Training in progress, epoch 2

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f2f70337ee8d8eb4d0ba14d8ecf184668cdd7e94967940986bd292c18cde0de7
+oid sha256:f04cd5c5635e1e12a21d2bdca8b80a3c7fc566d02f6805de36f1e28b68475ddf
 size 4938985352

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:58845f8bc794f9732cb8776960d321fbc477c5551c90fbb7e0e83f656c19a589
+oid sha256:f8916961aa9cb7280bf8f9201445808a69d82824c221dd72ebe20874a537c438
 size 4947390880

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b24522ba21ab49196c767e9418d4ae15524d2942ecdffcd766628f27dac1715
+oid sha256:fa56db8d950e99b336b36242091b12fb5b2014983e89ac14431c2b366d836d49
 size 3590488816

trainer_log.jsonl

@@ -11,3 +11,16 @@
 {"current_steps": 110, "total_steps": 360, "loss": 0.5361, "lr": 5e-06, "epoch": 0.9128630705394191, "percentage": 30.56, "elapsed_time": "1:38:57", "remaining_time": "3:44:55"}
 {"current_steps": 120, "total_steps": 360, "loss": 0.5354, "lr": 5e-06, "epoch": 0.995850622406639, "percentage": 33.33, "elapsed_time": "1:47:56", "remaining_time": "3:35:52"}
 {"current_steps": 120, "total_steps": 360, "eval_loss": 0.5283368229866028, "epoch": 0.995850622406639, "percentage": 33.33, "elapsed_time": "1:50:23", "remaining_time": "3:40:47"}
+{"current_steps": 130, "total_steps": 360, "loss": 0.5334, "lr": 5e-06, "epoch": 1.0788381742738589, "percentage": 36.11, "elapsed_time": "1:59:59", "remaining_time": "3:32:16"}
+{"current_steps": 140, "total_steps": 360, "loss": 0.5122, "lr": 5e-06, "epoch": 1.161825726141079, "percentage": 38.89, "elapsed_time": "2:08:58", "remaining_time": "3:22:39"}
+{"current_steps": 150, "total_steps": 360, "loss": 0.509, "lr": 5e-06, "epoch": 1.2448132780082988, "percentage": 41.67, "elapsed_time": "2:17:57", "remaining_time": "3:13:08"}
+{"current_steps": 160, "total_steps": 360, "loss": 0.5041, "lr": 5e-06, "epoch": 1.3278008298755186, "percentage": 44.44, "elapsed_time": "2:26:56", "remaining_time": "3:03:40"}
+{"current_steps": 170, "total_steps": 360, "loss": 0.4998, "lr": 5e-06, "epoch": 1.4107883817427385, "percentage": 47.22, "elapsed_time": "2:35:55", "remaining_time": "2:54:15"}
+{"current_steps": 180, "total_steps": 360, "loss": 0.5037, "lr": 5e-06, "epoch": 1.4937759336099585, "percentage": 50.0, "elapsed_time": "2:44:55", "remaining_time": "2:44:55"}
+{"current_steps": 190, "total_steps": 360, "loss": 0.4977, "lr": 5e-06, "epoch": 1.5767634854771784, "percentage": 52.78, "elapsed_time": "2:53:55", "remaining_time": "2:35:37"}
+{"current_steps": 200, "total_steps": 360, "loss": 0.4941, "lr": 5e-06, "epoch": 1.6597510373443982, "percentage": 55.56, "elapsed_time": "3:02:55", "remaining_time": "2:26:20"}
+{"current_steps": 210, "total_steps": 360, "loss": 0.4945, "lr": 5e-06, "epoch": 1.7427385892116183, "percentage": 58.33, "elapsed_time": "3:11:54", "remaining_time": "2:17:04"}
+{"current_steps": 220, "total_steps": 360, "loss": 0.4904, "lr": 5e-06, "epoch": 1.8257261410788381, "percentage": 61.11, "elapsed_time": "3:20:54", "remaining_time": "2:07:50"}
+{"current_steps": 230, "total_steps": 360, "loss": 0.491, "lr": 5e-06, "epoch": 1.908713692946058, "percentage": 63.89, "elapsed_time": "3:29:53", "remaining_time": "1:58:38"}
+{"current_steps": 240, "total_steps": 360, "loss": 0.4901, "lr": 5e-06, "epoch": 1.991701244813278, "percentage": 66.67, "elapsed_time": "3:38:53", "remaining_time": "1:49:26"}
+{"current_steps": 241, "total_steps": 360, "eval_loss": 0.5045989155769348, "epoch": 2.0, "percentage": 66.94, "elapsed_time": "3:41:45", "remaining_time": "1:49:29"}