Training in progress, epoch 3

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:beb194175a7c73d4a7f7c0ddfeee7cbba5f1b0076b8f0acac73e2d77e27a4e14
+oid sha256:508ea172c265d559b15d2fd9e2f5a2500241393c1f6e99a25abab57f291ad6cb
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:22fd86b99d2d0d922b9c82171f134420134b67f6f9672cc1475efd095556c37d
+oid sha256:9fdd688cc61e35b71346a3605cb982569326960c6aa1233c53597e93290e4d2e
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:df82761863653c0c50be800b587b183f49b7b1b7d38666f3c23547f9a32ccd00
+oid sha256:d5a5c6944e5c1f2ba40c569fc54f997c50cb382fe6df8b444ab944d6cf4b8f26
 size 4540516344

trainer_log.jsonl

@@ -34,3 +34,21 @@
 {"current_steps": 330, "total_steps": 504, "loss": 0.4738, "lr": 1.032945093404745e-06, "epoch": 1.9642857142857144, "percentage": 65.48, "elapsed_time": "3:03:45", "remaining_time": "1:36:53"}
 {"current_steps": 336, "total_steps": 504, "eval_loss": 0.061109066009521484, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "3:08:57", "remaining_time": "1:34:28"}
 {"current_steps": 340, "total_steps": 504, "loss": 0.4626, "lr": 9.80881438094648e-07, "epoch": 2.0238095238095237, "percentage": 67.46, "elapsed_time": "3:11:58", "remaining_time": "1:32:36"}
+{"current_steps": 350, "total_steps": 504, "loss": 0.4514, "lr": 9.302670913558731e-07, "epoch": 2.0833333333333335, "percentage": 69.44, "elapsed_time": "3:17:27", "remaining_time": "1:26:52"}
+{"current_steps": 360, "total_steps": 504, "loss": 0.4466, "lr": 8.813746312151494e-07, "epoch": 2.142857142857143, "percentage": 71.43, "elapsed_time": "3:22:55", "remaining_time": "1:21:10"}
+{"current_steps": 370, "total_steps": 504, "loss": 0.4461, "lr": 8.344673626673205e-07, "epoch": 2.2023809523809526, "percentage": 73.41, "elapsed_time": "3:28:24", "remaining_time": "1:15:28"}
+{"current_steps": 380, "total_steps": 504, "loss": 0.4491, "lr": 7.897978996751046e-07, "epoch": 2.261904761904762, "percentage": 75.4, "elapsed_time": "3:33:53", "remaining_time": "1:09:47"}
+{"current_steps": 390, "total_steps": 504, "loss": 0.4443, "lr": 7.476068047442332e-07, "epoch": 2.3214285714285716, "percentage": 77.38, "elapsed_time": "3:39:22", "remaining_time": "1:04:07"}
+{"current_steps": 400, "total_steps": 504, "loss": 0.4435, "lr": 7.081212934003984e-07, "epoch": 2.380952380952381, "percentage": 79.37, "elapsed_time": "3:44:49", "remaining_time": "0:58:27"}
+{"current_steps": 410, "total_steps": 504, "loss": 0.4428, "lr": 6.715540105448972e-07, "epoch": 2.4404761904761907, "percentage": 81.35, "elapsed_time": "3:50:17", "remaining_time": "0:52:47"}
+{"current_steps": 420, "total_steps": 504, "loss": 0.4427, "lr": 6.381018852787821e-07, "epoch": 2.5, "percentage": 83.33, "elapsed_time": "3:55:46", "remaining_time": "0:47:09"}
+{"current_steps": 430, "total_steps": 504, "loss": 0.4443, "lr": 6.079450703627314e-07, "epoch": 2.5595238095238093, "percentage": 85.32, "elapsed_time": "4:01:14", "remaining_time": "0:41:30"}
+{"current_steps": 440, "total_steps": 504, "loss": 0.4448, "lr": 5.812459720240681e-07, "epoch": 2.619047619047619, "percentage": 87.3, "elapsed_time": "4:06:43", "remaining_time": "0:35:53"}
+{"current_steps": 450, "total_steps": 504, "loss": 0.4425, "lr": 5.581483753357905e-07, "epoch": 2.678571428571429, "percentage": 89.29, "elapsed_time": "4:12:12", "remaining_time": "0:30:15"}
+{"current_steps": 460, "total_steps": 504, "loss": 0.4438, "lr": 5.387766698777935e-07, "epoch": 2.738095238095238, "percentage": 91.27, "elapsed_time": "4:17:41", "remaining_time": "0:24:38"}
+{"current_steps": 470, "total_steps": 504, "loss": 0.4423, "lr": 5.232351798503945e-07, "epoch": 2.7976190476190474, "percentage": 93.25, "elapsed_time": "4:23:10", "remaining_time": "0:19:02"}
+{"current_steps": 480, "total_steps": 504, "loss": 0.4416, "lr": 5.116076022477671e-07, "epoch": 2.857142857142857, "percentage": 95.24, "elapsed_time": "4:28:39", "remaining_time": "0:13:25"}
+{"current_steps": 490, "total_steps": 504, "loss": 0.4385, "lr": 5.039565561169362e-07, "epoch": 2.9166666666666665, "percentage": 97.22, "elapsed_time": "4:34:08", "remaining_time": "0:07:49"}
+{"current_steps": 500, "total_steps": 504, "loss": 0.4405, "lr": 5.003232453297512e-07, "epoch": 2.9761904761904763, "percentage": 99.21, "elapsed_time": "4:39:37", "remaining_time": "0:02:14"}
+{"current_steps": 504, "total_steps": 504, "eval_loss": 0.05991922318935394, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "4:44:35", "remaining_time": "0:00:00"}
+{"current_steps": 504, "total_steps": 504, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "4:45:17", "remaining_time": "0:00:00"}