467703913248467, "percentage": 92.34, "elapsed_time": "2:07:19", "remaining_time": "0:10:33", "throughput": 5168.68, "total_tokens": 39488256} {"current_steps": 39175, "total_steps": 42420, "loss": 0.3419, "lr": 1.773073742086928e-05, "epoch": 18.47006129184347, "percentage": 92.35, "elapsed_time": "2:07:20", "remaining_time": "0:10:32", "throughput": 5168.7, "total_tokens": 39493120} {"current_steps": 39180, "total_steps": 42420, "loss": 0.3899, "lr": 1.7676480049051367e-05, "epoch": 18.472418670438472, "percentage": 92.36, "elapsed_time": "2:07:21", "remaining_time": "0:10:31", "throughput": 5168.71, "total_tokens": 39497824} {"current_steps": 39185, "total_steps": 42420, "loss": 0.3336, "lr": 1.762230432641604e-05, "epoch": 18.474776049033476, "percentage": 92.37, "elapsed_time": "2:07:22", "remaining_time": "0:10:30", "throughput": 5168.71, "total_tokens": 39502368} {"current_steps": 39190, "total_steps": 42420, "loss": 0.3512, "lr": 1.7568210262134233e-05, "epoch": 18.477133427628477, "percentage": 92.39, "elapsed_time": "2:07:23", "remaining_time": "0:10:29", "throughput": 5168.75, "total_tokens": 39507584} {"current_steps": 39195, "total_steps": 42420, "loss": 0.3601, "lr": 1.7514197865363178e-05, "epoch": 18.47949080622348, "percentage": 92.4, "elapsed_time": "2:07:24", "remaining_time": "0:10:29", "throughput": 5168.83, "total_tokens": 39513504} {"current_steps": 39200, "total_steps": 42420, "loss": 0.3142, "lr": 1.7460267145246346e-05, "epoch": 18.48184818481848, "percentage": 92.41, "elapsed_time": "2:07:25", "remaining_time": "0:10:28", "throughput": 5168.85, "total_tokens": 39518208} {"current_steps": 39205, "total_steps": 42420, "loss": 0.3397, "lr": 1.7406418110913146e-05, "epoch": 18.484205563413486, "percentage": 92.42, "elapsed_time": "2:07:26", "remaining_time": "0:10:27", "throughput": 5168.87, "total_tokens": 39523296} {"current_steps": 39210, "total_steps": 42420, "loss": 0.3715, "lr": 1.7352650771479462e-05, "epoch": 18.486562942008486, "percentage": 92.43, "elapsed_time": "2:07:27", "remaining_time": "0:10:26", "throughput": 5168.94, "total_tokens": 39528992} {"current_steps": 39215, "total_steps": 42420, "loss": 0.3554, "lr": 1.7298965136047118e-05, "epoch": 18.48892032060349, "percentage": 92.44, "elapsed_time": "2:07:28", "remaining_time": "0:10:25", "throughput": 5168.98, "total_tokens": 39534368} {"current_steps": 39220, "total_steps": 42420, "loss": 0.331, "lr": 1.724536121370418e-05, "epoch": 18.49127769919849, "percentage": 92.46, "elapsed_time": "2:07:29", "remaining_time": "0:10:24", "throughput": 5168.99, "total_tokens": 39538816} {"current_steps": 39225, "total_steps": 42420, "loss": 0.3168, "lr": 1.7191839013524836e-05, "epoch": 18.493635077793492, "percentage": 92.47, "elapsed_time": "2:07:30", "remaining_time": "0:10:23", "throughput": 5169.02, "total_tokens": 39543776} {"current_steps": 39230, "total_steps": 42420, "loss": 0.3491, "lr": 1.7138398544569557e-05, "epoch": 18.495992456388496, "percentage": 92.48, "elapsed_time": "2:07:31", "remaining_time": "0:10:22", "throughput": 5169.08, "total_tokens": 39549344} {"current_steps": 39235, "total_steps": 42420, "loss": 0.3224, "lr": 1.7085039815884883e-05, "epoch": 18.498349834983497, "percentage": 92.49, "elapsed_time": "2:07:32", "remaining_time": "0:10:21", "throughput": 5169.11, "total_tokens": 39554304} {"current_steps": 39240, "total_steps": 42420, "loss": 0.3111, "lr": 1.703176283650354e-05, "epoch": 18.5007072135785, "percentage": 92.5, "elapsed_time": "2:07:32", "remaining_time": "0:10:20", "throughput": 5169.13, "total_tokens": 39558976} {"current_steps": 39245, "total_steps": 42420, "loss": 0.3565, "lr": 1.697856761544436e-05, "epoch": 18.5030645921735, "percentage": 92.52, "elapsed_time": "2:07:33", "remaining_time": "0:10:19", "throughput": 5169.13, "total_tokens": 39563456} {"current_steps": 39250, "total_steps": 42420, "loss": 0.3301, "lr": 1.6925454161712484e-05, "epoch": 18.505421970768506, "percentage": 92.53, "elapsed_time": "2:07:34", "remaining_time": "0:10:18", "throughput": 5169.2, "total_tokens": 39568992} {"current_steps": 39255, "total_steps": 42420, "loss": 0.3214, "lr": 1.6872422484299156e-05, "epoch": 18.507779349363506, "percentage": 92.54, "elapsed_time": "2:07:35", "remaining_time": "0:10:17", "throughput": 5169.21, "total_tokens": 39573472} {"current_steps": 39260, "total_steps": 42420, "loss": 0.319, "lr": 1.6819472592181585e-05, "epoch": 18.51013672795851, "percentage": 92.55, "elapsed_time": "2:07:36", "remaining_time": "0:10:16", "throughput": 5169.28, "total_tokens": 39579392} {"current_steps": 39265, "total_steps": 42420, "loss": 0.3574, "lr": 1.6766604494323377e-05, "epoch": 18.51249410655351, "percentage": 92.56, "elapsed_time": "2:07:37", "remaining_time": "0:10:15", "throughput": 5169.31, "total_tokens": 39584320} {"current_steps": 39270, "total_steps": 42420, "loss": 0.3272, "lr": 1.6713818199674212e-05, "epoch": 18.514851485148515, "percentage": 92.57, "elapsed_time": "2:07:38", "remaining_time": "0:10:14", "throughput": 5169.33, "total_tokens": 39589472} {"current_steps": 39275, "total_steps": 42420, "loss": 0.3178, "lr": 1.6661113717169828e-05, "epoch": 18.517208863743516, "percentage": 92.59, "elapsed_time": "2:07:39", "remaining_time": "0:10:13", "throughput": 5169.34, "total_tokens": 39594048} {"current_steps": 39280, "total_steps": 42420, "loss": 0.3173, "lr": 1.6608491055732312e-05, "epoch": 18.51956624233852, "percentage": 92.6, "elapsed_time": "2:07:40", "remaining_time": "0:10:12", "throughput": 5169.37, "total_tokens": 39598880} {"current_steps": 39285, "total_steps": 42420, "loss": 0.3567, "lr": 1.6555950224269756e-05, "epoch": 18.52192362093352, "percentage": 92.61, "elapsed_time": "2:07:41", "remaining_time": "0:10:11", "throughput": 5169.42, "total_tokens": 39604352} {"current_steps": 39290, "total_steps": 42420, "loss": 0.3271, "lr": 1.6503491231676383e-05, "epoch": 18.524280999528525, "percentage": 92.62, "elapsed_time": "2:07:42", "remaining_time": "0:10:10", "throughput": 5169.49, "total_tokens": 39610304} {"current_steps": 39295, "total_steps": 42420, "loss": 0.3392, "lr": 1.6451114086832696e-05, "epoch": 18.526638378123526, "percentage": 92.63, "elapsed_time": "2:07:43", "remaining_time": "0:10:09", "throughput": 5169.5, "total_tokens": 39615040} {"current_steps": 39300, "total_steps": 42420, "loss": 0.3421, "lr": 1.6398818798605165e-05, "epoch": 18.52899575671853, "percentage": 92.64, "elapsed_time": "2:07:44", "remaining_time": "0:10:08", "throughput": 5169.55, "total_tokens": 39620384} {"current_steps": 39305, "total_steps": 42420, "loss": 0.3423, "lr": 1.634660537584648e-05, "epoch": 18.53135313531353, "percentage": 92.66, "elapsed_time": "2:07:45", "remaining_time": "0:10:07", "throughput": 5169.66, "total_tokens": 39627360} {"current_steps": 39310, "total_steps": 42420, "loss": 0.3362, "lr": 1.6294473827395573e-05, "epoch": 18.533710513908535, "percentage": 92.67, "elapsed_time": "2:07:46", "remaining_time": "0:10:06", "throughput": 5169.68, "total_tokens": 39632000} {"current_steps": 39315, "total_steps": 42420, "loss": 0.3484, "lr": 1.624242416207744e-05, "epoch": 18.536067892503535, "percentage": 92.68, "elapsed_time": "2:07:47", "remaining_time": "0:10:05", "throughput": 5169.69, "total_tokens": 39636576} {"current_steps": 39320, "total_steps": 42420, "loss": 0.3344, "lr": 1.619045638870309e-05, "epoch": 18.53842527109854, "percentage": 92.69, "elapsed_time": "2:07:48", "remaining_time": "0:10:04", "throughput": 5169.72, "total_tokens": 39641440} {"current_steps": 39325, "total_steps": 42420, "loss": 0.2866, "lr": 1.6138570516069873e-05, "epoch": 18.54078264969354, "percentage": 92.7, "elapsed_time": "2:07:48", "remaining_time": "0:10:03", "throughput": 5169.72, "total_tokens": 39645824} {"current_steps": 39330, "total_steps": 42420, "loss": 0.3503, "lr": 1.6086766552961208e-05, "epoch": 18.543140028288544, "percentage": 92.72, "elapsed_time": "2:07:49", "remaining_time": "0:10:02", "throughput": 5169.72, "total_tokens": 39650176} {"current_steps": 39335, "total_steps": 42420, "loss": 0.3655, "lr": 1.603504450814658e-05, "epoch": 18.545497406883545, "percentage": 92.73, "elapsed_time": "2:07:50", "remaining_time": "0:10:01", "throughput": 5169.8, "total_tokens": 39656448} {"current_steps": 39340, "total_steps": 42420, "loss": 0.339, "lr": 1.5983404390381705e-05, "epoch": 18.54785478547855, "percentage": 92.74, "elapsed_time": "2:07:51", "remaining_time": "0:10:00", "throughput": 5169.8, "total_tokens": 39660992} {"current_steps": 39345, "total_steps": 42420, "loss": 0.3352, "lr": 1.5931846208408264e-05, "epoch": 18.55021216407355, "percentage": 92.75, "elapsed_time": "2:07:52", "remaining_time": "0:09:59", "throughput": 5169.84, "total_tokens": 39666208} {"current_steps": 39350, "total_steps": 42420, "loss": 0.3565, "lr": 1.5880369970954267e-05, "epoch": 18.552569542668554, "percentage": 92.76, "elapsed_time": "2:07:53", "remaining_time": "0:09:58", "throughput": 5169.82, "total_tokens": 39670208} {"current_steps": 39355, "total_steps": 42420, "loss": 0.3468, "lr": 1.582897568673375e-05, "epoch": 18.554926921263554, "percentage": 92.77, "elapsed_time": "2:07:54", "remaining_time": "0:09:57", "throughput": 5169.9, "total_tokens": 39675904} {"current_steps": 39360, "total_steps": 42420, "loss": 0.3145, "lr": 1.5777663364446858e-05, "epoch": 18.55728429985856, "percentage": 92.79, "elapsed_time": "2:07:55", "remaining_time": "0:09:56", "throughput": 5169.96, "total_tokens": 39681600} {"current_steps": 39365, "total_steps": 42420, "loss": 0.3525, "lr": 1.5726433012779927e-05, "epoch": 18.55964167845356, "percentage": 92.8, "elapsed_time": "2:07:56", "remaining_time": "0:09:55", "throughput": 5170.09, "total_tokens": 39688864} {"current_steps": 39370, "total_steps": 42420, "loss": 0.3186, "lr": 1.5675284640405462e-05, "epoch": 18.561999057048563, "percentage": 92.81, "elapsed_time": "2:07:57", "remaining_time": "0:09:54", "throughput": 5170.09, "total_tokens": 39693312} {"current_steps": 39375, "total_steps": 42420, "loss": 0.321, "lr": 1.562421825598187e-05, "epoch": 18.564356435643564, "percentage": 92.82, "elapsed_time": "2:07:58", "remaining_time": "0:09:53", "throughput": 5170.17, "total_tokens": 39699616} {"current_steps": 39380, "total_steps": 42420, "loss": 0.313, "lr": 1.5573233868153957e-05, "epoch": 18.566713814238568, "percentage": 92.83, "elapsed_time": "2:07:59", "remaining_time": "0:09:52", "throughput": 5170.21, "total_tokens": 39704864} {"current_steps": 39385, "total_steps": 42420, "loss": 0.3593, "lr": 1.5522331485552432e-05, "epoch": 18.56907119283357, "percentage": 92.85, "elapsed_time": "2:08:00", "remaining_time": "0:09:51", "throughput": 5170.21, "total_tokens": 39709152} {"current_steps": 39390, "total_steps": 42420, "loss": 0.319, "lr": 1.547151111679418e-05, "epoch": 18.571428571428573, "percentage": 92.86, "elapsed_time": "2:08:01", "remaining_time": "0:09:50", "throughput": 5170.24, "total_tokens": 39714112} {"current_steps": 39395, "total_steps": 42420, "loss": 0.3185, "lr": 1.5420772770482317e-05, "epoch": 18.573785950023574, "percentage": 92.87, "elapsed_time": "2:08:02", "remaining_time": "0:09:49", "throughput": 5170.25, "total_tokens": 39719136} {"current_steps": 39400, "total_steps": 42420, "loss": 0.3192, "lr": 1.537011645520592e-05, "epoch": 18.576143328618578, "percentage": 92.88, "elapsed_time": "2:08:03", "remaining_time": "0:09:48", "throughput": 5170.31, "total_tokens": 39724608} {"current_steps": 39405, "total_steps": 42420, "loss": 0.3605, "lr": 1.5319542179540237e-05, "epoch": 18.57850070721358, "percentage": 92.89, "elapsed_time": "2:08:04", "remaining_time": "0:09:47", "throughput": 5170.32, "total_tokens": 39728960} {"current_steps": 39410, "total_steps": 42420, "loss": 0.3665, "lr": 1.52690499520467e-05, "epoch": 18.580858085808583, "percentage": 92.9, "elapsed_time": "2:08:05", "remaining_time": "0:09:46", "throughput": 5170.38, "total_tokens": 39734944} {"current_steps": 39415, "total_steps": 42420, "loss": 0.2808, "lr": 1.5218639781272693e-05, "epoch": 18.583215464403583, "percentage": 92.92, "elapsed_time": "2:08:05", "remaining_time": "0:09:45", "throughput": 5170.39, "total_tokens": 39739328} {"current_steps": 39420, "total_steps": 42420, "loss": 0.3683, "lr": 1.5168311675751945e-05, "epoch": 18.585572842998587, "percentage": 92.93, "elapsed_time": "2:08:06", "remaining_time": "0:09:44", "throughput": 5170.43, "total_tokens": 39744512} {"current_steps": 39425, "total_steps": 42420, "loss": 0.3173, "lr": 1.5118065644003976e-05, "epoch": 18.587930221593588, "percentage": 92.94, "elapsed_time": "2:08:07", "remaining_time": "0:09:44", "throughput": 5170.41, "total_tokens": 39748480} {"current_steps": 39430, "total_steps": 42420, "loss": 0.3298, "lr": 1.5067901694534702e-05, "epoch": 18.59028760018859, "percentage": 92.95, "elapsed_time": "2:08:08", "remaining_time": "0:09:43", "throughput": 5170.43, "total_tokens": 39753088} {"current_steps": 39435, "total_steps": 42420, "loss": 0.2995, "lr": 1.5017819835835945e-05, "epoch": 18.592644978783593, "percentage": 92.96, "elapsed_time": "2:08:09", "remaining_time": "0:09:42", "throughput": 5170.44, "total_tokens": 39757696} {"current_steps": 39440, "total_steps": 42420, "loss": 0.3106, "lr": 1.4967820076385808e-05, "epoch": 18.595002357378593, "percentage": 92.98, "elapsed_time": "2:08:10", "remaining_time": "0:09:41", "throughput": 5170.5, "total_tokens": 39763520} {"current_steps": 39445, "total_steps": 42420, "loss": 0.3205, "lr": 1.4917902424648356e-05, "epoch": 18.597359735973598, "percentage": 92.99, "elapsed_time": "2:08:11", "remaining_time": "0:09:40", "throughput": 5170.5, "total_tokens": 39768000} {"current_steps": 39450, "total_steps": 42420, "loss": 0.363, "lr": 1.4868066889073772e-05, "epoch": 18.599717114568598, "percentage": 93.0, "elapsed_time": "2:08:12", "remaining_time": "0:09:39", "throughput": 5170.54, "total_tokens": 39773152} {"current_steps": 39455, "total_steps": 42420, "loss": 0.3501, "lr": 1.4818313478098366e-05, "epoch": 18.602074493163602, "percentage": 93.01, "elapsed_time": "2:08:13", "remaining_time": "0:09:38", "throughput": 5170.57, "total_tokens": 39777888} {"current_steps": 39460, "total_steps": 42420, "loss": 0.3549, "lr": 1.4768642200144678e-05, "epoch": 18.604431871758603, "percentage": 93.02, "elapsed_time": "2:08:14", "remaining_time": "0:09:37", "throughput": 5170.62, "total_tokens": 39783616} {"current_steps": 39465, "total_steps": 42420, "loss": 0.3382, "lr": 1.4719053063621035e-05, "epoch": 18.606789250353607, "percentage": 93.03, "elapsed_time": "2:08:15", "remaining_time": "0:09:36", "throughput": 5170.66, "total_tokens": 39788864} {"current_steps": 39470, "total_steps": 42420, "loss": 0.3099, "lr": 1.4669546076922113e-05, "epoch": 18.609146628948608, "percentage": 93.05, "elapsed_time": "2:08:15", "remaining_time": "0:09:35", "throughput": 5170.64, "total_tokens": 39792768} {"current_steps": 39475, "total_steps": 42420, "loss": 0.3421, "lr": 1.4620121248428541e-05, "epoch": 18.611504007543612, "percentage": 93.06, "elapsed_time": "2:08:16", "remaining_time": "0:09:34", "throughput": 5170.68, "total_tokens": 39797792} {"current_steps": 39480, "total_steps": 42420, "loss": 0.3069, "lr": 1.4570778586507238e-05, "epoch": 18.613861386138613, "percentage": 93.07, "elapsed_time": "2:08:17", "remaining_time": "0:09:33", "throughput": 5170.67, "total_tokens": 39801792} {"current_steps": 39485, "total_steps": 42420, "loss": 0.3679, "lr": 1.4521518099510966e-05, "epoch": 18.616218764733617, "percentage": 93.08, "elapsed_time": "2:08:18", "remaining_time": "0:09:32", "throughput": 5170.71, "total_tokens": 39807264} {"current_steps": 39490, "total_steps": 42420, "loss": 0.329, "lr": 1.4472339795778721e-05, "epoch": 18.618576143328617, "percentage": 93.09, "elapsed_time": "2:08:19", "remaining_time": "0:09:31", "throughput": 5170.73, "total_tokens": 39811904} {"current_steps": 39495, "total_steps": 42420, "loss": 0.3298, "lr": 1.4423243683635567e-05, "epoch": 18.62093352192362, "percentage": 93.1, "elapsed_time": "2:08:20", "remaining_time": "0:09:30", "throughput": 5170.75, "total_tokens": 39816832} {"current_steps": 39500, "total_steps": 42420, "loss": 0.3554, "lr": 1.4374229771392688e-05, "epoch": 18.623290900518622, "percentage": 93.12, "elapsed_time": "2:08:21", "remaining_time": "0:09:29", "throughput": 5170.75, "total_tokens": 39821120} {"current_steps": 39505, "total_steps": 42420, "loss": 0.3829, "lr": 1.432529806734717e-05, "epoch": 18.625648279113626, "percentage": 93.13, "elapsed_time": "2:08:22", "remaining_time": "0:09:28", "throughput": 5170.84, "total_tokens": 39827072} {"current_steps": 39510, "total_steps": 42420, "loss": 0.2992, "lr": 1.4276448579782442e-05, "epoch": 18.628005657708627, "percentage": 93.14, "elapsed_time": "2:08:23", "remaining_time": "0:09:27", "throughput": 5170.84, "total_tokens": 39831584} {"current_steps": 39515, "total_steps": 42420, "loss": 0.3763, "lr": 1.4227681316967834e-05, "epoch": 18.63036303630363, "percentage": 93.15, "elapsed_time": "2:08:24", "remaining_time": "0:09:26", "throughput": 5170.92, "total_tokens": 39837632} {"current_steps": 39520, "total_steps": 42420, "loss": 0.3413, "lr": 1.4178996287158852e-05, "epoch": 18.632720414898632, "percentage": 93.16, "elapsed_time": "2:08:25", "remaining_time": "0:09:25", "throughput": 5170.97, "total_tokens": 39843040} {"current_steps": 39525, "total_steps": 42420, "loss": 0.3134, "lr": 1.4130393498597017e-05, "epoch": 18.635077793493636, "percentage": 93.18, "elapsed_time": "2:08:26", "remaining_time": "0:09:24", "throughput": 5170.99, "total_tokens": 39848064} {"current_steps": 39530, "total_steps": 42420, "loss": 0.3132, "lr": 1.4081872959509968e-05, "epoch": 18.637435172088637, "percentage": 93.19, "elapsed_time": "2:08:26", "remaining_time": "0:09:23", "throughput": 5171.0, "total_tokens": 39852672} {"current_steps": 39535, "total_steps": 42420, "loss": 0.3084, "lr": 1.4033434678111413e-05, "epoch": 18.63979255068364, "percentage": 93.2, "elapsed_time": "2:08:27", "remaining_time": "0:09:22", "throughput": 5171.01, "total_tokens": 39857152} {"current_steps": 39540, "total_steps": 42420, "loss": 0.315, "lr": 1.398507866260118e-05, "epoch": 18.64214992927864, "percentage": 93.21, "elapsed_time": "2:08:28", "remaining_time": "0:09:21", "throughput": 5171.0, "total_tokens": 39861568} {"current_steps": 39545, "total_steps": 42420, "loss": 0.3726, "lr": 1.3936804921165003e-05, "epoch": 18.644507307873646, "percentage": 93.22, "elapsed_time": "2:08:29", "remaining_time": "0:09:20", "throughput": 5171.01, "total_tokens": 39865952} {"current_steps": 39550, "total_steps": 42420, "loss": 0.289, "lr": 1.388861346197484e-05, "epoch": 18.646864686468646, "percentage": 93.23, "elapsed_time": "2:08:30", "remaining_time": "0:09:19", "throughput": 5171.03, "total_tokens": 39870656} {"current_steps": 39555, "total_steps": 42420, "loss": 0.3029, "lr": 1.3840504293188727e-05, "epoch": 18.64922206506365, "percentage": 93.25, "elapsed_time": "2:08:31", "remaining_time": "0:09:18", "throughput": 5171.07, "total_tokens": 39875968} {"current_steps": 39560, "total_steps": 42420, "loss": 0.2911, "lr": 1.3792477422950755e-05, "epoch": 18.65157944365865, "percentage": 93.26, "elapsed_time": "2:08:32", "remaining_time": "0:09:17", "throughput": 5171.12, "total_tokens": 39881088} {"current_steps": 39565, "total_steps": 42420, "loss": 0.3734, "lr": 1.3744532859391034e-05, "epoch": 18.653936822253655, "percentage": 93.27, "elapsed_time": "2:08:33", "remaining_time": "0:09:16", "throughput": 5171.18, "total_tokens": 39886656} {"current_steps": 39570, "total_steps": 42420, "loss": 0.3058, "lr": 1.3696670610625738e-05, "epoch": 18.656294200848656, "percentage": 93.28, "elapsed_time": "2:08:34", "remaining_time": "0:09:15", "throughput": 5171.2, "total_tokens": 39891616} {"current_steps": 39575, "total_steps": 42420, "loss": 0.3805, "lr": 1.3648890684757164e-05, "epoch": 18.65865157944366, "percentage": 93.29, "elapsed_time": "2:08:35", "remaining_time": "0:09:14", "throughput": 5171.2, "total_tokens": 39896064} {"current_steps": 39580, "total_steps": 42420, "loss": 0.313, "lr": 1.3601193089873566e-05, "epoch": 18.66100895803866, "percentage": 93.31, "elapsed_time": "2:08:36", "remaining_time": "0:09:13", "throughput": 5171.26, "total_tokens": 39901696} {"current_steps": 39585, "total_steps": 42420, "loss": 0.3074, "lr": 1.3553577834049424e-05, "epoch": 18.663366336633665, "percentage": 93.32, "elapsed_time": "2:08:36", "remaining_time": "0:09:12", "throughput": 5171.26, "total_tokens": 39906368} {"current_steps": 39590, "total_steps": 42420, "loss": 0.3703, "lr": 1.3506044925345185e-05, "epoch": 18.665723715228665, "percentage": 93.33, "elapsed_time": "2:08:37", "remaining_time": "0:09:11", "throughput": 5171.29, "total_tokens": 39911264} {"current_steps": 39595, "total_steps": 42420, "loss": 0.3175, "lr": 1.3458594371807354e-05, "epoch": 18.66808109382367, "percentage": 93.34, "elapsed_time": "2:08:38", "remaining_time": "0:09:10", "throughput": 5171.34, "total_tokens": 39916992} {"current_steps": 39600, "total_steps": 42420, "loss": 0.342, "lr": 1.341122618146845e-05, "epoch": 18.67043847241867, "percentage": 93.35, "elapsed_time": "2:08:39", "remaining_time": "0:09:09", "throughput": 5171.37, "total_tokens": 39921984} {"current_steps": 39605, "total_steps": 42420, "loss": 0.3501, "lr": 1.3363940362347171e-05, "epoch": 18.672795851013674, "percentage": 93.36, "elapsed_time": "2:08:40", "remaining_time": "0:09:08", "throughput": 5171.4, "total_tokens": 39927168} {"current_steps": 39610, "total_steps": 42420, "loss": 0.3029, "lr": 1.331673692244817e-05, "epoch": 18.675153229608675, "percentage": 93.38, "elapsed_time": "2:08:41", "remaining_time": "0:09:07", "throughput": 5171.42, "total_tokens": 39931904} {"current_steps": 39615, "total_steps": 42420, "loss": 0.3707, "lr": 1.3269615869762163e-05, "epoch": 18.677510608203676, "percentage": 93.39, "elapsed_time": "2:08:42", "remaining_time": "0:09:06", "throughput": 5171.43, "total_tokens": 39936384} {"current_steps": 39620, "total_steps": 42420, "loss": 0.3218, "lr": 1.3222577212265996e-05, "epoch": 18.67986798679868, "percentage": 93.4, "elapsed_time": "2:08:43", "remaining_time": "0:09:05", "throughput": 5171.43, "total_tokens": 39940864} {"current_steps": 39625, "total_steps": 42420, "loss": 0.3562, "lr": 1.3175620957922519e-05, "epoch": 18.68222536539368, "percentage": 93.41, "elapsed_time": "2:08:44", "remaining_time": "0:09:04", "throughput": 5171.51, "total_tokens": 39947008} {"current_steps": 39630, "total_steps": 42420, "loss": 0.3529, "lr": 1.312874711468065e-05, "epoch": 18.684582743988685, "percentage": 93.42, "elapsed_time": "2:08:45", "remaining_time": "0:09:03", "throughput": 5171.55, "total_tokens": 39952032} {"current_steps": 39635, "total_steps": 42420, "loss": 0.3252, "lr": 1.3081955690475212e-05, "epoch": 18.686940122583685, "percentage": 93.43, "elapsed_time": "2:08:46", "remaining_time": "0:09:02", "throughput": 5171.53, "total_tokens": 39956096} {"current_steps": 39640, "total_steps": 42420, "loss": 0.3209, "lr": 1.303524669322731e-05, "epoch": 18.68929750117869, "percentage": 93.45, "elapsed_time": "2:08:47", "remaining_time": "0:09:01", "throughput": 5171.64, "total_tokens": 39962880} {"current_steps": 39645, "total_steps": 42420, "loss": 0.3422, "lr": 1.2988620130843954e-05, "epoch": 18.69165487977369, "percentage": 93.46, "elapsed_time": "2:08:48", "remaining_time": "0:09:00", "throughput": 5171.7, "total_tokens": 39968352} {"current_steps": 39650, "total_steps": 42420, "loss": 0.3314, "lr": 1.2942076011218218e-05, "epoch": 18.694012258368694, "percentage": 93.47, "elapsed_time": "2:08:49", "remaining_time": "0:08:59", "throughput": 5171.73, "total_tokens": 39973600} {"current_steps": 39655, "total_steps": 42420, "loss": 0.346, "lr": 1.2895614342229245e-05, "epoch": 18.696369636963695, "percentage": 93.48, "elapsed_time": "2:08:50", "remaining_time": "0:08:58", "throughput": 5171.73, "total_tokens": 39977824} {"current_steps": 39660, "total_steps": 42420, "loss": 0.3366, "lr": 1.2849235131742187e-05, "epoch": 18.6987270155587, "percentage": 93.49, "elapsed_time": "2:08:51", "remaining_time": "0:08:58", "throughput": 5171.76, "total_tokens": 39983264} {"current_steps": 39665, "total_steps": 42420, "loss": 0.3141, "lr": 1.2802938387608265e-05, "epoch": 18.7010843941537, "percentage": 93.51, "elapsed_time": "2:08:51", "remaining_time": "0:08:57", "throughput": 5171.76, "total_tokens": 39987488} {"current_steps": 39670, "total_steps": 42420, "loss": 0.3333, "lr": 1.275672411766482e-05, "epoch": 18.703441772748704, "percentage": 93.52, "elapsed_time": "2:08:52", "remaining_time": "0:08:56", "throughput": 5171.79, "total_tokens": 39992448} {"current_steps": 39675, "total_steps": 42420, "loss": 0.2509, "lr": 1.2710592329734983e-05, "epoch": 18.705799151343705, "percentage": 93.53, "elapsed_time": "2:08:53", "remaining_time": "0:08:55", "throughput": 5171.79, "total_tokens": 39997152} {"current_steps": 39680, "total_steps": 42420, "loss": 0.3465, "lr": 1.2664543031628173e-05, "epoch": 18.70815652993871, "percentage": 93.54, "elapsed_time": "2:08:54", "remaining_time": "0:08:54", "throughput": 5171.76, "total_tokens": 40000864} {"current_steps": 39685, "total_steps": 42420, "loss": 0.3084, "lr": 1.2618576231139766e-05, "epoch": 18.71051390853371, "percentage": 93.55, "elapsed_time": "2:08:55", "remaining_time": "0:08:53", "throughput": 5171.81, "total_tokens": 40006304} {"current_steps": 39690, "total_steps": 42420, "loss": 0.3338, "lr": 1.2572691936051095e-05, "epoch": 18.712871287128714, "percentage": 93.56, "elapsed_time": "2:08:56", "remaining_time": "0:08:52", "throughput": 5171.79, "total_tokens": 40010496} {"current_steps": 39695, "total_steps": 42420, "loss": 0.3282, "lr": 1.2526890154129667e-05, "epoch": 18.715228665723714, "percentage": 93.58, "elapsed_time": "2:08:57", "remaining_time": "0:08:51", "throughput": 5171.84, "total_tokens": 40016000} {"current_steps": 39700, "total_steps": 42420, "loss": 0.3083, "lr": 1.248117089312889e-05, "epoch": 18.71758604431872, "percentage": 93.59, "elapsed_time": "2:08:58", "remaining_time": "0:08:50", "throughput": 5171.83, "total_tokens": 40020224} {"current_steps": 39705, "total_steps": 42420, "loss": 0.3824, "lr": 1.2435534160788298e-05, "epoch": 18.71994342291372, "percentage": 93.6, "elapsed_time": "2:08:59", "remaining_time": "0:08:49", "throughput": 5171.86, "total_tokens": 40025376} {"current_steps": 39710, "total_steps": 42420, "loss": 0.3265, "lr": 1.2389979964833432e-05, "epoch": 18.722300801508723, "percentage": 93.61, "elapsed_time": "2:08:59", "remaining_time": "0:08:48", "throughput": 5171.86, "total_tokens": 40029728} {"current_steps": 39715, "total_steps": 42420, "loss": 0.3566, "lr": 1.2344508312975788e-05, "epoch": 18.724658180103724, "percentage": 93.62, "elapsed_time": "2:09:00", "remaining_time": "0:08:47", "throughput": 5171.86, "total_tokens": 40033920} {"current_steps": 39720, "total_steps": 42420, "loss": 0.3967, "lr": 1.2299119212912935e-05, "epoch": 18.727015558698728, "percentage": 93.64, "elapsed_time": "2:09:01", "remaining_time": "0:08:46", "throughput": 5171.87, "total_tokens": 40038720} {"current_steps": 39725, "total_steps": 42420, "loss": 0.3994, "lr": 1.2253812672328557e-05, "epoch": 18.72937293729373, "percentage": 93.65, "elapsed_time": "2:09:02", "remaining_time": "0:08:45", "throughput": 5171.94, "total_tokens": 40044608} {"current_steps": 39730, "total_steps": 42420, "loss": 0.3642, "lr": 1.2208588698892186e-05, "epoch": 18.731730315888733, "percentage": 93.66, "elapsed_time": "2:09:03", "remaining_time": "0:08:44", "throughput": 5171.99, "total_tokens": 40050016} {"current_steps": 39735, "total_steps": 42420, "loss": 0.3871, "lr": 1.2163447300259534e-05, "epoch": 18.734087694483733, "percentage": 93.67, "elapsed_time": "2:09:04", "remaining_time": "0:08:43", "throughput": 5171.99, "total_tokens": 40054336} {"current_steps": 39740, "total_steps": 42420, "loss": 0.3081, "lr": 1.2118388484072263e-05, "epoch": 18.736445073078738, "percentage": 93.68, "elapsed_time": "2:09:05", "remaining_time": "0:08:42", "throughput": 5172.0, "total_tokens": 40058560} {"current_steps": 39745, "total_steps": 42420, "loss": 0.3206, "lr": 1.2073412257958105e-05, "epoch": 18.738802451673738, "percentage": 93.69, "elapsed_time": "2:09:06", "remaining_time": "0:08:41", "throughput": 5172.0, "total_tokens": 40062976} {"current_steps": 39750, "total_steps": 42420, "loss": 0.3641, "lr": 1.2028518629530749e-05, "epoch": 18.741159830268742, "percentage": 93.71, "elapsed_time": "2:09:06", "remaining_time": "0:08:40", "throughput": 5171.98, "total_tokens": 40067104} {"current_steps": 39755, "total_steps": 42420, "loss": 0.3221, "lr": 1.1983707606389838e-05, "epoch": 18.743517208863743, "percentage": 93.72, "elapsed_time": "2:09:07", "remaining_time": "0:08:39", "throughput": 5172.01, "total_tokens": 40071904} {"current_steps": 39760, "total_steps": 42420, "loss": 0.3293, "lr": 1.1938979196121192e-05, "epoch": 18.745874587458747, "percentage": 93.73, "elapsed_time": "2:09:08", "remaining_time": "0:08:38", "throughput": 5172.01, "total_tokens": 40076320} {"current_steps": 39765, "total_steps": 42420, "loss": 0.3777, "lr": 1.1894333406296586e-05, "epoch": 18.748231966053748, "percentage": 93.74, "elapsed_time": "2:09:09", "remaining_time": "0:08:37", "throughput": 5171.99, "total_tokens": 40080480} {"current_steps": 39770, "total_steps": 42420, "loss": 0.3201, "lr": 1.1849770244473701e-05, "epoch": 18.750589344648752, "percentage": 93.75, "elapsed_time": "2:09:10", "remaining_time": "0:08:36", "throughput": 5172.01, "total_tokens": 40085088} {"current_steps": 39775, "total_steps": 42420, "loss": 0.3335, "lr": 1.1805289718196499e-05, "epoch": 18.752946723243753, "percentage": 93.76, "elapsed_time": "2:09:11", "remaining_time": "0:08:35", "throughput": 5172.09, "total_tokens": 40091232} {"current_steps": 39780, "total_steps": 42420, "loss": 0.3228, "lr": 1.176089183499457e-05, "epoch": 18.755304101838757, "percentage": 93.78, "elapsed_time": "2:09:12", "remaining_time": "0:08:34", "throughput": 5172.08, "total_tokens": 40095552} {"current_steps": 39785, "total_steps": 42420, "loss": 0.3018, "lr": 1.17165766023839e-05, "epoch": 18.757661480433757, "percentage": 93.79, "elapsed_time": "2:09:13", "remaining_time": "0:08:33", "throughput": 5172.09, "total_tokens": 40100032} {"current_steps": 39790, "total_steps": 42420, "loss": 0.3261, "lr": 1.1672344027866266e-05, "epoch": 18.76001885902876, "percentage": 93.8, "elapsed_time": "2:09:14", "remaining_time": "0:08:32", "throughput": 5172.11, "total_tokens": 40104832} {"current_steps": 39795, "total_steps": 42420, "loss": 0.343, "lr": 1.1628194118929403e-05, "epoch": 18.762376237623762, "percentage": 93.81, "elapsed_time": "2:09:14", "remaining_time": "0:08:31", "throughput": 5172.14, "total_tokens": 40109728} {"current_steps": 39800, "total_steps": 42420, "loss": 0.3113, "lr": 1.1584126883047219e-05, "epoch": 18.764733616218766, "percentage": 93.82, "elapsed_time": "2:09:15", "remaining_time": "0:08:30", "throughput": 5172.17, "total_tokens": 40114656} {"current_steps": 39805, "total_steps": 42420, "loss": 0.3195, "lr": 1.1540142327679525e-05, "epoch": 18.767090994813767, "percentage": 93.84, "elapsed_time": "2:09:16", "remaining_time": "0:08:29", "throughput": 5172.2, "total_tokens": 40119808} {"current_steps": 39810, "total_steps": 42420, "loss": 0.3645, "lr": 1.1496240460272089e-05, "epoch": 18.76944837340877, "percentage": 93.85, "elapsed_time": "2:09:17", "remaining_time": "0:08:28", "throughput": 5172.23, "total_tokens": 40124736} {"current_steps": 39815, "total_steps": 42420, "loss": 0.3305, "lr": 1.1452421288256853e-05, "epoch": 18.77180575200377, "percentage": 93.86, "elapsed_time": "2:09:18", "remaining_time": "0:08:27", "throughput": 5172.22, "total_tokens": 40128896} {"current_steps": 39820, "total_steps": 42420, "loss": 0.3379, "lr": 1.1408684819051662e-05, "epoch": 18.774163130598772, "percentage": 93.87, "elapsed_time": "2:09:19", "remaining_time": "0:08:26", "throughput": 5172.26, "total_tokens": 40134240} {"current_steps": 39825, "total_steps": 42420, "loss": 0.3372, "lr": 1.1365031060060316e-05, "epoch": 18.776520509193777, "percentage": 93.88, "elapsed_time": "2:09:20", "remaining_time": "0:08:25", "throughput": 5172.27, "total_tokens": 40138656} {"current_steps": 39830, "total_steps": 42420, "loss": 0.3068, "lr": 1.1321460018672625e-05, "epoch": 18.778877887788777, "percentage": 93.89, "elapsed_time": "2:09:21", "remaining_time": "0:08:24", "throughput": 5172.32, "total_tokens": 40143872} {"current_steps": 39835, "total_steps": 42420, "loss": 0.298, "lr": 1.1277971702264466e-05, "epoch": 18.78123526638378, "percentage": 93.91, "elapsed_time": "2:09:22", "remaining_time": "0:08:23", "throughput": 5172.31, "total_tokens": 40148032} {"current_steps": 39840, "total_steps": 42420, "loss": 0.3116, "lr": 1.1234566118197731e-05, "epoch": 18.783592644978782, "percentage": 93.92, "elapsed_time": "2:09:22", "remaining_time": "0:08:22", "throughput": 5172.31, "total_tokens": 40152480} {"current_steps": 39845, "total_steps": 42420, "loss": 0.3345, "lr": 1.1191243273820151e-05, "epoch": 18.785950023573786, "percentage": 93.93, "elapsed_time": "2:09:23", "remaining_time": "0:08:21", "throughput": 5172.32, "total_tokens": 40156896} {"current_steps": 39850, "total_steps": 42420, "loss": 0.3358, "lr": 1.1148003176465581e-05, "epoch": 18.788307402168787, "percentage": 93.94, "elapsed_time": "2:09:24", "remaining_time": "0:08:20", "throughput": 5172.32, "total_tokens": 40161376} {"current_steps": 39855, "total_steps": 42420, "loss": 0.3456, "lr": 1.1104845833453836e-05, "epoch": 18.79066478076379, "percentage": 93.95, "elapsed_time": "2:09:25", "remaining_time": "0:08:19", "throughput": 5172.41, "total_tokens": 40167936} {"current_steps": 39860, "total_steps": 42420, "loss": 0.3481, "lr": 1.106177125209068e-05, "epoch": 18.79302215935879, "percentage": 93.97, "elapsed_time": "2:09:26", "remaining_time": "0:08:18", "throughput": 5172.43, "total_tokens": 40172704} {"current_steps": 39865, "total_steps": 42420, "loss": 0.3809, "lr": 1.1018779439668002e-05, "epoch": 18.795379537953796, "percentage": 93.98, "elapsed_time": "2:09:27", "remaining_time": "0:08:17", "throughput": 5172.48, "total_tokens": 40178048} {"current_steps": 39870, "total_steps": 42420, "loss": 0.2955, "lr": 1.0975870403463538e-05, "epoch": 18.797736916548796, "percentage": 93.99, "elapsed_time": "2:09:28", "remaining_time": "0:08:16", "throughput": 5172.5, "total_tokens": 40182848} {"current_steps": 39875, "total_steps": 42420, "loss": 0.3314, "lr": 1.0933044150741033e-05, "epoch": 18.8000942951438, "percentage": 94.0, "elapsed_time": "2:09:29", "remaining_time": "0:08:15", "throughput": 5172.51, "total_tokens": 40187424} {"current_steps": 39880, "total_steps": 42420, "loss": 0.3212, "lr": 1.0890300688750355e-05, "epoch": 18.8024516737388, "percentage": 94.01, "elapsed_time": "2:09:30", "remaining_time": "0:08:14", "throughput": 5172.5, "total_tokens": 40191520} {"current_steps": 39885, "total_steps": 42420, "loss": 0.3206, "lr": 1.0847640024727101e-05, "epoch": 18.804809052333805, "percentage": 94.02, "elapsed_time": "2:09:31", "remaining_time": "0:08:13", "throughput": 5172.51, "total_tokens": 40196320} {"current_steps": 39890, "total_steps": 42420, "loss": 0.3889, "lr": 1.0805062165892999e-05, "epoch": 18.807166430928806, "percentage": 94.04, "elapsed_time": "2:09:32", "remaining_time": "0:08:12", "throughput": 5172.57, "total_tokens": 40201632} {"current_steps": 39895, "total_steps": 42420, "loss": 0.3629, "lr": 1.076256711945589e-05, "epoch": 18.80952380952381, "percentage": 94.05, "elapsed_time": "2:09:32", "remaining_time": "0:08:11", "throughput": 5172.56, "total_tokens": 40205888} {"current_steps": 39900, "total_steps": 42420, "loss": 0.373, "lr": 1.0720154892609358e-05, "epoch": 18.81188118811881, "percentage": 94.06, "elapsed_time": "2:09:33", "remaining_time": "0:08:10", "throughput": 5172.64, "total_tokens": 40212000} {"current_steps": 39905, "total_steps": 42420, "loss": 0.3427, "lr": 1.0677825492533099e-05, "epoch": 18.814238566713815, "percentage": 94.07, "elapsed_time": "2:09:34", "remaining_time": "0:08:10", "throughput": 5172.67, "total_tokens": 40216736} {"current_steps": 39910, "total_steps": 42420, "loss": 0.2903, "lr": 1.0635578926392775e-05, "epoch": 18.816595945308816, "percentage": 94.08, "elapsed_time": "2:09:35", "remaining_time": "0:08:09", "throughput": 5172.69, "total_tokens": 40221280} {"current_steps": 39915, "total_steps": 42420, "loss": 0.3258, "lr": 1.0593415201339996e-05, "epoch": 18.81895332390382, "percentage": 94.09, "elapsed_time": "2:09:36", "remaining_time": "0:08:08", "throughput": 5172.68, "total_tokens": 40225728} {"current_steps": 39920, "total_steps": 42420, "loss": 0.3313, "lr": 1.055133432451244e-05, "epoch": 18.82131070249882, "percentage": 94.11, "elapsed_time": "2:09:37", "remaining_time": "0:08:07", "throughput": 5172.72, "total_tokens": 40230848} {"current_steps": 39925, "total_steps": 42420, "loss": 0.3167, "lr": 1.0509336303033523e-05, "epoch": 18.823668081093825, "percentage": 94.12, "elapsed_time": "2:09:38", "remaining_time": "0:08:06", "throughput": 5172.76, "total_tokens": 40236128} {"current_steps": 39930, "total_steps": 42420, "loss": 0.2996, "lr": 1.0467421144012945e-05, "epoch": 18.826025459688825, "percentage": 94.13, "elapsed_time": "2:09:39", "remaining_time": "0:08:05", "throughput": 5172.8, "total_tokens": 40241248} {"current_steps": 39935, "total_steps": 42420, "loss": 0.3359, "lr": 1.0425588854546142e-05, "epoch": 18.82838283828383, "percentage": 94.14, "elapsed_time": "2:09:40", "remaining_time": "0:08:04", "throughput": 5172.83, "total_tokens": 40246080} {"current_steps": 39940, "total_steps": 42420, "loss": 0.3522, "lr": 1.038383944171467e-05, "epoch": 18.83074021687883, "percentage": 94.15, "elapsed_time": "2:09:41", "remaining_time": "0:08:03", "throughput": 5172.87, "total_tokens": 40251008} {"current_steps": 39945, "total_steps": 42420, "loss": 0.3382, "lr": 1.0342172912585934e-05, "epoch": 18.833097595473834, "percentage": 94.17, "elapsed_time": "2:09:42", "remaining_time": "0:08:02", "throughput": 5172.88, "total_tokens": 40255616} {"current_steps": 39950, "total_steps": 42420, "loss": 0.3565, "lr": 1.0300589274213456e-05, "epoch": 18.835454974068835, "percentage": 94.18, "elapsed_time": "2:09:42", "remaining_time": "0:08:01", "throughput": 5172.92, "total_tokens": 40260768} {"current_steps": 39955, "total_steps": 42420, "loss": 0.3469, "lr": 1.025908853363655e-05, "epoch": 18.83781235266384, "percentage": 94.19, "elapsed_time": "2:09:43", "remaining_time": "0:08:00", "throughput": 5172.93, "total_tokens": 40265280} {"current_steps": 39960, "total_steps": 42420, "loss": 0.3373, "lr": 1.021767069788071e-05, "epoch": 18.84016973125884, "percentage": 94.2, "elapsed_time": "2:09:44", "remaining_time": "0:07:59", "throughput": 5172.9, "total_tokens": 40269088} {"current_steps": 39965, "total_steps": 42420, "loss": 0.3083, "lr": 1.0176335773957157e-05, "epoch": 18.842527109853844, "percentage": 94.21, "elapsed_time": "2:09:45", "remaining_time": "0:07:58", "throughput": 5172.9, "total_tokens": 40273760} {"current_steps": 39970, "total_steps": 42420, "loss": 0.286, "lr": 1.0135083768863184e-05, "epoch": 18.844884488448844, "percentage": 94.22, "elapsed_time": "2:09:46", "remaining_time": "0:07:57", "throughput": 5172.93, "total_tokens": 40278688} {"current_steps": 39975, "total_steps": 42420, "loss": 0.3508, "lr": 1.0093914689582096e-05, "epoch": 18.84724186704385, "percentage": 94.24, "elapsed_time": "2:09:47", "remaining_time": "0:07:56", "throughput": 5173.0, "total_tokens": 40284544} {"current_steps": 39980, "total_steps": 42420, "loss": 0.3186, "lr": 1.0052828543083148e-05, "epoch": 18.84959924563885, "percentage": 94.25, "elapsed_time": "2:09:48", "remaining_time": "0:07:55", "throughput": 5173.01, "total_tokens": 40289248} {"current_steps": 39985, "total_steps": 42420, "loss": 0.3699, "lr": 1.0011825336321445e-05, "epoch": 18.851956624233853, "percentage": 94.26, "elapsed_time": "2:09:49", "remaining_time": "0:07:54", "throughput": 5173.1, "total_tokens": 40295520} {"current_steps": 39990, "total_steps": 42420, "loss": 0.336, "lr": 9.970905076238268e-06, "epoch": 18.854314002828854, "percentage": 94.27, "elapsed_time": "2:09:50", "remaining_time": "0:07:53", "throughput": 5173.18, "total_tokens": 40301632} {"current_steps": 39995, "total_steps": 42420, "loss": 0.3225, "lr": 9.930067769760576e-06, "epoch": 18.85667138142386, "percentage": 94.28, "elapsed_time": "2:09:51", "remaining_time": "0:07:52", "throughput": 5173.22, "total_tokens": 40306656} {"current_steps": 40000, "total_steps": 42420, "loss": 0.3605, "lr": 9.88931342380156e-06, "epoch": 18.85902876001886, "percentage": 94.3, "elapsed_time": "2:09:52", "remaining_time": "0:07:51", "throughput": 5173.28, "total_tokens": 40312512} {"current_steps": 40005, "total_steps": 42420, "loss": 0.2783, "lr": 9.848642045260148e-06, "epoch": 18.861386138613863, "percentage": 94.31, "elapsed_time": "2:09:53", "remaining_time": "0:07:50", "throughput": 5173.36, "total_tokens": 40318336} {"current_steps": 40010, "total_steps": 42420, "loss": 0.3253, "lr": 9.808053641021331e-06, "epoch": 18.863743517208864, "percentage": 94.32, "elapsed_time": "2:09:54", "remaining_time": "0:07:49", "throughput": 5173.36, "total_tokens": 40322880} {"current_steps": 40015, "total_steps": 42420, "loss": 0.3596, "lr": 9.767548217956057e-06, "epoch": 18.866100895803868, "percentage": 94.33, "elapsed_time": "2:09:55", "remaining_time": "0:07:48", "throughput": 5173.37, "total_tokens": 40327488} {"current_steps": 40020, "total_steps": 42420, "loss": 0.3657, "lr": 9.727125782921176e-06, "epoch": 18.86845827439887, "percentage": 94.34, "elapsed_time": "2:09:56", "remaining_time": "0:07:47", "throughput": 5173.35, "total_tokens": 40331648} {"current_steps": 40025, "total_steps": 42420, "loss": 0.3172, "lr": 9.686786342759546e-06, "epoch": 18.87081565299387, "percentage": 94.35, "elapsed_time": "2:09:56", "remaining_time": "0:07:46", "throughput": 5173.39, "total_tokens": 40336800} {"current_steps": 40030, "total_steps": 42420, "loss": 0.3754, "lr": 9.646529904299928e-06, "epoch": 18.873173031588873, "percentage": 94.37, "elapsed_time": "2:09:57", "remaining_time": "0:07:45", "throughput": 5173.44, "total_tokens": 40342016} {"current_steps": 40035, "total_steps": 42420, "loss": 0.3523, "lr": 9.606356474357092e-06, "epoch": 18.875530410183874, "percentage": 94.38, "elapsed_time": "2:09:58", "remaining_time": "0:07:44", "throughput": 5173.47, "total_tokens": 40346976} {"current_steps": 40040, "total_steps": 42420, "loss": 0.3392, "lr": 9.56626605973171e-06, "epoch": 18.877887788778878, "percentage": 94.39, "elapsed_time": "2:09:59", "remaining_time": "0:07:43", "throughput": 5173.52, "total_tokens": 40352288} {"current_steps": 40045, "total_steps": 42420, "loss": 0.3871, "lr": 9.526258667210352e-06, "epoch": 18.88024516737388, "percentage": 94.4, "elapsed_time": "2:10:00", "remaining_time": "0:07:42", "throughput": 5173.58, "total_tokens": 40357824} {"current_steps": 40050, "total_steps": 42420, "loss": 0.3399, "lr": 9.486334303565713e-06, "epoch": 18.882602545968883, "percentage": 94.41, "elapsed_time": "2:10:01", "remaining_time": "0:07:41", "throughput": 5173.6, "total_tokens": 40362688} {"current_steps": 40055, "total_steps": 42420, "loss": 0.3297, "lr": 9.446492975556165e-06, "epoch": 18.884959924563884, "percentage": 94.42, "elapsed_time": "2:10:02", "remaining_time": "0:07:40", "throughput": 5173.65, "total_tokens": 40367936} {"current_steps": 40060, "total_steps": 42420, "loss": 0.3368, "lr": 9.406734689926256e-06, "epoch": 18.887317303158888, "percentage": 94.44, "elapsed_time": "2:10:03", "remaining_time": "0:07:39", "throughput": 5173.68, "total_tokens": 40373184} {"current_steps": 40065, "total_steps": 42420, "loss": 0.3024, "lr": 9.367059453406323e-06, "epoch": 18.88967468175389, "percentage": 94.45, "elapsed_time": "2:10:04", "remaining_time": "0:07:38", "throughput": 5173.71, "total_tokens": 40377856} {"current_steps": 40070, "total_steps": 42420, "loss": 0.3046, "lr": 9.327467272712776e-06, "epoch": 18.892032060348892, "percentage": 94.46, "elapsed_time": "2:10:05", "remaining_time": "0:07:37", "throughput": 5173.73, "total_tokens": 40382592} {"current_steps": 40075, "total_steps": 42420, "loss": 0.3646, "lr": 9.28795815454786e-06, "epoch": 18.894389438943893, "percentage": 94.47, "elapsed_time": "2:10:06", "remaining_time": "0:07:36", "throughput": 5173.73, "total_tokens": 40386976} {"current_steps": 40080, "total_steps": 42420, "loss": 0.3522, "lr": 9.248532105599839e-06, "epoch": 18.896746817538897, "percentage": 94.48, "elapsed_time": "2:10:07", "remaining_time": "0:07:35", "throughput": 5173.78, "total_tokens": 40392256} {"current_steps": 40085, "total_steps": 42420, "loss": 0.3479, "lr": 9.209189132542817e-06, "epoch": 18.899104196133898, "percentage": 94.5, "elapsed_time": "2:10:08", "remaining_time": "0:07:34", "throughput": 5173.84, "total_tokens": 40397920} {"current_steps": 40090, "total_steps": 42420, "loss": 0.3198, "lr": 9.169929242036967e-06, "epoch": 18.901461574728902, "percentage": 94.51, "elapsed_time": "2:10:09", "remaining_time": "0:07:33", "throughput": 5173.92, "total_tokens": 40403872} {"current_steps": 40095, "total_steps": 42420, "loss": 0.3083, "lr": 9.130752440728252e-06, "epoch": 18.903818953323903, "percentage": 94.52, "elapsed_time": "2:10:10", "remaining_time": "0:07:32", "throughput": 5173.94, "total_tokens": 40408672} {"current_steps": 40100, "total_steps": 42420, "loss": 0.2998, "lr": 9.091658735248642e-06, "epoch": 18.906176331918907, "percentage": 94.53, "elapsed_time": "2:10:10", "remaining_time": "0:07:31", "throughput": 5173.94, "total_tokens": 40412992} {"current_steps": 40105, "total_steps": 42420, "loss": 0.3476, "lr": 9.052648132216068e-06, "epoch": 18.908533710513908, "percentage": 94.54, "elapsed_time": "2:10:12", "remaining_time": "0:07:30", "throughput": 5174.06, "total_tokens": 40419808} {"current_steps": 40110, "total_steps": 42420, "loss": 0.3391, "lr": 9.013720638234357e-06, "epoch": 18.91089108910891, "percentage": 94.55, "elapsed_time": "2:10:12", "remaining_time": "0:07:29", "throughput": 5174.11, "total_tokens": 40425312} {"current_steps": 40115, "total_steps": 42420, "loss": 0.3789, "lr": 8.974876259893293e-06, "epoch": 18.913248467703912, "percentage": 94.57, "elapsed_time": "2:10:13", "remaining_time": "0:07:28", "throughput": 5174.12, "total_tokens": 40430176} {"current_steps": 40120, "total_steps": 42420, "loss": 0.3433, "lr": 8.936115003768507e-06, "epoch": 18.915605846298917, "percentage": 94.58, "elapsed_time": "2:10:14", "remaining_time": "0:07:28", "throughput": 5174.14, "total_tokens": 40434752} {"current_steps": 40125, "total_steps": 42420, "loss": 0.3441, "lr": 8.897436876421639e-06, "epoch": 18.917963224893917, "percentage": 94.59, "elapsed_time": "2:10:15", "remaining_time": "0:07:27", "throughput": 5174.14, "total_tokens": 40439040} {"current_steps": 40130, "total_steps": 42420, "loss": 0.3097, "lr": 8.858841884400393e-06, "epoch": 18.92032060348892, "percentage": 94.6, "elapsed_time": "2:10:16", "remaining_time": "0:07:26", "throughput": 5174.13, "total_tokens": 40442944} {"current_steps": 40135, "total_steps": 42420, "loss": 0.3443, "lr": 8.82033003423799e-06, "epoch": 18.922677982083922, "percentage": 94.61, "elapsed_time": "2:10:17", "remaining_time": "0:07:25", "throughput": 5174.14, "total_tokens": 40447552} {"current_steps": 40140, "total_steps": 42420, "loss": 0.2812, "lr": 8.78190133245399e-06, "epoch": 18.925035360678926, "percentage": 94.63, "elapsed_time": "2:10:18", "remaining_time": "0:07:24", "throughput": 5174.19, "total_tokens": 40452640} {"current_steps": 40145, "total_steps": 42420, "loss": 0.3494, "lr": 8.743555785553692e-06, "epoch": 18.927392739273927, "percentage": 94.64, "elapsed_time": "2:10:19", "remaining_time": "0:07:23", "throughput": 5174.19, "total_tokens": 40457024} {"current_steps": 40150, "total_steps": 42420, "loss": 0.3466, "lr": 8.705293400028291e-06, "epoch": 18.92975011786893, "percentage": 94.65, "elapsed_time": "2:10:19", "remaining_time": "0:07:22", "throughput": 5174.24, "total_tokens": 40462304} {"current_steps": 40155, "total_steps": 42420, "loss": 0.3371, "lr": 8.667114182355052e-06, "epoch": 18.93210749646393, "percentage": 94.66, "elapsed_time": "2:10:20", "remaining_time": "0:07:21", "throughput": 5174.27, "total_tokens": 40467424} {"current_steps": 40160, "total_steps": 42420, "loss": 0.3535, "lr": 8.629018138997025e-06, "epoch": 18.934464875058936, "percentage": 94.67, "elapsed_time": "2:10:21", "remaining_time": "0:07:20", "throughput": 5174.37, "total_tokens": 40473856} {"current_steps": 40165, "total_steps": 42420, "loss": 0.3627, "lr": 8.591005276403219e-06, "epoch": 18.936822253653936, "percentage": 94.68, "elapsed_time": "2:10:22", "remaining_time": "0:07:19", "throughput": 5174.42, "total_tokens": 40479488} {"current_steps": 40170, "total_steps": 42420, "loss": 0.3867, "lr": 8.553075601008597e-06, "epoch": 18.93917963224894, "percentage": 94.7, "elapsed_time": "2:10:24", "remaining_time": "0:07:18", "throughput": 5174.53, "total_tokens": 40486368} {"current_steps": 40175, "total_steps": 42420, "loss": 0.2888, "lr": 8.515229119233969e-06, "epoch": 18.94153701084394, "percentage": 94.71, "elapsed_time": "2:10:25", "remaining_time": "0:07:17", "throughput": 5174.54, "total_tokens": 40490976} {"current_steps": 40180, "total_steps": 42420, "loss": 0.3359, "lr": 8.477465837486097e-06, "epoch": 18.943894389438945, "percentage": 94.72, "elapsed_time": "2:10:26", "remaining_time": "0:07:16", "throughput": 5174.6, "total_tokens": 40496480} {"current_steps": 40185, "total_steps": 42420, "loss": 0.3182, "lr": 8.439785762157648e-06, "epoch": 18.946251768033946, "percentage": 94.73, "elapsed_time": "2:10:26", "remaining_time": "0:07:15", "throughput": 5174.62, "total_tokens": 40501312} {"current_steps": 40190, "total_steps": 42420, "loss": 0.3121, "lr": 8.402188899627294e-06, "epoch": 18.94860914662895, "percentage": 94.74, "elapsed_time": "2:10:27", "remaining_time": "0:07:14", "throughput": 5174.66, "total_tokens": 40506656} {"current_steps": 40195, "total_steps": 42420, "loss": 0.3195, "lr": 8.364675256259503e-06, "epoch": 18.95096652522395, "percentage": 94.75, "elapsed_time": "2:10:28", "remaining_time": "0:07:13", "throughput": 5174.67, "total_tokens": 40511360} {"current_steps": 40200, "total_steps": 42420, "loss": 0.3549, "lr": 8.327244838404691e-06, "epoch": 18.953323903818955, "percentage": 94.77, "elapsed_time": "2:10:29", "remaining_time": "0:07:12", "throughput": 5174.7, "total_tokens": 40516352} {"current_steps": 40205, "total_steps": 42420, "loss": 0.3022, "lr": 8.289897652399237e-06, "epoch": 18.955681282413956, "percentage": 94.78, "elapsed_time": "2:10:30", "remaining_time": "0:07:11", "throughput": 5174.74, "total_tokens": 40521248} {"current_steps": 40210, "total_steps": 42420, "loss": 0.3118, "lr": 8.252633704565304e-06, "epoch": 18.95803866100896, "percentage": 94.79, "elapsed_time": "2:10:31", "remaining_time": "0:07:10", "throughput": 5174.89, "total_tokens": 40528960} {"current_steps": 40215, "total_steps": 42420, "loss": 0.3121, "lr": 8.215453001211126e-06, "epoch": 18.96039603960396, "percentage": 94.8, "elapsed_time": "2:10:32", "remaining_time": "0:07:09", "throughput": 5174.97, "total_tokens": 40534976} {"current_steps": 40220, "total_steps": 42420, "loss": 0.3607, "lr": 8.178355548630666e-06, "epoch": 18.962753418198965, "percentage": 94.81, "elapsed_time": "2:10:33", "remaining_time": "0:07:08", "throughput": 5175.0, "total_tokens": 40539680} {"current_steps": 40225, "total_steps": 42420, "loss": 0.2958, "lr": 8.141341353104016e-06, "epoch": 18.965110796793965, "percentage": 94.83, "elapsed_time": "2:10:34", "remaining_time": "0:07:07", "throughput": 5174.99, "total_tokens": 40543776} {"current_steps": 40230, "total_steps": 42420, "loss": 0.3367, "lr": 8.104410420896935e-06, "epoch": 18.967468175388966, "percentage": 94.84, "elapsed_time": "2:10:35", "remaining_time": "0:07:06", "throughput": 5175.04, "total_tokens": 40549120} {"current_steps": 40235, "total_steps": 42420, "loss": 0.2909, "lr": 8.067562758261316e-06, "epoch": 18.96982555398397, "percentage": 94.85, "elapsed_time": "2:10:36", "remaining_time": "0:07:05", "throughput": 5175.04, "total_tokens": 40553280} {"current_steps": 40240, "total_steps": 42420, "loss": 0.3226, "lr": 8.030798371434778e-06, "epoch": 18.97218293257897, "percentage": 94.86, "elapsed_time": "2:10:37", "remaining_time": "0:07:04", "throughput": 5175.17, "total_tokens": 40560672} {"current_steps": 40245, "total_steps": 42420, "loss": 0.3181, "lr": 7.994117266640898e-06, "epoch": 18.974540311173975, "percentage": 94.87, "elapsed_time": "2:10:38", "remaining_time": "0:07:03", "throughput": 5175.22, "total_tokens": 40565920} {"current_steps": 40250, "total_steps": 42420, "loss": 0.3023, "lr": 7.957519450089158e-06, "epoch": 18.976897689768975, "percentage": 94.88, "elapsed_time": "2:10:39", "remaining_time": "0:07:02", "throughput": 5175.28, "total_tokens": 40571552} {"current_steps": 40255, "total_steps": 42420, "loss": 0.3238, "lr": 7.921004927974984e-06, "epoch": 18.97925506836398, "percentage": 94.9, "elapsed_time": "2:10:40", "remaining_time": "0:07:01", "throughput": 5175.3, "total_tokens": 40576608} {"current_steps": 40260, "total_steps": 42420, "loss": 0.3703, "lr": 7.884573706479714e-06, "epoch": 18.98161244695898, "percentage": 94.91, "elapsed_time": "2:10:41", "remaining_time": "0:07:00", "throughput": 5175.34, "total_tokens": 40581856} {"current_steps": 40265, "total_steps": 42420, "loss": 0.3277, "lr": 7.848225791770414e-06, "epoch": 18.983969825553984, "percentage": 94.92, "elapsed_time": "2:10:42", "remaining_time": "0:06:59", "throughput": 5175.36, "total_tokens": 40586784} {"current_steps": 40270, "total_steps": 42420, "loss": 0.35, "lr": 7.811961190000216e-06, "epoch": 18.986327204148985, "percentage": 94.93, "elapsed_time": "2:10:43", "remaining_time": "0:06:58", "throughput": 5175.38, "total_tokens": 40591264} {"current_steps": 40275, "total_steps": 42420, "loss": 0.3011, "lr": 7.7757799073081e-06, "epoch": 18.98868458274399, "percentage": 94.94, "elapsed_time": "2:10:44", "remaining_time": "0:06:57", "throughput": 5175.42, "total_tokens": 40596608} {"current_steps": 40280, "total_steps": 42420, "loss": 0.2623, "lr": 7.739681949818945e-06, "epoch": 18.99104196133899, "percentage": 94.96, "elapsed_time": "2:10:45", "remaining_time": "0:06:56", "throughput": 5175.48, "total_tokens": 40602304} {"current_steps": 40285, "total_steps": 42420, "loss": 0.3074, "lr": 7.703667323643581e-06, "epoch": 18.993399339933994, "percentage": 94.97, "elapsed_time": "2:10:46", "remaining_time": "0:06:55", "throughput": 5175.51, "total_tokens": 40607200} {"current_steps": 40290, "total_steps": 42420, "loss": 0.2935, "lr": 7.66773603487858e-06, "epoch": 18.995756718528995, "percentage": 94.98, "elapsed_time": "2:10:47", "remaining_time": "0:06:54", "throughput": 5175.61, "total_tokens": 40613856} {"current_steps": 40295, "total_steps": 42420, "loss": 0.348, "lr": 7.631888089606576e-06, "epoch": 18.998114097124, "percentage": 94.99, "elapsed_time": "2:10:48", "remaining_time": "0:06:53", "throughput": 5175.61, "total_tokens": 40618336} {"current_steps": 40299, "total_steps": 42420, "eval_loss": 0.3382989168167114, "epoch": 19.0, "percentage": 95.0, "elapsed_time": "2:11:08", "remaining_time": "0:06:54", "throughput": 5163.14, "total_tokens": 40624896} {"current_steps": 40300, "total_steps": 42420, "loss": 0.3037, "lr": 7.59612349389599e-06, "epoch": 19.000471475719, "percentage": 95.0, "elapsed_time": "2:11:11", "remaining_time": "0:06:54", "throughput": 5161.37, "total_tokens": 40626080} {"current_steps": 40305, "total_steps": 42420, "loss": 0.3073, "lr": 7.560442253801148e-06, "epoch": 19.002828854314004, "percentage": 95.01, "elapsed_time": "2:11:11", "remaining_time": "0:06:53", "throughput": 5161.35, "total_tokens": 40630080} {"current_steps": 40310, "total_steps": 42420, "loss": 0.3312, "lr": 7.524844375362272e-06, "epoch": 19.005186232909004, "percentage": 95.03, "elapsed_time": "2:11:12", "remaining_time": "0:06:52", "throughput": 5161.37, "total_tokens": 40634592} {"current_steps": 40315, "total_steps": 42420, "loss": 0.3276, "lr": 7.4893298646054875e-06, "epoch": 19.00754361150401, "percentage": 95.04, "elapsed_time": "2:11:13", "remaining_time": "0:06:51", "throughput": 5161.38, "total_tokens": 40639104} {"current_steps": 40320, "total_steps": 42420, "loss": 0.2729, "lr": 7.4538987275428186e-06, "epoch": 19.00990099009901, "percentage": 95.05, "elapsed_time": "2:11:14", "remaining_time": "0:06:50", "throughput": 5161.35, "total_tokens": 40643264} {"current_steps": 40325, "total_steps": 42420, "loss": 0.3642, "lr": 7.4185509701720775e-06, "epoch": 19.012258368694013, "percentage": 95.06, "elapsed_time": "2:11:15", "remaining_time": "0:06:49", "throughput": 5161.36, "total_tokens": 40647808} {"current_steps": 40330, "total_steps": 42420, "loss": 0.3607, "lr": 7.3832865984771455e-06, "epoch": 19.014615747289014, "percentage": 95.07, "elapsed_time": "2:11:16", "remaining_time": "0:06:48", "throughput": 5161.33, "total_tokens": 40651552} {"current_steps": 40335, "total_steps": 42420, "loss": 0.3722, "lr": 7.348105618427636e-06, "epoch": 19.016973125884018, "percentage": 95.08, "elapsed_time": "2:11:17", "remaining_time": "0:06:47", "throughput": 5161.34, "total_tokens": 40655968} {"current_steps": 40340, "total_steps": 42420, "loss": 0.3297, "lr": 7.313008035979174e-06, "epoch": 19.01933050447902, "percentage": 95.1, "elapsed_time": "2:11:17", "remaining_time": "0:06:46", "throughput": 5161.34, "total_tokens": 40660128} {"current_steps": 40345, "total_steps": 42420, "loss": 0.3105, "lr": 7.277993857073006e-06, "epoch": 19.021687883074023, "percentage": 95.11, "elapsed_time": "2:11:18", "remaining_time": "0:06:45", "throughput": 5161.35, "total_tokens": 40664608} {"current_steps": 40350, "total_steps": 42420, "loss": 0.31, "lr": 7.243063087636559e-06, "epoch": 19.024045261669023, "percentage": 95.12, "elapsed_time": "2:11:19", "remaining_time": "0:06:44", "throughput": 5161.36, "total_tokens": 40668992} {"current_steps": 40355, "total_steps": 42420, "loss": 0.3477, "lr": 7.20821573358299e-06, "epoch": 19.026402640264028, "percentage": 95.13, "elapsed_time": "2:11:20", "remaining_time": "0:06:43", "throughput": 5161.39, "total_tokens": 40674240} {"current_steps": 40360, "total_steps": 42420, "loss": 0.3371, "lr": 7.173451800811359e-06, "epoch": 19.02876001885903, "percentage": 95.14, "elapsed_time": "2:11:21", "remaining_time": "0:06:42", "throughput": 5161.47, "total_tokens": 40680032} {"current_steps": 40365, "total_steps": 42420, "loss": 0.3425, "lr": 7.138771295206625e-06, "epoch": 19.031117397454032, "percentage": 95.16, "elapsed_time": "2:11:22", "remaining_time": "0:06:41", "throughput": 5161.47, "total_tokens": 40684384} {"current_steps": 40370, "total_steps": 42420, "loss": 0.3, "lr": 7.1041742226395924e-06, "epoch": 19.033474776049033, "percentage": 95.17, "elapsed_time": "2:11:23", "remaining_time": "0:06:40", "throughput": 5161.49, "total_tokens": 40689184} {"current_steps": 40375, "total_steps": 42420, "loss": 0.289, "lr": 7.069660588966964e-06, "epoch": 19.035832154644037, "percentage": 95.18, "elapsed_time": "2:11:24", "remaining_time": "0:06:39", "throughput": 5161.5, "total_tokens": 40693632} {"current_steps": 40380, "total_steps": 42420, "loss": 0.3797, "lr": 7.0352304000314005e-06, "epoch": 19.038189533239038, "percentage": 95.19, "elapsed_time": "2:11:25", "remaining_time": "0:06:38", "throughput": 5161.56, "total_tokens": 40699136} {"current_steps": 40385, "total_steps": 42420, "loss": 0.2922, "lr": 7.000883661661184e-06, "epoch": 19.040546911834042, "percentage": 95.2, "elapsed_time": "2:11:26", "remaining_time": "0:06:37", "throughput": 5161.62, "total_tokens": 40704544} {"current_steps": 40390, "total_steps": 42420, "loss": 0.3157, "lr": 6.966620379670719e-06, "epoch": 19.042904290429043, "percentage": 95.21, "elapsed_time": "2:11:26", "remaining_time": "0:06:36", "throughput": 5161.65, "total_tokens": 40709280} {"current_steps": 40395, "total_steps": 42420, "loss": 0.3161, "lr": 6.932440559860198e-06, "epoch": 19.045261669024047, "percentage": 95.23, "elapsed_time": "2:11:27", "remaining_time": "0:06:35", "throughput": 5161.68, "total_tokens": 40714336} {"current_steps": 40400, "total_steps": 42420, "loss": 0.3131, "lr": 6.898344208015661e-06, "epoch": 19.047619047619047, "percentage": 95.24, "elapsed_time": "2:11:28", "remaining_time": "0:06:34", "throughput": 5161.66, "total_tokens": 40718400} {"current_steps": 40405, "total_steps": 42420, "loss": 0.3964, "lr": 6.864331329909101e-06, "epoch": 19.04997642621405, "percentage": 95.25, "elapsed_time": "2:11:29", "remaining_time": "0:06:33", "throughput": 5161.67, "total_tokens": 40723104} {"current_steps": 40410, "total_steps": 42420, "loss": 0.3743, "lr": 6.830401931298246e-06, "epoch": 19.052333804809052, "percentage": 95.26, "elapsed_time": "2:11:30", "remaining_time": "0:06:32", "throughput": 5161.68, "total_tokens": 40727776} {"current_steps": 40415, "total_steps": 42420, "loss": 0.3349, "lr": 6.796556017926836e-06, "epoch": 19.054691183404056, "percentage": 95.27, "elapsed_time": "2:11:31", "remaining_time": "0:06:31", "throughput": 5161.75, "total_tokens": 40733376} {"current_steps": 40420, "total_steps": 42420, "loss": 0.3617, "lr": 6.762793595524397e-06, "epoch": 19.057048561999057, "percentage": 95.29, "elapsed_time": "2:11:32", "remaining_time": "0:06:30", "throughput": 5161.76, "total_tokens": 40738016} {"current_steps": 40425, "total_steps": 42420, "loss": 0.3048, "lr": 6.729114669806247e-06, "epoch": 19.059405940594058, "percentage": 95.3, "elapsed_time": "2:11:33", "remaining_time": "0:06:29", "throughput": 5161.79, "total_tokens": 40743008} {"current_steps": 40430, "total_steps": 42420, "loss": 0.3422, "lr": 6.6955192464737695e-06, "epoch": 19.061763319189062, "percentage": 95.31, "elapsed_time": "2:11:34", "remaining_time": "0:06:28", "throughput": 5161.9, "total_tokens": 40749824} {"current_steps": 40435, "total_steps": 42420, "loss": 0.3584, "lr": 6.662007331214026e-06, "epoch": 19.064120697784062, "percentage": 95.32, "elapsed_time": "2:11:35", "remaining_time": "0:06:27", "throughput": 5161.98, "total_tokens": 40755872} {"current_steps": 40440, "total_steps": 42420, "loss": 0.3163, "lr": 6.628578929700035e-06, "epoch": 19.066478076379067, "percentage": 95.33, "elapsed_time": "2:11:36", "remaining_time": "0:06:26", "throughput": 5162.05, "total_tokens": 40761440} {"current_steps": 40445, "total_steps": 42420, "loss": 0.3641, "lr": 6.595234047590659e-06, "epoch": 19.068835454974067, "percentage": 95.34, "elapsed_time": "2:11:37", "remaining_time": "0:06:25", "throughput": 5162.04, "total_tokens": 40765760} {"current_steps": 40450, "total_steps": 42420, "loss": 0.3245, "lr": 6.56197269053066e-06, "epoch": 19.07119283356907, "percentage": 95.36, "elapsed_time": "2:11:38", "remaining_time": "0:06:24", "throughput": 5162.07, "total_tokens": 40770656} {"current_steps": 40455, "total_steps": 42420, "loss": 0.3293, "lr": 6.528794864150533e-06, "epoch": 19.073550212164072, "percentage": 95.37, "elapsed_time": "2:11:39", "remaining_time": "0:06:23", "throughput": 5162.13, "total_tokens": 40776192} {"current_steps": 40460, "total_steps": 42420, "loss": 0.3595, "lr": 6.495700574066843e-06, "epoch": 19.075907590759076, "percentage": 95.38, "elapsed_time": "2:11:40", "remaining_time": "0:06:22", "throughput": 5162.17, "total_tokens": 40781632} {"current_steps": 40465, "total_steps": 42420, "loss": 0.3459, "lr": 6.462689825881774e-06, "epoch": 19.078264969354077, "percentage": 95.39, "elapsed_time": "2:11:41", "remaining_time": "0:06:21", "throughput": 5162.23, "total_tokens": 40787232} {"current_steps": 40470, "total_steps": 42420, "loss": 0.3212, "lr": 6.429762625183522e-06, "epoch": 19.08062234794908, "percentage": 95.4, "elapsed_time": "2:11:42", "remaining_time": "0:06:20", "throughput": 5162.29, "total_tokens": 40792544} {"current_steps": 40475, "total_steps": 42420, "loss": 0.2677, "lr": 6.396918977546129e-06, "epoch": 19.08297972654408, "percentage": 95.41, "elapsed_time": "2:11:43", "remaining_time": "0:06:19", "throughput": 5162.41, "total_tokens": 40799616} {"current_steps": 40480, "total_steps": 42420, "loss": 0.3123, "lr": 6.364158888529425e-06, "epoch": 19.085337105139086, "percentage": 95.43, "elapsed_time": "2:11:44", "remaining_time": "0:06:18", "throughput": 5162.43, "total_tokens": 40804288} {"current_steps": 40485, "total_steps": 42420, "loss": 0.3171, "lr": 6.331482363679142e-06, "epoch": 19.087694483734087, "percentage": 95.44, "elapsed_time": "2:11:44", "remaining_time": "0:06:17", "throughput": 5162.45, "total_tokens": 40808832} {"current_steps": 40490, "total_steps": 42420, "loss": 0.3657, "lr": 6.298889408526909e-06, "epoch": 19.09005186232909, "percentage": 95.45, "elapsed_time": "2:11:45", "remaining_time": "0:06:16", "throughput": 5162.49, "total_tokens": 40813952} {"current_steps": 40495, "total_steps": 42420, "loss": 0.3154, "lr": 6.266380028590146e-06, "epoch": 19.09240924092409, "percentage": 95.46, "elapsed_time": "2:11:46", "remaining_time": "0:06:15", "throughput": 5162.51, "total_tokens": 40818528} {"current_steps": 40500, "total_steps": 42420, "loss": 0.4104, "lr": 6.233954229372118e-06, "epoch": 19.094766619519095, "percentage": 95.47, "elapsed_time": "2:11:47", "remaining_time": "0:06:14", "throughput": 5162.52, "total_tokens": 40823072} {"current_steps": 40505, "total_steps": 42420, "loss": 0.2794, "lr": 6.20161201636199e-06, "epoch": 19.097123998114096, "percentage": 95.49, "elapsed_time": "2:11:48", "remaining_time": "0:06:13", "throughput": 5162.49, "total_tokens": 40827040} {"current_steps": 40510, "total_steps": 42420, "loss": 0.3616, "lr": 6.169353395034716e-06, "epoch": 19.0994813767091, "percentage": 95.5, "elapsed_time": "2:11:49", "remaining_time": "0:06:12", "throughput": 5162.5, "total_tokens": 40831776} {"current_steps": 40515, "total_steps": 42420, "loss": 0.345, "lr": 6.137178370851148e-06, "epoch": 19.1018387553041, "percentage": 95.51, "elapsed_time": "2:11:50", "remaining_time": "0:06:11", "throughput": 5162.54, "total_tokens": 40836832} {"current_steps": 40520, "total_steps": 42420, "loss": 0.3461, "lr": 6.105086949257932e-06, "epoch": 19.104196133899105, "percentage": 95.52, "elapsed_time": "2:11:51", "remaining_time": "0:06:10", "throughput": 5162.61, "total_tokens": 40842432} {"current_steps": 40525, "total_steps": 42420, "loss": 0.3079, "lr": 6.073079135687664e-06, "epoch": 19.106553512494106, "percentage": 95.53, "elapsed_time": "2:11:52", "remaining_time": "0:06:09", "throughput": 5162.61, "total_tokens": 40847040} {"current_steps": 40530, "total_steps": 42420, "loss": 0.3437, "lr": 6.04115493555868e-06, "epoch": 19.10891089108911, "percentage": 95.54, "elapsed_time": "2:11:53", "remaining_time": "0:06:09", "throughput": 5162.66, "total_tokens": 40852352} {"current_steps": 40535, "total_steps": 42420, "loss": 0.3064, "lr": 6.009314354275208e-06, "epoch": 19.11126826968411, "percentage": 95.56, "elapsed_time": "2:11:53", "remaining_time": "0:06:08", "throughput": 5162.69, "total_tokens": 40857120} {"current_steps": 40540, "total_steps": 42420, "loss": 0.3653, "lr": 5.977557397227385e-06, "epoch": 19.113625648279115, "percentage": 95.57, "elapsed_time": "2:11:54", "remaining_time": "0:06:07", "throughput": 5162.72, "total_tokens": 40861760} {"current_steps": 40545, "total_steps": 42420, "loss": 0.2888, "lr": 5.9458840697910185e-06, "epoch": 19.115983026874115, "percentage": 95.58, "elapsed_time": "2:11:55", "remaining_time": "0:06:06", "throughput": 5162.81, "total_tokens": 40868096} {"current_steps": 40550, "total_steps": 42420, "loss": 0.3124, "lr": 5.914294377327989e-06, "epoch": 19.11834040546912, "percentage": 95.59, "elapsed_time": "2:11:56", "remaining_time": "0:06:05", "throughput": 5162.85, "total_tokens": 40873280} {"current_steps": 40555, "total_steps": 42420, "loss": 0.3414, "lr": 5.88278832518574e-06, "epoch": 19.12069778406412, "percentage": 95.6, "elapsed_time": "2:11:57", "remaining_time": "0:06:04", "throughput": 5162.87, "total_tokens": 40878272} {"current_steps": 40560, "total_steps": 42420, "loss": 0.2963, "lr": 5.851365918697837e-06, "epoch": 19.123055162659124, "percentage": 95.62, "elapsed_time": "2:11:58", "remaining_time": "0:06:03", "throughput": 5162.97, "total_tokens": 40884864} {"current_steps": 40565, "total_steps": 42420, "loss": 0.3947, "lr": 5.820027163183528e-06, "epoch": 19.125412541254125, "percentage": 95.63, "elapsed_time": "2:11:59", "remaining_time": "0:06:02", "throughput": 5162.98, "total_tokens": 40889344} {"current_steps": 40570, "total_steps": 42420, "loss": 0.3338, "lr": 5.7887720639479005e-06, "epoch": 19.12776991984913, "percentage": 95.64, "elapsed_time": "2:12:00", "remaining_time": "0:06:01", "throughput": 5163.01, "total_tokens": 40894048} {"current_steps": 40575, "total_steps": 42420, "loss": 0.3287, "lr": 5.757600626281889e-06, "epoch": 19.13012729844413, "percentage": 95.65, "elapsed_time": "2:12:01", "remaining_time": "0:06:00", "throughput": 5163.04, "total_tokens": 40898912} {"current_steps": 40580, "total_steps": 42420, "loss": 0.3202, "lr": 5.726512855462385e-06, "epoch": 19.132484677039134, "percentage": 95.66, "elapsed_time": "2:12:02", "remaining_time": "0:05:59", "throughput": 5163.11, "total_tokens": 40904864} {"current_steps": 40585, "total_steps": 42420, "loss": 0.3647, "lr": 5.69550875675201e-06, "epoch": 19.134842055634135, "percentage": 95.67, "elapsed_time": "2:12:03", "remaining_time": "0:05:58", "throughput": 5163.13, "total_tokens": 40909472} {"current_steps": 40590, "total_steps": 42420, "loss": 0.3435, "lr": 5.664588335399123e-06, "epoch": 19.13719943422914, "percentage": 95.69, "elapsed_time": "2:12:04", "remaining_time": "0:05:57", "throughput": 5163.13, "total_tokens": 40914016} {"current_steps": 40595, "total_steps": 42420, "loss": 0.325, "lr": 5.633751596638148e-06, "epoch": 19.13955681282414, "percentage": 95.7, "elapsed_time": "2:12:05", "remaining_time": "0:05:56", "throughput": 5163.14, "total_tokens": 40918688} {"current_steps": 40600, "total_steps": 42420, "loss": 0.3411, "lr": 5.6029985456891304e-06, "epoch": 19.141914191419144, "percentage": 95.71, "elapsed_time": "2:12:06", "remaining_time": "0:05:55", "throughput": 5163.23, "total_tokens": 40925120} {"current_steps": 40605, "total_steps": 42420, "loss": 0.3521, "lr": 5.572329187758074e-06, "epoch": 19.144271570014144, "percentage": 95.72, "elapsed_time": "2:12:07", "remaining_time": "0:05:54", "throughput": 5163.29, "total_tokens": 40930528} {"current_steps": 40610, "total_steps": 42420, "loss": 0.3508, "lr": 5.541743528036769e-06, "epoch": 19.14662894860915, "percentage": 95.73, "elapsed_time": "2:12:08", "remaining_time": "0:05:53", "throughput": 5163.32, "total_tokens": 40935296} {"current_steps": 40615, "total_steps": 42420, "loss": 0.2866, "lr": 5.511241571702852e-06, "epoch": 19.14898632720415, "percentage": 95.74, "elapsed_time": "2:12:09", "remaining_time": "0:05:52", "throughput": 5163.36, "total_tokens": 40940256} {"current_steps": 40620, "total_steps": 42420, "loss": 0.328, "lr": 5.480823323919859e-06, "epoch": 19.15134370579915, "percentage": 95.76, "elapsed_time": "2:12:09", "remaining_time": "0:05:51", "throughput": 5163.37, "total_tokens": 40944928} {"current_steps": 40625, "total_steps": 42420, "loss": 0.3692, "lr": 5.4504887898370045e-06, "epoch": 19.153701084394154, "percentage": 95.77, "elapsed_time": "2:12:10", "remaining_time": "0:05:50", "throughput": 5163.4, "total_tokens": 40949824} {"current_steps": 40630, "total_steps": 42420, "loss": 0.331, "lr": 5.420237974589404e-06, "epoch": 19.156058462989154, "percentage": 95.78, "elapsed_time": "2:12:11", "remaining_time": "0:05:49", "throughput": 5163.45, "total_tokens": 40955456} {"current_steps": 40635, "total_steps": 42420, "loss": 0.305, "lr": 5.390070883298015e-06, "epoch": 19.15841584158416, "percentage": 95.79, "elapsed_time": "2:12:12", "remaining_time": "0:05:48", "throughput": 5163.46, "total_tokens": 40959936} {"current_steps": 40640, "total_steps": 42420, "loss": 0.2813, "lr": 5.3599875210696425e-06, "epoch": 19.16077322017916, "percentage": 95.8, "elapsed_time": "2:12:13", "remaining_time": "0:05:47", "throughput": 5163.56, "total_tokens": 40966336} {"current_steps": 40645, "total_steps": 42420, "loss": 0.3795, "lr": 5.329987892996879e-06, "epoch": 19.163130598774163, "percentage": 95.82, "elapsed_time": "2:12:14", "remaining_time": "0:05:46", "throughput": 5163.6, "total_tokens": 40971296} {"current_steps": 40650, "total_steps": 42420, "loss": 0.3503, "lr": 5.300072004158107e-06, "epoch": 19.165487977369164, "percentage": 95.83, "elapsed_time": "2:12:15", "remaining_time": "0:05:45", "throughput": 5163.68, "total_tokens": 40977376} {"current_steps": 40655, "total_steps": 42420, "loss": 0.3985, "lr": 5.270239859617665e-06, "epoch": 19.167845355964168, "percentage": 95.84, "elapsed_time": "2:12:16", "remaining_time": "0:05:44", "throughput": 5163.74, "total_tokens": 40983264} {"current_steps": 40660, "total_steps": 42420, "loss": 0.3419, "lr": 5.2404914644255676e-06, "epoch": 19.17020273455917, "percentage": 95.85, "elapsed_time": "2:12:17", "remaining_time": "0:05:43", "throughput": 5163.74, "total_tokens": 40987776} {"current_steps": 40665, "total_steps": 42420, "loss": 0.3455, "lr": 5.2108268236177315e-06, "epoch": 19.172560113154173, "percentage": 95.86, "elapsed_time": "2:12:18", "remaining_time": "0:05:42", "throughput": 5163.74, "total_tokens": 40992160} {"current_steps": 40670, "total_steps": 42420, "loss": 0.3201, "lr": 5.181245942215862e-06, "epoch": 19.174917491749174, "percentage": 95.87, "elapsed_time": "2:12:19", "remaining_time": "0:05:41", "throughput": 5163.84, "total_tokens": 40998624} {"current_steps": 40675, "total_steps": 42420, "loss": 0.2751, "lr": 5.1517488252275644e-06, "epoch": 19.177274870344178, "percentage": 95.89, "elapsed_time": "2:12:20", "remaining_time": "0:05:40", "throughput": 5163.91, "total_tokens": 41004416} {"current_steps": 40680, "total_steps": 42420, "loss": 0.2646, "lr": 5.1223354776460675e-06, "epoch": 19.17963224893918, "percentage": 95.9, "elapsed_time": "2:12:21", "remaining_time": "0:05:39", "throughput": 5163.9, "total_tokens": 41008640} {"current_steps": 40685, "total_steps": 42420, "loss": 0.2949, "lr": 5.093005904450665e-06, "epoch": 19.181989627534183, "percentage": 95.91, "elapsed_time": "2:12:22", "remaining_time": "0:05:38", "throughput": 5163.92, "total_tokens": 41013696} {"current_steps": 40690, "total_steps": 42420, "loss": 0.3482, "lr": 5.06376011060633e-06, "epoch": 19.184347006129183, "percentage": 95.92, "elapsed_time": "2:12:23", "remaining_time": "0:05:37", "throughput": 5163.96, "total_tokens": 41018816} {"current_steps": 40695, "total_steps": 42420, "loss": 0.2824, "lr": 5.0345981010638805e-06, "epoch": 19.186704384724187, "percentage": 95.93, "elapsed_time": "2:12:24", "remaining_time": "0:05:36", "throughput": 5163.98, "total_tokens": 41023808} {"current_steps": 40700, "total_steps": 42420, "loss": 0.3496, "lr": 5.005519880759868e-06, "epoch": 19.189061763319188, "percentage": 95.95, "elapsed_time": "2:12:25", "remaining_time": "0:05:35", "throughput": 5163.97, "total_tokens": 41027904} {"current_steps": 40705, "total_steps": 42420, "loss": 0.3669, "lr": 4.976525454616853e-06, "epoch": 19.191419141914192, "percentage": 95.96, "elapsed_time": "2:12:25", "remaining_time": "0:05:34", "throughput": 5164.0, "total_tokens": 41033024} {"current_steps": 40710, "total_steps": 42420, "loss": 0.3402, "lr": 4.947614827543079e-06, "epoch": 19.193776520509193, "percentage": 95.97, "elapsed_time": "2:12:26", "remaining_time": "0:05:33", "throughput": 5164.0, "total_tokens": 41037312} {"current_steps": 40715, "total_steps": 42420, "loss": 0.3999, "lr": 4.918788004432628e-06, "epoch": 19.196133899104197, "percentage": 95.98, "elapsed_time": "2:12:27", "remaining_time": "0:05:32", "throughput": 5164.06, "total_tokens": 41043008} {"current_steps": 40720, "total_steps": 42420, "loss": 0.3623, "lr": 4.890044990165321e-06, "epoch": 19.198491277699198, "percentage": 95.99, "elapsed_time": "2:12:28", "remaining_time": "0:05:31", "throughput": 5164.07, "total_tokens": 41047616} {"current_steps": 40725, "total_steps": 42420, "loss": 0.3664, "lr": 4.861385789606876e-06, "epoch": 19.200848656294202, "percentage": 96.0, "elapsed_time": "2:12:29", "remaining_time": "0:05:30", "throughput": 5164.2, "total_tokens": 41054976} {"current_steps": 40730, "total_steps": 42420, "loss": 0.3552, "lr": 4.8328104076089115e-06, "epoch": 19.203206034889202, "percentage": 96.02, "elapsed_time": "2:12:31", "remaining_time": "0:05:29", "throughput": 5164.3, "total_tokens": 41062304} {"current_steps": 40735, "total_steps": 42420, "loss": 0.3204, "lr": 4.804318849008615e-06, "epoch": 19.205563413484207, "percentage": 96.03, "elapsed_time": "2:12:32", "remaining_time": "0:05:28", "throughput": 5164.31, "total_tokens": 41066816} {"current_steps": 40740, "total_steps": 42420, "loss": 0.2822, "lr": 4.775911118629239e-06, "epoch": 19.207920792079207, "percentage": 96.04, "elapsed_time": "2:12:32", "remaining_time": "0:05:27", "throughput": 5164.35, "total_tokens": 41072064} {"current_steps": 40745, "total_steps": 42420, "loss": 0.3562, "lr": 4.747587221279659e-06, "epoch": 19.21027817067421, "percentage": 96.05, "elapsed_time": "2:12:33", "remaining_time": "0:05:26", "throughput": 5164.42, "total_tokens": 41077472} {"current_steps": 40750, "total_steps": 42420, "loss": 0.3296, "lr": 4.71934716175465e-06, "epoch": 19.212635549269212, "percentage": 96.06, "elapsed_time": "2:12:34", "remaining_time": "0:05:26", "throughput": 5164.41, "total_tokens": 41081856} {"current_steps": 40755, "total_steps": 42420, "loss": 0.3305, "lr": 4.691190944834833e-06, "epoch": 19.214992927864216, "percentage": 96.07, "elapsed_time": "2:12:35", "remaining_time": "0:05:25", "throughput": 5164.4, "total_tokens": 41086272} {"current_steps": 40760, "total_steps": 42420, "loss": 0.2961, "lr": 4.663118575286507e-06, "epoch": 19.217350306459217, "percentage": 96.09, "elapsed_time": "2:12:36", "remaining_time": "0:05:24", "throughput": 5164.41, "total_tokens": 41090752} {"current_steps": 40765, "total_steps": 42420, "loss": 0.306, "lr": 4.6351300578618695e-06, "epoch": 19.21970768505422, "percentage": 96.1, "elapsed_time": "2:12:37", "remaining_time": "0:05:23", "throughput": 5164.46, "total_tokens": 41096064} {"current_steps": 40770, "total_steps": 42420, "loss": 0.3041, "lr": 4.607225397298853e-06, "epoch": 19.22206506364922, "percentage": 96.11, "elapsed_time": "2:12:38", "remaining_time": "0:05:22", "throughput": 5164.47, "total_tokens": 41100544} {"current_steps": 40775, "total_steps": 42420, "loss": 0.3319, "lr": 4.57940459832129e-06, "epoch": 19.224422442244226, "percentage": 96.12, "elapsed_time": "2:12:39", "remaining_time": "0:05:21", "throughput": 5164.47, "total_tokens": 41104640} {"current_steps": 40780, "total_steps": 42420, "loss": 0.3275, "lr": 4.551667665638803e-06, "epoch": 19.226779820839226, "percentage": 96.13, "elapsed_time": "2:12:40", "remaining_time": "0:05:20", "throughput": 5164.53, "total_tokens": 41110560} {"current_steps": 40785, "total_steps": 42420, "loss": 0.3231, "lr": 4.5240146039467465e-06, "epoch": 19.22913719943423, "percentage": 96.15, "elapsed_time": "2:12:41", "remaining_time": "0:05:19", "throughput": 5164.55, "total_tokens": 41115104} {"current_steps": 40790, "total_steps": 42420, "loss": 0.3521, "lr": 4.496445417926265e-06, "epoch": 19.23149457802923, "percentage": 96.16, "elapsed_time": "2:12:42", "remaining_time": "0:05:18", "throughput": 5164.6, "total_tokens": 41120800} {"current_steps": 40795, "total_steps": 42420, "loss": 0.3535, "lr": 4.468960112244458e-06, "epoch": 19.233851956624235, "percentage": 96.17, "elapsed_time": "2:12:42", "remaining_time": "0:05:17", "throughput": 5164.62, "total_tokens": 41125472} {"current_steps": 40800, "total_steps": 42420, "loss": 0.2957, "lr": 4.441558691554049e-06, "epoch": 19.236209335219236, "percentage": 96.18, "elapsed_time": "2:12:43", "remaining_time": "0:05:16", "throughput": 5164.63, "total_tokens": 41129856} {"current_steps": 40805, "total_steps": 42420, "loss": 0.3287, "lr": 4.4142411604936595e-06, "epoch": 19.23856671381424, "percentage": 96.19, "elapsed_time": "2:12:44", "remaining_time": "0:05:15", "throughput": 5164.69, "total_tokens": 41135520} {"current_steps": 40810, "total_steps": 42420, "loss": 0.3487, "lr": 4.3870075236876476e-06, "epoch": 19.24092409240924, "percentage": 96.2, "elapsed_time": "2:12:45", "remaining_time": "0:05:14", "throughput": 5164.72, "total_tokens": 41140512} {"current_steps": 40815, "total_steps": 42420, "loss": 0.3847, "lr": 4.359857785746213e-06, "epoch": 19.243281471004245, "percentage": 96.22, "elapsed_time": "2:12:46", "remaining_time": "0:05:13", "throughput": 5164.77, "total_tokens": 41146048} {"current_steps": 40820, "total_steps": 42420, "loss": 0.3111, "lr": 4.332791951265402e-06, "epoch": 19.245638849599246, "percentage": 96.23, "elapsed_time": "2:12:47", "remaining_time": "0:05:12", "throughput": 5164.77, "total_tokens": 41150368} {"current_steps": 40825, "total_steps": 42420, "loss": 0.2907, "lr": 4.305810024826939e-06, "epoch": 19.247996228194246, "percentage": 96.24, "elapsed_time": "2:12:48", "remaining_time": "0:05:11", "throughput": 5164.81, "total_tokens": 41155168} {"current_steps": 40830, "total_steps": 42420, "loss": 0.3483, "lr": 4.278912010998448e-06, "epoch": 19.25035360678925, "percentage": 96.25, "elapsed_time": "2:12:49", "remaining_time": "0:05:10", "throughput": 5164.86, "total_tokens": 41160896} {"current_steps": 40835, "total_steps": 42420, "loss": 0.2958, "lr": 4.2520979143333415e-06, "epoch": 19.25271098538425, "percentage": 96.26, "elapsed_time": "2:12:50", "remaining_time": "0:05:09", "throughput": 5164.88, "total_tokens": 41165696} {"current_steps": 40840, "total_steps": 42420, "loss": 0.2844, "lr": 4.225367739370656e-06, "epoch": 19.255068363979255, "percentage": 96.28, "elapsed_time": "2:12:51", "remaining_time": "0:05:08", "throughput": 5164.87, "total_tokens": 41170208} {"current_steps": 40845, "total_steps": 42420, "loss": 0.3216, "lr": 4.198721490635493e-06, "epoch": 19.257425742574256, "percentage": 96.29, "elapsed_time": "2:12:52", "remaining_time": "0:05:07", "throughput": 5164.88, "total_tokens": 41174816} {"current_steps": 40850, "total_steps": 42420, "loss": 0.3341, "lr": 4.172159172638523e-06, "epoch": 19.25978312116926, "percentage": 96.3, "elapsed_time": "2:12:52", "remaining_time": "0:05:06", "throughput": 5164.89, "total_tokens": 41179232} {"current_steps": 40855, "total_steps": 42420, "loss": 0.2988, "lr": 4.145680789876316e-06, "epoch": 19.26214049976426, "percentage": 96.31, "elapsed_time": "2:12:53", "remaining_time": "0:05:05", "throughput": 5164.91, "total_tokens": 41184064} {"current_steps": 40860, "total_steps": 42420, "loss": 0.2788, "lr": 4.119286346831231e-06, "epoch": 19.264497878359265, "percentage": 96.32, "elapsed_time": "2:12:54", "remaining_time": "0:05:04", "throughput": 5164.95, "total_tokens": 41189312} {"current_steps": 40865, "total_steps": 42420, "loss": 0.3131, "lr": 4.092975847971414e-06, "epoch": 19.266855256954265, "percentage": 96.33, "elapsed_time": "2:12:55", "remaining_time": "0:05:03", "throughput": 5164.99, "total_tokens": 41194720} {"current_steps": 40870, "total_steps": 42420, "loss": 0.3351, "lr": 4.066749297750749e-06, "epoch": 19.26921263554927, "percentage": 96.35, "elapsed_time": "2:12:56", "remaining_time": "0:05:02", "throughput": 5165.02, "total_tokens": 41199520} {"current_steps": 40875, "total_steps": 42420, "loss": 0.3229, "lr": 4.040606700608962e-06, "epoch": 19.27157001414427, "percentage": 96.36, "elapsed_time": "2:12:57", "remaining_time": "0:05:01", "throughput": 5165.01, "total_tokens": 41203584} {"current_steps": 40880, "total_steps": 42420, "loss": 0.3579, "lr": 4.014548060971568e-06, "epoch": 19.273927392739274, "percentage": 96.37, "elapsed_time": "2:12:58", "remaining_time": "0:05:00", "throughput": 5165.02, "total_tokens": 41208192} {"current_steps": 40885, "total_steps": 42420, "loss": 0.3039, "lr": 3.988573383249816e-06, "epoch": 19.276284771334275, "percentage": 96.38, "elapsed_time": "2:12:59", "remaining_time": "0:04:59", "throughput": 5165.02, "total_tokens": 41212576} {"current_steps": 40890, "total_steps": 42420, "loss": 0.3514, "lr": 3.9626826718408e-06, "epoch": 19.27864214992928, "percentage": 96.39, "elapsed_time": "2:13:00", "remaining_time": "0:04:58", "throughput": 5165.05, "total_tokens": 41217376} {"current_steps": 40895, "total_steps": 42420, "loss": 0.2815, "lr": 3.9368759311273484e-06, "epoch": 19.28099952852428, "percentage": 96.4, "elapsed_time": "2:13:00", "remaining_time": "0:04:57", "throughput": 5165.08, "total_tokens": 41222432} {"current_steps": 40900, "total_steps": 42420, "loss": 0.3114, "lr": 3.9111531654781875e-06, "epoch": 19.283356907119284, "percentage": 96.42, "elapsed_time": "2:13:01", "remaining_time": "0:04:56", "throughput": 5165.1, "total_tokens": 41226976} {"current_steps": 40905, "total_steps": 42420, "loss": 0.3172, "lr": 3.885514379247612e-06, "epoch": 19.285714285714285, "percentage": 96.43, "elapsed_time": "2:13:02", "remaining_time": "0:04:55", "throughput": 5165.17, "total_tokens": 41232960} {"current_steps": 40910, "total_steps": 42420, "loss": 0.3663, "lr": 3.859959576775984e-06, "epoch": 19.28807166430929, "percentage": 96.44, "elapsed_time": "2:13:03", "remaining_time": "0:04:54", "throughput": 5165.19, "total_tokens": 41237952} {"current_steps": 40915, "total_steps": 42420, "loss": 0.4145, "lr": 3.834488762389177e-06, "epoch": 19.29042904290429, "percentage": 96.45, "elapsed_time": "2:13:04", "remaining_time": "0:04:53", "throughput": 5165.23, "total_tokens": 41242912} {"current_steps": 40920, "total_steps": 42420, "loss": 0.3018, "lr": 3.809101940399018e-06, "epoch": 19.292786421499294, "percentage": 96.46, "elapsed_time": "2:13:05", "remaining_time": "0:04:52", "throughput": 5165.26, "total_tokens": 41248000} {"current_steps": 40925, "total_steps": 42420, "loss": 0.3503, "lr": 3.7837991151030703e-06, "epoch": 19.295143800094294, "percentage": 96.48, "elapsed_time": "2:13:06", "remaining_time": "0:04:51", "throughput": 5165.28, "total_tokens": 41252544} {"current_steps": 40930, "total_steps": 42420, "loss": 0.3435, "lr": 3.7585802907846856e-06, "epoch": 19.2975011786893, "percentage": 96.49, "elapsed_time": "2:13:07", "remaining_time": "0:04:50", "throughput": 5165.29, "total_tokens": 41256992} {"current_steps": 40935, "total_steps": 42420, "loss": 0.3116, "lr": 3.733445471712893e-06, "epoch": 19.2998585572843, "percentage": 96.5, "elapsed_time": "2:13:08", "remaining_time": "0:04:49", "throughput": 5165.29, "total_tokens": 41261312} {"current_steps": 40940, "total_steps": 42420, "loss": 0.3246, "lr": 3.708394662142678e-06, "epoch": 19.302215935879303, "percentage": 96.51, "elapsed_time": "2:13:08", "remaining_time": "0:04:48", "throughput": 5165.27, "total_tokens": 41265216} {"current_steps": 40945, "total_steps": 42420, "loss": 0.3776, "lr": 3.6834278663147035e-06, "epoch": 19.304573314474304, "percentage": 96.52, "elapsed_time": "2:13:09", "remaining_time": "0:04:47", "throughput": 5165.3, "total_tokens": 41270080} {"current_steps": 40950, "total_steps": 42420, "loss": 0.3256, "lr": 3.658545088455423e-06, "epoch": 19.306930693069308, "percentage": 96.53, "elapsed_time": "2:13:10", "remaining_time": "0:04:46", "throughput": 5165.3, "total_tokens": 41274592} {"current_steps": 40955, "total_steps": 42420, "loss": 0.3383, "lr": 3.6337463327770214e-06, "epoch": 19.30928807166431, "percentage": 96.55, "elapsed_time": "2:13:11", "remaining_time": "0:04:45", "throughput": 5165.3, "total_tokens": 41279200} {"current_steps": 40960, "total_steps": 42420, "loss": 0.3757, "lr": 3.609031603477586e-06, "epoch": 19.311645450259313, "percentage": 96.56, "elapsed_time": "2:13:12", "remaining_time": "0:04:44", "throughput": 5165.34, "total_tokens": 41284384} {"current_steps": 40965, "total_steps": 42420, "loss": 0.3156, "lr": 3.584400904740881e-06, "epoch": 19.314002828854314, "percentage": 96.57, "elapsed_time": "2:13:13", "remaining_time": "0:04:43", "throughput": 5165.36, "total_tokens": 41289056} {"current_steps": 40970, "total_steps": 42420, "loss": 0.3196, "lr": 3.559854240736404e-06, "epoch": 19.316360207449318, "percentage": 96.58, "elapsed_time": "2:13:14", "remaining_time": "0:04:42", "throughput": 5165.37, "total_tokens": 41293632} {"current_steps": 40975, "total_steps": 42420, "loss": 0.3449, "lr": 3.5353916156195545e-06, "epoch": 19.31871758604432, "percentage": 96.59, "elapsed_time": "2:13:15", "remaining_time": "0:04:41", "throughput": 5165.4, "total_tokens": 41298496} {"current_steps": 40980, "total_steps": 42420, "loss": 0.3384, "lr": 3.5110130335314074e-06, "epoch": 19.321074964639323, "percentage": 96.61, "elapsed_time": "2:13:16", "remaining_time": "0:04:40", "throughput": 5165.43, "total_tokens": 41303616} {"current_steps": 40985, "total_steps": 42420, "loss": 0.317, "lr": 3.4867184985988286e-06, "epoch": 19.323432343234323, "percentage": 96.62, "elapsed_time": "2:13:17", "remaining_time": "0:04:39", "throughput": 5165.46, "total_tokens": 41308448} {"current_steps": 40990, "total_steps": 42420, "loss": 0.3384, "lr": 3.462508014934529e-06, "epoch": 19.325789721829327, "percentage": 96.63, "elapsed_time": "2:13:17", "remaining_time": "0:04:39", "throughput": 5165.47, "total_tokens": 41313152} {"current_steps": 40995, "total_steps": 42420, "loss": 0.3166, "lr": 3.4383815866368408e-06, "epoch": 19.328147100424328, "percentage": 96.64, "elapsed_time": "2:13:18", "remaining_time": "0:04:38", "throughput": 5165.49, "total_tokens": 41318016} {"current_steps": 41000, "total_steps": 42420, "loss": 0.3573, "lr": 3.414339217790052e-06, "epoch": 19.330504479019332, "percentage": 96.65, "elapsed_time": "2:13:19", "remaining_time": "0:04:37", "throughput": 5165.51, "total_tokens": 41322880} {"current_steps": 41005, "total_steps": 42420, "loss": 0.3586, "lr": 3.3903809124641304e-06, "epoch": 19.332861857614333, "percentage": 96.66, "elapsed_time": "2:13:20", "remaining_time": "0:04:36", "throughput": 5165.55, "total_tokens": 41328512} {"current_steps": 41010, "total_steps": 42420, "loss": 0.307, "lr": 3.3665066747147198e-06, "epoch": 19.335219236209337, "percentage": 96.68, "elapsed_time": "2:13:21", "remaining_time": "0:04:35", "throughput": 5165.53, "total_tokens": 41332448} {"current_steps": 41015, "total_steps": 42420, "loss": 0.3512, "lr": 3.3427165085833655e-06, "epoch": 19.337576614804338, "percentage": 96.69, "elapsed_time": "2:13:22", "remaining_time": "0:04:34", "throughput": 5165.57, "total_tokens": 41337504} {"current_steps": 41020, "total_steps": 42420, "loss": 0.3354, "lr": 3.319010418097346e-06, "epoch": 19.33993399339934, "percentage": 96.7, "elapsed_time": "2:13:23", "remaining_time": "0:04:33", "throughput": 5165.66, "total_tokens": 41343744} {"current_steps": 41025, "total_steps": 42420, "loss": 0.3388, "lr": 3.295388407269728e-06, "epoch": 19.342291371994342, "percentage": 96.71, "elapsed_time": "2:13:24", "remaining_time": "0:04:32", "throughput": 5165.68, "total_tokens": 41348256} {"current_steps": 41030, "total_steps": 42420, "loss": 0.3079, "lr": 3.2718504800992586e-06, "epoch": 19.344648750589343, "percentage": 96.72, "elapsed_time": "2:13:25", "remaining_time": "0:04:31", "throughput": 5165.69, "total_tokens": 41352608} {"current_steps": 41035, "total_steps": 42420, "loss": 0.3816, "lr": 3.2483966405705277e-06, "epoch": 19.347006129184347, "percentage": 96.74, "elapsed_time": "2:13:26", "remaining_time": "0:04:30", "throughput": 5165.73, "total_tokens": 41357664} {"current_steps": 41040, "total_steps": 42420, "loss": 0.3045, "lr": 3.225026892653915e-06, "epoch": 19.349363507779348, "percentage": 96.75, "elapsed_time": "2:13:27", "remaining_time": "0:04:29", "throughput": 5165.77, "total_tokens": 41362560} {"current_steps": 41045, "total_steps": 42420, "loss": 0.3159, "lr": 3.2017412403055335e-06, "epoch": 19.351720886374352, "percentage": 96.76, "elapsed_time": "2:13:28", "remaining_time": "0:04:28", "throughput": 5165.81, "total_tokens": 41368000} {"current_steps": 41050, "total_steps": 42420, "loss": 0.3494, "lr": 3.178539687467119e-06, "epoch": 19.354078264969353, "percentage": 96.77, "elapsed_time": "2:13:29", "remaining_time": "0:04:27", "throughput": 5165.88, "total_tokens": 41373696} {"current_steps": 41055, "total_steps": 42420, "loss": 0.4023, "lr": 3.1554222380664187e-06, "epoch": 19.356435643564357, "percentage": 96.78, "elapsed_time": "2:13:29", "remaining_time": "0:04:26", "throughput": 5165.91, "total_tokens": 41378720} {"current_steps": 41060, "total_steps": 42420, "loss": 0.3478, "lr": 3.1323888960168024e-06, "epoch": 19.358793022159357, "percentage": 96.79, "elapsed_time": "2:13:30", "remaining_time": "0:04:25", "throughput": 5165.93, "total_tokens": 41383328} {"current_steps": 41065, "total_steps": 42420, "loss": 0.273, "lr": 3.1094396652174283e-06, "epoch": 19.36115040075436, "percentage": 96.81, "elapsed_time": "2:13:31", "remaining_time": "0:04:24", "throughput": 5165.97, "total_tokens": 41388544} {"current_steps": 41070, "total_steps": 42420, "loss": 0.3142, "lr": 3.0865745495531893e-06, "epoch": 19.363507779349362, "percentage": 96.82, "elapsed_time": "2:13:32", "remaining_time": "0:04:23", "throughput": 5166.01, "total_tokens": 41393920} {"current_steps": 41075, "total_steps": 42420, "loss": 0.3084, "lr": 3.0637935528947113e-06, "epoch": 19.365865157944366, "percentage": 96.83, "elapsed_time": "2:13:33", "remaining_time": "0:04:22", "throughput": 5166.05, "total_tokens": 41399072} {"current_steps": 41080, "total_steps": 42420, "loss": 0.3421, "lr": 3.0410966790985763e-06, "epoch": 19.368222536539367, "percentage": 96.84, "elapsed_time": "2:13:34", "remaining_time": "0:04:21", "throughput": 5166.11, "total_tokens": 41404704} {"current_steps": 41085, "total_steps": 42420, "loss": 0.3203, "lr": 3.0184839320068768e-06, "epoch": 19.37057991513437, "percentage": 96.85, "elapsed_time": "2:13:35", "remaining_time": "0:04:20", "throughput": 5166.12, "total_tokens": 41409184} {"current_steps": 41090, "total_steps": 42420, "loss": 0.3043, "lr": 2.9959553154475515e-06, "epoch": 19.372937293729372, "percentage": 96.86, "elapsed_time": "2:13:36", "remaining_time": "0:04:19", "throughput": 5166.19, "total_tokens": 41415168} {"current_steps": 41095, "total_steps": 42420, "loss": 0.3177, "lr": 2.9735108332343276e-06, "epoch": 19.375294672324376, "percentage": 96.88, "elapsed_time": "2:13:37", "remaining_time": "0:04:18", "throughput": 5166.23, "total_tokens": 41420352} {"current_steps": 41100, "total_steps": 42420, "loss": 0.2748, "lr": 2.9511504891667207e-06, "epoch": 19.377652050919377, "percentage": 96.89, "elapsed_time": "2:13:38", "remaining_time": "0:04:17", "throughput": 5166.32, "total_tokens": 41426912} {"current_steps": 41105, "total_steps": 42420, "loss": 0.2928, "lr": 2.9288742870299257e-06, "epoch": 19.38000942951438, "percentage": 96.9, "elapsed_time": "2:13:39", "remaining_time": "0:04:16", "throughput": 5166.35, "total_tokens": 41431904} {"current_steps": 41110, "total_steps": 42420, "loss": 0.338, "lr": 2.90668223059487e-06, "epoch": 19.38236680810938, "percentage": 96.91, "elapsed_time": "2:13:40", "remaining_time": "0:04:15", "throughput": 5166.39, "total_tokens": 41437024} {"current_steps": 41115, "total_steps": 42420, "loss": 0.3175, "lr": 2.8845743236184387e-06, "epoch": 19.384724186704386, "percentage": 96.92, "elapsed_time": "2:13:41", "remaining_time": "0:04:14", "throughput": 5166.37, "total_tokens": 41440928} {"current_steps": 41120, "total_steps": 42420, "loss": 0.3307, "lr": 2.86255056984297e-06, "epoch": 19.387081565299386, "percentage": 96.94, "elapsed_time": "2:13:42", "remaining_time": "0:04:13", "throughput": 5166.41, "total_tokens": 41446144} {"current_steps": 41125, "total_steps": 42420, "loss": 0.2745, "lr": 2.8406109729968154e-06, "epoch": 19.38943894389439, "percentage": 96.95, "elapsed_time": "2:13:43", "remaining_time": "0:04:12", "throughput": 5166.47, "total_tokens": 41451776} {"current_steps": 41130, "total_steps": 42420, "loss": 0.292, "lr": 2.8187555367939486e-06, "epoch": 19.39179632248939, "percentage": 96.96, "elapsed_time": "2:13:44", "remaining_time": "0:04:11", "throughput": 5166.49, "total_tokens": 41456480} {"current_steps": 41135, "total_steps": 42420, "loss": 0.3685, "lr": 2.796984264934077e-06, "epoch": 19.394153701084395, "percentage": 96.97, "elapsed_time": "2:13:45", "remaining_time": "0:04:10", "throughput": 5166.53, "total_tokens": 41461760} {"current_steps": 41140, "total_steps": 42420, "loss": 0.3341, "lr": 2.7752971611027524e-06, "epoch": 19.396511079679396, "percentage": 96.98, "elapsed_time": "2:13:45", "remaining_time": "0:04:09", "throughput": 5166.5, "total_tokens": 41465536} {"current_steps": 41145, "total_steps": 42420, "loss": 0.2892, "lr": 2.753694228971204e-06, "epoch": 19.3988684582744, "percentage": 96.99, "elapsed_time": "2:13:46", "remaining_time": "0:04:08", "throughput": 5166.57, "total_tokens": 41471392} {"current_steps": 41150, "total_steps": 42420, "loss": 0.3581, "lr": 2.7321754721965077e-06, "epoch": 19.4012258368694, "percentage": 97.01, "elapsed_time": "2:13:47", "remaining_time": "0:04:07", "throughput": 5166.58, "total_tokens": 41476224} {"current_steps": 41155, "total_steps": 42420, "loss": 0.3171, "lr": 2.710740894421304e-06, "epoch": 19.403583215464405, "percentage": 97.02, "elapsed_time": "2:13:48", "remaining_time": "0:04:06", "throughput": 5166.61, "total_tokens": 41481216} {"current_steps": 41160, "total_steps": 42420, "loss": 0.3304, "lr": 2.689390499274247e-06, "epoch": 19.405940594059405, "percentage": 97.03, "elapsed_time": "2:13:49", "remaining_time": "0:04:05", "throughput": 5166.66, "total_tokens": 41486336} {"current_steps": 41165, "total_steps": 42420, "loss": 0.3385, "lr": 2.6681242903695e-06, "epoch": 19.40829797265441, "percentage": 97.04, "elapsed_time": "2:13:50", "remaining_time": "0:04:04", "throughput": 5166.7, "total_tokens": 41491360} {"current_steps": 41170, "total_steps": 42420, "loss": 0.3477, "lr": 2.6469422713070733e-06, "epoch": 19.41065535124941, "percentage": 97.05, "elapsed_time": "2:13:51", "remaining_time": "0:04:03", "throughput": 5166.8, "total_tokens": 41497888} {"current_steps": 41175, "total_steps": 42420, "loss": 0.3153, "lr": 2.625844445672765e-06, "epoch": 19.413012729844414, "percentage": 97.07, "elapsed_time": "2:13:52", "remaining_time": "0:04:02", "throughput": 5166.83, "total_tokens": 41503040} {"current_steps": 41180, "total_steps": 42420, "loss": 0.3533, "lr": 2.6048308170381062e-06, "epoch": 19.415370108439415, "percentage": 97.08, "elapsed_time": "2:13:53", "remaining_time": "0:04:01", "throughput": 5166.91, "total_tokens": 41508992} {"current_steps": 41185, "total_steps": 42420, "loss": 0.3644, "lr": 2.5839013889602523e-06, "epoch": 19.41772748703442, "percentage": 97.09, "elapsed_time": "2:13:54", "remaining_time": "0:04:00", "throughput": 5166.95, "total_tokens": 41514176} {"current_steps": 41190, "total_steps": 42420, "loss": 0.381, "lr": 2.5630561649822028e-06, "epoch": 19.42008486562942, "percentage": 97.1, "elapsed_time": "2:13:55", "remaining_time": "0:03:59", "throughput": 5166.97, "total_tokens": 41519008} {"current_steps": 41195, "total_steps": 42420, "loss": 0.3575, "lr": 2.5422951486328005e-06, "epoch": 19.422442244224424, "percentage": 97.11, "elapsed_time": "2:13:56", "remaining_time": "0:03:58", "throughput": 5166.99, "total_tokens": 41523840} {"current_steps": 41200, "total_steps": 42420, "loss": 0.2869, "lr": 2.5216183434265127e-06, "epoch": 19.424799622819425, "percentage": 97.12, "elapsed_time": "2:13:57", "remaining_time": "0:03:57", "throughput": 5166.99, "total_tokens": 41528224} {"current_steps": 41205, "total_steps": 42420, "loss": 0.3055, "lr": 2.5010257528634837e-06, "epoch": 19.42715700141443, "percentage": 97.14, "elapsed_time": "2:13:58", "remaining_time": "0:03:57", "throughput": 5167.08, "total_tokens": 41534560} {"current_steps": 41210, "total_steps": 42420, "loss": 0.3251, "lr": 2.480517380429759e-06, "epoch": 19.42951438000943, "percentage": 97.15, "elapsed_time": "2:13:59", "remaining_time": "0:03:56", "throughput": 5167.1, "total_tokens": 41539392} {"current_steps": 41215, "total_steps": 42420, "loss": 0.3064, "lr": 2.4600932295970603e-06, "epoch": 19.431871758604434, "percentage": 97.16, "elapsed_time": "2:14:00", "remaining_time": "0:03:55", "throughput": 5167.12, "total_tokens": 41544000} {"current_steps": 41220, "total_steps": 42420, "loss": 0.3533, "lr": 2.439753303822789e-06, "epoch": 19.434229137199434, "percentage": 97.17, "elapsed_time": "2:14:01", "remaining_time": "0:03:54", "throughput": 5167.17, "total_tokens": 41549344} {"current_steps": 41225, "total_steps": 42420, "loss": 0.3674, "lr": 2.419497606550192e-06, "epoch": 19.436586515794435, "percentage": 97.18, "elapsed_time": "2:14:01", "remaining_time": "0:03:53", "throughput": 5167.19, "total_tokens": 41554304} {"current_steps": 41230, "total_steps": 42420, "loss": 0.3634, "lr": 2.3993261412082468e-06, "epoch": 19.43894389438944, "percentage": 97.19, "elapsed_time": "2:14:02", "remaining_time": "0:03:52", "throughput": 5167.2, "total_tokens": 41558976} {"current_steps": 41235, "total_steps": 42420, "loss": 0.3667, "lr": 2.379238911211612e-06, "epoch": 19.44130127298444, "percentage": 97.21, "elapsed_time": "2:14:03", "remaining_time": "0:03:51", "throughput": 5167.21, "total_tokens": 41563648} {"current_steps": 41240, "total_steps": 42420, "loss": 0.3715, "lr": 2.3592359199606783e-06, "epoch": 19.443658651579444, "percentage": 97.22, "elapsed_time": "2:14:04", "remaining_time": "0:03:50", "throughput": 5167.23, "total_tokens": 41568512} {"current_steps": 41245, "total_steps": 42420, "loss": 0.3336, "lr": 2.339317170841626e-06, "epoch": 19.446016030174444, "percentage": 97.23, "elapsed_time": "2:14:05", "remaining_time": "0:03:49", "throughput": 5167.39, "total_tokens": 41576544} {"current_steps": 41250, "total_steps": 42420, "loss": 0.2935, "lr": 2.3194826672264245e-06, "epoch": 19.44837340876945, "percentage": 97.24, "elapsed_time": "2:14:06", "remaining_time": "0:03:48", "throughput": 5167.45, "total_tokens": 41582048} {"current_steps": 41255, "total_steps": 42420, "loss": 0.3085, "lr": 2.299732412472666e-06, "epoch": 19.45073078736445, "percentage": 97.25, "elapsed_time": "2:14:07", "remaining_time": "0:03:47", "throughput": 5167.47, "total_tokens": 41586784} {"current_steps": 41260, "total_steps": 42420, "loss": 0.2888, "lr": 2.280066409923787e-06, "epoch": 19.453088165959453, "percentage": 97.27, "elapsed_time": "2:14:08", "remaining_time": "0:03:46", "throughput": 5167.54, "total_tokens": 41592416} {"current_steps": 41265, "total_steps": 42420, "loss": 0.3305, "lr": 2.2604846629087904e-06, "epoch": 19.455445544554454, "percentage": 97.28, "elapsed_time": "2:14:09", "remaining_time": "0:03:45", "throughput": 5167.53, "total_tokens": 41596832} {"current_steps": 41270, "total_steps": 42420, "loss": 0.3148, "lr": 2.240987174742637e-06, "epoch": 19.45780292314946, "percentage": 97.29, "elapsed_time": "2:14:10", "remaining_time": "0:03:44", "throughput": 5167.57, "total_tokens": 41602112} {"current_steps": 41275, "total_steps": 42420, "loss": 0.3373, "lr": 2.2215739487258526e-06, "epoch": 19.46016030174446, "percentage": 97.3, "elapsed_time": "2:14:11", "remaining_time": "0:03:43", "throughput": 5167.56, "total_tokens": 41606272} {"current_steps": 41280, "total_steps": 42420, "loss": 0.3192, "lr": 2.2022449881448083e-06, "epoch": 19.462517680339463, "percentage": 97.31, "elapsed_time": "2:14:12", "remaining_time": "0:03:42", "throughput": 5167.55, "total_tokens": 41610368} {"current_steps": 41285, "total_steps": 42420, "loss": 0.2928, "lr": 2.183000296271609e-06, "epoch": 19.464875058934464, "percentage": 97.32, "elapsed_time": "2:14:13", "remaining_time": "0:03:41", "throughput": 5167.6, "total_tokens": 41615808} {"current_steps": 41290, "total_steps": 42420, "loss": 0.3691, "lr": 2.1638398763639266e-06, "epoch": 19.467232437529468, "percentage": 97.34, "elapsed_time": "2:14:14", "remaining_time": "0:03:40", "throughput": 5167.61, "total_tokens": 41620480} {"current_steps": 41295, "total_steps": 42420, "loss": 0.3189, "lr": 2.1447637316655e-06, "epoch": 19.46958981612447, "percentage": 97.35, "elapsed_time": "2:14:15", "remaining_time": "0:03:39", "throughput": 5167.65, "total_tokens": 41625632} {"current_steps": 41300, "total_steps": 42420, "loss": 0.3486, "lr": 2.125771865405357e-06, "epoch": 19.471947194719473, "percentage": 97.36, "elapsed_time": "2:14:16", "remaining_time": "0:03:38", "throughput": 5167.73, "total_tokens": 41632032} {"current_steps": 41305, "total_steps": 42420, "loss": 0.3273, "lr": 2.1068642807986484e-06, "epoch": 19.474304573314473, "percentage": 97.37, "elapsed_time": "2:14:17", "remaining_time": "0:03:37", "throughput": 5167.76, "total_tokens": 41636960} {"current_steps": 41310, "total_steps": 42420, "loss": 0.3094, "lr": 2.088040981046091e-06, "epoch": 19.476661951909477, "percentage": 97.38, "elapsed_time": "2:14:17", "remaining_time": "0:03:36", "throughput": 5167.8, "total_tokens": 41641984} {"current_steps": 41315, "total_steps": 42420, "loss": 0.312, "lr": 2.0693019693341364e-06, "epoch": 19.479019330504478, "percentage": 97.4, "elapsed_time": "2:14:18", "remaining_time": "0:03:35", "throughput": 5167.82, "total_tokens": 41646752} {"current_steps": 41320, "total_steps": 42420, "loss": 0.3289, "lr": 2.0506472488349693e-06, "epoch": 19.481376709099482, "percentage": 97.41, "elapsed_time": "2:14:19", "remaining_time": "0:03:34", "throughput": 5167.84, "total_tokens": 41651488} {"current_steps": 41325, "total_steps": 42420, "loss": 0.3338, "lr": 2.032076822706508e-06, "epoch": 19.483734087694483, "percentage": 97.42, "elapsed_time": "2:14:20", "remaining_time": "0:03:33", "throughput": 5167.89, "total_tokens": 41657024} {"current_steps": 41330, "total_steps": 42420, "loss": 0.322, "lr": 2.013590694092404e-06, "epoch": 19.486091466289487, "percentage": 97.43, "elapsed_time": "2:14:21", "remaining_time": "0:03:32", "throughput": 5167.9, "total_tokens": 41661696} {"current_steps": 41335, "total_steps": 42420, "loss": 0.3667, "lr": 1.9951888661221553e-06, "epoch": 19.488448844884488, "percentage": 97.44, "elapsed_time": "2:14:22", "remaining_time": "0:03:31", "throughput": 5167.97, "total_tokens": 41667968} {"current_steps": 41340, "total_steps": 42420, "loss": 0.3385, "lr": 1.976871341910713e-06, "epoch": 19.490806223479492, "percentage": 97.45, "elapsed_time": "2:14:23", "remaining_time": "0:03:30", "throughput": 5167.97, "total_tokens": 41672128} {"current_steps": 41345, "total_steps": 42420, "loss": 0.3176, "lr": 1.958638124559042e-06, "epoch": 19.493163602074493, "percentage": 97.47, "elapsed_time": "2:14:24", "remaining_time": "0:03:29", "throughput": 5168.01, "total_tokens": 41677312} {"current_steps": 41350, "total_steps": 42420, "loss": 0.311, "lr": 1.940489217153618e-06, "epoch": 19.495520980669497, "percentage": 97.48, "elapsed_time": "2:14:25", "remaining_time": "0:03:28", "throughput": 5168.06, "total_tokens": 41683200} {"current_steps": 41355, "total_steps": 42420, "loss": 0.3303, "lr": 1.9224246227668716e-06, "epoch": 19.497878359264497, "percentage": 97.49, "elapsed_time": "2:14:26", "remaining_time": "0:03:27", "throughput": 5168.09, "total_tokens": 41688128} {"current_steps": 41360, "total_steps": 42420, "loss": 0.3198, "lr": 1.9044443444566906e-06, "epoch": 19.5002357378595, "percentage": 97.5, "elapsed_time": "2:14:27", "remaining_time": "0:03:26", "throughput": 5168.13, "total_tokens": 41693472} {"current_steps": 41365, "total_steps": 42420, "loss": 0.3532, "lr": 1.8865483852669729e-06, "epoch": 19.502593116454502, "percentage": 97.51, "elapsed_time": "2:14:28", "remaining_time": "0:03:25", "throughput": 5168.14, "total_tokens": 41697952} {"current_steps": 41370, "total_steps": 42420, "loss": 0.3091, "lr": 1.8687367482271288e-06, "epoch": 19.504950495049506, "percentage": 97.52, "elapsed_time": "2:14:29", "remaining_time": "0:03:24", "throughput": 5168.18, "total_tokens": 41703136} {"current_steps": 41375, "total_steps": 42420, "loss": 0.3204, "lr": 1.8510094363523023e-06, "epoch": 19.507307873644507, "percentage": 97.54, "elapsed_time": "2:14:30", "remaining_time": "0:03:23", "throughput": 5168.24, "total_tokens": 41708800} {"current_steps": 41380, "total_steps": 42420, "loss": 0.302, "lr": 1.8333664526435368e-06, "epoch": 19.50966525223951, "percentage": 97.55, "elapsed_time": "2:14:31", "remaining_time": "0:03:22", "throughput": 5168.28, "total_tokens": 41714048} {"current_steps": 41385, "total_steps": 42420, "loss": 0.3367, "lr": 1.8158078000874434e-06, "epoch": 19.51202263083451, "percentage": 97.56, "elapsed_time": "2:14:32", "remaining_time": "0:03:21", "throughput": 5168.31, "total_tokens": 41719200} {"current_steps": 41390, "total_steps": 42420, "loss": 0.3736, "lr": 1.7983334816563667e-06, "epoch": 19.514380009429516, "percentage": 97.57, "elapsed_time": "2:14:33", "remaining_time": "0:03:20", "throughput": 5168.36, "total_tokens": 41724416} {"current_steps": 41395, "total_steps": 42420, "loss": 0.2938, "lr": 1.7809435003084407e-06, "epoch": 19.516737388024517, "percentage": 97.58, "elapsed_time": "2:14:33", "remaining_time": "0:03:19", "throughput": 5168.39, "total_tokens": 41729184} {"current_steps": 41400, "total_steps": 42420, "loss": 0.3024, "lr": 1.7636378589875325e-06, "epoch": 19.51909476661952, "percentage": 97.6, "elapsed_time": "2:14:34", "remaining_time": "0:03:18", "throughput": 5168.41, "total_tokens": 41734016} {"current_steps": 41405, "total_steps": 42420, "loss": 0.3287, "lr": 1.7464165606231874e-06, "epoch": 19.52145214521452, "percentage": 97.61, "elapsed_time": "2:14:35", "remaining_time": "0:03:17", "throughput": 5168.43, "total_tokens": 41738880} {"current_steps": 41410, "total_steps": 42420, "loss": 0.3627, "lr": 1.7292796081306295e-06, "epoch": 19.523809523809526, "percentage": 97.62, "elapsed_time": "2:14:36", "remaining_time": "0:03:16", "throughput": 5168.42, "total_tokens": 41743168} {"current_steps": 41415, "total_steps": 42420, "loss": 0.3053, "lr": 1.7122270044108711e-06, "epoch": 19.526166902404526, "percentage": 97.63, "elapsed_time": "2:14:37", "remaining_time": "0:03:16", "throughput": 5168.46, "total_tokens": 41748192} {"current_steps": 41420, "total_steps": 42420, "loss": 0.3519, "lr": 1.6952587523506591e-06, "epoch": 19.528524280999527, "percentage": 97.64, "elapsed_time": "2:14:38", "remaining_time": "0:03:15", "throughput": 5168.44, "total_tokens": 41752064} {"current_steps": 41425, "total_steps": 42420, "loss": 0.3548, "lr": 1.678374854822362e-06, "epoch": 19.53088165959453, "percentage": 97.65, "elapsed_time": "2:14:39", "remaining_time": "0:03:14", "throughput": 5168.45, "total_tokens": 41756768} {"current_steps": 41430, "total_steps": 42420, "loss": 0.3373, "lr": 1.661575314684194e-06, "epoch": 19.53323903818953, "percentage": 97.67, "elapsed_time": "2:14:40", "remaining_time": "0:03:13", "throughput": 5168.49, "total_tokens": 41761888} {"current_steps": 41435, "total_steps": 42420, "loss": 0.3447, "lr": 1.6448601347800463e-06, "epoch": 19.535596416784536, "percentage": 97.68, "elapsed_time": "2:14:41", "remaining_time": "0:03:12", "throughput": 5168.54, "total_tokens": 41767584} {"current_steps": 41440, "total_steps": 42420, "loss": 0.345, "lr": 1.6282293179394892e-06, "epoch": 19.537953795379536, "percentage": 97.69, "elapsed_time": "2:14:42", "remaining_time": "0:03:11", "throughput": 5168.57, "total_tokens": 41772800} {"current_steps": 41445, "total_steps": 42420, "loss": 0.2817, "lr": 1.6116828669778261e-06, "epoch": 19.54031117397454, "percentage": 97.7, "elapsed_time": "2:14:42", "remaining_time": "0:03:10", "throughput": 5168.58, "total_tokens": 41777344} {"current_steps": 41450, "total_steps": 42420, "loss": 0.3239, "lr": 1.5952207846960942e-06, "epoch": 19.54266855256954, "percentage": 97.71, "elapsed_time": "2:14:43", "remaining_time": "0:03:09", "throughput": 5168.59, "total_tokens": 41781696} {"current_steps": 41455, "total_steps": 42420, "loss": 0.3049, "lr": 1.5788430738810645e-06, "epoch": 19.545025931164545, "percentage": 97.73, "elapsed_time": "2:14:44", "remaining_time": "0:03:08", "throughput": 5168.66, "total_tokens": 41787648} {"current_steps": 41460, "total_steps": 42420, "loss": 0.358, "lr": 1.5625497373051855e-06, "epoch": 19.547383309759546, "percentage": 97.74, "elapsed_time": "2:14:45", "remaining_time": "0:03:07", "throughput": 5168.69, "total_tokens": 41792672} {"current_steps": 41465, "total_steps": 42420, "loss": 0.3595, "lr": 1.5463407777266403e-06, "epoch": 19.54974068835455, "percentage": 97.75, "elapsed_time": "2:14:46", "remaining_time": "0:03:06", "throughput": 5168.72, "total_tokens": 41797568} {"current_steps": 41470, "total_steps": 42420, "loss": 0.3146, "lr": 1.5302161978894002e-06, "epoch": 19.55209806694955, "percentage": 97.76, "elapsed_time": "2:14:47", "remaining_time": "0:03:05", "throughput": 5168.79, "total_tokens": 41803488} {"current_steps": 41475, "total_steps": 42420, "loss": 0.3547, "lr": 1.5141760005229488e-06, "epoch": 19.554455445544555, "percentage": 97.77, "elapsed_time": "2:14:48", "remaining_time": "0:03:04", "throughput": 5168.84, "total_tokens": 41808864} {"current_steps": 41480, "total_steps": 42420, "loss": 0.3757, "lr": 1.4982201883427249e-06, "epoch": 19.556812824139556, "percentage": 97.78, "elapsed_time": "2:14:49", "remaining_time": "0:03:03", "throughput": 5168.82, "total_tokens": 41812768} {"current_steps": 41485, "total_steps": 42420, "loss": 0.2981, "lr": 1.4823487640497347e-06, "epoch": 19.55917020273456, "percentage": 97.8, "elapsed_time": "2:14:50", "remaining_time": "0:03:02", "throughput": 5168.82, "total_tokens": 41816928} {"current_steps": 41490, "total_steps": 42420, "loss": 0.3527, "lr": 1.4665617303307733e-06, "epoch": 19.56152758132956, "percentage": 97.81, "elapsed_time": "2:14:51", "remaining_time": "0:03:01", "throughput": 5168.86, "total_tokens": 41822240} {"current_steps": 41495, "total_steps": 42420, "loss": 0.3869, "lr": 1.4508590898582586e-06, "epoch": 19.563884959924565, "percentage": 97.82, "elapsed_time": "2:14:52", "remaining_time": "0:03:00", "throughput": 5168.89, "total_tokens": 41827296} {"current_steps": 41500, "total_steps": 42420, "loss": 0.3345, "lr": 1.435240845290453e-06, "epoch": 19.566242338519565, "percentage": 97.83, "elapsed_time": "2:14:53", "remaining_time": "0:02:59", "throughput": 5168.95, "total_tokens": 41833120} {"current_steps": 41505, "total_steps": 42420, "loss": 0.3203, "lr": 1.4197069992711865e-06, "epoch": 19.56859971711457, "percentage": 97.84, "elapsed_time": "2:14:54", "remaining_time": "0:02:58", "throughput": 5168.98, "total_tokens": 41838080} {"current_steps": 41510, "total_steps": 42420, "loss": 0.344, "lr": 1.4042575544301329e-06, "epoch": 19.57095709570957, "percentage": 97.85, "elapsed_time": "2:14:54", "remaining_time": "0:02:57", "throughput": 5168.99, "total_tokens": 41842944} {"current_steps": 41515, "total_steps": 42420, "loss": 0.3623, "lr": 1.3888925133825891e-06, "epoch": 19.573314474304574, "percentage": 97.87, "elapsed_time": "2:14:55", "remaining_time": "0:02:56", "throughput": 5169.02, "total_tokens": 41847872} {"current_steps": 41520, "total_steps": 42420, "loss": 0.3034, "lr": 1.3736118787296413e-06, "epoch": 19.575671852899575, "percentage": 97.88, "elapsed_time": "2:14:56", "remaining_time": "0:02:55", "throughput": 5169.01, "total_tokens": 41852160} {"current_steps": 41525, "total_steps": 42420, "loss": 0.3244, "lr": 1.3584156530579984e-06, "epoch": 19.57802923149458, "percentage": 97.89, "elapsed_time": "2:14:57", "remaining_time": "0:02:54", "throughput": 5169.04, "total_tokens": 41857120} {"current_steps": 41530, "total_steps": 42420, "loss": 0.2958, "lr": 1.343303838940102e-06, "epoch": 19.58038661008958, "percentage": 97.9, "elapsed_time": "2:14:58", "remaining_time": "0:02:53", "throughput": 5169.05, "total_tokens": 41861664} {"current_steps": 41535, "total_steps": 42420, "loss": 0.3622, "lr": 1.3282764389341839e-06, "epoch": 19.582743988684584, "percentage": 97.91, "elapsed_time": "2:14:59", "remaining_time": "0:02:52", "throughput": 5169.07, "total_tokens": 41866688} {"current_steps": 41540, "total_steps": 42420, "loss": 0.3125, "lr": 1.3133334555840982e-06, "epoch": 19.585101367279584, "percentage": 97.93, "elapsed_time": "2:15:00", "remaining_time": "0:02:51", "throughput": 5169.11, "total_tokens": 41871904} {"current_steps": 41545, "total_steps": 42420, "loss": 0.3225, "lr": 1.2984748914194323e-06, "epoch": 19.58745874587459, "percentage": 97.94, "elapsed_time": "2:15:01", "remaining_time": "0:02:50", "throughput": 5169.12, "total_tokens": 41876448} {"current_steps": 41550, "total_steps": 42420, "loss": 0.3071, "lr": 1.2837007489555074e-06, "epoch": 19.58981612446959, "percentage": 97.95, "elapsed_time": "2:15:02", "remaining_time": "0:02:49", "throughput": 5169.13, "total_tokens": 41881152} {"current_steps": 41555, "total_steps": 42420, "loss": 0.3637, "lr": 1.2690110306933234e-06, "epoch": 19.592173503064593, "percentage": 97.96, "elapsed_time": "2:15:03", "remaining_time": "0:02:48", "throughput": 5169.15, "total_tokens": 41886592} {"current_steps": 41560, "total_steps": 42420, "loss": 0.3065, "lr": 1.2544057391196129e-06, "epoch": 19.594530881659594, "percentage": 97.97, "elapsed_time": "2:15:04", "remaining_time": "0:02:47", "throughput": 5169.18, "total_tokens": 41891776} {"current_steps": 41565, "total_steps": 42420, "loss": 0.3812, "lr": 1.2398848767067871e-06, "epoch": 19.596888260254598, "percentage": 97.98, "elapsed_time": "2:15:04", "remaining_time": "0:02:46", "throughput": 5169.18, "total_tokens": 41895968} {"current_steps": 41570, "total_steps": 42420, "loss": 0.3197, "lr": 1.225448445912991e-06, "epoch": 19.5992456388496, "percentage": 98.0, "elapsed_time": "2:15:05", "remaining_time": "0:02:45", "throughput": 5169.19, "total_tokens": 41900448} {"current_steps": 41575, "total_steps": 42420, "loss": 0.3459, "lr": 1.2110964491820476e-06, "epoch": 19.601603017444603, "percentage": 98.01, "elapsed_time": "2:15:06", "remaining_time": "0:02:44", "throughput": 5169.22, "total_tokens": 41905664} {"current_steps": 41580, "total_steps": 42420, "loss": 0.3392, "lr": 1.196828888943513e-06, "epoch": 19.603960396039604, "percentage": 98.02, "elapsed_time": "2:15:07", "remaining_time": "0:02:43", "throughput": 5169.25, "total_tokens": 41910336} {"current_steps": 41585, "total_steps": 42420, "loss": 0.3384, "lr": 1.1826457676126778e-06, "epoch": 19.606317774634608, "percentage": 98.03, "elapsed_time": "2:15:08", "remaining_time": "0:02:42", "throughput": 5169.33, "total_tokens": 41916256} {"current_steps": 41590, "total_steps": 42420, "loss": 0.3648, "lr": 1.1685470875904547e-06, "epoch": 19.60867515322961, "percentage": 98.04, "elapsed_time": "2:15:09", "remaining_time": "0:02:41", "throughput": 5169.36, "total_tokens": 41921152} {"current_steps": 41595, "total_steps": 42420, "loss": 0.3904, "lr": 1.1545328512635456e-06, "epoch": 19.611032531824613, "percentage": 98.06, "elapsed_time": "2:15:10", "remaining_time": "0:02:40", "throughput": 5169.41, "total_tokens": 41926432} {"current_steps": 41600, "total_steps": 42420, "loss": 0.3361, "lr": 1.1406030610043306e-06, "epoch": 19.613389910419613, "percentage": 98.07, "elapsed_time": "2:15:11", "remaining_time": "0:02:39", "throughput": 5169.38, "total_tokens": 41930464} {"current_steps": 41605, "total_steps": 42420, "loss": 0.2911, "lr": 1.1267577191708122e-06, "epoch": 19.615747289014617, "percentage": 98.08, "elapsed_time": "2:15:12", "remaining_time": "0:02:38", "throughput": 5169.44, "total_tokens": 41935840} {"current_steps": 41610, "total_steps": 42420, "loss": 0.3621, "lr": 1.1129968281068936e-06, "epoch": 19.618104667609618, "percentage": 98.09, "elapsed_time": "2:15:13", "remaining_time": "0:02:37", "throughput": 5169.48, "total_tokens": 41940736} {"current_steps": 41615, "total_steps": 42420, "loss": 0.2948, "lr": 1.0993203901419336e-06, "epoch": 19.620462046204622, "percentage": 98.1, "elapsed_time": "2:15:13", "remaining_time": "0:02:36", "throughput": 5169.48, "total_tokens": 41944864} {"current_steps": 41620, "total_steps": 42420, "loss": 0.3417, "lr": 1.0857284075911912e-06, "epoch": 19.622819424799623, "percentage": 98.11, "elapsed_time": "2:15:14", "remaining_time": "0:02:35", "throughput": 5169.55, "total_tokens": 41950688} {"current_steps": 41625, "total_steps": 42420, "loss": 0.3859, "lr": 1.072220882755548e-06, "epoch": 19.625176803394623, "percentage": 98.13, "elapsed_time": "2:15:15", "remaining_time": "0:02:35", "throughput": 5169.6, "total_tokens": 41955872} {"current_steps": 41630, "total_steps": 42420, "loss": 0.3383, "lr": 1.0587978179216197e-06, "epoch": 19.627534181989628, "percentage": 98.14, "elapsed_time": "2:15:16", "remaining_time": "0:02:34", "throughput": 5169.61, "total_tokens": 41960288} {"current_steps": 41635, "total_steps": 42420, "loss": 0.3703, "lr": 1.0454592153616993e-06, "epoch": 19.62989156058463, "percentage": 98.15, "elapsed_time": "2:15:17", "remaining_time": "0:02:33", "throughput": 5169.63, "total_tokens": 41965216} {"current_steps": 41640, "total_steps": 42420, "loss": 0.3044, "lr": 1.0322050773337589e-06, "epoch": 19.632248939179632, "percentage": 98.16, "elapsed_time": "2:15:18", "remaining_time": "0:02:32", "throughput": 5169.67, "total_tokens": 41970240} {"current_steps": 41645, "total_steps": 42420, "loss": 0.3124, "lr": 1.0190354060814477e-06, "epoch": 19.634606317774633, "percentage": 98.17, "elapsed_time": "2:15:19", "remaining_time": "0:02:31", "throughput": 5169.74, "total_tokens": 41975936} {"current_steps": 41650, "total_steps": 42420, "loss": 0.3376, "lr": 1.0059502038342604e-06, "epoch": 19.636963696369637, "percentage": 98.18, "elapsed_time": "2:15:20", "remaining_time": "0:02:30", "throughput": 5169.79, "total_tokens": 41980992} {"current_steps": 41655, "total_steps": 42420, "loss": 0.3512, "lr": 9.92949472807314e-07, "epoch": 19.639321074964638, "percentage": 98.2, "elapsed_time": "2:15:21", "remaining_time": "0:02:29", "throughput": 5169.8, "total_tokens": 41985600} {"current_steps": 41660, "total_steps": 42420, "loss": 0.3772, "lr": 9.800332152012925e-07, "epoch": 19.641678453559642, "percentage": 98.21, "elapsed_time": "2:15:22", "remaining_time": "0:02:28", "throughput": 5169.82, "total_tokens": 41990432} {"current_steps": 41665, "total_steps": 42420, "loss": 0.3575, "lr": 9.672014332028356e-07, "epoch": 19.644035832154643, "percentage": 98.22, "elapsed_time": "2:15:23", "remaining_time": "0:02:27", "throughput": 5169.8, "total_tokens": 41994400} {"current_steps": 41670, "total_steps": 42420, "loss": 0.3167, "lr": 9.544541289840392e-07, "epoch": 19.646393210749647, "percentage": 98.23, "elapsed_time": "2:15:24", "remaining_time": "0:02:26", "throughput": 5169.88, "total_tokens": 42000672} {"current_steps": 41675, "total_steps": 42420, "loss": 0.3581, "lr": 9.417913047028992e-07, "epoch": 19.648750589344647, "percentage": 98.24, "elapsed_time": "2:15:25", "remaining_time": "0:02:25", "throughput": 5169.91, "total_tokens": 42005696} {"current_steps": 41680, "total_steps": 42420, "loss": 0.3116, "lr": 9.292129625029233e-07, "epoch": 19.65110796793965, "percentage": 98.26, "elapsed_time": "2:15:25", "remaining_time": "0:02:24", "throughput": 5169.95, "total_tokens": 42010848} {"current_steps": 41685, "total_steps": 42420, "loss": 0.2809, "lr": 9.167191045134083e-07, "epoch": 19.653465346534652, "percentage": 98.27, "elapsed_time": "2:15:26", "remaining_time": "0:02:23", "throughput": 5170.02, "total_tokens": 42016544} {"current_steps": 41690, "total_steps": 42420, "loss": 0.3116, "lr": 9.043097328494398e-07, "epoch": 19.655822725129656, "percentage": 98.28, "elapsed_time": "2:15:28", "remaining_time": "0:02:22", "throughput": 5170.09, "total_tokens": 42022656} {"current_steps": 41695, "total_steps": 42420, "loss": 0.3417, "lr": 8.919848496117266e-07, "epoch": 19.658180103724657, "percentage": 98.29, "elapsed_time": "2:15:29", "remaining_time": "0:02:21", "throughput": 5170.18, "total_tokens": 42028896} {"current_steps": 41700, "total_steps": 42420, "loss": 0.3042, "lr": 8.797444568865442e-07, "epoch": 19.66053748231966, "percentage": 98.3, "elapsed_time": "2:15:29", "remaining_time": "0:02:20", "throughput": 5170.2, "total_tokens": 42033536} {"current_steps": 41705, "total_steps": 42420, "loss": 0.3598, "lr": 8.675885567460129e-07, "epoch": 19.662894860914662, "percentage": 98.31, "elapsed_time": "2:15:30", "remaining_time": "0:02:19", "throughput": 5170.24, "total_tokens": 42038496} {"current_steps": 41710, "total_steps": 42420, "loss": 0.3557, "lr": 8.55517151248042e-07, "epoch": 19.665252239509666, "percentage": 98.33, "elapsed_time": "2:15:31", "remaining_time": "0:02:18", "throughput": 5170.26, "total_tokens": 42043712} {"current_steps": 41715, "total_steps": 42420, "loss": 0.389, "lr": 8.435302424359969e-07, "epoch": 19.667609618104667, "percentage": 98.34, "elapsed_time": "2:15:32", "remaining_time": "0:02:17", "throughput": 5170.32, "total_tokens": 42049760} {"current_steps": 41720, "total_steps": 42420, "loss": 0.3468, "lr": 8.316278323390325e-07, "epoch": 19.66996699669967, "percentage": 98.35, "elapsed_time": "2:15:33", "remaining_time": "0:02:16", "throughput": 5170.3, "total_tokens": 42053792} {"current_steps": 41725, "total_steps": 42420, "loss": 0.3713, "lr": 8.198099229721478e-07, "epoch": 19.67232437529467, "percentage": 98.36, "elapsed_time": "2:15:34", "remaining_time": "0:02:15", "throughput": 5170.41, "total_tokens": 42061024} {"current_steps": 41730, "total_steps": 42420, "loss": 0.3264, "lr": 8.080765163357984e-07, "epoch": 19.674681753889676, "percentage": 98.37, "elapsed_time": "2:15:35", "remaining_time": "0:02:14", "throughput": 5170.44, "total_tokens": 42066080} {"current_steps": 41735, "total_steps": 42420, "loss": 0.3347, "lr": 7.964276144163396e-07, "epoch": 19.677039132484676, "percentage": 98.39, "elapsed_time": "2:15:36", "remaining_time": "0:02:13", "throughput": 5170.5, "total_tokens": 42072224} {"current_steps": 41740, "total_steps": 42420, "loss": 0.3917, "lr": 7.848632191856387e-07, "epoch": 19.67939651107968, "percentage": 98.4, "elapsed_time": "2:15:37", "remaining_time": "0:02:12", "throughput": 5170.51, "total_tokens": 42076768} {"current_steps": 41745, "total_steps": 42420, "loss": 0.3258, "lr": 7.733833326014627e-07, "epoch": 19.68175388967468, "percentage": 98.41, "elapsed_time": "2:15:38", "remaining_time": "0:02:11", "throughput": 5170.57, "total_tokens": 42082560} {"current_steps": 41750, "total_steps": 42420, "loss": 0.3203, "lr": 7.619879566070909e-07, "epoch": 19.684111268269685, "percentage": 98.42, "elapsed_time": "2:15:39", "remaining_time": "0:02:10", "throughput": 5170.63, "total_tokens": 42087968} {"current_steps": 41755, "total_steps": 42420, "loss": 0.395, "lr": 7.506770931315354e-07, "epoch": 19.686468646864686, "percentage": 98.43, "elapsed_time": "2:15:40", "remaining_time": "0:02:09", "throughput": 5170.67, "total_tokens": 42093152} {"current_steps": 41760, "total_steps": 42420, "loss": 0.2969, "lr": 7.394507440895981e-07, "epoch": 19.68882602545969, "percentage": 98.44, "elapsed_time": "2:15:41", "remaining_time": "0:02:08", "throughput": 5170.67, "total_tokens": 42097536} {"current_steps": 41765, "total_steps": 42420, "loss": 0.3445, "lr": 7.283089113817032e-07, "epoch": 19.69118340405469, "percentage": 98.46, "elapsed_time": "2:15:42", "remaining_time": "0:02:07", "throughput": 5170.64, "total_tokens": 42101280} {"current_steps": 41770, "total_steps": 42420, "loss": 0.2924, "lr": 7.17251596893953e-07, "epoch": 19.693540782649695, "percentage": 98.47, "elapsed_time": "2:15:43", "remaining_time": "0:02:06", "throughput": 5170.68, "total_tokens": 42106528} {"current_steps": 41775, "total_steps": 42420, "loss": 0.3495, "lr": 7.062788024981837e-07, "epoch": 19.695898161244696, "percentage": 98.48, "elapsed_time": "2:15:44", "remaining_time": "0:02:05", "throughput": 5170.67, "total_tokens": 42110560} {"current_steps": 41780, "total_steps": 42420, "loss": 0.3382, "lr": 6.953905300518537e-07, "epoch": 19.6982555398397, "percentage": 98.49, "elapsed_time": "2:15:45", "remaining_time": "0:02:04", "throughput": 5170.69, "total_tokens": 42115424} {"current_steps": 41785, "total_steps": 42420, "loss": 0.3355, "lr": 6.845867813982109e-07, "epoch": 19.7006129184347, "percentage": 98.5, "elapsed_time": "2:15:46", "remaining_time": "0:02:03", "throughput": 5170.73, "total_tokens": 42120960} {"current_steps": 41790, "total_steps": 42420, "loss": 0.3161, "lr": 6.738675583661258e-07, "epoch": 19.702970297029704, "percentage": 98.51, "elapsed_time": "2:15:47", "remaining_time": "0:02:02", "throughput": 5170.78, "total_tokens": 42126400} {"current_steps": 41795, "total_steps": 42420, "loss": 0.3164, "lr": 6.632328627702022e-07, "epoch": 19.705327675624705, "percentage": 98.53, "elapsed_time": "2:15:47", "remaining_time": "0:02:01", "throughput": 5170.81, "total_tokens": 42131456} {"current_steps": 41800, "total_steps": 42420, "loss": 0.408, "lr": 6.526826964107224e-07, "epoch": 19.70768505421971, "percentage": 98.54, "elapsed_time": "2:15:48", "remaining_time": "0:02:00", "throughput": 5170.83, "total_tokens": 42135968} {"current_steps": 41805, "total_steps": 42420, "loss": 0.3712, "lr": 6.422170610735911e-07, "epoch": 19.71004243281471, "percentage": 98.55, "elapsed_time": "2:15:49", "remaining_time": "0:01:59", "throughput": 5170.83, "total_tokens": 42140384} {"current_steps": 41810, "total_steps": 42420, "loss": 0.3259, "lr": 6.318359585305023e-07, "epoch": 19.712399811409714, "percentage": 98.56, "elapsed_time": "2:15:50", "remaining_time": "0:01:58", "throughput": 5170.82, "total_tokens": 42144512} {"current_steps": 41815, "total_steps": 42420, "loss": 0.354, "lr": 6.215393905388278e-07, "epoch": 19.714757190004715, "percentage": 98.57, "elapsed_time": "2:15:51", "remaining_time": "0:01:57", "throughput": 5170.81, "total_tokens": 42148608} {"current_steps": 41820, "total_steps": 42420, "loss": 0.3307, "lr": 6.113273588415624e-07, "epoch": 19.71711456859972, "percentage": 98.59, "elapsed_time": "2:15:52", "remaining_time": "0:01:56", "throughput": 5170.85, "total_tokens": 42154048} {"current_steps": 41825, "total_steps": 42420, "loss": 0.2933, "lr": 6.011998651673789e-07, "epoch": 19.71947194719472, "percentage": 98.6, "elapsed_time": "2:15:53", "remaining_time": "0:01:55", "throughput": 5170.9, "total_tokens": 42159648} {"current_steps": 41830, "total_steps": 42420, "loss": 0.2875, "lr": 5.911569112307946e-07, "epoch": 19.72182932578972, "percentage": 98.61, "elapsed_time": "2:15:54", "remaining_time": "0:01:55", "throughput": 5170.96, "total_tokens": 42165472} {"current_steps": 41835, "total_steps": 42420, "loss": 0.3261, "lr": 5.811984987318942e-07, "epoch": 19.724186704384724, "percentage": 98.62, "elapsed_time": "2:15:55", "remaining_time": "0:01:54", "throughput": 5171.03, "total_tokens": 42171520} {"current_steps": 41840, "total_steps": 42420, "loss": 0.3002, "lr": 5.713246293563845e-07, "epoch": 19.726544082979725, "percentage": 98.63, "elapsed_time": "2:15:56", "remaining_time": "0:01:53", "throughput": 5171.1, "total_tokens": 42177344} {"current_steps": 41845, "total_steps": 42420, "loss": 0.2987, "lr": 5.615353047758176e-07, "epoch": 19.72890146157473, "percentage": 98.64, "elapsed_time": "2:15:57", "remaining_time": "0:01:52", "throughput": 5171.14, "total_tokens": 42182560} {"current_steps": 41850, "total_steps": 42420, "loss": 0.3168, "lr": 5.518305266473677e-07, "epoch": 19.73125884016973, "percentage": 98.66, "elapsed_time": "2:15:58", "remaining_time": "0:01:51", "throughput": 5171.16, "total_tokens": 42187264} {"current_steps": 41855, "total_steps": 42420, "loss": 0.3289, "lr": 5.42210296613832e-07, "epoch": 19.733616218764734, "percentage": 98.67, "elapsed_time": "2:15:59", "remaining_time": "0:01:50", "throughput": 5171.21, "total_tokens": 42192448} {"current_steps": 41860, "total_steps": 42420, "loss": 0.3307, "lr": 5.326746163037966e-07, "epoch": 19.735973597359735, "percentage": 98.68, "elapsed_time": "2:15:59", "remaining_time": "0:01:49", "throughput": 5171.22, "total_tokens": 42196896} {"current_steps": 41865, "total_steps": 42420, "loss": 0.2864, "lr": 5.2322348733147e-07, "epoch": 19.73833097595474, "percentage": 98.69, "elapsed_time": "2:16:00", "remaining_time": "0:01:48", "throughput": 5171.23, "total_tokens": 42201344} {"current_steps": 41870, "total_steps": 42420, "loss": 0.3044, "lr": 5.13856911296795e-07, "epoch": 19.74068835454974, "percentage": 98.7, "elapsed_time": "2:16:01", "remaining_time": "0:01:47", "throughput": 5171.25, "total_tokens": 42206048} {"current_steps": 41875, "total_steps": 42420, "loss": 0.342, "lr": 5.045748897853364e-07, "epoch": 19.743045733144744, "percentage": 98.72, "elapsed_time": "2:16:02", "remaining_time": "0:01:46", "throughput": 5171.28, "total_tokens": 42210720} {"current_steps": 41880, "total_steps": 42420, "loss": 0.3435, "lr": 4.953774243683928e-07, "epoch": 19.745403111739744, "percentage": 98.73, "elapsed_time": "2:16:03", "remaining_time": "0:01:45", "throughput": 5171.3, "total_tokens": 42215616} {"current_steps": 41885, "total_steps": 42420, "loss": 0.3219, "lr": 4.862645166029412e-07, "epoch": 19.74776049033475, "percentage": 98.74, "elapsed_time": "2:16:04", "remaining_time": "0:01:44", "throughput": 5171.36, "total_tokens": 42221120} {"current_steps": 41890, "total_steps": 42420, "loss": 0.347, "lr": 4.772361680316361e-07, "epoch": 19.75011786892975, "percentage": 98.75, "elapsed_time": "2:16:05", "remaining_time": "0:01:43", "throughput": 5171.37, "total_tokens": 42225728} {"current_steps": 41895, "total_steps": 42420, "loss": 0.3056, "lr": 4.6829238018286644e-07, "epoch": 19.752475247524753, "percentage": 98.76, "elapsed_time": "2:16:06", "remaining_time": "0:01:42", "throughput": 5171.45, "total_tokens": 42232032} {"current_steps": 41900, "total_steps": 42420, "loss": 0.3101, "lr": 4.5943315457064317e-07, "epoch": 19.754832626119754, "percentage": 98.77, "elapsed_time": "2:16:07", "remaining_time": "0:01:41", "throughput": 5171.45, "total_tokens": 42236320} {"current_steps": 41905, "total_steps": 42420, "loss": 0.3168, "lr": 4.5065849269465555e-07, "epoch": 19.757190004714758, "percentage": 98.79, "elapsed_time": "2:16:08", "remaining_time": "0:01:40", "throughput": 5171.47, "total_tokens": 42241152} {"current_steps": 41910, "total_steps": 42420, "loss": 0.3513, "lr": 4.4196839604027095e-07, "epoch": 19.75954738330976, "percentage": 98.8, "elapsed_time": "2:16:09", "remaining_time": "0:01:39", "throughput": 5171.51, "total_tokens": 42246368} {"current_steps": 41915, "total_steps": 42420, "loss": 0.3116, "lr": 4.3336286607870146e-07, "epoch": 19.761904761904763, "percentage": 98.81, "elapsed_time": "2:16:09", "remaining_time": "0:01:38", "throughput": 5171.51, "total_tokens": 42250656} {"current_steps": 41920, "total_steps": 42420, "loss": 0.2857, "lr": 4.248419042665597e-07, "epoch": 19.764262140499763, "percentage": 98.82, "elapsed_time": "2:16:10", "remaining_time": "0:01:37", "throughput": 5171.56, "total_tokens": 42256192} {"current_steps": 41925, "total_steps": 42420, "loss": 0.2916, "lr": 4.1640551204646937e-07, "epoch": 19.766619519094768, "percentage": 98.83, "elapsed_time": "2:16:11", "remaining_time": "0:01:36", "throughput": 5171.59, "total_tokens": 42261376} {"current_steps": 41930, "total_steps": 42420, "loss": 0.3229, "lr": 4.080536908463994e-07, "epoch": 19.768976897689768, "percentage": 98.84, "elapsed_time": "2:16:12", "remaining_time": "0:01:35", "throughput": 5171.59, "total_tokens": 42265632} {"current_steps": 41935, "total_steps": 42420, "loss": 0.3201, "lr": 3.997864420802744e-07, "epoch": 19.771334276284772, "percentage": 98.86, "elapsed_time": "2:16:13", "remaining_time": "0:01:34", "throughput": 5171.62, "total_tokens": 42270784} {"current_steps": 41940, "total_steps": 42420, "loss": 0.3021, "lr": 3.9160376714758585e-07, "epoch": 19.773691654879773, "percentage": 98.87, "elapsed_time": "2:16:14", "remaining_time": "0:01:33", "throughput": 5171.64, "total_tokens": 42275424} {"current_steps": 41945, "total_steps": 42420, "loss": 0.3293, "lr": 3.8350566743350355e-07, "epoch": 19.776049033474777, "percentage": 98.88, "elapsed_time": "2:16:15", "remaining_time": "0:01:32", "throughput": 5171.65, "total_tokens": 42280224} {"current_steps": 41950, "total_steps": 42420, "loss": 0.3256, "lr": 3.7549214430893095e-07, "epoch": 19.778406412069778, "percentage": 98.89, "elapsed_time": "2:16:16", "remaining_time": "0:01:31", "throughput": 5171.64, "total_tokens": 42284288} {"current_steps": 41955, "total_steps": 42420, "loss": 0.3684, "lr": 3.6756319913039405e-07, "epoch": 19.780763790664782, "percentage": 98.9, "elapsed_time": "2:16:17", "remaining_time": "0:01:30", "throughput": 5171.64, "total_tokens": 42288608} {"current_steps": 41960, "total_steps": 42420, "loss": 0.2925, "lr": 3.597188332400969e-07, "epoch": 19.783121169259783, "percentage": 98.92, "elapsed_time": "2:16:18", "remaining_time": "0:01:29", "throughput": 5171.73, "total_tokens": 42295136} {"current_steps": 41965, "total_steps": 42420, "loss": 0.3156, "lr": 3.5195904796603284e-07, "epoch": 19.785478547854787, "percentage": 98.93, "elapsed_time": "2:16:19", "remaining_time": "0:01:28", "throughput": 5171.79, "total_tokens": 42300576} {"current_steps": 41970, "total_steps": 42420, "loss": 0.3717, "lr": 3.4428384462176224e-07, "epoch": 19.787835926449787, "percentage": 98.94, "elapsed_time": "2:16:19", "remaining_time": "0:01:27", "throughput": 5171.78, "total_tokens": 42304864} {"current_steps": 41975, "total_steps": 42420, "loss": 0.3541, "lr": 3.366932245065235e-07, "epoch": 19.79019330504479, "percentage": 98.95, "elapsed_time": "2:16:20", "remaining_time": "0:01:26", "throughput": 5171.8, "total_tokens": 42309696} {"current_steps": 41980, "total_steps": 42420, "loss": 0.3368, "lr": 3.291871889053444e-07, "epoch": 19.792550683639792, "percentage": 98.96, "elapsed_time": "2:16:21", "remaining_time": "0:01:25", "throughput": 5171.85, "total_tokens": 42314976} {"current_steps": 41985, "total_steps": 42420, "loss": 0.2956, "lr": 3.217657390888751e-07, "epoch": 19.794908062234796, "percentage": 98.97, "elapsed_time": "2:16:22", "remaining_time": "0:01:24", "throughput": 5171.9, "total_tokens": 42320416} {"current_steps": 41990, "total_steps": 42420, "loss": 0.3386, "lr": 3.1442887631333294e-07, "epoch": 19.797265440829797, "percentage": 98.99, "elapsed_time": "2:16:23", "remaining_time": "0:01:23", "throughput": 5171.97, "total_tokens": 42326368} {"current_steps": 41995, "total_steps": 42420, "loss": 0.3184, "lr": 3.07176601820891e-07, "epoch": 19.7996228194248, "percentage": 99.0, "elapsed_time": "2:16:24", "remaining_time": "0:01:22", "throughput": 5171.96, "total_tokens": 42330592} {"current_steps": 42000, "total_steps": 42420, "loss": 0.3341, "lr": 3.000089168391229e-07, "epoch": 19.801980198019802, "percentage": 99.01, "elapsed_time": "2:16:25", "remaining_time": "0:01:21", "throughput": 5172.01, "total_tokens": 42336192} {"current_steps": 42005, "total_steps": 42420, "loss": 0.3258, "lr": 2.9292582258139135e-07, "epoch": 19.804337576614806, "percentage": 99.02, "elapsed_time": "2:16:26", "remaining_time": "0:01:20", "throughput": 5172.03, "total_tokens": 42340960} {"current_steps": 42010, "total_steps": 42420, "loss": 0.3532, "lr": 2.8592732024679267e-07, "epoch": 19.806694955209807, "percentage": 99.03, "elapsed_time": "2:16:27", "remaining_time": "0:01:19", "throughput": 5172.13, "total_tokens": 42347200} {"current_steps": 42015, "total_steps": 42420, "loss": 0.2724, "lr": 2.79013411020046e-07, "epoch": 19.809052333804807, "percentage": 99.05, "elapsed_time": "2:16:28", "remaining_time": "0:01:18", "throughput": 5172.14, "total_tokens": 42351872} {"current_steps": 42020, "total_steps": 42420, "loss": 0.3239, "lr": 2.7218409607154827e-07, "epoch": 19.81140971239981, "percentage": 99.06, "elapsed_time": "2:16:29", "remaining_time": "0:01:17", "throughput": 5172.15, "total_tokens": 42356384} {"current_steps": 42025, "total_steps": 42420, "loss": 0.2649, "lr": 2.654393765573748e-07, "epoch": 19.813767090994816, "percentage": 99.07, "elapsed_time": "2:16:30", "remaining_time": "0:01:16", "throughput": 5172.2, "total_tokens": 42362048} {"current_steps": 42030, "total_steps": 42420, "loss": 0.3316, "lr": 2.5877925361933454e-07, "epoch": 19.816124469589816, "percentage": 99.08, "elapsed_time": "2:16:31", "remaining_time": "0:01:16", "throughput": 5172.24, "total_tokens": 42367328} {"current_steps": 42035, "total_steps": 42420, "loss": 0.3112, "lr": 2.5220372838480333e-07, "epoch": 19.818481848184817, "percentage": 99.09, "elapsed_time": "2:16:32", "remaining_time": "0:01:15", "throughput": 5172.26, "total_tokens": 42372288} {"current_steps": 42040, "total_steps": 42420, "loss": 0.339, "lr": 2.4571280196694633e-07, "epoch": 19.82083922677982, "percentage": 99.1, "elapsed_time": "2:16:33", "remaining_time": "0:01:14", "throughput": 5172.33, "total_tokens": 42378528} {"current_steps": 42045, "total_steps": 42420, "loss": 0.3372, "lr": 2.393064754646068e-07, "epoch": 19.82319660537482, "percentage": 99.12, "elapsed_time": "2:16:34", "remaining_time": "0:01:13", "throughput": 5172.38, "total_tokens": 42384160} {"current_steps": 42050, "total_steps": 42420, "loss": 0.3294, "lr": 2.329847499621951e-07, "epoch": 19.825553983969826, "percentage": 99.13, "elapsed_time": "2:16:35", "remaining_time": "0:01:12", "throughput": 5172.38, "total_tokens": 42388544} {"current_steps": 42055, "total_steps": 42420, "loss": 0.3636, "lr": 2.2674762652991065e-07, "epoch": 19.827911362564826, "percentage": 99.14, "elapsed_time": "2:16:36", "remaining_time": "0:01:11", "throughput": 5172.41, "total_tokens": 42393632} {"current_steps": 42060, "total_steps": 42420, "loss": 0.3247, "lr": 2.205951062236311e-07, "epoch": 19.83026874115983, "percentage": 99.15, "elapsed_time": "2:16:37", "remaining_time": "0:01:10", "throughput": 5172.44, "total_tokens": 42398816} {"current_steps": 42065, "total_steps": 42420, "loss": 0.3127, "lr": 2.1452719008480115e-07, "epoch": 19.83262611975483, "percentage": 99.16, "elapsed_time": "2:16:37", "remaining_time": "0:01:09", "throughput": 5172.48, "total_tokens": 42403744} {"current_steps": 42070, "total_steps": 42420, "loss": 0.3894, "lr": 2.0854387914059913e-07, "epoch": 19.834983498349835, "percentage": 99.17, "elapsed_time": "2:16:38", "remaining_time": "0:01:08", "throughput": 5172.51, "total_tokens": 42408896} {"current_steps": 42075, "total_steps": 42420, "loss": 0.3087, "lr": 2.0264517440399254e-07, "epoch": 19.837340876944836, "percentage": 99.19, "elapsed_time": "2:16:39", "remaining_time": "0:01:07", "throughput": 5172.49, "total_tokens": 42412992} {"current_steps": 42080, "total_steps": 42420, "loss": 0.373, "lr": 1.9683107687346046e-07, "epoch": 19.83969825553984, "percentage": 99.2, "elapsed_time": "2:16:40", "remaining_time": "0:01:06", "throughput": 5172.52, "total_tokens": 42418112} {"current_steps": 42085, "total_steps": 42420, "loss": 0.3769, "lr": 1.911015875332711e-07, "epoch": 19.84205563413484, "percentage": 99.21, "elapsed_time": "2:16:41", "remaining_time": "0:01:05", "throughput": 5172.54, "total_tokens": 42423232} {"current_steps": 42090, "total_steps": 42420, "loss": 0.3009, "lr": 1.8545670735325982e-07, "epoch": 19.844413012729845, "percentage": 99.22, "elapsed_time": "2:16:42", "remaining_time": "0:01:04", "throughput": 5172.56, "total_tokens": 42428352} {"current_steps": 42095, "total_steps": 42420, "loss": 0.321, "lr": 1.7989643728910654e-07, "epoch": 19.846770391324846, "percentage": 99.23, "elapsed_time": "2:16:43", "remaining_time": "0:01:03", "throughput": 5172.55, "total_tokens": 42432448} {"current_steps": 42100, "total_steps": 42420, "loss": 0.3173, "lr": 1.7442077828200288e-07, "epoch": 19.84912776991985, "percentage": 99.25, "elapsed_time": "2:16:44", "remaining_time": "0:01:02", "throughput": 5172.55, "total_tokens": 42436544} {"current_steps": 42105, "total_steps": 42420, "loss": 0.3045, "lr": 1.690297312589295e-07, "epoch": 19.85148514851485, "percentage": 99.26, "elapsed_time": "2:16:45", "remaining_time": "0:01:01", "throughput": 5172.6, "total_tokens": 42442048} {"current_steps": 42110, "total_steps": 42420, "loss": 0.3535, "lr": 1.6372329713248978e-07, "epoch": 19.853842527109855, "percentage": 99.27, "elapsed_time": "2:16:46", "remaining_time": "0:01:00", "throughput": 5172.63, "total_tokens": 42446784} {"current_steps": 42115, "total_steps": 42420, "loss": 0.37, "lr": 1.5850147680090965e-07, "epoch": 19.856199905704855, "percentage": 99.28, "elapsed_time": "2:16:46", "remaining_time": "0:00:59", "throughput": 5172.67, "total_tokens": 42452064} {"current_steps": 42120, "total_steps": 42420, "loss": 0.3449, "lr": 1.5336427114820416e-07, "epoch": 19.85855728429986, "percentage": 99.29, "elapsed_time": "2:16:47", "remaining_time": "0:00:58", "throughput": 5172.66, "total_tokens": 42456064} {"current_steps": 42125, "total_steps": 42420, "loss": 0.3635, "lr": 1.4831168104406656e-07, "epoch": 19.86091466289486, "percentage": 99.3, "elapsed_time": "2:16:48", "remaining_time": "0:00:57", "throughput": 5172.75, "total_tokens": 42462656} {"current_steps": 42130, "total_steps": 42420, "loss": 0.3319, "lr": 1.433437073437571e-07, "epoch": 19.863272041489864, "percentage": 99.32, "elapsed_time": "2:16:49", "remaining_time": "0:00:56", "throughput": 5172.81, "total_tokens": 42468736} {"current_steps": 42135, "total_steps": 42420, "loss": 0.3141, "lr": 1.384603508882143e-07, "epoch": 19.865629420084865, "percentage": 99.33, "elapsed_time": "2:16:50", "remaining_time": "0:00:55", "throughput": 5172.81, "total_tokens": 42473152} {"current_steps": 42140, "total_steps": 42420, "loss": 0.3541, "lr": 1.3366161250422116e-07, "epoch": 19.86798679867987, "percentage": 99.34, "elapsed_time": "2:16:51", "remaining_time": "0:00:54", "throughput": 5172.85, "total_tokens": 42478176} {"current_steps": 42145, "total_steps": 42420, "loss": 0.3293, "lr": 1.289474930040724e-07, "epoch": 19.87034417727487, "percentage": 99.35, "elapsed_time": "2:16:52", "remaining_time": "0:00:53", "throughput": 5172.87, "total_tokens": 42482944} {"current_steps": 42150, "total_steps": 42420, "loss": 0.3649, "lr": 1.2431799318574076e-07, "epoch": 19.872701555869874, "percentage": 99.36, "elapsed_time": "2:16:53", "remaining_time": "0:00:52", "throughput": 5172.89, "total_tokens": 42487616} {"current_steps": 42155, "total_steps": 42420, "loss": 0.3103, "lr": 1.1977311383298828e-07, "epoch": 19.875058934464874, "percentage": 99.38, "elapsed_time": "2:16:54", "remaining_time": "0:00:51", "throughput": 5172.94, "total_tokens": 42493056} {"current_steps": 42160, "total_steps": 42420, "loss": 0.3187, "lr": 1.1531285571519945e-07, "epoch": 19.87741631305988, "percentage": 99.39, "elapsed_time": "2:16:55", "remaining_time": "0:00:50", "throughput": 5172.97, "total_tokens": 42498112} {"current_steps": 42165, "total_steps": 42420, "loss": 0.3406, "lr": 1.1093721958732595e-07, "epoch": 19.87977369165488, "percentage": 99.4, "elapsed_time": "2:16:56", "remaining_time": "0:00:49", "throughput": 5173.0, "total_tokens": 42502976} {"current_steps": 42170, "total_steps": 42420, "loss": 0.3387, "lr": 1.0664620619021959e-07, "epoch": 19.882131070249883, "percentage": 99.41, "elapsed_time": "2:16:57", "remaining_time": "0:00:48", "throughput": 5173.03, "total_tokens": 42508096} {"current_steps": 42175, "total_steps": 42420, "loss": 0.2803, "lr": 1.0243981625013276e-07, "epoch": 19.884488448844884, "percentage": 99.42, "elapsed_time": "2:16:58", "remaining_time": "0:00:47", "throughput": 5173.07, "total_tokens": 42513504} {"current_steps": 42180, "total_steps": 42420, "loss": 0.3649, "lr": 9.831805047921805e-08, "epoch": 19.88684582743989, "percentage": 99.43, "elapsed_time": "2:16:59", "remaining_time": "0:00:46", "throughput": 5173.08, "total_tokens": 42518112} {"current_steps": 42185, "total_steps": 42420, "loss": 0.2847, "lr": 9.428090957525059e-08, "epoch": 19.88920320603489, "percentage": 99.45, "elapsed_time": "2:17:00", "remaining_time": "0:00:45", "throughput": 5173.17, "total_tokens": 42524032} {"current_steps": 42190, "total_steps": 42420, "loss": 0.3084, "lr": 9.032839422157269e-08, "epoch": 19.891560584629893, "percentage": 99.46, "elapsed_time": "2:17:00", "remaining_time": "0:00:44", "throughput": 5173.18, "total_tokens": 42528576} {"current_steps": 42195, "total_steps": 42420, "loss": 0.36, "lr": 8.646050508737125e-08, "epoch": 19.893917963224894, "percentage": 99.47, "elapsed_time": "2:17:01", "remaining_time": "0:00:43", "throughput": 5173.18, "total_tokens": 42532928} {"current_steps": 42200, "total_steps": 42420, "loss": 0.357, "lr": 8.267724282728928e-08, "epoch": 19.896275341819898, "percentage": 99.48, "elapsed_time": "2:17:02", "remaining_time": "0:00:42", "throughput": 5173.26, "total_tokens": 42539104} {"current_steps": 42205, "total_steps": 42420, "loss": 0.3446, "lr": 7.897860808186996e-08, "epoch": 19.8986327204149, "percentage": 99.49, "elapsed_time": "2:17:03", "remaining_time": "0:00:41", "throughput": 5173.28, "total_tokens": 42543904} {"current_steps": 42210, "total_steps": 42420, "loss": 0.3127, "lr": 7.536460147722357e-08, "epoch": 19.900990099009903, "percentage": 99.5, "elapsed_time": "2:17:04", "remaining_time": "0:00:40", "throughput": 5173.36, "total_tokens": 42549888} {"current_steps": 42215, "total_steps": 42420, "loss": 0.3038, "lr": 7.183522362502747e-08, "epoch": 19.903347477604903, "percentage": 99.52, "elapsed_time": "2:17:05", "remaining_time": "0:00:39", "throughput": 5173.37, "total_tokens": 42554848} {"current_steps": 42220, "total_steps": 42420, "loss": 0.3361, "lr": 6.839047512291475e-08, "epoch": 19.905704856199904, "percentage": 99.53, "elapsed_time": "2:17:06", "remaining_time": "0:00:38", "throughput": 5173.46, "total_tokens": 42560896} {"current_steps": 42225, "total_steps": 42420, "loss": 0.3727, "lr": 6.503035655391898e-08, "epoch": 19.908062234794908, "percentage": 99.54, "elapsed_time": "2:17:07", "remaining_time": "0:00:37", "throughput": 5173.52, "total_tokens": 42566592} {"current_steps": 42230, "total_steps": 42420, "loss": 0.3734, "lr": 6.175486848686295e-08, "epoch": 19.91041961338991, "percentage": 99.55, "elapsed_time": "2:17:08", "remaining_time": "0:00:37", "throughput": 5173.59, "total_tokens": 42572384} {"current_steps": 42235, "total_steps": 42420, "loss": 0.3493, "lr": 5.856401147619206e-08, "epoch": 19.912776991984913, "percentage": 99.56, "elapsed_time": "2:17:09", "remaining_time": "0:00:36", "throughput": 5173.61, "total_tokens": 42577248} {"current_steps": 42240, "total_steps": 42420, "loss": 0.3528, "lr": 5.5457786062196316e-08, "epoch": 19.915134370579914, "percentage": 99.58, "elapsed_time": "2:17:10", "remaining_time": "0:00:35", "throughput": 5173.67, "total_tokens": 42583232} {"current_steps": 42245, "total_steps": 42420, "loss": 0.3062, "lr": 5.243619277056633e-08, "epoch": 19.917491749174918, "percentage": 99.59, "elapsed_time": "2:17:11", "remaining_time": "0:00:34", "throughput": 5173.68, "total_tokens": 42588000} {"current_steps": 42250, "total_steps": 42420, "loss": 0.314, "lr": 4.949923211289287e-08, "epoch": 19.91984912776992, "percentage": 99.6, "elapsed_time": "2:17:12", "remaining_time": "0:00:33", "throughput": 5173.69, "total_tokens": 42592416} {"current_steps": 42255, "total_steps": 42420, "loss": 0.3531, "lr": 4.664690458627829e-08, "epoch": 19.922206506364923, "percentage": 99.61, "elapsed_time": "2:17:13", "remaining_time": "0:00:32", "throughput": 5173.73, "total_tokens": 42597504} {"current_steps": 42260, "total_steps": 42420, "loss": 0.288, "lr": 4.387921067366962e-08, "epoch": 19.924563884959923, "percentage": 99.62, "elapsed_time": "2:17:14", "remaining_time": "0:00:31", "throughput": 5173.74, "total_tokens": 42602336} {"current_steps": 42265, "total_steps": 42420, "loss": 0.3291, "lr": 4.119615084352546e-08, "epoch": 19.926921263554927, "percentage": 99.63, "elapsed_time": "2:17:15", "remaining_time": "0:00:30", "throughput": 5173.75, "total_tokens": 42607136} {"current_steps": 42270, "total_steps": 42420, "loss": 0.3889, "lr": 3.8597725550038044e-08, "epoch": 19.929278642149928, "percentage": 99.65, "elapsed_time": "2:17:16", "remaining_time": "0:00:29", "throughput": 5173.8, "total_tokens": 42612512} {"current_steps": 42275, "total_steps": 42420, "loss": 0.3284, "lr": 3.608393523313325e-08, "epoch": 19.931636020744932, "percentage": 99.66, "elapsed_time": "2:17:17", "remaining_time": "0:00:28", "throughput": 5173.88, "total_tokens": 42618784} {"current_steps": 42280, "total_steps": 42420, "loss": 0.3094, "lr": 3.365478031824854e-08, "epoch": 19.933993399339933, "percentage": 99.67, "elapsed_time": "2:17:18", "remaining_time": "0:00:27", "throughput": 5173.9, "total_tokens": 42623648} {"current_steps": 42285, "total_steps": 42420, "loss": 0.3511, "lr": 3.131026121672153e-08, "epoch": 19.936350777934937, "percentage": 99.68, "elapsed_time": "2:17:19", "remaining_time": "0:00:26", "throughput": 5173.92, "total_tokens": 42628352} {"current_steps": 42290, "total_steps": 42420, "loss": 0.3338, "lr": 2.905037832534596e-08, "epoch": 19.938708156529938, "percentage": 99.69, "elapsed_time": "2:17:19", "remaining_time": "0:00:25", "throughput": 5173.93, "total_tokens": 42632960} {"current_steps": 42295, "total_steps": 42420, "loss": 0.3289, "lr": 2.6875132026760173e-08, "epoch": 19.94106553512494, "percentage": 99.71, "elapsed_time": "2:17:20", "remaining_time": "0:00:24", "throughput": 5173.91, "total_tokens": 42637088} {"current_steps": 42300, "total_steps": 42420, "loss": 0.2916, "lr": 2.478452268911413e-08, "epoch": 19.943422913719942, "percentage": 99.72, "elapsed_time": "2:17:21", "remaining_time": "0:00:23", "throughput": 5173.96, "total_tokens": 42642304} {"current_steps": 42305, "total_steps": 42420, "loss": 0.3711, "lr": 2.277855066634693e-08, "epoch": 19.945780292314947, "percentage": 99.73, "elapsed_time": "2:17:22", "remaining_time": "0:00:22", "throughput": 5174.01, "total_tokens": 42648000} {"current_steps": 42310, "total_steps": 42420, "loss": 0.3874, "lr": 2.08572162980758e-08, "epoch": 19.948137670909947, "percentage": 99.74, "elapsed_time": "2:17:23", "remaining_time": "0:00:21", "throughput": 5174.01, "total_tokens": 42652288} {"current_steps": 42315, "total_steps": 42420, "loss": 0.3662, "lr": 1.9020519909485058e-08, "epoch": 19.95049504950495, "percentage": 99.75, "elapsed_time": "2:17:24", "remaining_time": "0:00:20", "throughput": 5174.01, "total_tokens": 42656640} {"current_steps": 42320, "total_steps": 42420, "loss": 0.3525, "lr": 1.7268461811548174e-08, "epoch": 19.952852428099952, "percentage": 99.76, "elapsed_time": "2:17:25", "remaining_time": "0:00:19", "throughput": 5174.01, "total_tokens": 42661120} {"current_steps": 42325, "total_steps": 42420, "loss": 0.3568, "lr": 1.5601042300861234e-08, "epoch": 19.955209806694956, "percentage": 99.78, "elapsed_time": "2:17:26", "remaining_time": "0:00:18", "throughput": 5174.03, "total_tokens": 42666528} {"current_steps": 42330, "total_steps": 42420, "loss": 0.3358, "lr": 1.4018261659587417e-08, "epoch": 19.957567185289957, "percentage": 99.79, "elapsed_time": "2:17:27", "remaining_time": "0:00:17", "throughput": 5174.03, "total_tokens": 42670880} {"current_steps": 42335, "total_steps": 42420, "loss": 0.3628, "lr": 1.2520120155845583e-08, "epoch": 19.95992456388496, "percentage": 99.8, "elapsed_time": "2:17:27", "remaining_time": "0:00:16", "throughput": 5174.03, "total_tokens": 42675232} {"current_steps": 42340, "total_steps": 42420, "loss": 0.3374, "lr": 1.1106618043044137e-08, "epoch": 19.96228194247996, "percentage": 99.81, "elapsed_time": "2:17:28", "remaining_time": "0:00:15", "throughput": 5174.06, "total_tokens": 42680384} {"current_steps": 42345, "total_steps": 42420, "loss": 0.3073, "lr": 9.777755560602674e-09, "epoch": 19.964639321074966, "percentage": 99.82, "elapsed_time": "2:17:29", "remaining_time": "0:00:14", "throughput": 5174.1, "total_tokens": 42685440} {"current_steps": 42350, "total_steps": 42420, "loss": 0.3485, "lr": 8.533532933452382e-09, "epoch": 19.966996699669966, "percentage": 99.83, "elapsed_time": "2:17:30", "remaining_time": "0:00:13", "throughput": 5174.11, "total_tokens": 42690208} {"current_steps": 42355, "total_steps": 42420, "loss": 0.2954, "lr": 7.373950372202564e-09, "epoch": 19.96935407826497, "percentage": 99.85, "elapsed_time": "2:17:31", "remaining_time": "0:00:12", "throughput": 5174.14, "total_tokens": 42695200} {"current_steps": 42360, "total_steps": 42420, "loss": 0.3422, "lr": 6.2990080731406555e-09, "epoch": 19.97171145685997, "percentage": 99.86, "elapsed_time": "2:17:32", "remaining_time": "0:00:11", "throughput": 5174.11, "total_tokens": 42699296} {"current_steps": 42365, "total_steps": 42420, "loss": 0.3384, "lr": 5.308706218232207e-09, "epoch": 19.974068835454975, "percentage": 99.87, "elapsed_time": "2:17:33", "remaining_time": "0:00:10", "throughput": 5174.14, "total_tokens": 42704000} {"current_steps": 42370, "total_steps": 42420, "loss": 0.3307, "lr": 4.403044975176407e-09, "epoch": 19.976426214049976, "percentage": 99.88, "elapsed_time": "2:17:34", "remaining_time": "0:00:09", "throughput": 5174.16, "total_tokens": 42708896} {"current_steps": 42375, "total_steps": 42420, "loss": 0.3298, "lr": 3.5820244971840334e-09, "epoch": 19.97878359264498, "percentage": 99.89, "elapsed_time": "2:17:35", "remaining_time": "0:00:08", "throughput": 5174.16, "total_tokens": 42713440} {"current_steps": 42380, "total_steps": 42420, "loss": 0.3003, "lr": 2.845644923366031e-09, "epoch": 19.98114097123998, "percentage": 99.91, "elapsed_time": "2:17:35", "remaining_time": "0:00:07", "throughput": 5174.15, "total_tokens": 42717600} {"current_steps": 42385, "total_steps": 42420, "loss": 0.3374, "lr": 2.193906378289423e-09, "epoch": 19.983498349834985, "percentage": 99.92, "elapsed_time": "2:17:36", "remaining_time": "0:00:06", "throughput": 5174.17, "total_tokens": 42722432} {"current_steps": 42390, "total_steps": 42420, "loss": 0.3379, "lr": 1.6268089722548673e-09, "epoch": 19.985855728429986, "percentage": 99.93, "elapsed_time": "2:17:37", "remaining_time": "0:00:05", "throughput": 5174.24, "total_tokens": 42728224} {"current_steps": 42395, "total_steps": 42420, "loss": 0.3332, "lr": 1.1443528013521664e-09, "epoch": 19.98821310702499, "percentage": 99.94, "elapsed_time": "2:17:38", "remaining_time": "0:00:04", "throughput": 5174.24, "total_tokens": 42732608} {"current_steps": 42400, "total_steps": 42420, "loss": 0.3678, "lr": 7.465379472382239e-10, "epoch": 19.99057048561999, "percentage": 99.95, "elapsed_time": "2:17:39", "remaining_time": "0:00:03", "throughput": 5174.26, "total_tokens": 42737504} {"current_steps": 42405, "total_steps": 42420, "loss": 0.3804, "lr": 4.333644771925549e-10, "epoch": 19.992927864214995, "percentage": 99.96, "elapsed_time": "2:17:40", "remaining_time": "0:00:02", "throughput": 5174.29, "total_tokens": 42742496} {"current_steps": 42410, "total_steps": 42420, "loss": 0.3997, "lr": 2.0483244428382009e-10, "epoch": 19.995285242809995, "percentage": 99.98, "elapsed_time": "2:17:41", "remaining_time": "0:00:01", "throughput": 5174.32, "total_tokens": 42747712} {"current_steps": 42415, "total_steps": 42420, "loss": 0.3356, "lr": 6.094188714778071e-11, "epoch": 19.997642621405, "percentage": 99.99, "elapsed_time": "2:17:42", "remaining_time": "0:00:00", "throughput": 5174.34, "total_tokens": 42752704} {"current_steps": 42420, "total_steps": 42420, "loss": 0.2996, "lr": 1.692830209343299e-12, "epoch": 20.0, "percentage": 100.0, "elapsed_time": "2:17:43", "remaining_time": "0:00:00", "throughput": 5174.33, "total_tokens": 42758400} {"current_steps": 42420, "total_steps": 42420, "eval_loss": 0.3377535045146942, "epoch": 20.0, "percentage": 100.0, "elapsed_time": "2:18:02", "remaining_time": "0:00:00", "throughput": 5162.43, "total_tokens": 42758400} {"current_steps": 42420, "total_steps": 42420, "epoch": 20.0, "percentage": 100.0, "elapsed_time": "2:18:04", "remaining_time": "0:00:00", "throughput": 5161.12, "total_tokens": 42758400}