Training in progress, epoch 1

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:50aa4b147c22806458f4c468980fb9769bcb0d9bbc6d5d01c315a52cb0b8bc3b
+oid sha256:75b84c1c263fc2c000a96b480e7c2e6f845e1ca917925035e70062a3f61436aa
 size 4976698672

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:68e8a576bc7c87ccb982d2d424d5e57a136c74cff616b297ea40095222cc98e8
+oid sha256:eef6618b18e1934c6dc5ac9e58d43f5c43d872c3549db23c16e20ffd97f20ae8
 size 4999802720

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5853464c684e87157a84708fac4e79089b02bceaabd89c6ef1f34e1d6cb2a004
+oid sha256:8df1124d1eb3272384ae11a342e03879d3109de75935c6a1e4ecf0c3820adbc8
 size 4915916176

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5cc1ced797fc375b3d03b420a5c1ca3b3d427613a540ff93a5b5bbfcbffef4a3
+oid sha256:9ec1f03af7554ec9f94b51f8fd38540327bcdfd1a552595b039262d1872a4edf
 size 1168138808

trainer_log.jsonl

@@ -8,3 +8,13 @@
 {"current_steps": 80, "total_steps": 276, "loss": 0.9609, "lr": 5e-06, "epoch": 0.8672086720867209, "percentage": 28.99, "elapsed_time": "1:15:36", "remaining_time": "3:05:13"}
 {"current_steps": 90, "total_steps": 276, "loss": 0.967, "lr": 5e-06, "epoch": 0.975609756097561, "percentage": 32.61, "elapsed_time": "1:25:02", "remaining_time": "2:55:45"}
 {"current_steps": 92, "total_steps": 276, "eval_loss": 0.9647226333618164, "epoch": 0.997289972899729, "percentage": 33.33, "elapsed_time": "1:28:47", "remaining_time": "2:57:34"}
+{"current_steps": 100, "total_steps": 276, "loss": 1.0085, "lr": 5e-06, "epoch": 1.084010840108401, "percentage": 36.23, "elapsed_time": "1:37:15", "remaining_time": "2:51:10"}
+{"current_steps": 110, "total_steps": 276, "loss": 0.9266, "lr": 5e-06, "epoch": 1.1924119241192412, "percentage": 39.86, "elapsed_time": "1:46:41", "remaining_time": "2:41:00"}
+{"current_steps": 120, "total_steps": 276, "loss": 0.9176, "lr": 5e-06, "epoch": 1.3008130081300813, "percentage": 43.48, "elapsed_time": "1:56:08", "remaining_time": "2:30:59"}
+{"current_steps": 130, "total_steps": 276, "loss": 0.9165, "lr": 5e-06, "epoch": 1.4092140921409215, "percentage": 47.1, "elapsed_time": "2:05:34", "remaining_time": "2:21:01"}
+{"current_steps": 140, "total_steps": 276, "loss": 0.9178, "lr": 5e-06, "epoch": 1.5176151761517616, "percentage": 50.72, "elapsed_time": "2:15:00", "remaining_time": "2:11:08"}
+{"current_steps": 150, "total_steps": 276, "loss": 0.9095, "lr": 5e-06, "epoch": 1.6260162601626016, "percentage": 54.35, "elapsed_time": "2:24:27", "remaining_time": "2:01:20"}
+{"current_steps": 160, "total_steps": 276, "loss": 0.922, "lr": 5e-06, "epoch": 1.7344173441734418, "percentage": 57.97, "elapsed_time": "2:33:53", "remaining_time": "1:51:34"}
+{"current_steps": 170, "total_steps": 276, "loss": 0.9161, "lr": 5e-06, "epoch": 1.8428184281842819, "percentage": 61.59, "elapsed_time": "2:43:18", "remaining_time": "1:41:49"}
+{"current_steps": 180, "total_steps": 276, "loss": 0.917, "lr": 5e-06, "epoch": 1.951219512195122, "percentage": 65.22, "elapsed_time": "2:52:45", "remaining_time": "1:32:08"}
+{"current_steps": 184, "total_steps": 276, "eval_loss": 0.9558340907096863, "epoch": 1.994579945799458, "percentage": 66.67, "elapsed_time": "2:58:37", "remaining_time": "1:29:18"}