Training in progress, step 1130

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2126eddefc37d9b1a63fd6add450340a72c449bf2bdd5ac4baa7422a04cce858
+oid sha256:cc7c0f5cb8ec05fa5071e901fa220f530bd1254391533da35f2baba0a6b8b04d
 size 8388736

trainer_log.jsonl

@@ -236,3 +236,11 @@
 {"current_steps": 1085, "total_steps": 1130, "loss": 0.0004, "lr": 2.5197220116515987e-07, "epoch": 9.601769911504425, "percentage": 96.02, "elapsed_time": "0:02:25", "remaining_time": "0:00:06", "throughput": 2017.94, "total_tokens": 294352}
 {"current_steps": 1090, "total_steps": 1130, "loss": 0.0019, "lr": 2.0024189799641435e-07, "epoch": 9.646017699115044, "percentage": 96.46, "elapsed_time": "0:02:26", "remaining_time": "0:00:05", "throughput": 2021.24, "total_tokens": 295776}
 {"current_steps": 1095, "total_steps": 1130, "loss": 0.0004, "lr": 1.5442771053230665e-07, "epoch": 9.690265486725664, "percentage": 96.9, "elapsed_time": "0:02:26", "remaining_time": "0:00:04", "throughput": 2023.42, "total_tokens": 297008}
+{"current_steps": 1100, "total_steps": 1130, "loss": 0.0008, "lr": 1.1454056799370672e-07, "epoch": 9.734513274336283, "percentage": 97.35, "elapsed_time": "0:02:27", "remaining_time": "0:00:04", "throughput": 2026.26, "total_tokens": 298320}
+{"current_steps": 1105, "total_steps": 1130, "loss": 0.002, "lr": 8.058998567304633e-08, "epoch": 9.778761061946902, "percentage": 97.79, "elapsed_time": "0:02:27", "remaining_time": "0:00:03", "throughput": 2028.79, "total_tokens": 299568}
+{"current_steps": 1110, "total_steps": 1130, "loss": 0.0008, "lr": 5.2584062664368105e-08, "epoch": 9.823008849557523, "percentage": 98.23, "elapsed_time": "0:02:28", "remaining_time": "0:00:02", "throughput": 2030.75, "total_tokens": 300720}
+{"current_steps": 1115, "total_steps": 1130, "loss": 0.0013, "lr": 3.0529479931265535e-08, "epoch": 9.867256637168142, "percentage": 98.67, "elapsed_time": "0:02:28", "remaining_time": "0:00:01", "throughput": 2032.84, "total_tokens": 301888}
+{"current_steps": 1120, "total_steps": 1130, "loss": 0.0007, "lr": 1.443149871309113e-08, "epoch": 9.91150442477876, "percentage": 99.12, "elapsed_time": "0:02:28", "remaining_time": "0:00:01", "throughput": 2036.84, "total_tokens": 303456}
+{"current_steps": 1125, "total_steps": 1130, "loss": 0.0003, "lr": 4.293959269863201e-09, "epoch": 9.955752212389381, "percentage": 99.56, "elapsed_time": "0:02:29", "remaining_time": "0:00:00", "throughput": 2040.9, "total_tokens": 305024}
+{"current_steps": 1130, "total_steps": 1130, "loss": 0.0011, "lr": 1.192799661514643e-10, "epoch": 10.0, "percentage": 100.0, "elapsed_time": "0:02:29", "remaining_time": "0:00:00", "throughput": 2041.53, "total_tokens": 306152}
+{"current_steps": 1130, "total_steps": 1130, "epoch": 10.0, "percentage": 100.0, "elapsed_time": "0:02:31", "remaining_time": "0:00:00", "throughput": 2021.02, "total_tokens": 306152}