Training in progress, epoch 2

model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:72dc1425ee6ebbd2154137fd30453c6901ff0467c64a4bea43b61bbdaac1b68c
+oid sha256:45a81fa55ed6857430077a5c0c93224f49f3089c175cf86d3264371f6af94824
 size 4943162336

model-00002-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7a02d3a384bb48ac53d764cfb0053ea6ea229721faced968754147616644f739
+oid sha256:4dbb4db81d9124ad28c6faa456f842bd52c7ca5d208050109f70bd87696d5efd
 size 4999819336

model-00003-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17f587c13c1343b8363cf6f3c42440fb30d39630f978edc8a19d2bca27a3ee5c
+oid sha256:404625968838f13ac14d663bce6727753aeb7357d496a5d6e41e5a3f3dbaf7a8
 size 4540516344

trainer_log.jsonl

@@ -98,3 +98,52 @@
 {"current_steps": 970, "total_steps": 1479, "loss": 0.4866, "learning_rate": 1.9120961682945154e-06, "epoch": 1.9645569620253165, "percentage": 65.58, "elapsed_time": "15:15:11", "remaining_time": "8:00:14"}
 {"current_steps": 980, "total_steps": 1479, "loss": 0.4856, "learning_rate": 1.8745304282494367e-06, "epoch": 1.9848101265822784, "percentage": 66.26, "elapsed_time": "15:24:31", "remaining_time": "7:50:45"}
 {"current_steps": 987, "total_steps": 1479, "eval_loss": 0.06770560145378113, "epoch": 1.998987341772152, "percentage": 66.73, "elapsed_time": "15:39:49", "remaining_time": "7:48:28"}
+{"current_steps": 990, "total_steps": 1479, "loss": 0.4629, "learning_rate": 1.8369646882043577e-06, "epoch": 2.0050632911392405, "percentage": 66.94, "elapsed_time": "15:43:18", "remaining_time": "7:45:56"}
+{"current_steps": 1000, "total_steps": 1479, "loss": 0.4148, "learning_rate": 1.799398948159279e-06, "epoch": 2.0253164556962027, "percentage": 67.61, "elapsed_time": "15:52:39", "remaining_time": "7:36:19"}
+{"current_steps": 1010, "total_steps": 1479, "loss": 0.4141, "learning_rate": 1.7618332081142001e-06, "epoch": 2.0455696202531644, "percentage": 68.29, "elapsed_time": "16:02:00", "remaining_time": "7:26:42"}
+{"current_steps": 1020, "total_steps": 1479, "loss": 0.4097, "learning_rate": 1.724267468069121e-06, "epoch": 2.0658227848101265, "percentage": 68.97, "elapsed_time": "16:11:22", "remaining_time": "7:17:06"}
+{"current_steps": 1030, "total_steps": 1479, "loss": 0.4098, "learning_rate": 1.6867017280240422e-06, "epoch": 2.0860759493670886, "percentage": 69.64, "elapsed_time": "16:20:42", "remaining_time": "7:07:30"}
+{"current_steps": 1040, "total_steps": 1479, "loss": 0.4063, "learning_rate": 1.6491359879789634e-06, "epoch": 2.1063291139240508, "percentage": 70.32, "elapsed_time": "16:30:01", "remaining_time": "6:57:54"}
+{"current_steps": 1050, "total_steps": 1479, "loss": 0.4148, "learning_rate": 1.6115702479338842e-06, "epoch": 2.1265822784810124, "percentage": 70.99, "elapsed_time": "16:39:21", "remaining_time": "6:48:18"}
+{"current_steps": 1060, "total_steps": 1479, "loss": 0.4095, "learning_rate": 1.5740045078888055e-06, "epoch": 2.1468354430379746, "percentage": 71.67, "elapsed_time": "16:48:42", "remaining_time": "6:38:43"}
+{"current_steps": 1070, "total_steps": 1479, "loss": 0.4124, "learning_rate": 1.5364387678437267e-06, "epoch": 2.1670886075949367, "percentage": 72.35, "elapsed_time": "16:58:00", "remaining_time": "6:29:07"}
+{"current_steps": 1080, "total_steps": 1479, "loss": 0.4106, "learning_rate": 1.4988730277986477e-06, "epoch": 2.187341772151899, "percentage": 73.02, "elapsed_time": "17:07:21", "remaining_time": "6:19:32"}
+{"current_steps": 1090, "total_steps": 1479, "loss": 0.4079, "learning_rate": 1.461307287753569e-06, "epoch": 2.207594936708861, "percentage": 73.7, "elapsed_time": "17:16:39", "remaining_time": "6:09:57"}
+{"current_steps": 1100, "total_steps": 1479, "loss": 0.4166, "learning_rate": 1.4237415477084902e-06, "epoch": 2.2278481012658227, "percentage": 74.37, "elapsed_time": "17:26:00", "remaining_time": "6:00:23"}
+{"current_steps": 1110, "total_steps": 1479, "loss": 0.4123, "learning_rate": 1.386175807663411e-06, "epoch": 2.248101265822785, "percentage": 75.05, "elapsed_time": "17:35:17", "remaining_time": "5:50:48"}
+{"current_steps": 1120, "total_steps": 1479, "loss": 0.413, "learning_rate": 1.3486100676183322e-06, "epoch": 2.268354430379747, "percentage": 75.73, "elapsed_time": "17:44:35", "remaining_time": "5:41:14"}
+{"current_steps": 1130, "total_steps": 1479, "loss": 0.4142, "learning_rate": 1.3110443275732533e-06, "epoch": 2.2886075949367086, "percentage": 76.4, "elapsed_time": "17:53:56", "remaining_time": "5:31:41"}
+{"current_steps": 1140, "total_steps": 1479, "loss": 0.409, "learning_rate": 1.2734785875281743e-06, "epoch": 2.3088607594936708, "percentage": 77.08, "elapsed_time": "18:03:17", "remaining_time": "5:22:08"}
+{"current_steps": 1150, "total_steps": 1479, "loss": 0.4157, "learning_rate": 1.2359128474830955e-06, "epoch": 2.329113924050633, "percentage": 77.76, "elapsed_time": "18:12:37", "remaining_time": "5:12:35"}
+{"current_steps": 1160, "total_steps": 1479, "loss": 0.4155, "learning_rate": 1.1983471074380167e-06, "epoch": 2.349367088607595, "percentage": 78.43, "elapsed_time": "18:21:58", "remaining_time": "5:03:02"}
+{"current_steps": 1170, "total_steps": 1479, "loss": 0.4143, "learning_rate": 1.1607813673929378e-06, "epoch": 2.369620253164557, "percentage": 79.11, "elapsed_time": "18:31:18", "remaining_time": "4:53:30"}
+{"current_steps": 1180, "total_steps": 1479, "loss": 0.4143, "learning_rate": 1.1232156273478588e-06, "epoch": 2.389873417721519, "percentage": 79.78, "elapsed_time": "18:40:37", "remaining_time": "4:43:57"}
+{"current_steps": 1190, "total_steps": 1479, "loss": 0.4134, "learning_rate": 1.08564988730278e-06, "epoch": 2.410126582278481, "percentage": 80.46, "elapsed_time": "18:49:58", "remaining_time": "4:34:25"}
+{"current_steps": 1200, "total_steps": 1479, "loss": 0.4113, "learning_rate": 1.048084147257701e-06, "epoch": 2.430379746835443, "percentage": 81.14, "elapsed_time": "18:59:19", "remaining_time": "4:24:53"}
+{"current_steps": 1210, "total_steps": 1479, "loss": 0.4107, "learning_rate": 1.010518407212622e-06, "epoch": 2.4506329113924052, "percentage": 81.81, "elapsed_time": "19:08:40", "remaining_time": "4:15:22"}
+{"current_steps": 1220, "total_steps": 1479, "loss": 0.4137, "learning_rate": 9.729526671675433e-07, "epoch": 2.470886075949367, "percentage": 82.49, "elapsed_time": "19:18:01", "remaining_time": "4:05:50"}
+{"current_steps": 1230, "total_steps": 1479, "loss": 0.4123, "learning_rate": 9.353869271224644e-07, "epoch": 2.491139240506329, "percentage": 83.16, "elapsed_time": "19:27:22", "remaining_time": "3:56:19"}
+{"current_steps": 1240, "total_steps": 1479, "loss": 0.4162, "learning_rate": 8.978211870773855e-07, "epoch": 2.511392405063291, "percentage": 83.84, "elapsed_time": "19:36:41", "remaining_time": "3:46:47"}
+{"current_steps": 1250, "total_steps": 1479, "loss": 0.4094, "learning_rate": 8.602554470323066e-07, "epoch": 2.5316455696202533, "percentage": 84.52, "elapsed_time": "19:46:02", "remaining_time": "3:37:16"}
+{"current_steps": 1260, "total_steps": 1479, "loss": 0.4041, "learning_rate": 8.226897069872278e-07, "epoch": 2.5518987341772155, "percentage": 85.19, "elapsed_time": "19:55:23", "remaining_time": "3:27:46"}
+{"current_steps": 1270, "total_steps": 1479, "loss": 0.411, "learning_rate": 7.851239669421488e-07, "epoch": 2.572151898734177, "percentage": 85.87, "elapsed_time": "20:04:44", "remaining_time": "3:18:15"}
+{"current_steps": 1280, "total_steps": 1479, "loss": 0.4127, "learning_rate": 7.4755822689707e-07, "epoch": 2.5924050632911393, "percentage": 86.54, "elapsed_time": "20:14:04", "remaining_time": "3:08:45"}
+{"current_steps": 1290, "total_steps": 1479, "loss": 0.4181, "learning_rate": 7.09992486851991e-07, "epoch": 2.6126582278481014, "percentage": 87.22, "elapsed_time": "20:23:25", "remaining_time": "2:59:14"}
+{"current_steps": 1300, "total_steps": 1479, "loss": 0.4114, "learning_rate": 6.724267468069122e-07, "epoch": 2.632911392405063, "percentage": 87.9, "elapsed_time": "20:32:46", "remaining_time": "2:49:44"}
+{"current_steps": 1310, "total_steps": 1479, "loss": 0.4104, "learning_rate": 6.348610067618332e-07, "epoch": 2.6531645569620252, "percentage": 88.57, "elapsed_time": "20:42:06", "remaining_time": "2:40:14"}
+{"current_steps": 1320, "total_steps": 1479, "loss": 0.4119, "learning_rate": 5.972952667167544e-07, "epoch": 2.6734177215189874, "percentage": 89.25, "elapsed_time": "20:51:26", "remaining_time": "2:30:44"}
+{"current_steps": 1330, "total_steps": 1479, "loss": 0.4145, "learning_rate": 5.597295266716755e-07, "epoch": 2.6936708860759495, "percentage": 89.93, "elapsed_time": "21:00:48", "remaining_time": "2:21:14"}
+{"current_steps": 1340, "total_steps": 1479, "loss": 0.4124, "learning_rate": 5.221637866265966e-07, "epoch": 2.7139240506329116, "percentage": 90.6, "elapsed_time": "21:10:09", "remaining_time": "2:11:45"}
+{"current_steps": 1350, "total_steps": 1479, "loss": 0.4104, "learning_rate": 4.845980465815176e-07, "epoch": 2.7341772151898733, "percentage": 91.28, "elapsed_time": "21:19:31", "remaining_time": "2:02:15"}
+{"current_steps": 1360, "total_steps": 1479, "loss": 0.4063, "learning_rate": 4.4703230653643883e-07, "epoch": 2.7544303797468355, "percentage": 91.95, "elapsed_time": "21:28:50", "remaining_time": "1:52:46"}
+{"current_steps": 1370, "total_steps": 1479, "loss": 0.4114, "learning_rate": 4.094665664913599e-07, "epoch": 2.7746835443037976, "percentage": 92.63, "elapsed_time": "21:38:12", "remaining_time": "1:43:17"}
+{"current_steps": 1380, "total_steps": 1479, "loss": 0.4092, "learning_rate": 3.7190082644628103e-07, "epoch": 2.7949367088607593, "percentage": 93.31, "elapsed_time": "21:47:34", "remaining_time": "1:33:48"}
+{"current_steps": 1390, "total_steps": 1479, "loss": 0.4121, "learning_rate": 3.343350864012021e-07, "epoch": 2.8151898734177214, "percentage": 93.98, "elapsed_time": "21:56:54", "remaining_time": "1:24:19"}
+{"current_steps": 1400, "total_steps": 1479, "loss": 0.4045, "learning_rate": 2.9676934635612324e-07, "epoch": 2.8354430379746836, "percentage": 94.66, "elapsed_time": "22:06:13", "remaining_time": "1:14:50"}
+{"current_steps": 1410, "total_steps": 1479, "loss": 0.4132, "learning_rate": 2.5920360631104436e-07, "epoch": 2.8556962025316457, "percentage": 95.33, "elapsed_time": "22:15:33", "remaining_time": "1:05:21"}
+{"current_steps": 1420, "total_steps": 1479, "loss": 0.4077, "learning_rate": 2.2163786626596544e-07, "epoch": 2.875949367088608, "percentage": 96.01, "elapsed_time": "22:24:52", "remaining_time": "0:55:52"}
+{"current_steps": 1430, "total_steps": 1479, "loss": 0.4089, "learning_rate": 1.8407212622088657e-07, "epoch": 2.8962025316455695, "percentage": 96.69, "elapsed_time": "22:34:13", "remaining_time": "0:46:24"}
+{"current_steps": 1440, "total_steps": 1479, "loss": 0.4108, "learning_rate": 1.4650638617580767e-07, "epoch": 2.9164556962025316, "percentage": 97.36, "elapsed_time": "22:43:34", "remaining_time": "0:36:55"}
+{"current_steps": 1450, "total_steps": 1479, "loss": 0.4193, "learning_rate": 1.0894064613072878e-07, "epoch": 2.9367088607594938, "percentage": 98.04, "elapsed_time": "22:52:56", "remaining_time": "0:27:27"}
+{"current_steps": 1460, "total_steps": 1479, "loss": 0.4063, "learning_rate": 7.13749060856499e-08, "epoch": 2.9569620253164555, "percentage": 98.72, "elapsed_time": "23:02:16", "remaining_time": "0:17:59"}
+{"current_steps": 1470, "total_steps": 1479, "loss": 0.4105, "learning_rate": 3.3809166040571e-08, "epoch": 2.9772151898734176, "percentage": 99.39, "elapsed_time": "23:11:37", "remaining_time": "0:08:31"}