Training in progress, epoch 2

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b91ec591bd30bb634f7dbbb7ac6bc9035a86997e07ff63bd54e51df10298ec26
+oid sha256:1a10406810c1df8cb712499735672853c177235cf36c7f1d43b93e41fb5852a2
 size 4903351912

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f4a89b8145c49907bd24ace62d7fd4ed41f2b5f0bd8f09fcece38cfde671b1e
+oid sha256:6c6a5ce0ecadcd85fab4ca01cc85be650c83698b25670bb3d9eafbcb7ae9cba7
 size 4947570872

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bc2baff7fc26f7b33986dca510ce9267b0eb52f8b68056ce35e1d70936d120b9
+oid sha256:db254854d3abb7ea8f0e9b0695bd7e10e648a93efa82b13097e283b1ed196ecd
 size 4962221464

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7973798768decabb825fed8aa2e827d61d8225aa54aac2d3988ab0469ee17a07
+oid sha256:fe6270e042942a0cac0e93a3047abd854a2094f44b9b03559cbe32b542183dfd
 size 3670322200

trainer_log.jsonl

@@ -88,3 +88,47 @@
 {"current_steps": 870, "total_steps": 1329, "loss": 0.5457, "learning_rate": 4.795735609559657e-06, "epoch": 1.9633286318758816, "percentage": 65.46, "elapsed_time": "21:43:28", "remaining_time": "11:27:41"}
 {"current_steps": 880, "total_steps": 1329, "loss": 0.5439, "learning_rate": 4.793531180296065e-06, "epoch": 1.9858956276445698, "percentage": 66.22, "elapsed_time": "21:58:21", "remaining_time": "11:12:39"}
 {"current_steps": 886, "total_steps": 1329, "eval_loss": 0.5869531631469727, "epoch": 1.9994358251057829, "percentage": 66.67, "elapsed_time": "22:19:06", "remaining_time": "11:09:33"}
+{"current_steps": 890, "total_steps": 1329, "loss": 0.5679, "learning_rate": 4.7913297881339085e-06, "epoch": 2.008462623413258, "percentage": 66.97, "elapsed_time": "22:26:04", "remaining_time": "11:03:57"}
+{"current_steps": 900, "total_steps": 1329, "loss": 0.4798, "learning_rate": 4.789131426105757e-06, "epoch": 2.0310296191819464, "percentage": 67.72, "elapsed_time": "22:40:55", "remaining_time": "10:48:42"}
+{"current_steps": 910, "total_steps": 1329, "loss": 0.4817, "learning_rate": 4.786936087266542e-06, "epoch": 2.0535966149506346, "percentage": 68.47, "elapsed_time": "22:55:46", "remaining_time": "10:33:27"}
+{"current_steps": 920, "total_steps": 1329, "loss": 0.4741, "learning_rate": 4.784743764693455e-06, "epoch": 2.076163610719323, "percentage": 69.22, "elapsed_time": "23:10:37", "remaining_time": "10:18:13"}
+{"current_steps": 930, "total_steps": 1329, "loss": 0.4814, "learning_rate": 4.7825544514858655e-06, "epoch": 2.098730606488011, "percentage": 69.98, "elapsed_time": "23:25:26", "remaining_time": "10:02:58"}
+{"current_steps": 940, "total_steps": 1329, "loss": 0.4856, "learning_rate": 4.780368140765222e-06, "epoch": 2.1212976022566994, "percentage": 70.73, "elapsed_time": "23:40:17", "remaining_time": "9:47:45"}
+{"current_steps": 950, "total_steps": 1329, "loss": 0.4873, "learning_rate": 4.778184825674966e-06, "epoch": 2.143864598025388, "percentage": 71.48, "elapsed_time": "23:55:07", "remaining_time": "9:32:32"}
+{"current_steps": 960, "total_steps": 1329, "loss": 0.4837, "learning_rate": 4.776004499380439e-06, "epoch": 2.1664315937940763, "percentage": 72.23, "elapsed_time": "1 day, 0:09:55", "remaining_time": "9:17:19"}
+{"current_steps": 970, "total_steps": 1329, "loss": 0.4818, "learning_rate": 4.773827155068793e-06, "epoch": 2.1889985895627646, "percentage": 72.99, "elapsed_time": "1 day, 0:24:45", "remaining_time": "9:02:06"}
+{"current_steps": 980, "total_steps": 1329, "loss": 0.4767, "learning_rate": 4.771652785948902e-06, "epoch": 2.211565585331453, "percentage": 73.74, "elapsed_time": "1 day, 0:39:35", "remaining_time": "8:46:54"}
+{"current_steps": 990, "total_steps": 1329, "loss": 0.4765, "learning_rate": 4.769481385251275e-06, "epoch": 2.234132581100141, "percentage": 74.49, "elapsed_time": "1 day, 0:54:25", "remaining_time": "8:31:43"}
+{"current_steps": 1000, "total_steps": 1329, "loss": 0.4855, "learning_rate": 4.767312946227961e-06, "epoch": 2.2566995768688294, "percentage": 75.24, "elapsed_time": "1 day, 1:09:15", "remaining_time": "8:16:32"}
+{"current_steps": 1010, "total_steps": 1329, "loss": 0.4919, "learning_rate": 4.765147462152471e-06, "epoch": 2.2792665726375176, "percentage": 76.0, "elapsed_time": "1 day, 1:24:02", "remaining_time": "8:01:21"}
+{"current_steps": 1020, "total_steps": 1329, "loss": 0.4861, "learning_rate": 4.762984926319677e-06, "epoch": 2.301833568406206, "percentage": 76.75, "elapsed_time": "1 day, 1:38:51", "remaining_time": "7:46:11"}
+{"current_steps": 1030, "total_steps": 1329, "loss": 0.4905, "learning_rate": 4.760825332045738e-06, "epoch": 2.324400564174894, "percentage": 77.5, "elapsed_time": "1 day, 1:53:41", "remaining_time": "7:31:01"}
+{"current_steps": 1040, "total_steps": 1329, "loss": 0.4883, "learning_rate": 4.758668672668006e-06, "epoch": 2.3469675599435824, "percentage": 78.25, "elapsed_time": "1 day, 2:08:31", "remaining_time": "7:15:52"}
+{"current_steps": 1050, "total_steps": 1329, "loss": 0.4861, "learning_rate": 4.756514941544941e-06, "epoch": 2.3695345557122707, "percentage": 79.01, "elapsed_time": "1 day, 2:23:22", "remaining_time": "7:00:43"}
+{"current_steps": 1060, "total_steps": 1329, "loss": 0.4855, "learning_rate": 4.754364132056025e-06, "epoch": 2.392101551480959, "percentage": 79.76, "elapsed_time": "1 day, 2:38:11", "remaining_time": "6:45:34"}
+{"current_steps": 1070, "total_steps": 1329, "loss": 0.4869, "learning_rate": 4.752216237601676e-06, "epoch": 2.414668547249647, "percentage": 80.51, "elapsed_time": "1 day, 2:53:02", "remaining_time": "6:30:26"}
+{"current_steps": 1080, "total_steps": 1329, "loss": 0.49, "learning_rate": 4.750071251603165e-06, "epoch": 2.4372355430183354, "percentage": 81.26, "elapsed_time": "1 day, 3:07:49", "remaining_time": "6:15:18"}
+{"current_steps": 1090, "total_steps": 1329, "loss": 0.488, "learning_rate": 4.7479291675025314e-06, "epoch": 2.459802538787024, "percentage": 82.02, "elapsed_time": "1 day, 3:22:38", "remaining_time": "6:00:10"}
+{"current_steps": 1100, "total_steps": 1329, "loss": 0.491, "learning_rate": 4.745789978762496e-06, "epoch": 2.4823695345557124, "percentage": 82.77, "elapsed_time": "1 day, 3:37:27", "remaining_time": "5:45:03"}
+{"current_steps": 1110, "total_steps": 1329, "loss": 0.4943, "learning_rate": 4.7436536788663765e-06, "epoch": 2.5049365303244007, "percentage": 83.52, "elapsed_time": "1 day, 3:52:18", "remaining_time": "5:29:56"}
+{"current_steps": 1120, "total_steps": 1329, "loss": 0.4848, "learning_rate": 4.74152026131801e-06, "epoch": 2.527503526093089, "percentage": 84.27, "elapsed_time": "1 day, 4:07:07", "remaining_time": "5:14:49"}
+{"current_steps": 1130, "total_steps": 1329, "loss": 0.4899, "learning_rate": 4.739389719641665e-06, "epoch": 2.550070521861777, "percentage": 85.03, "elapsed_time": "1 day, 4:21:58", "remaining_time": "4:59:43"}
+{"current_steps": 1140, "total_steps": 1329, "loss": 0.4883, "learning_rate": 4.7372620473819615e-06, "epoch": 2.5726375176304654, "percentage": 85.78, "elapsed_time": "1 day, 4:36:48", "remaining_time": "4:44:37"}
+{"current_steps": 1150, "total_steps": 1329, "loss": 0.489, "learning_rate": 4.735137238103785e-06, "epoch": 2.5952045133991537, "percentage": 86.53, "elapsed_time": "1 day, 4:51:38", "remaining_time": "4:29:31"}
+{"current_steps": 1160, "total_steps": 1329, "loss": 0.4927, "learning_rate": 4.7330152853922064e-06, "epoch": 2.617771509167842, "percentage": 87.28, "elapsed_time": "1 day, 5:06:26", "remaining_time": "4:14:26"}
+{"current_steps": 1170, "total_steps": 1329, "loss": 0.4817, "learning_rate": 4.730896182852409e-06, "epoch": 2.64033850493653, "percentage": 88.04, "elapsed_time": "1 day, 5:21:16", "remaining_time": "3:59:21"}
+{"current_steps": 1180, "total_steps": 1329, "loss": 0.4924, "learning_rate": 4.72877992410959e-06, "epoch": 2.6629055007052185, "percentage": 88.79, "elapsed_time": "1 day, 5:36:04", "remaining_time": "3:44:16"}
+{"current_steps": 1190, "total_steps": 1329, "loss": 0.4932, "learning_rate": 4.7266665028088985e-06, "epoch": 2.685472496473907, "percentage": 89.54, "elapsed_time": "1 day, 5:50:54", "remaining_time": "3:29:11"}
+{"current_steps": 1200, "total_steps": 1329, "loss": 0.4917, "learning_rate": 4.72455591261534e-06, "epoch": 2.7080394922425954, "percentage": 90.29, "elapsed_time": "1 day, 6:05:43", "remaining_time": "3:14:06"}
+{"current_steps": 1210, "total_steps": 1329, "loss": 0.493, "learning_rate": 4.722448147213712e-06, "epoch": 2.7306064880112837, "percentage": 91.05, "elapsed_time": "1 day, 6:20:33", "remaining_time": "2:59:02"}
+{"current_steps": 1220, "total_steps": 1329, "loss": 0.4896, "learning_rate": 4.720343200308507e-06, "epoch": 2.753173483779972, "percentage": 91.8, "elapsed_time": "1 day, 6:35:22", "remaining_time": "2:43:58"}
+{"current_steps": 1230, "total_steps": 1329, "loss": 0.4961, "learning_rate": 4.7182410656238484e-06, "epoch": 2.77574047954866, "percentage": 92.55, "elapsed_time": "1 day, 6:50:11", "remaining_time": "2:28:55"}
+{"current_steps": 1240, "total_steps": 1329, "loss": 0.4944, "learning_rate": 4.716141736903407e-06, "epoch": 2.7983074753173485, "percentage": 93.3, "elapsed_time": "1 day, 7:05:00", "remaining_time": "2:13:51"}
+{"current_steps": 1250, "total_steps": 1329, "loss": 0.4945, "learning_rate": 4.714045207910318e-06, "epoch": 2.8208744710860367, "percentage": 94.06, "elapsed_time": "1 day, 7:19:50", "remaining_time": "1:58:48"}
+{"current_steps": 1260, "total_steps": 1329, "loss": 0.4963, "learning_rate": 4.71195147242711e-06, "epoch": 2.843441466854725, "percentage": 94.81, "elapsed_time": "1 day, 7:34:40", "remaining_time": "1:43:45"}
+{"current_steps": 1270, "total_steps": 1329, "loss": 0.4932, "learning_rate": 4.709860524255622e-06, "epoch": 2.8660084626234132, "percentage": 95.56, "elapsed_time": "1 day, 7:49:29", "remaining_time": "1:28:42"}
+{"current_steps": 1280, "total_steps": 1329, "loss": 0.4912, "learning_rate": 4.707772357216934e-06, "epoch": 2.8885754583921015, "percentage": 96.31, "elapsed_time": "1 day, 8:04:18", "remaining_time": "1:13:39"}
+{"current_steps": 1290, "total_steps": 1329, "loss": 0.4949, "learning_rate": 4.705686965151282e-06, "epoch": 2.9111424541607898, "percentage": 97.07, "elapsed_time": "1 day, 8:19:08", "remaining_time": "0:58:37"}
+{"current_steps": 1300, "total_steps": 1329, "loss": 0.4908, "learning_rate": 4.703604341917987e-06, "epoch": 2.933709449929478, "percentage": 97.82, "elapsed_time": "1 day, 8:33:58", "remaining_time": "0:43:35"}
+{"current_steps": 1310, "total_steps": 1329, "loss": 0.4946, "learning_rate": 4.701524481395374e-06, "epoch": 2.9562764456981663, "percentage": 98.57, "elapsed_time": "1 day, 8:48:48", "remaining_time": "0:28:33"}
+{"current_steps": 1320, "total_steps": 1329, "loss": 0.4909, "learning_rate": 4.699447377480703e-06, "epoch": 2.9788434414668545, "percentage": 99.32, "elapsed_time": "1 day, 9:03:37", "remaining_time": "0:13:31"}