Training in progress, step 1800

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:02f092c8551b39f72b2480ed050635fbcee05ba27bc1999c55e842ece4cdbe8b
+oid sha256:703d6c7e98f4bf68731cde74c5b20cc8e4bb82cbcfea9af1822815e13d9fd381
 size 29034840

trainer_log.jsonl

@@ -395,3 +395,14 @@
 {"current_steps": 1745, "total_steps": 6770, "loss": 0.2691, "lr": 8.86626255549189e-05, "epoch": 0.5155096011816839, "percentage": 25.78, "elapsed_time": "0:29:50", "remaining_time": "1:25:54", "throughput": 10120.67, "total_tokens": 18116880}
 {"current_steps": 1750, "total_steps": 6770, "loss": 0.3173, "lr": 8.858506984460383e-05, "epoch": 0.51698670605613, "percentage": 25.85, "elapsed_time": "0:31:17", "remaining_time": "1:29:44", "throughput": 9679.72, "total_tokens": 18169344}
 {"current_steps": 1750, "total_steps": 6770, "eval_loss": 0.44908422231674194, "epoch": 0.51698670605613, "percentage": 25.85, "elapsed_time": "0:31:36", "remaining_time": "1:30:40", "throughput": 9580.54, "total_tokens": 18169344}
+{"current_steps": 1755, "total_steps": 6770, "loss": 0.2966, "lr": 8.850728393614902e-05, "epoch": 0.518463810930576, "percentage": 25.92, "elapsed_time": "0:33:07", "remaining_time": "1:34:40", "throughput": 9166.33, "total_tokens": 18221144}
+{"current_steps": 1760, "total_steps": 6770, "loss": 0.3101, "lr": 8.842926829362446e-05, "epoch": 0.5199409158050221, "percentage": 26.0, "elapsed_time": "0:34:34", "remaining_time": "1:38:24", "throughput": 8808.93, "total_tokens": 18272752}
+{"current_steps": 1765, "total_steps": 6770, "loss": 0.2545, "lr": 8.835102338247064e-05, "epoch": 0.5214180206794683, "percentage": 26.07, "elapsed_time": "0:36:00", "remaining_time": "1:42:07", "throughput": 8481.26, "total_tokens": 18325888}
+{"current_steps": 1770, "total_steps": 6770, "loss": 0.3223, "lr": 8.827254966949593e-05, "epoch": 0.5228951255539144, "percentage": 26.14, "elapsed_time": "0:37:27", "remaining_time": "1:45:48", "throughput": 8177.92, "total_tokens": 18378016}
+{"current_steps": 1775, "total_steps": 6770, "loss": 0.2714, "lr": 8.819384762287373e-05, "epoch": 0.5243722304283605, "percentage": 26.22, "elapsed_time": "0:38:55", "remaining_time": "1:49:31", "throughput": 7892.96, "total_tokens": 18431240}
+{"current_steps": 1780, "total_steps": 6770, "loss": 0.3438, "lr": 8.811491771213964e-05, "epoch": 0.5258493353028065, "percentage": 26.29, "elapsed_time": "0:40:21", "remaining_time": "1:53:07", "throughput": 7633.32, "total_tokens": 18482832}
+{"current_steps": 1785, "total_steps": 6770, "loss": 0.2324, "lr": 8.803576040818873e-05, "epoch": 0.5273264401772526, "percentage": 26.37, "elapsed_time": "0:41:49", "remaining_time": "1:56:48", "throughput": 7385.4, "total_tokens": 18534992}
+{"current_steps": 1790, "total_steps": 6770, "loss": 0.2259, "lr": 8.795637618327269e-05, "epoch": 0.5288035450516987, "percentage": 26.44, "elapsed_time": "0:43:17", "remaining_time": "2:00:25", "throughput": 7156.95, "total_tokens": 18587752}
+{"current_steps": 1795, "total_steps": 6770, "loss": 0.2468, "lr": 8.7876765510997e-05, "epoch": 0.5302806499261448, "percentage": 26.51, "elapsed_time": "0:44:46", "remaining_time": "2:04:05", "throughput": 6939.35, "total_tokens": 18640440}
+{"current_steps": 1800, "total_steps": 6770, "loss": 0.3127, "lr": 8.779692886631812e-05, "epoch": 0.5317577548005908, "percentage": 26.59, "elapsed_time": "0:46:12", "remaining_time": "2:07:35", "throughput": 6741.35, "total_tokens": 18691928}
+{"current_steps": 1800, "total_steps": 6770, "eval_loss": 0.3499237596988678, "epoch": 0.5317577548005908, "percentage": 26.59, "elapsed_time": "0:46:32", "remaining_time": "2:08:29", "throughput": 6694.66, "total_tokens": 18691928}