cooking_sft_fail_new_mem / trainer_log.jsonl
izzcw's picture
Training in progress, step 3751
48e379b verified
{"current_steps": 10, "total_steps": 3751, "loss": 0.5957, "lr": 2.6595744680851066e-07, "epoch": 0.002665911325124548, "percentage": 0.27, "elapsed_time": "0:02:24", "remaining_time": "14:58:16"}
{"current_steps": 20, "total_steps": 3751, "loss": 0.5163, "lr": 5.319148936170213e-07, "epoch": 0.005331822650249096, "percentage": 0.53, "elapsed_time": "0:04:37", "remaining_time": "14:23:41"}
{"current_steps": 30, "total_steps": 3751, "loss": 0.4334, "lr": 7.97872340425532e-07, "epoch": 0.007997733975373645, "percentage": 0.8, "elapsed_time": "0:07:00", "remaining_time": "14:30:12"}
{"current_steps": 40, "total_steps": 3751, "loss": 0.3824, "lr": 1.0638297872340427e-06, "epoch": 0.010663645300498192, "percentage": 1.07, "elapsed_time": "0:09:17", "remaining_time": "14:21:59"}
{"current_steps": 50, "total_steps": 3751, "loss": 0.3821, "lr": 1.3297872340425533e-06, "epoch": 0.01332955662562274, "percentage": 1.33, "elapsed_time": "0:11:36", "remaining_time": "14:19:22"}
{"current_steps": 50, "total_steps": 3751, "eval_loss": 0.4735161364078522, "epoch": 0.01332955662562274, "percentage": 1.33, "elapsed_time": "0:15:59", "remaining_time": "19:43:59"}
{"current_steps": 60, "total_steps": 3751, "loss": 0.3431, "lr": 1.595744680851064e-06, "epoch": 0.01599546795074729, "percentage": 1.6, "elapsed_time": "0:18:18", "remaining_time": "18:45:49"}
{"current_steps": 70, "total_steps": 3751, "loss": 0.3327, "lr": 1.8617021276595745e-06, "epoch": 0.018661379275871838, "percentage": 1.87, "elapsed_time": "0:20:36", "remaining_time": "18:03:51"}
{"current_steps": 80, "total_steps": 3751, "loss": 0.3079, "lr": 2.1276595744680853e-06, "epoch": 0.021327290600996383, "percentage": 2.13, "elapsed_time": "0:22:54", "remaining_time": "17:31:07"}
{"current_steps": 90, "total_steps": 3751, "loss": 0.306, "lr": 2.393617021276596e-06, "epoch": 0.023993201926120932, "percentage": 2.4, "elapsed_time": "0:25:15", "remaining_time": "17:07:34"}
{"current_steps": 100, "total_steps": 3751, "loss": 0.302, "lr": 2.6595744680851065e-06, "epoch": 0.02665911325124548, "percentage": 2.67, "elapsed_time": "0:27:33", "remaining_time": "16:46:04"}
{"current_steps": 100, "total_steps": 3751, "eval_loss": 0.3177907168865204, "epoch": 0.02665911325124548, "percentage": 2.67, "elapsed_time": "0:31:56", "remaining_time": "19:26:13"}
{"current_steps": 110, "total_steps": 3751, "loss": 0.2978, "lr": 2.9255319148936174e-06, "epoch": 0.02932502457637003, "percentage": 2.93, "elapsed_time": "0:34:14", "remaining_time": "18:53:34"}
{"current_steps": 120, "total_steps": 3751, "loss": 0.3046, "lr": 3.191489361702128e-06, "epoch": 0.03199093590149458, "percentage": 3.2, "elapsed_time": "0:36:33", "remaining_time": "18:26:07"}
{"current_steps": 130, "total_steps": 3751, "loss": 0.3074, "lr": 3.457446808510639e-06, "epoch": 0.034656847226619124, "percentage": 3.47, "elapsed_time": "0:38:49", "remaining_time": "18:01:18"}
{"current_steps": 140, "total_steps": 3751, "loss": 0.3147, "lr": 3.723404255319149e-06, "epoch": 0.037322758551743676, "percentage": 3.73, "elapsed_time": "0:41:07", "remaining_time": "17:40:42"}
{"current_steps": 150, "total_steps": 3751, "loss": 0.2988, "lr": 3.98936170212766e-06, "epoch": 0.03998866987686822, "percentage": 4.0, "elapsed_time": "0:43:25", "remaining_time": "17:22:41"}
{"current_steps": 150, "total_steps": 3751, "eval_loss": 0.32532769441604614, "epoch": 0.03998866987686822, "percentage": 4.0, "elapsed_time": "0:47:48", "remaining_time": "19:07:55"}
{"current_steps": 160, "total_steps": 3751, "loss": 0.3049, "lr": 4.255319148936171e-06, "epoch": 0.04265458120199277, "percentage": 4.27, "elapsed_time": "0:50:09", "remaining_time": "18:45:35"}
{"current_steps": 170, "total_steps": 3751, "loss": 0.3077, "lr": 4.521276595744681e-06, "epoch": 0.04532049252711732, "percentage": 4.53, "elapsed_time": "0:52:26", "remaining_time": "18:24:31"}
{"current_steps": 180, "total_steps": 3751, "loss": 0.2957, "lr": 4.787234042553192e-06, "epoch": 0.047986403852241864, "percentage": 4.8, "elapsed_time": "0:54:46", "remaining_time": "18:06:31"}
{"current_steps": 190, "total_steps": 3751, "loss": 0.2993, "lr": 5.053191489361703e-06, "epoch": 0.05065231517736641, "percentage": 5.07, "elapsed_time": "0:57:06", "remaining_time": "17:50:12"}
{"current_steps": 200, "total_steps": 3751, "loss": 0.3054, "lr": 5.319148936170213e-06, "epoch": 0.05331822650249096, "percentage": 5.33, "elapsed_time": "0:59:28", "remaining_time": "17:36:07"}
{"current_steps": 200, "total_steps": 3751, "eval_loss": 0.3250272572040558, "epoch": 0.05331822650249096, "percentage": 5.33, "elapsed_time": "1:03:52", "remaining_time": "18:53:58"}
{"current_steps": 210, "total_steps": 3751, "loss": 0.3117, "lr": 5.5851063829787235e-06, "epoch": 0.05598413782761551, "percentage": 5.6, "elapsed_time": "1:06:05", "remaining_time": "18:34:26"}
{"current_steps": 220, "total_steps": 3751, "loss": 0.3102, "lr": 5.851063829787235e-06, "epoch": 0.05865004915274006, "percentage": 5.87, "elapsed_time": "1:08:24", "remaining_time": "18:17:54"}
{"current_steps": 230, "total_steps": 3751, "loss": 0.3096, "lr": 6.117021276595745e-06, "epoch": 0.061315960477864605, "percentage": 6.13, "elapsed_time": "1:10:43", "remaining_time": "18:02:42"}
{"current_steps": 240, "total_steps": 3751, "loss": 0.3046, "lr": 6.382978723404256e-06, "epoch": 0.06398187180298916, "percentage": 6.4, "elapsed_time": "1:13:00", "remaining_time": "17:47:59"}
{"current_steps": 250, "total_steps": 3751, "loss": 0.2967, "lr": 6.648936170212767e-06, "epoch": 0.0666477831281137, "percentage": 6.66, "elapsed_time": "1:15:20", "remaining_time": "17:35:08"}
{"current_steps": 250, "total_steps": 3751, "eval_loss": 0.3232134282588959, "epoch": 0.0666477831281137, "percentage": 6.66, "elapsed_time": "1:19:43", "remaining_time": "18:36:31"}
{"current_steps": 260, "total_steps": 3751, "loss": 0.3112, "lr": 6.914893617021278e-06, "epoch": 0.06931369445323825, "percentage": 6.93, "elapsed_time": "1:22:01", "remaining_time": "18:21:19"}
{"current_steps": 270, "total_steps": 3751, "loss": 0.314, "lr": 7.1808510638297875e-06, "epoch": 0.07197960577836279, "percentage": 7.2, "elapsed_time": "1:24:18", "remaining_time": "18:06:57"}
{"current_steps": 280, "total_steps": 3751, "loss": 0.3041, "lr": 7.446808510638298e-06, "epoch": 0.07464551710348735, "percentage": 7.46, "elapsed_time": "1:26:38", "remaining_time": "17:54:04"}
{"current_steps": 290, "total_steps": 3751, "loss": 0.3136, "lr": 7.71276595744681e-06, "epoch": 0.0773114284286119, "percentage": 7.73, "elapsed_time": "1:28:59", "remaining_time": "17:42:08"}
{"current_steps": 300, "total_steps": 3751, "loss": 0.3137, "lr": 7.97872340425532e-06, "epoch": 0.07997733975373644, "percentage": 8.0, "elapsed_time": "1:31:19", "remaining_time": "17:30:30"}
{"current_steps": 300, "total_steps": 3751, "eval_loss": 0.3207298815250397, "epoch": 0.07997733975373644, "percentage": 8.0, "elapsed_time": "1:35:42", "remaining_time": "18:20:55"}
{"current_steps": 310, "total_steps": 3751, "loss": 0.3092, "lr": 8.24468085106383e-06, "epoch": 0.08264325107886099, "percentage": 8.26, "elapsed_time": "1:37:58", "remaining_time": "18:07:26"}
{"current_steps": 320, "total_steps": 3751, "loss": 0.3255, "lr": 8.510638297872341e-06, "epoch": 0.08530916240398553, "percentage": 8.53, "elapsed_time": "1:40:14", "remaining_time": "17:54:48"}
{"current_steps": 330, "total_steps": 3751, "loss": 0.3219, "lr": 8.776595744680852e-06, "epoch": 0.08797507372911008, "percentage": 8.8, "elapsed_time": "1:42:34", "remaining_time": "17:43:19"}
{"current_steps": 340, "total_steps": 3751, "loss": 0.3093, "lr": 9.042553191489362e-06, "epoch": 0.09064098505423464, "percentage": 9.06, "elapsed_time": "1:44:51", "remaining_time": "17:32:01"}
{"current_steps": 350, "total_steps": 3751, "loss": 0.3221, "lr": 9.308510638297872e-06, "epoch": 0.09330689637935918, "percentage": 9.33, "elapsed_time": "1:47:12", "remaining_time": "17:21:41"}
{"current_steps": 350, "total_steps": 3751, "eval_loss": 0.3211060166358948, "epoch": 0.09330689637935918, "percentage": 9.33, "elapsed_time": "1:51:35", "remaining_time": "18:04:16"}
{"current_steps": 360, "total_steps": 3751, "loss": 0.3209, "lr": 9.574468085106385e-06, "epoch": 0.09597280770448373, "percentage": 9.6, "elapsed_time": "1:53:50", "remaining_time": "17:52:15"}
{"current_steps": 370, "total_steps": 3751, "loss": 0.3188, "lr": 9.840425531914895e-06, "epoch": 0.09863871902960827, "percentage": 9.86, "elapsed_time": "1:56:12", "remaining_time": "17:41:53"}
{"current_steps": 380, "total_steps": 3751, "loss": 0.3202, "lr": 9.999965341346946e-06, "epoch": 0.10130463035473282, "percentage": 10.13, "elapsed_time": "1:58:33", "remaining_time": "17:31:41"}
{"current_steps": 390, "total_steps": 3751, "loss": 0.3143, "lr": 9.999575437018172e-06, "epoch": 0.10397054167985738, "percentage": 10.4, "elapsed_time": "2:00:56", "remaining_time": "17:22:19"}
{"current_steps": 400, "total_steps": 3751, "loss": 0.3188, "lr": 9.998752338940612e-06, "epoch": 0.10663645300498192, "percentage": 10.66, "elapsed_time": "2:03:18", "remaining_time": "17:13:01"}
{"current_steps": 400, "total_steps": 3751, "eval_loss": 0.3204084634780884, "epoch": 0.10663645300498192, "percentage": 10.66, "elapsed_time": "2:07:41", "remaining_time": "17:49:45"}
{"current_steps": 410, "total_steps": 3751, "loss": 0.3026, "lr": 9.997496118432509e-06, "epoch": 0.10930236433010647, "percentage": 10.93, "elapsed_time": "2:09:59", "remaining_time": "17:39:15"}
{"current_steps": 420, "total_steps": 3751, "loss": 0.3192, "lr": 9.995806884340483e-06, "epoch": 0.11196827565523101, "percentage": 11.2, "elapsed_time": "2:12:16", "remaining_time": "17:29:02"}
{"current_steps": 430, "total_steps": 3751, "loss": 0.3143, "lr": 9.99368478303009e-06, "epoch": 0.11463418698035556, "percentage": 11.46, "elapsed_time": "2:14:36", "remaining_time": "17:19:33"}
{"current_steps": 440, "total_steps": 3751, "loss": 0.3189, "lr": 9.991129998373145e-06, "epoch": 0.11730009830548012, "percentage": 11.73, "elapsed_time": "2:16:56", "remaining_time": "17:10:27"}
{"current_steps": 450, "total_steps": 3751, "loss": 0.308, "lr": 9.988142751731797e-06, "epoch": 0.11996600963060466, "percentage": 12.0, "elapsed_time": "2:19:14", "remaining_time": "17:01:22"}
{"current_steps": 450, "total_steps": 3751, "eval_loss": 0.31486886739730835, "epoch": 0.11996600963060466, "percentage": 12.0, "elapsed_time": "2:23:37", "remaining_time": "17:33:31"}
{"current_steps": 460, "total_steps": 3751, "loss": 0.3172, "lr": 9.984723301939337e-06, "epoch": 0.12263192095572921, "percentage": 12.26, "elapsed_time": "2:25:55", "remaining_time": "17:24:02"}
{"current_steps": 470, "total_steps": 3751, "loss": 0.3147, "lr": 9.980871945277777e-06, "epoch": 0.12529783228085375, "percentage": 12.53, "elapsed_time": "2:28:14", "remaining_time": "17:14:49"}
{"current_steps": 480, "total_steps": 3751, "loss": 0.3211, "lr": 9.976589015452178e-06, "epoch": 0.12796374360597831, "percentage": 12.8, "elapsed_time": "2:30:32", "remaining_time": "17:05:51"}
{"current_steps": 490, "total_steps": 3751, "loss": 0.3109, "lr": 9.97187488356174e-06, "epoch": 0.13062965493110285, "percentage": 13.06, "elapsed_time": "2:32:51", "remaining_time": "16:57:15"}
{"current_steps": 500, "total_steps": 3751, "loss": 0.3123, "lr": 9.966729958067638e-06, "epoch": 0.1332955662562274, "percentage": 13.33, "elapsed_time": "2:35:08", "remaining_time": "16:48:43"}
{"current_steps": 500, "total_steps": 3751, "eval_loss": 0.3106406331062317, "epoch": 0.1332955662562274, "percentage": 13.33, "elapsed_time": "2:39:31", "remaining_time": "17:17:13"}
{"current_steps": 510, "total_steps": 3751, "loss": 0.3032, "lr": 9.961154684757636e-06, "epoch": 0.13596147758135196, "percentage": 13.6, "elapsed_time": "2:42:53", "remaining_time": "17:15:08"}
{"current_steps": 520, "total_steps": 3751, "loss": 0.3048, "lr": 9.955149546707465e-06, "epoch": 0.1386273889064765, "percentage": 13.86, "elapsed_time": "2:45:10", "remaining_time": "17:06:19"}
{"current_steps": 530, "total_steps": 3751, "loss": 0.2947, "lr": 9.948715064238956e-06, "epoch": 0.14129330023160105, "percentage": 14.13, "elapsed_time": "2:47:33", "remaining_time": "16:58:20"}
{"current_steps": 540, "total_steps": 3751, "loss": 0.2998, "lr": 9.941851794874969e-06, "epoch": 0.14395921155672559, "percentage": 14.4, "elapsed_time": "2:49:49", "remaining_time": "16:49:51"}
{"current_steps": 550, "total_steps": 3751, "loss": 0.3138, "lr": 9.934560333291077e-06, "epoch": 0.14662512288185015, "percentage": 14.66, "elapsed_time": "2:52:08", "remaining_time": "16:41:50"}
{"current_steps": 550, "total_steps": 3751, "eval_loss": 0.3050285875797272, "epoch": 0.14662512288185015, "percentage": 14.66, "elapsed_time": "2:56:31", "remaining_time": "17:07:21"}
{"current_steps": 560, "total_steps": 3751, "loss": 0.3014, "lr": 9.926841311264037e-06, "epoch": 0.1492910342069747, "percentage": 14.93, "elapsed_time": "2:58:47", "remaining_time": "16:58:49"}
{"current_steps": 570, "total_steps": 3751, "loss": 0.3076, "lr": 9.918695397617064e-06, "epoch": 0.15195694553209924, "percentage": 15.2, "elapsed_time": "3:01:08", "remaining_time": "16:50:52"}
{"current_steps": 580, "total_steps": 3751, "loss": 0.3013, "lr": 9.91012329816186e-06, "epoch": 0.1546228568572238, "percentage": 15.46, "elapsed_time": "3:03:30", "remaining_time": "16:43:19"}
{"current_steps": 590, "total_steps": 3751, "loss": 0.3017, "lr": 9.901125755637473e-06, "epoch": 0.15728876818234833, "percentage": 15.73, "elapsed_time": "3:05:50", "remaining_time": "16:35:37"}
{"current_steps": 600, "total_steps": 3751, "loss": 0.3032, "lr": 9.89170354964594e-06, "epoch": 0.15995467950747289, "percentage": 16.0, "elapsed_time": "3:08:08", "remaining_time": "16:28:04"}
{"current_steps": 600, "total_steps": 3751, "eval_loss": 0.3046238422393799, "epoch": 0.15995467950747289, "percentage": 16.0, "elapsed_time": "3:12:31", "remaining_time": "16:51:06"}
{"current_steps": 610, "total_steps": 3751, "loss": 0.3027, "lr": 9.881857496584726e-06, "epoch": 0.16262059083259742, "percentage": 16.26, "elapsed_time": "3:14:53", "remaining_time": "16:43:29"}
{"current_steps": 620, "total_steps": 3751, "loss": 0.296, "lr": 9.871588449575999e-06, "epoch": 0.16528650215772198, "percentage": 16.53, "elapsed_time": "3:17:12", "remaining_time": "16:35:52"}
{"current_steps": 630, "total_steps": 3751, "loss": 0.3031, "lr": 9.860897298392712e-06, "epoch": 0.16795241348284654, "percentage": 16.8, "elapsed_time": "3:19:28", "remaining_time": "16:28:09"}
{"current_steps": 640, "total_steps": 3751, "loss": 0.3006, "lr": 9.849784969381488e-06, "epoch": 0.17061832480797107, "percentage": 17.06, "elapsed_time": "3:21:47", "remaining_time": "16:20:53"}
{"current_steps": 650, "total_steps": 3751, "loss": 0.2827, "lr": 9.83825242538238e-06, "epoch": 0.17328423613309563, "percentage": 17.33, "elapsed_time": "3:24:06", "remaining_time": "16:13:46"}
{"current_steps": 650, "total_steps": 3751, "eval_loss": 0.30168309807777405, "epoch": 0.17328423613309563, "percentage": 17.33, "elapsed_time": "3:28:29", "remaining_time": "16:34:41"}
{"current_steps": 660, "total_steps": 3751, "loss": 0.2814, "lr": 9.826300665645432e-06, "epoch": 0.17595014745822016, "percentage": 17.6, "elapsed_time": "3:30:46", "remaining_time": "16:27:07"}
{"current_steps": 670, "total_steps": 3751, "loss": 0.2852, "lr": 9.813930725744095e-06, "epoch": 0.17861605878334472, "percentage": 17.86, "elapsed_time": "3:33:09", "remaining_time": "16:20:11"}
{"current_steps": 680, "total_steps": 3751, "loss": 0.3016, "lr": 9.801143677485509e-06, "epoch": 0.18128197010846928, "percentage": 18.13, "elapsed_time": "3:35:31", "remaining_time": "16:13:19"}
{"current_steps": 690, "total_steps": 3751, "loss": 0.3023, "lr": 9.787940628817627e-06, "epoch": 0.1839478814335938, "percentage": 18.4, "elapsed_time": "3:37:50", "remaining_time": "16:06:23"}
{"current_steps": 700, "total_steps": 3751, "loss": 0.2953, "lr": 9.774322723733216e-06, "epoch": 0.18661379275871837, "percentage": 18.66, "elapsed_time": "3:40:06", "remaining_time": "15:59:22"}
{"current_steps": 700, "total_steps": 3751, "eval_loss": 0.2970203757286072, "epoch": 0.18661379275871837, "percentage": 18.66, "elapsed_time": "3:44:29", "remaining_time": "16:18:29"}
{"current_steps": 710, "total_steps": 3751, "loss": 0.2819, "lr": 9.760291142170739e-06, "epoch": 0.1892797040838429, "percentage": 18.93, "elapsed_time": "3:46:49", "remaining_time": "16:11:32"}
{"current_steps": 720, "total_steps": 3751, "loss": 0.2922, "lr": 9.745847099912116e-06, "epoch": 0.19194561540896746, "percentage": 19.19, "elapsed_time": "3:49:09", "remaining_time": "16:04:41"}
{"current_steps": 730, "total_steps": 3751, "loss": 0.2936, "lr": 9.73099184847738e-06, "epoch": 0.19461152673409202, "percentage": 19.46, "elapsed_time": "3:51:24", "remaining_time": "15:57:37"}
{"current_steps": 740, "total_steps": 3751, "loss": 0.2852, "lr": 9.715726675016238e-06, "epoch": 0.19727743805921655, "percentage": 19.73, "elapsed_time": "3:53:43", "remaining_time": "15:51:02"}
{"current_steps": 750, "total_steps": 3751, "loss": 0.2854, "lr": 9.700052902196541e-06, "epoch": 0.1999433493843411, "percentage": 19.99, "elapsed_time": "3:56:05", "remaining_time": "15:44:41"}
{"current_steps": 750, "total_steps": 3751, "eval_loss": 0.292442262172699, "epoch": 0.1999433493843411, "percentage": 19.99, "elapsed_time": "4:00:28", "remaining_time": "16:02:13"}
{"current_steps": 760, "total_steps": 3751, "loss": 0.2911, "lr": 9.68397188808969e-06, "epoch": 0.20260926070946564, "percentage": 20.26, "elapsed_time": "4:02:44", "remaining_time": "15:55:18"}
{"current_steps": 770, "total_steps": 3751, "loss": 0.3003, "lr": 9.667485026052956e-06, "epoch": 0.2052751720345902, "percentage": 20.53, "elapsed_time": "4:04:59", "remaining_time": "15:48:29"}
{"current_steps": 780, "total_steps": 3751, "loss": 0.2954, "lr": 9.650593744608754e-06, "epoch": 0.20794108335971476, "percentage": 20.79, "elapsed_time": "4:07:26", "remaining_time": "15:42:29"}
{"current_steps": 790, "total_steps": 3751, "loss": 0.2921, "lr": 9.633299507320862e-06, "epoch": 0.2106069946848393, "percentage": 21.06, "elapsed_time": "4:09:50", "remaining_time": "15:36:25"}
{"current_steps": 800, "total_steps": 3751, "loss": 0.2872, "lr": 9.615603812667618e-06, "epoch": 0.21327290600996385, "percentage": 21.33, "elapsed_time": "4:12:06", "remaining_time": "15:29:57"}
{"current_steps": 800, "total_steps": 3751, "eval_loss": 0.2895732522010803, "epoch": 0.21327290600996385, "percentage": 21.33, "elapsed_time": "4:16:29", "remaining_time": "15:46:08"}
{"current_steps": 810, "total_steps": 3751, "loss": 0.2907, "lr": 9.597508193912077e-06, "epoch": 0.21593881733508838, "percentage": 21.59, "elapsed_time": "4:18:46", "remaining_time": "15:39:34"}
{"current_steps": 820, "total_steps": 3751, "loss": 0.2867, "lr": 9.579014218969158e-06, "epoch": 0.21860472866021294, "percentage": 21.86, "elapsed_time": "4:21:10", "remaining_time": "15:33:31"}
{"current_steps": 830, "total_steps": 3751, "loss": 0.2942, "lr": 9.560123490269795e-06, "epoch": 0.2212706399853375, "percentage": 22.13, "elapsed_time": "4:23:30", "remaining_time": "15:27:22"}
{"current_steps": 840, "total_steps": 3751, "loss": 0.2832, "lr": 9.540837644622091e-06, "epoch": 0.22393655131046203, "percentage": 22.39, "elapsed_time": "4:25:51", "remaining_time": "15:21:19"}
{"current_steps": 850, "total_steps": 3751, "loss": 0.2866, "lr": 9.521158353069494e-06, "epoch": 0.2266024626355866, "percentage": 22.66, "elapsed_time": "4:28:09", "remaining_time": "15:15:13"}
{"current_steps": 850, "total_steps": 3751, "eval_loss": 0.28362876176834106, "epoch": 0.2266024626355866, "percentage": 22.66, "elapsed_time": "4:32:32", "remaining_time": "15:30:11"}
{"current_steps": 860, "total_steps": 3751, "loss": 0.2877, "lr": 9.501087320746007e-06, "epoch": 0.22926837396071112, "percentage": 22.93, "elapsed_time": "4:34:52", "remaining_time": "15:24:01"}
{"current_steps": 870, "total_steps": 3751, "loss": 0.2857, "lr": 9.480626286728445e-06, "epoch": 0.23193428528583568, "percentage": 23.19, "elapsed_time": "4:37:10", "remaining_time": "15:17:52"}
{"current_steps": 880, "total_steps": 3751, "loss": 0.2839, "lr": 9.459777023885754e-06, "epoch": 0.23460019661096024, "percentage": 23.46, "elapsed_time": "4:39:25", "remaining_time": "15:11:35"}
{"current_steps": 890, "total_steps": 3751, "loss": 0.2833, "lr": 9.438541338725397e-06, "epoch": 0.23726610793608477, "percentage": 23.73, "elapsed_time": "4:41:45", "remaining_time": "15:05:43"}
{"current_steps": 900, "total_steps": 3751, "loss": 0.2925, "lr": 9.416921071236821e-06, "epoch": 0.23993201926120933, "percentage": 23.99, "elapsed_time": "4:44:05", "remaining_time": "14:59:54"}
{"current_steps": 900, "total_steps": 3751, "eval_loss": 0.2794356942176819, "epoch": 0.23993201926120933, "percentage": 23.99, "elapsed_time": "4:48:28", "remaining_time": "15:13:48"}
{"current_steps": 910, "total_steps": 3751, "loss": 0.2846, "lr": 9.394918094732044e-06, "epoch": 0.24259793058633386, "percentage": 24.26, "elapsed_time": "4:50:53", "remaining_time": "15:08:10"}
{"current_steps": 920, "total_steps": 3751, "loss": 0.2826, "lr": 9.37253431568332e-06, "epoch": 0.24526384191145842, "percentage": 24.53, "elapsed_time": "4:53:11", "remaining_time": "15:02:12"}
{"current_steps": 930, "total_steps": 3751, "loss": 0.2856, "lr": 9.349771673557966e-06, "epoch": 0.24792975323658298, "percentage": 24.79, "elapsed_time": "4:55:31", "remaining_time": "14:56:24"}
{"current_steps": 940, "total_steps": 3751, "loss": 0.2886, "lr": 9.326632140650311e-06, "epoch": 0.2505956645617075, "percentage": 25.06, "elapsed_time": "4:57:50", "remaining_time": "14:50:39"}
{"current_steps": 950, "total_steps": 3751, "loss": 0.2843, "lr": 9.303117721910801e-06, "epoch": 0.25326157588683207, "percentage": 25.33, "elapsed_time": "5:00:09", "remaining_time": "14:44:59"}
{"current_steps": 950, "total_steps": 3751, "eval_loss": 0.28232333064079285, "epoch": 0.25326157588683207, "percentage": 25.33, "elapsed_time": "5:04:32", "remaining_time": "14:57:55"}
{"current_steps": 960, "total_steps": 3751, "loss": 0.2758, "lr": 9.279230454772282e-06, "epoch": 0.25592748721195663, "percentage": 25.59, "elapsed_time": "5:06:53", "remaining_time": "14:52:14"}
{"current_steps": 970, "total_steps": 3751, "loss": 0.2789, "lr": 9.25497240897346e-06, "epoch": 0.2585933985370812, "percentage": 25.86, "elapsed_time": "5:09:12", "remaining_time": "14:46:30"}
{"current_steps": 980, "total_steps": 3751, "loss": 0.2854, "lr": 9.23034568637957e-06, "epoch": 0.2612593098622057, "percentage": 26.13, "elapsed_time": "5:11:32", "remaining_time": "14:40:52"}
{"current_steps": 990, "total_steps": 3751, "loss": 0.2857, "lr": 9.205352420800253e-06, "epoch": 0.26392522118733025, "percentage": 26.39, "elapsed_time": "5:13:49", "remaining_time": "14:35:13"}
{"current_steps": 1000, "total_steps": 3751, "loss": 0.292, "lr": 9.179994777804677e-06, "epoch": 0.2665911325124548, "percentage": 26.66, "elapsed_time": "5:16:08", "remaining_time": "14:29:41"}
{"current_steps": 1000, "total_steps": 3751, "eval_loss": 0.27887627482414246, "epoch": 0.2665911325124548, "percentage": 26.66, "elapsed_time": "5:20:31", "remaining_time": "14:41:44"}
{"current_steps": 1010, "total_steps": 3751, "loss": 0.2863, "lr": 9.154274954533895e-06, "epoch": 0.26925704383757937, "percentage": 26.93, "elapsed_time": "5:24:11", "remaining_time": "14:39:49"}
{"current_steps": 1020, "total_steps": 3751, "loss": 0.279, "lr": 9.128195179510466e-06, "epoch": 0.2719229551627039, "percentage": 27.19, "elapsed_time": "5:26:32", "remaining_time": "14:34:18"}
{"current_steps": 1030, "total_steps": 3751, "loss": 0.2743, "lr": 9.101757712445369e-06, "epoch": 0.27458886648782843, "percentage": 27.46, "elapsed_time": "5:28:53", "remaining_time": "14:28:50"}
{"current_steps": 1040, "total_steps": 3751, "loss": 0.2757, "lr": 9.07496484404221e-06, "epoch": 0.277254777812953, "percentage": 27.73, "elapsed_time": "5:31:12", "remaining_time": "14:23:23"}
{"current_steps": 1050, "total_steps": 3751, "loss": 0.2775, "lr": 9.04781889579873e-06, "epoch": 0.27992068913807755, "percentage": 27.99, "elapsed_time": "5:33:30", "remaining_time": "14:17:55"}
{"current_steps": 1050, "total_steps": 3751, "eval_loss": 0.27627232670783997, "epoch": 0.27992068913807755, "percentage": 27.99, "elapsed_time": "5:37:53", "remaining_time": "14:29:11"}
{"current_steps": 1060, "total_steps": 3751, "loss": 0.2803, "lr": 9.020322219805674e-06, "epoch": 0.2825866004632021, "percentage": 28.26, "elapsed_time": "5:40:16", "remaining_time": "14:23:50"}
{"current_steps": 1070, "total_steps": 3751, "loss": 0.2831, "lr": 8.99247719854297e-06, "epoch": 0.28525251178832667, "percentage": 28.53, "elapsed_time": "5:42:37", "remaining_time": "14:18:29"}
{"current_steps": 1080, "total_steps": 3751, "loss": 0.2794, "lr": 8.964286244673315e-06, "epoch": 0.28791842311345117, "percentage": 28.79, "elapsed_time": "5:44:54", "remaining_time": "14:13:00"}
{"current_steps": 1090, "total_steps": 3751, "loss": 0.2728, "lr": 8.935751800833117e-06, "epoch": 0.29058433443857573, "percentage": 29.06, "elapsed_time": "5:47:12", "remaining_time": "14:07:38"}
{"current_steps": 1100, "total_steps": 3751, "loss": 0.2652, "lr": 8.906876339420851e-06, "epoch": 0.2932502457637003, "percentage": 29.33, "elapsed_time": "5:49:33", "remaining_time": "14:02:25"}
{"current_steps": 1100, "total_steps": 3751, "eval_loss": 0.27173659205436707, "epoch": 0.2932502457637003, "percentage": 29.33, "elapsed_time": "5:53:56", "remaining_time": "14:12:59"}
{"current_steps": 1110, "total_steps": 3751, "loss": 0.2731, "lr": 8.877662362382844e-06, "epoch": 0.29591615708882485, "percentage": 29.59, "elapsed_time": "5:56:14", "remaining_time": "14:07:36"}
{"current_steps": 1120, "total_steps": 3751, "loss": 0.2669, "lr": 8.848112400996473e-06, "epoch": 0.2985820684139494, "percentage": 29.86, "elapsed_time": "5:58:38", "remaining_time": "14:02:29"}
{"current_steps": 1130, "total_steps": 3751, "loss": 0.274, "lr": 8.818229015650862e-06, "epoch": 0.3012479797390739, "percentage": 30.13, "elapsed_time": "6:00:59", "remaining_time": "13:57:18"}
{"current_steps": 1140, "total_steps": 3751, "loss": 0.275, "lr": 8.788014795625018e-06, "epoch": 0.30391389106419847, "percentage": 30.39, "elapsed_time": "6:03:19", "remaining_time": "13:52:07"}
{"current_steps": 1150, "total_steps": 3751, "loss": 0.27, "lr": 8.757472358863481e-06, "epoch": 0.30657980238932303, "percentage": 30.66, "elapsed_time": "6:05:37", "remaining_time": "13:46:57"}
{"current_steps": 1150, "total_steps": 3751, "eval_loss": 0.2712327241897583, "epoch": 0.30657980238932303, "percentage": 30.66, "elapsed_time": "6:10:00", "remaining_time": "13:56:52"}
{"current_steps": 1160, "total_steps": 3751, "loss": 0.2676, "lr": 8.726604351749503e-06, "epoch": 0.3092457137144476, "percentage": 30.93, "elapsed_time": "6:12:19", "remaining_time": "13:51:37"}
{"current_steps": 1170, "total_steps": 3751, "loss": 0.2779, "lr": 8.69541344887573e-06, "epoch": 0.3119116250395721, "percentage": 31.19, "elapsed_time": "6:14:38", "remaining_time": "13:46:26"}
{"current_steps": 1180, "total_steps": 3751, "loss": 0.2693, "lr": 8.66390235281248e-06, "epoch": 0.31457753636469665, "percentage": 31.46, "elapsed_time": "6:16:55", "remaining_time": "13:41:13"}
{"current_steps": 1190, "total_steps": 3751, "loss": 0.2709, "lr": 8.632073793873548e-06, "epoch": 0.3172434476898212, "percentage": 31.72, "elapsed_time": "6:19:18", "remaining_time": "13:36:18"}
{"current_steps": 1200, "total_steps": 3751, "loss": 0.277, "lr": 8.599930529879669e-06, "epoch": 0.31990935901494577, "percentage": 31.99, "elapsed_time": "6:21:43", "remaining_time": "13:31:27"}
{"current_steps": 1200, "total_steps": 3751, "eval_loss": 0.2749159336090088, "epoch": 0.31990935901494577, "percentage": 31.99, "elapsed_time": "6:26:05", "remaining_time": "13:40:46"}
{"current_steps": 1210, "total_steps": 3751, "loss": 0.2629, "lr": 8.567475345919532e-06, "epoch": 0.32257527034007033, "percentage": 32.26, "elapsed_time": "6:28:27", "remaining_time": "13:35:46"}
{"current_steps": 1220, "total_steps": 3751, "loss": 0.2747, "lr": 8.534711054108487e-06, "epoch": 0.32524118166519483, "percentage": 32.52, "elapsed_time": "6:30:46", "remaining_time": "13:30:41"}
{"current_steps": 1230, "total_steps": 3751, "loss": 0.2644, "lr": 8.501640493344866e-06, "epoch": 0.3279070929903194, "percentage": 32.79, "elapsed_time": "6:33:02", "remaining_time": "13:25:34"}
{"current_steps": 1240, "total_steps": 3751, "loss": 0.263, "lr": 8.468266529064025e-06, "epoch": 0.33057300431544395, "percentage": 33.06, "elapsed_time": "6:35:17", "remaining_time": "13:20:28"}
{"current_steps": 1250, "total_steps": 3751, "loss": 0.2681, "lr": 8.434592052990044e-06, "epoch": 0.3332389156405685, "percentage": 33.32, "elapsed_time": "6:37:34", "remaining_time": "13:15:27"}
{"current_steps": 1250, "total_steps": 3751, "eval_loss": 0.2708764672279358, "epoch": 0.3332389156405685, "percentage": 33.32, "elapsed_time": "6:41:57", "remaining_time": "13:24:13"}
{"current_steps": 1260, "total_steps": 3751, "loss": 0.272, "lr": 8.400619982885183e-06, "epoch": 0.33590482696569307, "percentage": 33.59, "elapsed_time": "6:44:14", "remaining_time": "13:19:11"}
{"current_steps": 1270, "total_steps": 3751, "loss": 0.2713, "lr": 8.366353262297069e-06, "epoch": 0.3385707382908176, "percentage": 33.86, "elapsed_time": "6:46:34", "remaining_time": "13:14:16"}
{"current_steps": 1280, "total_steps": 3751, "loss": 0.2792, "lr": 8.331794860303644e-06, "epoch": 0.34123664961594213, "percentage": 34.12, "elapsed_time": "6:48:52", "remaining_time": "13:09:19"}
{"current_steps": 1290, "total_steps": 3751, "loss": 0.2628, "lr": 8.296947771255905e-06, "epoch": 0.3439025609410667, "percentage": 34.39, "elapsed_time": "6:51:10", "remaining_time": "13:04:24"}
{"current_steps": 1300, "total_steps": 3751, "loss": 0.2699, "lr": 8.261815014518465e-06, "epoch": 0.34656847226619125, "percentage": 34.66, "elapsed_time": "6:53:28", "remaining_time": "12:59:32"}
{"current_steps": 1300, "total_steps": 3751, "eval_loss": 0.27181142568588257, "epoch": 0.34656847226619125, "percentage": 34.66, "elapsed_time": "6:57:50", "remaining_time": "13:07:48"}
{"current_steps": 1310, "total_steps": 3751, "loss": 0.2734, "lr": 8.226399634207929e-06, "epoch": 0.3492343835913158, "percentage": 34.92, "elapsed_time": "7:00:10", "remaining_time": "13:02:56"}
{"current_steps": 1320, "total_steps": 3751, "loss": 0.2638, "lr": 8.190704698929128e-06, "epoch": 0.3519002949164403, "percentage": 35.19, "elapsed_time": "7:02:29", "remaining_time": "12:58:05"}
{"current_steps": 1330, "total_steps": 3751, "loss": 0.2595, "lr": 8.154733301509249e-06, "epoch": 0.3545662062415649, "percentage": 35.46, "elapsed_time": "7:04:48", "remaining_time": "12:53:15"}
{"current_steps": 1340, "total_steps": 3751, "loss": 0.2663, "lr": 8.118488558729846e-06, "epoch": 0.35723211756668943, "percentage": 35.72, "elapsed_time": "7:07:10", "remaining_time": "12:48:36"}
{"current_steps": 1350, "total_steps": 3751, "loss": 0.2682, "lr": 8.081973611056784e-06, "epoch": 0.359898028891814, "percentage": 35.99, "elapsed_time": "7:09:31", "remaining_time": "12:43:54"}
{"current_steps": 1350, "total_steps": 3751, "eval_loss": 0.267572283744812, "epoch": 0.359898028891814, "percentage": 35.99, "elapsed_time": "7:13:54", "remaining_time": "12:51:42"}
{"current_steps": 1360, "total_steps": 3751, "loss": 0.2593, "lr": 8.045191622368128e-06, "epoch": 0.36256394021693855, "percentage": 36.26, "elapsed_time": "7:16:12", "remaining_time": "12:46:53"}
{"current_steps": 1370, "total_steps": 3751, "loss": 0.2639, "lr": 8.008145779680011e-06, "epoch": 0.36522985154206306, "percentage": 36.52, "elapsed_time": "7:18:32", "remaining_time": "12:42:09"}
{"current_steps": 1380, "total_steps": 3751, "loss": 0.2669, "lr": 7.970839292870488e-06, "epoch": 0.3678957628671876, "percentage": 36.79, "elapsed_time": "7:20:47", "remaining_time": "12:37:20"}
{"current_steps": 1390, "total_steps": 3751, "loss": 0.2558, "lr": 7.933275394401407e-06, "epoch": 0.3705616741923122, "percentage": 37.06, "elapsed_time": "7:23:06", "remaining_time": "12:32:38"}
{"current_steps": 1400, "total_steps": 3751, "loss": 0.2668, "lr": 7.89545733903834e-06, "epoch": 0.37322758551743673, "percentage": 37.32, "elapsed_time": "7:25:22", "remaining_time": "12:27:54"}
{"current_steps": 1400, "total_steps": 3751, "eval_loss": 0.26621583104133606, "epoch": 0.37322758551743673, "percentage": 37.32, "elapsed_time": "7:29:45", "remaining_time": "12:35:16"}
{"current_steps": 1410, "total_steps": 3751, "loss": 0.2562, "lr": 7.857388403568564e-06, "epoch": 0.3758934968425613, "percentage": 37.59, "elapsed_time": "7:32:05", "remaining_time": "12:30:35"}
{"current_steps": 1420, "total_steps": 3751, "loss": 0.2641, "lr": 7.819071886517134e-06, "epoch": 0.3785594081676858, "percentage": 37.86, "elapsed_time": "7:34:23", "remaining_time": "12:25:55"}
{"current_steps": 1430, "total_steps": 3751, "loss": 0.2574, "lr": 7.780511107861095e-06, "epoch": 0.38122531949281036, "percentage": 38.12, "elapsed_time": "7:36:40", "remaining_time": "12:21:12"}
{"current_steps": 1440, "total_steps": 3751, "loss": 0.2523, "lr": 7.741709408741804e-06, "epoch": 0.3838912308179349, "percentage": 38.39, "elapsed_time": "7:38:57", "remaining_time": "12:16:34"}
{"current_steps": 1450, "total_steps": 3751, "loss": 0.2615, "lr": 7.702670151175435e-06, "epoch": 0.3865571421430595, "percentage": 38.66, "elapsed_time": "7:41:17", "remaining_time": "12:12:00"}
{"current_steps": 1450, "total_steps": 3751, "eval_loss": 0.26893940567970276, "epoch": 0.3865571421430595, "percentage": 38.66, "elapsed_time": "7:45:39", "remaining_time": "12:18:57"}
{"current_steps": 1460, "total_steps": 3751, "loss": 0.2598, "lr": 7.663396717761687e-06, "epoch": 0.38922305346818403, "percentage": 38.92, "elapsed_time": "7:48:00", "remaining_time": "12:14:23"}
{"current_steps": 1470, "total_steps": 3751, "loss": 0.2595, "lr": 7.6238925113906715e-06, "epoch": 0.39188896479330854, "percentage": 39.19, "elapsed_time": "7:50:15", "remaining_time": "12:09:42"}
{"current_steps": 1480, "total_steps": 3751, "loss": 0.2706, "lr": 7.5841609549480854e-06, "epoch": 0.3945548761184331, "percentage": 39.46, "elapsed_time": "7:52:32", "remaining_time": "12:05:06"}
{"current_steps": 1490, "total_steps": 3751, "loss": 0.252, "lr": 7.544205491018626e-06, "epoch": 0.39722078744355765, "percentage": 39.72, "elapsed_time": "7:54:54", "remaining_time": "12:00:38"}
{"current_steps": 1500, "total_steps": 3751, "loss": 0.2501, "lr": 7.5040295815877e-06, "epoch": 0.3998866987686822, "percentage": 39.99, "elapsed_time": "7:57:13", "remaining_time": "11:56:09"}
{"current_steps": 1500, "total_steps": 3751, "eval_loss": 0.2582685649394989, "epoch": 0.3998866987686822, "percentage": 39.99, "elapsed_time": "8:01:36", "remaining_time": "12:02:44"}
{"current_steps": 1510, "total_steps": 3751, "loss": 0.2604, "lr": 7.463636707741458e-06, "epoch": 0.4025526100938068, "percentage": 40.26, "elapsed_time": "8:04:56", "remaining_time": "11:59:42"}
{"current_steps": 1520, "total_steps": 3751, "loss": 0.264, "lr": 7.423030369365175e-06, "epoch": 0.4052185214189313, "percentage": 40.52, "elapsed_time": "8:07:14", "remaining_time": "11:55:09"}
{"current_steps": 1530, "total_steps": 3751, "loss": 0.2565, "lr": 7.382214084839993e-06, "epoch": 0.40788443274405584, "percentage": 40.79, "elapsed_time": "8:09:34", "remaining_time": "11:50:41"}
{"current_steps": 1540, "total_steps": 3751, "loss": 0.2638, "lr": 7.341191390738073e-06, "epoch": 0.4105503440691804, "percentage": 41.06, "elapsed_time": "8:11:52", "remaining_time": "11:46:11"}
{"current_steps": 1550, "total_steps": 3751, "loss": 0.2545, "lr": 7.299965841516164e-06, "epoch": 0.41321625539430495, "percentage": 41.32, "elapsed_time": "8:14:10", "remaining_time": "11:41:43"}
{"current_steps": 1550, "total_steps": 3751, "eval_loss": 0.25675299763679504, "epoch": 0.41321625539430495, "percentage": 41.32, "elapsed_time": "8:18:33", "remaining_time": "11:47:56"}
{"current_steps": 1560, "total_steps": 3751, "loss": 0.2637, "lr": 7.2585410092076154e-06, "epoch": 0.4158821667194295, "percentage": 41.59, "elapsed_time": "8:20:57", "remaining_time": "11:43:34"}
{"current_steps": 1570, "total_steps": 3751, "loss": 0.2489, "lr": 7.216920483112886e-06, "epoch": 0.418548078044554, "percentage": 41.86, "elapsed_time": "8:23:17", "remaining_time": "11:39:08"}
{"current_steps": 1580, "total_steps": 3751, "loss": 0.2545, "lr": 7.175107869488539e-06, "epoch": 0.4212139893696786, "percentage": 42.12, "elapsed_time": "8:25:35", "remaining_time": "11:34:42"}
{"current_steps": 1590, "total_steps": 3751, "loss": 0.2644, "lr": 7.133106791234771e-06, "epoch": 0.42387990069480314, "percentage": 42.39, "elapsed_time": "8:27:55", "remaining_time": "11:30:20"}
{"current_steps": 1600, "total_steps": 3751, "loss": 0.2618, "lr": 7.090920887581507e-06, "epoch": 0.4265458120199277, "percentage": 42.66, "elapsed_time": "8:30:19", "remaining_time": "11:26:03"}
{"current_steps": 1600, "total_steps": 3751, "eval_loss": 0.25225213170051575, "epoch": 0.4265458120199277, "percentage": 42.66, "elapsed_time": "8:34:42", "remaining_time": "11:31:57"}
{"current_steps": 1610, "total_steps": 3751, "loss": 0.246, "lr": 7.048553813773075e-06, "epoch": 0.42921172334505225, "percentage": 42.92, "elapsed_time": "8:37:06", "remaining_time": "11:27:39"}
{"current_steps": 1620, "total_steps": 3751, "loss": 0.248, "lr": 7.006009240751488e-06, "epoch": 0.43187763467017676, "percentage": 43.19, "elapsed_time": "8:39:24", "remaining_time": "11:23:15"}
{"current_steps": 1630, "total_steps": 3751, "loss": 0.2592, "lr": 6.963290854838376e-06, "epoch": 0.4345435459953013, "percentage": 43.46, "elapsed_time": "8:41:42", "remaining_time": "11:18:52"}
{"current_steps": 1640, "total_steps": 3751, "loss": 0.2502, "lr": 6.920402357415582e-06, "epoch": 0.4372094573204259, "percentage": 43.72, "elapsed_time": "8:44:03", "remaining_time": "11:14:33"}
{"current_steps": 1650, "total_steps": 3751, "loss": 0.2615, "lr": 6.877347464604446e-06, "epoch": 0.43987536864555044, "percentage": 43.99, "elapsed_time": "8:46:23", "remaining_time": "11:10:16"}
{"current_steps": 1650, "total_steps": 3751, "eval_loss": 0.2549818754196167, "epoch": 0.43987536864555044, "percentage": 43.99, "elapsed_time": "8:50:47", "remaining_time": "11:15:51"}
{"current_steps": 1660, "total_steps": 3751, "loss": 0.2468, "lr": 6.834129906943822e-06, "epoch": 0.442541279970675, "percentage": 44.25, "elapsed_time": "8:53:04", "remaining_time": "11:11:29"}
{"current_steps": 1670, "total_steps": 3751, "loss": 0.2522, "lr": 6.790753429066838e-06, "epoch": 0.4452071912957995, "percentage": 44.52, "elapsed_time": "8:55:21", "remaining_time": "11:07:06"}
{"current_steps": 1680, "total_steps": 3751, "loss": 0.2468, "lr": 6.7472217893764465e-06, "epoch": 0.44787310262092406, "percentage": 44.79, "elapsed_time": "8:57:35", "remaining_time": "11:02:42"}
{"current_steps": 1690, "total_steps": 3751, "loss": 0.2525, "lr": 6.70353875971976e-06, "epoch": 0.4505390139460486, "percentage": 45.05, "elapsed_time": "8:59:56", "remaining_time": "10:58:28"}
{"current_steps": 1700, "total_steps": 3751, "loss": 0.2512, "lr": 6.659708125061242e-06, "epoch": 0.4532049252711732, "percentage": 45.32, "elapsed_time": "9:02:16", "remaining_time": "10:54:14"}
{"current_steps": 1700, "total_steps": 3751, "eval_loss": 0.24882382154464722, "epoch": 0.4532049252711732, "percentage": 45.32, "elapsed_time": "9:06:39", "remaining_time": "10:59:31"}
{"current_steps": 1710, "total_steps": 3751, "loss": 0.2522, "lr": 6.615733683154762e-06, "epoch": 0.45587083659629773, "percentage": 45.59, "elapsed_time": "9:09:01", "remaining_time": "10:55:17"}
{"current_steps": 1720, "total_steps": 3751, "loss": 0.2505, "lr": 6.571619244214521e-06, "epoch": 0.45853674792142224, "percentage": 45.85, "elapsed_time": "9:11:22", "remaining_time": "10:51:03"}
{"current_steps": 1730, "total_steps": 3751, "loss": 0.2596, "lr": 6.527368630584919e-06, "epoch": 0.4612026592465468, "percentage": 46.12, "elapsed_time": "9:13:40", "remaining_time": "10:46:48"}
{"current_steps": 1740, "total_steps": 3751, "loss": 0.2494, "lr": 6.482985676409368e-06, "epoch": 0.46386857057167136, "percentage": 46.39, "elapsed_time": "9:15:57", "remaining_time": "10:42:32"}
{"current_steps": 1750, "total_steps": 3751, "loss": 0.245, "lr": 6.438474227298065e-06, "epoch": 0.4665344818967959, "percentage": 46.65, "elapsed_time": "9:18:14", "remaining_time": "10:38:18"}
{"current_steps": 1750, "total_steps": 3751, "eval_loss": 0.25038692355155945, "epoch": 0.4665344818967959, "percentage": 46.65, "elapsed_time": "9:22:36", "remaining_time": "10:43:18"}
{"current_steps": 1760, "total_steps": 3751, "loss": 0.2595, "lr": 6.393838139994797e-06, "epoch": 0.4692003932219205, "percentage": 46.92, "elapsed_time": "9:24:58", "remaining_time": "10:39:07"}
{"current_steps": 1770, "total_steps": 3751, "loss": 0.2516, "lr": 6.349081282042768e-06, "epoch": 0.471866304547045, "percentage": 47.19, "elapsed_time": "9:27:16", "remaining_time": "10:34:53"}
{"current_steps": 1780, "total_steps": 3751, "loss": 0.2469, "lr": 6.304207531449486e-06, "epoch": 0.47453221587216954, "percentage": 47.45, "elapsed_time": "9:29:35", "remaining_time": "10:30:42"}
{"current_steps": 1790, "total_steps": 3751, "loss": 0.2354, "lr": 6.259220776350746e-06, "epoch": 0.4771981271972941, "percentage": 47.72, "elapsed_time": "9:31:59", "remaining_time": "10:26:37"}
{"current_steps": 1800, "total_steps": 3751, "loss": 0.2503, "lr": 6.2141249146737545e-06, "epoch": 0.47986403852241866, "percentage": 47.99, "elapsed_time": "9:34:19", "remaining_time": "10:22:30"}
{"current_steps": 1800, "total_steps": 3751, "eval_loss": 0.24808603525161743, "epoch": 0.47986403852241866, "percentage": 47.99, "elapsed_time": "9:38:42", "remaining_time": "10:27:15"}
{"current_steps": 1810, "total_steps": 3751, "loss": 0.2466, "lr": 6.168923853799369e-06, "epoch": 0.4825299498475432, "percentage": 48.25, "elapsed_time": "9:41:02", "remaining_time": "10:23:06"}
{"current_steps": 1820, "total_steps": 3751, "loss": 0.2467, "lr": 6.123621510223552e-06, "epoch": 0.4851958611726677, "percentage": 48.52, "elapsed_time": "9:43:21", "remaining_time": "10:18:56"}
{"current_steps": 1830, "total_steps": 3751, "loss": 0.2523, "lr": 6.0782218092180164e-06, "epoch": 0.4878617724977923, "percentage": 48.79, "elapsed_time": "9:45:40", "remaining_time": "10:14:47"}
{"current_steps": 1840, "total_steps": 3751, "loss": 0.2428, "lr": 6.032728684490118e-06, "epoch": 0.49052768382291684, "percentage": 49.05, "elapsed_time": "9:47:56", "remaining_time": "10:10:37"}
{"current_steps": 1850, "total_steps": 3751, "loss": 0.2402, "lr": 5.987146077842015e-06, "epoch": 0.4931935951480414, "percentage": 49.32, "elapsed_time": "9:50:14", "remaining_time": "10:06:31"}
{"current_steps": 1850, "total_steps": 3751, "eval_loss": 0.24496783316135406, "epoch": 0.4931935951480414, "percentage": 49.32, "elapsed_time": "9:54:37", "remaining_time": "10:11:01"}
{"current_steps": 1860, "total_steps": 3751, "loss": 0.2526, "lr": 5.941477938829126e-06, "epoch": 0.49585950647316596, "percentage": 49.59, "elapsed_time": "9:56:55", "remaining_time": "10:06:52"}
{"current_steps": 1870, "total_steps": 3751, "loss": 0.2462, "lr": 5.8957282244179125e-06, "epoch": 0.49852541779829046, "percentage": 49.85, "elapsed_time": "9:59:17", "remaining_time": "10:02:49"}
{"current_steps": 1880, "total_steps": 3751, "loss": 0.2343, "lr": 5.84990089864303e-06, "epoch": 0.501191329123415, "percentage": 50.12, "elapsed_time": "10:01:37", "remaining_time": "9:58:44"}
{"current_steps": 1890, "total_steps": 3751, "loss": 0.2403, "lr": 5.803999932263859e-06, "epoch": 0.5038572404485396, "percentage": 50.39, "elapsed_time": "10:03:57", "remaining_time": "9:54:41"}
{"current_steps": 1900, "total_steps": 3751, "loss": 0.2346, "lr": 5.7580293024204455e-06, "epoch": 0.5065231517736641, "percentage": 50.65, "elapsed_time": "10:06:13", "remaining_time": "9:50:35"}
{"current_steps": 1900, "total_steps": 3751, "eval_loss": 0.24397991597652435, "epoch": 0.5065231517736641, "percentage": 50.65, "elapsed_time": "10:10:36", "remaining_time": "9:54:51"}
{"current_steps": 1910, "total_steps": 3751, "loss": 0.251, "lr": 5.7119929922889065e-06, "epoch": 0.5091890630987886, "percentage": 50.92, "elapsed_time": "10:12:57", "remaining_time": "9:50:49"}
{"current_steps": 1920, "total_steps": 3751, "loss": 0.2443, "lr": 5.665894990736301e-06, "epoch": 0.5118549744239133, "percentage": 51.19, "elapsed_time": "10:15:19", "remaining_time": "9:46:47"}
{"current_steps": 1930, "total_steps": 3751, "loss": 0.2492, "lr": 5.6197392919750095e-06, "epoch": 0.5145208857490378, "percentage": 51.45, "elapsed_time": "10:17:37", "remaining_time": "9:42:44"}
{"current_steps": 1940, "total_steps": 3751, "loss": 0.2472, "lr": 5.573529895216648e-06, "epoch": 0.5171867970741624, "percentage": 51.72, "elapsed_time": "10:19:57", "remaining_time": "9:38:43"}
{"current_steps": 1950, "total_steps": 3751, "loss": 0.2413, "lr": 5.5272708043255605e-06, "epoch": 0.5198527083992869, "percentage": 51.99, "elapsed_time": "10:22:16", "remaining_time": "9:34:43"}
{"current_steps": 1950, "total_steps": 3751, "eval_loss": 0.24250540137290955, "epoch": 0.5198527083992869, "percentage": 51.99, "elapsed_time": "10:26:39", "remaining_time": "9:38:46"}
{"current_steps": 1960, "total_steps": 3751, "loss": 0.237, "lr": 5.480966027471889e-06, "epoch": 0.5225186197244114, "percentage": 52.25, "elapsed_time": "10:28:59", "remaining_time": "9:34:45"}
{"current_steps": 1970, "total_steps": 3751, "loss": 0.2449, "lr": 5.434619576784288e-06, "epoch": 0.525184531049536, "percentage": 52.52, "elapsed_time": "10:31:19", "remaining_time": "9:30:45"}
{"current_steps": 1980, "total_steps": 3751, "loss": 0.2237, "lr": 5.388235468002286e-06, "epoch": 0.5278504423746605, "percentage": 52.79, "elapsed_time": "10:33:38", "remaining_time": "9:26:45"}
{"current_steps": 1990, "total_steps": 3751, "loss": 0.2454, "lr": 5.341817720128344e-06, "epoch": 0.5305163536997851, "percentage": 53.05, "elapsed_time": "10:35:55", "remaining_time": "9:22:44"}
{"current_steps": 2000, "total_steps": 3751, "loss": 0.24, "lr": 5.295370355079615e-06, "epoch": 0.5331822650249096, "percentage": 53.32, "elapsed_time": "10:38:17", "remaining_time": "9:18:49"}
{"current_steps": 2000, "total_steps": 3751, "eval_loss": 0.2383483648300171, "epoch": 0.5331822650249096, "percentage": 53.32, "elapsed_time": "10:42:40", "remaining_time": "9:22:39"}
{"current_steps": 2010, "total_steps": 3751, "loss": 0.2373, "lr": 5.248897397339462e-06, "epoch": 0.5358481763500341, "percentage": 53.59, "elapsed_time": "10:46:20", "remaining_time": "9:19:50"}
{"current_steps": 2020, "total_steps": 3751, "loss": 0.2355, "lr": 5.202402873608763e-06, "epoch": 0.5385140876751587, "percentage": 53.85, "elapsed_time": "10:48:42", "remaining_time": "9:15:53"}
{"current_steps": 2030, "total_steps": 3751, "loss": 0.2415, "lr": 5.155890812457e-06, "epoch": 0.5411799990002832, "percentage": 54.12, "elapsed_time": "10:51:01", "remaining_time": "9:11:55"}
{"current_steps": 2040, "total_steps": 3751, "loss": 0.2319, "lr": 5.1093652439732036e-06, "epoch": 0.5438459103254079, "percentage": 54.39, "elapsed_time": "10:53:19", "remaining_time": "9:07:57"}
{"current_steps": 2050, "total_steps": 3751, "loss": 0.2398, "lr": 5.062830199416764e-06, "epoch": 0.5465118216505324, "percentage": 54.65, "elapsed_time": "10:55:39", "remaining_time": "9:04:02"}
{"current_steps": 2050, "total_steps": 3751, "eval_loss": 0.24081052839756012, "epoch": 0.5465118216505324, "percentage": 54.65, "elapsed_time": "11:00:02", "remaining_time": "9:07:40"}
{"current_steps": 2060, "total_steps": 3751, "loss": 0.2393, "lr": 5.0162897108681375e-06, "epoch": 0.5491777329756569, "percentage": 54.92, "elapsed_time": "11:02:20", "remaining_time": "9:03:41"}
{"current_steps": 2070, "total_steps": 3751, "loss": 0.2447, "lr": 4.969747810879478e-06, "epoch": 0.5518436443007815, "percentage": 55.19, "elapsed_time": "11:04:36", "remaining_time": "8:59:42"}
{"current_steps": 2080, "total_steps": 3751, "loss": 0.2342, "lr": 4.923208532125235e-06, "epoch": 0.554509555625906, "percentage": 55.45, "elapsed_time": "11:06:54", "remaining_time": "8:55:46"}
{"current_steps": 2090, "total_steps": 3751, "loss": 0.2355, "lr": 4.876675907052745e-06, "epoch": 0.5571754669510306, "percentage": 55.72, "elapsed_time": "11:09:13", "remaining_time": "8:51:51"}
{"current_steps": 2100, "total_steps": 3751, "loss": 0.2473, "lr": 4.8301539675328205e-06, "epoch": 0.5598413782761551, "percentage": 55.99, "elapsed_time": "11:11:30", "remaining_time": "8:47:55"}
{"current_steps": 2100, "total_steps": 3751, "eval_loss": 0.23838448524475098, "epoch": 0.5598413782761551, "percentage": 55.99, "elapsed_time": "11:15:52", "remaining_time": "8:51:22"}
{"current_steps": 2110, "total_steps": 3751, "loss": 0.2202, "lr": 4.783646744510416e-06, "epoch": 0.5625072896012796, "percentage": 56.25, "elapsed_time": "11:18:09", "remaining_time": "8:47:25"}
{"current_steps": 2120, "total_steps": 3751, "loss": 0.2367, "lr": 4.737158267655358e-06, "epoch": 0.5651732009264042, "percentage": 56.52, "elapsed_time": "11:20:28", "remaining_time": "8:43:30"}
{"current_steps": 2130, "total_steps": 3751, "loss": 0.2355, "lr": 4.690692565013193e-06, "epoch": 0.5678391122515287, "percentage": 56.78, "elapsed_time": "11:22:45", "remaining_time": "8:39:35"}
{"current_steps": 2140, "total_steps": 3751, "loss": 0.2306, "lr": 4.644253662656167e-06, "epoch": 0.5705050235766533, "percentage": 57.05, "elapsed_time": "11:25:03", "remaining_time": "8:35:43"}
{"current_steps": 2150, "total_steps": 3751, "loss": 0.2423, "lr": 4.597845584334387e-06, "epoch": 0.5731709349017778, "percentage": 57.32, "elapsed_time": "11:27:23", "remaining_time": "8:31:52"}
{"current_steps": 2150, "total_steps": 3751, "eval_loss": 0.234757199883461, "epoch": 0.5731709349017778, "percentage": 57.32, "elapsed_time": "11:31:46", "remaining_time": "8:35:08"}
{"current_steps": 2160, "total_steps": 3751, "loss": 0.2273, "lr": 4.55147235112718e-06, "epoch": 0.5758368462269023, "percentage": 57.58, "elapsed_time": "11:34:03", "remaining_time": "8:31:13"}
{"current_steps": 2170, "total_steps": 3751, "loss": 0.2343, "lr": 4.505137981094675e-06, "epoch": 0.578502757552027, "percentage": 57.85, "elapsed_time": "11:36:23", "remaining_time": "8:27:22"}
{"current_steps": 2180, "total_steps": 3751, "loss": 0.2324, "lr": 4.458846488929663e-06, "epoch": 0.5811686688771515, "percentage": 58.12, "elapsed_time": "11:38:41", "remaining_time": "8:23:30"}
{"current_steps": 2190, "total_steps": 3751, "loss": 0.2345, "lr": 4.41260188560973e-06, "epoch": 0.5838345802022761, "percentage": 58.38, "elapsed_time": "11:40:58", "remaining_time": "8:19:38"}
{"current_steps": 2200, "total_steps": 3751, "loss": 0.2294, "lr": 4.366408178049728e-06, "epoch": 0.5865004915274006, "percentage": 58.65, "elapsed_time": "11:43:14", "remaining_time": "8:15:47"}
{"current_steps": 2200, "total_steps": 3751, "eval_loss": 0.23114009201526642, "epoch": 0.5865004915274006, "percentage": 58.65, "elapsed_time": "11:47:37", "remaining_time": "8:18:52"}
{"current_steps": 2210, "total_steps": 3751, "loss": 0.2276, "lr": 4.32026936875459e-06, "epoch": 0.5891664028525251, "percentage": 58.92, "elapsed_time": "11:49:56", "remaining_time": "8:15:01"}
{"current_steps": 2220, "total_steps": 3751, "loss": 0.2273, "lr": 4.274189455472529e-06, "epoch": 0.5918323141776497, "percentage": 59.18, "elapsed_time": "11:52:13", "remaining_time": "8:11:10"}
{"current_steps": 2230, "total_steps": 3751, "loss": 0.2288, "lr": 4.228172430848645e-06, "epoch": 0.5944982255027742, "percentage": 59.45, "elapsed_time": "11:54:29", "remaining_time": "8:07:19"}
{"current_steps": 2240, "total_steps": 3751, "loss": 0.2357, "lr": 4.182222282078983e-06, "epoch": 0.5971641368278988, "percentage": 59.72, "elapsed_time": "11:56:49", "remaining_time": "8:03:32"}
{"current_steps": 2250, "total_steps": 3751, "loss": 0.2403, "lr": 4.136342990565055e-06, "epoch": 0.5998300481530233, "percentage": 59.98, "elapsed_time": "11:59:07", "remaining_time": "7:59:44"}
{"current_steps": 2250, "total_steps": 3751, "eval_loss": 0.23227499425411224, "epoch": 0.5998300481530233, "percentage": 59.98, "elapsed_time": "12:03:30", "remaining_time": "8:02:39"}
{"current_steps": 2260, "total_steps": 3751, "loss": 0.2296, "lr": 4.090538531568867e-06, "epoch": 0.6024959594781478, "percentage": 60.25, "elapsed_time": "12:05:47", "remaining_time": "7:58:50"}
{"current_steps": 2270, "total_steps": 3751, "loss": 0.2328, "lr": 4.0448128738684775e-06, "epoch": 0.6051618708032724, "percentage": 60.52, "elapsed_time": "12:08:07", "remaining_time": "7:55:02"}
{"current_steps": 2280, "total_steps": 3751, "loss": 0.2325, "lr": 3.999169979414123e-06, "epoch": 0.6078277821283969, "percentage": 60.78, "elapsed_time": "12:10:25", "remaining_time": "7:51:15"}
{"current_steps": 2290, "total_steps": 3751, "loss": 0.2168, "lr": 3.9536138029849244e-06, "epoch": 0.6104936934535216, "percentage": 61.05, "elapsed_time": "12:12:45", "remaining_time": "7:47:29"}
{"current_steps": 2300, "total_steps": 3751, "loss": 0.2319, "lr": 3.908148291846225e-06, "epoch": 0.6131596047786461, "percentage": 61.32, "elapsed_time": "12:15:03", "remaining_time": "7:43:43"}
{"current_steps": 2300, "total_steps": 3751, "eval_loss": 0.22970426082611084, "epoch": 0.6131596047786461, "percentage": 61.32, "elapsed_time": "12:19:26", "remaining_time": "7:46:29"}
{"current_steps": 2310, "total_steps": 3751, "loss": 0.2341, "lr": 3.862777385407569e-06, "epoch": 0.6158255161037706, "percentage": 61.58, "elapsed_time": "12:21:41", "remaining_time": "7:42:40"}
{"current_steps": 2320, "total_steps": 3751, "loss": 0.2233, "lr": 3.817505014881378e-06, "epoch": 0.6184914274288952, "percentage": 61.85, "elapsed_time": "12:24:04", "remaining_time": "7:38:57"}
{"current_steps": 2330, "total_steps": 3751, "loss": 0.2267, "lr": 3.7723351029423143e-06, "epoch": 0.6211573387540197, "percentage": 62.12, "elapsed_time": "12:26:21", "remaining_time": "7:35:10"}
{"current_steps": 2340, "total_steps": 3751, "loss": 0.2217, "lr": 3.72727156338741e-06, "epoch": 0.6238232500791442, "percentage": 62.38, "elapsed_time": "12:28:40", "remaining_time": "7:31:26"}
{"current_steps": 2350, "total_steps": 3751, "loss": 0.222, "lr": 3.6823183007969375e-06, "epoch": 0.6264891614042688, "percentage": 62.65, "elapsed_time": "12:30:56", "remaining_time": "7:27:41"}
{"current_steps": 2350, "total_steps": 3751, "eval_loss": 0.22879666090011597, "epoch": 0.6264891614042688, "percentage": 62.65, "elapsed_time": "12:35:19", "remaining_time": "7:30:17"}
{"current_steps": 2360, "total_steps": 3751, "loss": 0.224, "lr": 3.637479210196102e-06, "epoch": 0.6291550727293933, "percentage": 62.92, "elapsed_time": "12:37:40", "remaining_time": "7:26:34"}
{"current_steps": 2370, "total_steps": 3751, "loss": 0.2324, "lr": 3.59275817671755e-06, "epoch": 0.6318209840545179, "percentage": 63.18, "elapsed_time": "12:39:57", "remaining_time": "7:22:49"}
{"current_steps": 2380, "total_steps": 3751, "loss": 0.2402, "lr": 3.548159075264738e-06, "epoch": 0.6344868953796424, "percentage": 63.45, "elapsed_time": "12:42:18", "remaining_time": "7:19:07"}
{"current_steps": 2390, "total_steps": 3751, "loss": 0.2296, "lr": 3.5036857701761857e-06, "epoch": 0.6371528067047669, "percentage": 63.72, "elapsed_time": "12:44:40", "remaining_time": "7:15:26"}
{"current_steps": 2400, "total_steps": 3751, "loss": 0.2193, "lr": 3.4593421148906523e-06, "epoch": 0.6398187180298915, "percentage": 63.98, "elapsed_time": "12:47:00", "remaining_time": "7:11:45"}
{"current_steps": 2400, "total_steps": 3751, "eval_loss": 0.2302982062101364, "epoch": 0.6398187180298915, "percentage": 63.98, "elapsed_time": "12:51:23", "remaining_time": "7:14:13"}
{"current_steps": 2410, "total_steps": 3751, "loss": 0.2241, "lr": 3.4151319516132414e-06, "epoch": 0.642484629355016, "percentage": 64.25, "elapsed_time": "12:53:45", "remaining_time": "7:10:32"}
{"current_steps": 2420, "total_steps": 3751, "loss": 0.228, "lr": 3.3710591109824954e-06, "epoch": 0.6451505406801407, "percentage": 64.52, "elapsed_time": "12:56:06", "remaining_time": "7:06:51"}
{"current_steps": 2430, "total_steps": 3751, "loss": 0.2248, "lr": 3.3271274117384834e-06, "epoch": 0.6478164520052652, "percentage": 64.78, "elapsed_time": "12:58:26", "remaining_time": "7:03:10"}
{"current_steps": 2440, "total_steps": 3751, "loss": 0.2255, "lr": 3.2833406603919243e-06, "epoch": 0.6504823633303897, "percentage": 65.05, "elapsed_time": "13:00:42", "remaining_time": "6:59:28"}
{"current_steps": 2450, "total_steps": 3751, "loss": 0.2252, "lr": 3.239702650894364e-06, "epoch": 0.6531482746555143, "percentage": 65.32, "elapsed_time": "13:03:00", "remaining_time": "6:55:47"}
{"current_steps": 2450, "total_steps": 3751, "eval_loss": 0.22466669976711273, "epoch": 0.6531482746555143, "percentage": 65.32, "elapsed_time": "13:07:23", "remaining_time": "6:58:07"}
{"current_steps": 2460, "total_steps": 3751, "loss": 0.2262, "lr": 3.1962171643094474e-06, "epoch": 0.6558141859806388, "percentage": 65.58, "elapsed_time": "13:09:41", "remaining_time": "6:54:25"}
{"current_steps": 2470, "total_steps": 3751, "loss": 0.2268, "lr": 3.152887968485303e-06, "epoch": 0.6584800973057634, "percentage": 65.85, "elapsed_time": "13:12:00", "remaining_time": "6:50:45"}
{"current_steps": 2480, "total_steps": 3751, "loss": 0.2307, "lr": 3.1097188177280735e-06, "epoch": 0.6611460086308879, "percentage": 66.12, "elapsed_time": "13:14:18", "remaining_time": "6:47:04"}
{"current_steps": 2490, "total_steps": 3751, "loss": 0.2306, "lr": 3.0667134524766173e-06, "epoch": 0.6638119199560124, "percentage": 66.38, "elapsed_time": "13:16:39", "remaining_time": "6:43:26"}
{"current_steps": 2500, "total_steps": 3751, "loss": 0.2304, "lr": 3.023875598978419e-06, "epoch": 0.666477831281137, "percentage": 66.65, "elapsed_time": "13:18:59", "remaining_time": "6:39:48"}
{"current_steps": 2500, "total_steps": 3751, "eval_loss": 0.22111880779266357, "epoch": 0.666477831281137, "percentage": 66.65, "elapsed_time": "13:23:22", "remaining_time": "6:42:00"}
{"current_steps": 2510, "total_steps": 3751, "loss": 0.2308, "lr": 2.981208968966721e-06, "epoch": 0.6691437426062615, "percentage": 66.92, "elapsed_time": "13:27:00", "remaining_time": "6:38:59"}
{"current_steps": 2520, "total_steps": 3751, "loss": 0.2279, "lr": 2.9387172593389148e-06, "epoch": 0.6718096539313861, "percentage": 67.18, "elapsed_time": "13:29:18", "remaining_time": "6:35:20"}
{"current_steps": 2530, "total_steps": 3751, "loss": 0.2269, "lr": 2.896404151836227e-06, "epoch": 0.6744755652565106, "percentage": 67.45, "elapsed_time": "13:31:39", "remaining_time": "6:31:43"}
{"current_steps": 2540, "total_steps": 3751, "loss": 0.2302, "lr": 2.8542733127247024e-06, "epoch": 0.6771414765816351, "percentage": 67.72, "elapsed_time": "13:33:58", "remaining_time": "6:28:04"}
{"current_steps": 2550, "total_steps": 3751, "loss": 0.2139, "lr": 2.812328392477536e-06, "epoch": 0.6798073879067598, "percentage": 67.98, "elapsed_time": "13:36:16", "remaining_time": "6:24:27"}
{"current_steps": 2550, "total_steps": 3751, "eval_loss": 0.2199471890926361, "epoch": 0.6798073879067598, "percentage": 67.98, "elapsed_time": "13:40:39", "remaining_time": "6:26:31"}
{"current_steps": 2560, "total_steps": 3751, "loss": 0.2237, "lr": 2.7705730254587802e-06, "epoch": 0.6824732992318843, "percentage": 68.25, "elapsed_time": "13:43:00", "remaining_time": "6:22:53"}
{"current_steps": 2570, "total_steps": 3751, "loss": 0.2166, "lr": 2.729010829608442e-06, "epoch": 0.6851392105570089, "percentage": 68.52, "elapsed_time": "13:45:17", "remaining_time": "6:19:14"}
{"current_steps": 2580, "total_steps": 3751, "loss": 0.2291, "lr": 2.6876454061289892e-06, "epoch": 0.6878051218821334, "percentage": 68.78, "elapsed_time": "13:47:36", "remaining_time": "6:15:37"}
{"current_steps": 2590, "total_steps": 3751, "loss": 0.2226, "lr": 2.646480339173337e-06, "epoch": 0.6904710332072579, "percentage": 69.05, "elapsed_time": "13:49:59", "remaining_time": "6:12:03"}
{"current_steps": 2600, "total_steps": 3751, "loss": 0.2186, "lr": 2.6055191955342886e-06, "epoch": 0.6931369445323825, "percentage": 69.31, "elapsed_time": "13:52:24", "remaining_time": "6:08:30"}
{"current_steps": 2600, "total_steps": 3751, "eval_loss": 0.219247967004776, "epoch": 0.6931369445323825, "percentage": 69.31, "elapsed_time": "13:56:47", "remaining_time": "6:10:26"}
{"current_steps": 2610, "total_steps": 3751, "loss": 0.2249, "lr": 2.564765524335478e-06, "epoch": 0.695802855857507, "percentage": 69.58, "elapsed_time": "13:59:05", "remaining_time": "6:06:49"}
{"current_steps": 2620, "total_steps": 3751, "loss": 0.2244, "lr": 2.524222856723869e-06, "epoch": 0.6984687671826316, "percentage": 69.85, "elapsed_time": "14:01:27", "remaining_time": "6:03:14"}
{"current_steps": 2630, "total_steps": 3751, "loss": 0.2168, "lr": 2.483894705563778e-06, "epoch": 0.7011346785077561, "percentage": 70.11, "elapsed_time": "14:03:45", "remaining_time": "5:59:38"}
{"current_steps": 2640, "total_steps": 3751, "loss": 0.2175, "lr": 2.4437845651325116e-06, "epoch": 0.7038005898328806, "percentage": 70.38, "elapsed_time": "14:06:05", "remaining_time": "5:56:03"}
{"current_steps": 2650, "total_steps": 3751, "loss": 0.2156, "lr": 2.403895910817593e-06, "epoch": 0.7064665011580052, "percentage": 70.65, "elapsed_time": "14:08:29", "remaining_time": "5:52:31"}
{"current_steps": 2650, "total_steps": 3751, "eval_loss": 0.21830175817012787, "epoch": 0.7064665011580052, "percentage": 70.65, "elapsed_time": "14:12:52", "remaining_time": "5:54:20"}
{"current_steps": 2660, "total_steps": 3751, "loss": 0.2111, "lr": 2.364232198815638e-06, "epoch": 0.7091324124831297, "percentage": 70.91, "elapsed_time": "14:15:09", "remaining_time": "5:50:44"}
{"current_steps": 2670, "total_steps": 3751, "loss": 0.2224, "lr": 2.3247968658328825e-06, "epoch": 0.7117983238082544, "percentage": 71.18, "elapsed_time": "14:17:28", "remaining_time": "5:47:09"}
{"current_steps": 2680, "total_steps": 3751, "loss": 0.213, "lr": 2.285593328787414e-06, "epoch": 0.7144642351333789, "percentage": 71.45, "elapsed_time": "14:19:47", "remaining_time": "5:43:35"}
{"current_steps": 2690, "total_steps": 3751, "loss": 0.2283, "lr": 2.246624984513099e-06, "epoch": 0.7171301464585034, "percentage": 71.71, "elapsed_time": "14:22:03", "remaining_time": "5:40:00"}
{"current_steps": 2700, "total_steps": 3751, "loss": 0.2187, "lr": 2.2078952094652705e-06, "epoch": 0.719796057783628, "percentage": 71.98, "elapsed_time": "14:24:19", "remaining_time": "5:36:26"}
{"current_steps": 2700, "total_steps": 3751, "eval_loss": 0.21585261821746826, "epoch": 0.719796057783628, "percentage": 71.98, "elapsed_time": "14:28:42", "remaining_time": "5:38:09"}
{"current_steps": 2710, "total_steps": 3751, "loss": 0.2221, "lr": 2.1694073594281663e-06, "epoch": 0.7224619691087525, "percentage": 72.25, "elapsed_time": "14:31:02", "remaining_time": "5:34:35"}
{"current_steps": 2720, "total_steps": 3751, "loss": 0.2192, "lr": 2.131164769224164e-06, "epoch": 0.7251278804338771, "percentage": 72.51, "elapsed_time": "14:33:19", "remaining_time": "5:31:01"}
{"current_steps": 2730, "total_steps": 3751, "loss": 0.2249, "lr": 2.0931707524248268e-06, "epoch": 0.7277937917590016, "percentage": 72.78, "elapsed_time": "14:35:36", "remaining_time": "5:27:28"}
{"current_steps": 2740, "total_steps": 3751, "loss": 0.2112, "lr": 2.0554286010638076e-06, "epoch": 0.7304597030841261, "percentage": 73.05, "elapsed_time": "14:37:56", "remaining_time": "5:23:56"}
{"current_steps": 2750, "total_steps": 3751, "loss": 0.222, "lr": 2.017941585351591e-06, "epoch": 0.7331256144092507, "percentage": 73.31, "elapsed_time": "14:40:14", "remaining_time": "5:20:24"}
{"current_steps": 2750, "total_steps": 3751, "eval_loss": 0.21738137304782867, "epoch": 0.7331256144092507, "percentage": 73.31, "elapsed_time": "14:44:37", "remaining_time": "5:22:00"}
{"current_steps": 2760, "total_steps": 3751, "loss": 0.2091, "lr": 1.98071295339216e-06, "epoch": 0.7357915257343752, "percentage": 73.58, "elapsed_time": "14:47:00", "remaining_time": "5:18:29"}
{"current_steps": 2770, "total_steps": 3751, "loss": 0.2133, "lr": 1.9437459309015426e-06, "epoch": 0.7384574370594998, "percentage": 73.85, "elapsed_time": "14:49:22", "remaining_time": "5:14:58"}
{"current_steps": 2780, "total_steps": 3751, "loss": 0.2102, "lr": 1.9070437209283304e-06, "epoch": 0.7411233483846243, "percentage": 74.11, "elapsed_time": "14:51:43", "remaining_time": "5:11:27"}
{"current_steps": 2790, "total_steps": 3751, "loss": 0.2138, "lr": 1.8706095035761418e-06, "epoch": 0.7437892597097489, "percentage": 74.38, "elapsed_time": "14:54:02", "remaining_time": "5:07:56"}
{"current_steps": 2800, "total_steps": 3751, "loss": 0.2162, "lr": 1.8344464357280722e-06, "epoch": 0.7464551710348735, "percentage": 74.65, "elapsed_time": "14:56:27", "remaining_time": "5:04:28"}
{"current_steps": 2800, "total_steps": 3751, "eval_loss": 0.21530871093273163, "epoch": 0.7464551710348735, "percentage": 74.65, "elapsed_time": "15:00:50", "remaining_time": "5:05:57"}
{"current_steps": 2810, "total_steps": 3751, "loss": 0.2245, "lr": 1.7985576507731744e-06, "epoch": 0.749121082359998, "percentage": 74.91, "elapsed_time": "15:03:08", "remaining_time": "5:02:26"}
{"current_steps": 2820, "total_steps": 3751, "loss": 0.2196, "lr": 1.762946258334951e-06, "epoch": 0.7517869936851226, "percentage": 75.18, "elapsed_time": "15:05:25", "remaining_time": "4:58:55"}
{"current_steps": 2830, "total_steps": 3751, "loss": 0.2119, "lr": 1.727615344001926e-06, "epoch": 0.7544529050102471, "percentage": 75.45, "elapsed_time": "15:07:44", "remaining_time": "4:55:25"}
{"current_steps": 2840, "total_steps": 3751, "loss": 0.2106, "lr": 1.6925679690602876e-06, "epoch": 0.7571188163353716, "percentage": 75.71, "elapsed_time": "15:10:02", "remaining_time": "4:51:55"}
{"current_steps": 2850, "total_steps": 3751, "loss": 0.2253, "lr": 1.6578071702286396e-06, "epoch": 0.7597847276604962, "percentage": 75.98, "elapsed_time": "15:12:25", "remaining_time": "4:48:27"}
{"current_steps": 2850, "total_steps": 3751, "eval_loss": 0.21321707963943481, "epoch": 0.7597847276604962, "percentage": 75.98, "elapsed_time": "15:16:49", "remaining_time": "4:49:50"}
{"current_steps": 2860, "total_steps": 3751, "loss": 0.22, "lr": 1.6233359593948777e-06, "epoch": 0.7624506389856207, "percentage": 76.25, "elapsed_time": "15:19:06", "remaining_time": "4:46:20"}
{"current_steps": 2870, "total_steps": 3751, "loss": 0.2133, "lr": 1.5891573233552315e-06, "epoch": 0.7651165503107453, "percentage": 76.51, "elapsed_time": "15:21:22", "remaining_time": "4:42:50"}
{"current_steps": 2880, "total_steps": 3751, "loss": 0.2161, "lr": 1.5552742235554551e-06, "epoch": 0.7677824616358698, "percentage": 76.78, "elapsed_time": "15:23:39", "remaining_time": "4:39:20"}
{"current_steps": 2890, "total_steps": 3751, "loss": 0.2129, "lr": 1.521689595834246e-06, "epoch": 0.7704483729609943, "percentage": 77.05, "elapsed_time": "15:25:57", "remaining_time": "4:35:51"}
{"current_steps": 2900, "total_steps": 3751, "loss": 0.2066, "lr": 1.4884063501688539e-06, "epoch": 0.773114284286119, "percentage": 77.31, "elapsed_time": "15:28:19", "remaining_time": "4:32:24"}
{"current_steps": 2900, "total_steps": 3751, "eval_loss": 0.21342259645462036, "epoch": 0.773114284286119, "percentage": 77.31, "elapsed_time": "15:32:42", "remaining_time": "4:33:42"}
{"current_steps": 2910, "total_steps": 3751, "loss": 0.2244, "lr": 1.4554273704229494e-06, "epoch": 0.7757801956112435, "percentage": 77.58, "elapsed_time": "15:35:02", "remaining_time": "4:30:13"}
{"current_steps": 2920, "total_steps": 3751, "loss": 0.2156, "lr": 1.4227555140967402e-06, "epoch": 0.7784461069363681, "percentage": 77.85, "elapsed_time": "15:37:17", "remaining_time": "4:26:44"}
{"current_steps": 2930, "total_steps": 3751, "loss": 0.2192, "lr": 1.3903936120793926e-06, "epoch": 0.7811120182614926, "percentage": 78.11, "elapsed_time": "15:39:38", "remaining_time": "4:23:17"}
{"current_steps": 2940, "total_steps": 3751, "loss": 0.2156, "lr": 1.3583444684037312e-06, "epoch": 0.7837779295866171, "percentage": 78.38, "elapsed_time": "15:41:59", "remaining_time": "4:19:50"}
{"current_steps": 2950, "total_steps": 3751, "loss": 0.2113, "lr": 1.3266108600032928e-06, "epoch": 0.7864438409117417, "percentage": 78.65, "elapsed_time": "15:44:17", "remaining_time": "4:16:23"}
{"current_steps": 2950, "total_steps": 3751, "eval_loss": 0.2107125222682953, "epoch": 0.7864438409117417, "percentage": 78.65, "elapsed_time": "15:48:40", "remaining_time": "4:17:35"}
{"current_steps": 2960, "total_steps": 3751, "loss": 0.2143, "lr": 1.2951955364717116e-06, "epoch": 0.7891097522368662, "percentage": 78.91, "elapsed_time": "15:50:59", "remaining_time": "4:14:08"}
{"current_steps": 2970, "total_steps": 3751, "loss": 0.2065, "lr": 1.2641012198244718e-06, "epoch": 0.7917756635619908, "percentage": 79.18, "elapsed_time": "15:53:18", "remaining_time": "4:10:41"}
{"current_steps": 2980, "total_steps": 3751, "loss": 0.2201, "lr": 1.2333306042630672e-06, "epoch": 0.7944415748871153, "percentage": 79.45, "elapsed_time": "15:55:38", "remaining_time": "4:07:14"}
{"current_steps": 2990, "total_steps": 3751, "loss": 0.2097, "lr": 1.202886355941546e-06, "epoch": 0.7971074862122398, "percentage": 79.71, "elapsed_time": "15:57:58", "remaining_time": "4:03:49"}
{"current_steps": 3000, "total_steps": 3751, "loss": 0.2107, "lr": 1.1727711127355118e-06, "epoch": 0.7997733975373644, "percentage": 79.98, "elapsed_time": "16:00:17", "remaining_time": "4:00:23"}
{"current_steps": 3000, "total_steps": 3751, "eval_loss": 0.20849083364009857, "epoch": 0.7997733975373644, "percentage": 79.98, "elapsed_time": "16:04:39", "remaining_time": "4:01:29"}
{"current_steps": 3010, "total_steps": 3751, "loss": 0.2126, "lr": 1.1429874840135492e-06, "epoch": 0.8024393088624889, "percentage": 80.25, "elapsed_time": "16:08:09", "remaining_time": "3:58:20"}
{"current_steps": 3020, "total_steps": 3751, "loss": 0.2024, "lr": 1.1135380504111476e-06, "epoch": 0.8051052201876135, "percentage": 80.51, "elapsed_time": "16:10:28", "remaining_time": "3:54:54"}
{"current_steps": 3030, "total_steps": 3751, "loss": 0.2103, "lr": 1.0844253636070805e-06, "epoch": 0.807771131512738, "percentage": 80.78, "elapsed_time": "16:12:43", "remaining_time": "3:51:27"}
{"current_steps": 3040, "total_steps": 3751, "loss": 0.2093, "lr": 1.0556519461023301e-06, "epoch": 0.8104370428378626, "percentage": 81.05, "elapsed_time": "16:15:02", "remaining_time": "3:48:02"}
{"current_steps": 3050, "total_steps": 3751, "loss": 0.2055, "lr": 1.0272202910015083e-06, "epoch": 0.8131029541629872, "percentage": 81.31, "elapsed_time": "16:17:22", "remaining_time": "3:44:38"}
{"current_steps": 3050, "total_steps": 3751, "eval_loss": 0.2096840888261795, "epoch": 0.8131029541629872, "percentage": 81.31, "elapsed_time": "16:21:45", "remaining_time": "3:45:38"}
{"current_steps": 3060, "total_steps": 3751, "loss": 0.2084, "lr": 9.991328617968509e-07, "epoch": 0.8157688654881117, "percentage": 81.58, "elapsed_time": "16:24:03", "remaining_time": "3:42:12"}
{"current_steps": 3070, "total_steps": 3751, "loss": 0.2076, "lr": 9.713920921547532e-07, "epoch": 0.8184347768132363, "percentage": 81.84, "elapsed_time": "16:26:19", "remaining_time": "3:38:47"}
{"current_steps": 3080, "total_steps": 3751, "loss": 0.2072, "lr": 9.440003857049173e-07, "epoch": 0.8211006881383608, "percentage": 82.11, "elapsed_time": "16:28:39", "remaining_time": "3:35:23"}
{"current_steps": 3090, "total_steps": 3751, "loss": 0.2252, "lr": 9.169601158320707e-07, "epoch": 0.8237665994634853, "percentage": 82.38, "elapsed_time": "16:30:56", "remaining_time": "3:31:58"}
{"current_steps": 3100, "total_steps": 3751, "loss": 0.2045, "lr": 8.902736254703347e-07, "epoch": 0.8264325107886099, "percentage": 82.64, "elapsed_time": "16:33:16", "remaining_time": "3:28:35"}
{"current_steps": 3100, "total_steps": 3751, "eval_loss": 0.20747578144073486, "epoch": 0.8264325107886099, "percentage": 82.64, "elapsed_time": "16:37:39", "remaining_time": "3:29:30"}
{"current_steps": 3110, "total_steps": 3751, "loss": 0.2038, "lr": 8.639432269002102e-07, "epoch": 0.8290984221137344, "percentage": 82.91, "elapsed_time": "16:39:57", "remaining_time": "3:26:05"}
{"current_steps": 3120, "total_steps": 3751, "loss": 0.2071, "lr": 8.379712015482333e-07, "epoch": 0.831764333438859, "percentage": 83.18, "elapsed_time": "16:42:16", "remaining_time": "3:22:42"}
{"current_steps": 3130, "total_steps": 3751, "loss": 0.1988, "lr": 8.123597997892918e-07, "epoch": 0.8344302447639835, "percentage": 83.44, "elapsed_time": "16:44:37", "remaining_time": "3:19:19"}
{"current_steps": 3140, "total_steps": 3751, "loss": 0.2181, "lr": 7.871112407516474e-07, "epoch": 0.837096156089108, "percentage": 83.71, "elapsed_time": "16:46:53", "remaining_time": "3:15:55"}
{"current_steps": 3150, "total_steps": 3751, "loss": 0.2172, "lr": 7.622277121246513e-07, "epoch": 0.8397620674142326, "percentage": 83.98, "elapsed_time": "16:49:09", "remaining_time": "3:12:32"}
{"current_steps": 3150, "total_steps": 3751, "eval_loss": 0.2062389850616455, "epoch": 0.8397620674142326, "percentage": 83.98, "elapsed_time": "16:53:32", "remaining_time": "3:13:22"}
{"current_steps": 3160, "total_steps": 3751, "loss": 0.2045, "lr": 7.377113699691879e-07, "epoch": 0.8424279787393572, "percentage": 84.24, "elapsed_time": "16:55:52", "remaining_time": "3:09:59"}
{"current_steps": 3170, "total_steps": 3751, "loss": 0.2162, "lr": 7.135643385308677e-07, "epoch": 0.8450938900644818, "percentage": 84.51, "elapsed_time": "16:58:13", "remaining_time": "3:06:37"}
{"current_steps": 3180, "total_steps": 3751, "loss": 0.2089, "lr": 6.897887100559608e-07, "epoch": 0.8477598013896063, "percentage": 84.78, "elapsed_time": "17:00:32", "remaining_time": "3:03:14"}
{"current_steps": 3190, "total_steps": 3751, "loss": 0.216, "lr": 6.663865446101192e-07, "epoch": 0.8504257127147308, "percentage": 85.04, "elapsed_time": "17:02:49", "remaining_time": "2:59:52"}
{"current_steps": 3200, "total_steps": 3751, "loss": 0.2138, "lr": 6.433598698998766e-07, "epoch": 0.8530916240398554, "percentage": 85.31, "elapsed_time": "17:05:06", "remaining_time": "2:56:30"}
{"current_steps": 3200, "total_steps": 3751, "eval_loss": 0.2074633538722992, "epoch": 0.8530916240398554, "percentage": 85.31, "elapsed_time": "17:09:29", "remaining_time": "2:57:15"}
{"current_steps": 3210, "total_steps": 3751, "loss": 0.2121, "lr": 6.207106810969577e-07, "epoch": 0.8557575353649799, "percentage": 85.58, "elapsed_time": "17:11:45", "remaining_time": "2:53:53"}
{"current_steps": 3220, "total_steps": 3751, "loss": 0.2141, "lr": 5.98440940665399e-07, "epoch": 0.8584234466901045, "percentage": 85.84, "elapsed_time": "17:14:03", "remaining_time": "2:50:31"}
{"current_steps": 3230, "total_steps": 3751, "loss": 0.1917, "lr": 5.765525781915172e-07, "epoch": 0.861089358015229, "percentage": 86.11, "elapsed_time": "17:16:20", "remaining_time": "2:47:09"}
{"current_steps": 3240, "total_steps": 3751, "loss": 0.213, "lr": 5.550474902167091e-07, "epoch": 0.8637552693403535, "percentage": 86.38, "elapsed_time": "17:18:37", "remaining_time": "2:43:48"}
{"current_steps": 3250, "total_steps": 3751, "loss": 0.194, "lr": 5.339275400731331e-07, "epoch": 0.8664211806654781, "percentage": 86.64, "elapsed_time": "17:20:55", "remaining_time": "2:40:27"}
{"current_steps": 3250, "total_steps": 3751, "eval_loss": 0.2050597220659256, "epoch": 0.8664211806654781, "percentage": 86.64, "elapsed_time": "17:25:18", "remaining_time": "2:41:08"}
{"current_steps": 3260, "total_steps": 3751, "loss": 0.2031, "lr": 5.131945577222485e-07, "epoch": 0.8690870919906026, "percentage": 86.91, "elapsed_time": "17:27:33", "remaining_time": "2:37:46"}
{"current_steps": 3270, "total_steps": 3751, "loss": 0.2061, "lr": 4.92850339596268e-07, "epoch": 0.8717530033157272, "percentage": 87.18, "elapsed_time": "17:29:51", "remaining_time": "2:34:25"}
{"current_steps": 3280, "total_steps": 3751, "loss": 0.2148, "lr": 4.728966484424913e-07, "epoch": 0.8744189146408518, "percentage": 87.44, "elapsed_time": "17:32:13", "remaining_time": "2:31:05"}
{"current_steps": 3290, "total_steps": 3751, "loss": 0.2181, "lr": 4.5333521317058207e-07, "epoch": 0.8770848259659763, "percentage": 87.71, "elapsed_time": "17:34:33", "remaining_time": "2:27:45"}
{"current_steps": 3300, "total_steps": 3751, "loss": 0.2133, "lr": 4.3416772870275295e-07, "epoch": 0.8797507372911009, "percentage": 87.98, "elapsed_time": "17:36:51", "remaining_time": "2:24:26"}
{"current_steps": 3300, "total_steps": 3751, "eval_loss": 0.20511919260025024, "epoch": 0.8797507372911009, "percentage": 87.98, "elapsed_time": "17:41:15", "remaining_time": "2:25:02"}
{"current_steps": 3310, "total_steps": 3751, "loss": 0.1987, "lr": 4.153958558269189e-07, "epoch": 0.8824166486162254, "percentage": 88.24, "elapsed_time": "17:43:39", "remaining_time": "2:21:42"}
{"current_steps": 3320, "total_steps": 3751, "loss": 0.2, "lr": 3.9702122105278405e-07, "epoch": 0.88508255994135, "percentage": 88.51, "elapsed_time": "17:46:05", "remaining_time": "2:18:23"}
{"current_steps": 3330, "total_steps": 3751, "loss": 0.1995, "lr": 3.7904541647092506e-07, "epoch": 0.8877484712664745, "percentage": 88.78, "elapsed_time": "17:48:24", "remaining_time": "2:15:04"}
{"current_steps": 3340, "total_steps": 3751, "loss": 0.1988, "lr": 3.614699996148285e-07, "epoch": 0.890414382591599, "percentage": 89.04, "elapsed_time": "17:50:41", "remaining_time": "2:11:45"}
{"current_steps": 3350, "total_steps": 3751, "loss": 0.2025, "lr": 3.442964933259474e-07, "epoch": 0.8930802939167236, "percentage": 89.31, "elapsed_time": "17:53:01", "remaining_time": "2:08:26"}
{"current_steps": 3350, "total_steps": 3751, "eval_loss": 0.20472171902656555, "epoch": 0.8930802939167236, "percentage": 89.31, "elapsed_time": "17:57:24", "remaining_time": "2:08:58"}
{"current_steps": 3360, "total_steps": 3751, "loss": 0.2066, "lr": 3.275263856217442e-07, "epoch": 0.8957462052418481, "percentage": 89.58, "elapsed_time": "17:59:41", "remaining_time": "2:05:38"}
{"current_steps": 3370, "total_steps": 3751, "loss": 0.2041, "lr": 3.1116112956677045e-07, "epoch": 0.8984121165669727, "percentage": 89.84, "elapsed_time": "18:02:00", "remaining_time": "2:02:19"}
{"current_steps": 3380, "total_steps": 3751, "loss": 0.2075, "lr": 2.952021431467522e-07, "epoch": 0.9010780278920972, "percentage": 90.11, "elapsed_time": "18:04:19", "remaining_time": "1:59:01"}
{"current_steps": 3390, "total_steps": 3751, "loss": 0.2102, "lr": 2.7965080914573786e-07, "epoch": 0.9037439392172217, "percentage": 90.38, "elapsed_time": "18:06:35", "remaining_time": "1:55:42"}
{"current_steps": 3400, "total_steps": 3751, "loss": 0.2088, "lr": 2.6450847502627883e-07, "epoch": 0.9064098505423464, "percentage": 90.64, "elapsed_time": "18:08:55", "remaining_time": "1:52:24"}
{"current_steps": 3400, "total_steps": 3751, "eval_loss": 0.20498156547546387, "epoch": 0.9064098505423464, "percentage": 90.64, "elapsed_time": "18:13:18", "remaining_time": "1:52:52"}
{"current_steps": 3410, "total_steps": 3751, "loss": 0.2127, "lr": 2.497764528126778e-07, "epoch": 0.9090757618674709, "percentage": 90.91, "elapsed_time": "18:15:37", "remaining_time": "1:49:33"}
{"current_steps": 3420, "total_steps": 3751, "loss": 0.2087, "lr": 2.3545601897731085e-07, "epoch": 0.9117416731925955, "percentage": 91.18, "elapsed_time": "18:17:56", "remaining_time": "1:46:15"}
{"current_steps": 3430, "total_steps": 3751, "loss": 0.208, "lr": 2.2154841433002062e-07, "epoch": 0.91440758451772, "percentage": 91.44, "elapsed_time": "18:20:11", "remaining_time": "1:42:57"}
{"current_steps": 3440, "total_steps": 3751, "loss": 0.197, "lr": 2.0805484391061003e-07, "epoch": 0.9170734958428445, "percentage": 91.71, "elapsed_time": "18:22:30", "remaining_time": "1:39:40"}
{"current_steps": 3450, "total_steps": 3751, "loss": 0.204, "lr": 1.9497647688442478e-07, "epoch": 0.9197394071679691, "percentage": 91.98, "elapsed_time": "18:24:49", "remaining_time": "1:36:23"}
{"current_steps": 3450, "total_steps": 3751, "eval_loss": 0.20443060994148254, "epoch": 0.9197394071679691, "percentage": 91.98, "elapsed_time": "18:29:12", "remaining_time": "1:36:46"}
{"current_steps": 3460, "total_steps": 3751, "loss": 0.2063, "lr": 1.8231444644105755e-07, "epoch": 0.9224053184930936, "percentage": 92.24, "elapsed_time": "18:31:30", "remaining_time": "1:33:28"}
{"current_steps": 3470, "total_steps": 3751, "loss": 0.2063, "lr": 1.7006984969615226e-07, "epoch": 0.9250712298182182, "percentage": 92.51, "elapsed_time": "18:33:48", "remaining_time": "1:30:11"}
{"current_steps": 3480, "total_steps": 3751, "loss": 0.2103, "lr": 1.5824374759635165e-07, "epoch": 0.9277371411433427, "percentage": 92.78, "elapsed_time": "18:36:09", "remaining_time": "1:26:55"}
{"current_steps": 3490, "total_steps": 3751, "loss": 0.1999, "lr": 1.4683716482736364e-07, "epoch": 0.9304030524684672, "percentage": 93.04, "elapsed_time": "18:38:26", "remaining_time": "1:23:38"}
{"current_steps": 3500, "total_steps": 3751, "loss": 0.2059, "lr": 1.358510897251808e-07, "epoch": 0.9330689637935918, "percentage": 93.31, "elapsed_time": "18:40:44", "remaining_time": "1:20:22"}
{"current_steps": 3500, "total_steps": 3751, "eval_loss": 0.20389488339424133, "epoch": 0.9330689637935918, "percentage": 93.31, "elapsed_time": "18:45:07", "remaining_time": "1:20:41"}
{"current_steps": 3510, "total_steps": 3751, "loss": 0.2041, "lr": 1.252864741904425e-07, "epoch": 0.9357348751187163, "percentage": 93.58, "elapsed_time": "18:48:47", "remaining_time": "1:17:30"}
{"current_steps": 3520, "total_steps": 3751, "loss": 0.2071, "lr": 1.1514423360595939e-07, "epoch": 0.938400786443841, "percentage": 93.84, "elapsed_time": "18:51:03", "remaining_time": "1:14:13"}
{"current_steps": 3530, "total_steps": 3751, "loss": 0.197, "lr": 1.0542524675739407e-07, "epoch": 0.9410666977689655, "percentage": 94.11, "elapsed_time": "18:53:21", "remaining_time": "1:10:57"}
{"current_steps": 3540, "total_steps": 3751, "loss": 0.2051, "lr": 9.613035575712304e-08, "epoch": 0.94373260909409, "percentage": 94.37, "elapsed_time": "18:55:40", "remaining_time": "1:07:41"}
{"current_steps": 3550, "total_steps": 3751, "loss": 0.2103, "lr": 8.726036597126619e-08, "epoch": 0.9463985204192146, "percentage": 94.64, "elapsed_time": "18:57:59", "remaining_time": "1:04:25"}
{"current_steps": 3550, "total_steps": 3751, "eval_loss": 0.20391830801963806, "epoch": 0.9463985204192146, "percentage": 94.64, "elapsed_time": "19:02:22", "remaining_time": "1:04:40"}
{"current_steps": 3560, "total_steps": 3751, "loss": 0.2028, "lr": 7.881604594990932e-08, "epoch": 0.9490644317443391, "percentage": 94.91, "elapsed_time": "19:04:43", "remaining_time": "1:01:24"}
{"current_steps": 3570, "total_steps": 3751, "loss": 0.2141, "lr": 7.079812736050906e-08, "epoch": 0.9517303430694637, "percentage": 95.17, "elapsed_time": "19:07:03", "remaining_time": "0:58:09"}
{"current_steps": 3580, "total_steps": 3751, "loss": 0.2008, "lr": 6.3207304924498e-08, "epoch": 0.9543962543945882, "percentage": 95.44, "elapsed_time": "19:09:21", "remaining_time": "0:54:53"}
{"current_steps": 3590, "total_steps": 3751, "loss": 0.2088, "lr": 5.604423635709122e-08, "epoch": 0.9570621657197127, "percentage": 95.71, "elapsed_time": "19:11:45", "remaining_time": "0:51:39"}
{"current_steps": 3600, "total_steps": 3751, "loss": 0.2102, "lr": 4.93095423102935e-08, "epoch": 0.9597280770448373, "percentage": 95.97, "elapsed_time": "19:13:59", "remaining_time": "0:48:24"}
{"current_steps": 3600, "total_steps": 3751, "eval_loss": 0.20385359227657318, "epoch": 0.9597280770448373, "percentage": 95.97, "elapsed_time": "19:18:22", "remaining_time": "0:48:35"}
{"current_steps": 3610, "total_steps": 3751, "loss": 0.1974, "lr": 4.3003806319127376e-08, "epoch": 0.9623939883699618, "percentage": 96.24, "elapsed_time": "19:20:46", "remaining_time": "0:45:20"}
{"current_steps": 3620, "total_steps": 3751, "loss": 0.2038, "lr": 3.712757475106854e-08, "epoch": 0.9650598996950864, "percentage": 96.51, "elapsed_time": "19:23:05", "remaining_time": "0:42:05"}
{"current_steps": 3630, "total_steps": 3751, "loss": 0.2033, "lr": 3.168135675870654e-08, "epoch": 0.9677258110202109, "percentage": 96.77, "elapsed_time": "19:25:22", "remaining_time": "0:38:50"}
{"current_steps": 3640, "total_steps": 3751, "loss": 0.2066, "lr": 2.6665624235629463e-08, "epoch": 0.9703917223453354, "percentage": 97.04, "elapsed_time": "19:27:40", "remaining_time": "0:35:36"}
{"current_steps": 3650, "total_steps": 3751, "loss": 0.2051, "lr": 2.2080811775535006e-08, "epoch": 0.97305763367046, "percentage": 97.31, "elapsed_time": "19:29:56", "remaining_time": "0:32:22"}
{"current_steps": 3650, "total_steps": 3751, "eval_loss": 0.20379623770713806, "epoch": 0.97305763367046, "percentage": 97.31, "elapsed_time": "19:34:19", "remaining_time": "0:32:29"}
{"current_steps": 3660, "total_steps": 3751, "loss": 0.2005, "lr": 1.7927316634573368e-08, "epoch": 0.9757235449955846, "percentage": 97.57, "elapsed_time": "19:36:36", "remaining_time": "0:29:15"}
{"current_steps": 3670, "total_steps": 3751, "loss": 0.198, "lr": 1.4205498696930332e-08, "epoch": 0.9783894563207092, "percentage": 97.84, "elapsed_time": "19:38:53", "remaining_time": "0:26:01"}
{"current_steps": 3680, "total_steps": 3751, "loss": 0.1995, "lr": 1.0915680443641663e-08, "epoch": 0.9810553676458337, "percentage": 98.11, "elapsed_time": "19:41:11", "remaining_time": "0:22:47"}
{"current_steps": 3690, "total_steps": 3751, "loss": 0.2048, "lr": 8.058146924651012e-09, "epoch": 0.9837212789709582, "percentage": 98.37, "elapsed_time": "19:43:31", "remaining_time": "0:19:33"}
{"current_steps": 3700, "total_steps": 3751, "loss": 0.2017, "lr": 5.633145734114665e-09, "epoch": 0.9863871902960828, "percentage": 98.64, "elapsed_time": "19:45:49", "remaining_time": "0:16:20"}
{"current_steps": 3700, "total_steps": 3751, "eval_loss": 0.203701451420784, "epoch": 0.9863871902960828, "percentage": 98.64, "elapsed_time": "19:50:12", "remaining_time": "0:16:24"}
{"current_steps": 3710, "total_steps": 3751, "loss": 0.2069, "lr": 3.640886988945935e-09, "epoch": 0.9890531016212073, "percentage": 98.91, "elapsed_time": "19:52:34", "remaining_time": "0:13:10"}
{"current_steps": 3720, "total_steps": 3751, "loss": 0.2021, "lr": 2.0815433106080583e-09, "epoch": 0.9917190129463319, "percentage": 99.17, "elapsed_time": "19:54:53", "remaining_time": "0:09:57"}
{"current_steps": 3730, "total_steps": 3751, "loss": 0.2134, "lr": 9.55249810161152e-10, "epoch": 0.9943849242714564, "percentage": 99.44, "elapsed_time": "19:57:12", "remaining_time": "0:06:44"}
{"current_steps": 3740, "total_steps": 3751, "loss": 0.211, "lr": 2.6210407655269386e-10, "epoch": 0.9970508355965809, "percentage": 99.71, "elapsed_time": "19:59:28", "remaining_time": "0:03:31"}
{"current_steps": 3750, "total_steps": 3751, "loss": 0.2088, "lr": 2.1661681620654963e-12, "epoch": 0.9997167469217055, "percentage": 99.97, "elapsed_time": "20:01:43", "remaining_time": "0:00:19"}
{"current_steps": 3750, "total_steps": 3751, "eval_loss": 0.2036588490009308, "epoch": 0.9997167469217055, "percentage": 99.97, "elapsed_time": "20:06:05", "remaining_time": "0:00:19"}
{"current_steps": 3751, "total_steps": 3751, "epoch": 0.999983338054218, "percentage": 100.0, "elapsed_time": "20:07:19", "remaining_time": "0:00:00"}