Training in progress, epoch 2

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:485a64510d6bad28827a5b75051783e8ae16d2f021f1ed0ebab0ac726b01f8b0
+oid sha256:8bfdc85045d899bbf99a7f549e0691c214d59e122275f7ef3663b3c3fecbc6d9
 size 4903351912

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9572e7caeed02b1ea96392522b82acbe4fbe29cd968f0e4de6a65f8eb6e99b3f
+oid sha256:92500fadbfa3c2e309efa4188207e60a3b5efcd9bbf1ee9b51a27c7fd07664a3
 size 4947570872

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1bf28aebe7fee46a5cd953565cbebbacbb048b86e8494282f11c5abd166b6bb8
+oid sha256:e4126c02246a16018b2de46b6942aa48e1046025db797f2edc6e912d8d4369c7
 size 4962221464

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8fa1eac0d99340eff2a58495e2d0175f779a4b57c8825f02310f7cf47849ab09
+oid sha256:6ddade9732b55dd56ee3971ce0dc289808a0c05ad290a1de3fd213470fbf3a72
 size 3670322200

trainer_log.jsonl

@@ -88,3 +88,47 @@
 {"current_steps": 870, "total_steps": 1329, "loss": 0.5469, "learning_rate": 1.818541996830428e-06, "epoch": 1.9633286318758816, "percentage": 65.46, "elapsed_time": "21:44:39", "remaining_time": "11:28:19"}
 {"current_steps": 880, "total_steps": 1329, "loss": 0.5447, "learning_rate": 1.7789223454833599e-06, "epoch": 1.9858956276445698, "percentage": 66.22, "elapsed_time": "21:59:30", "remaining_time": "11:13:15"}
 {"current_steps": 886, "total_steps": 1329, "eval_loss": 0.5841004252433777, "epoch": 1.9994358251057829, "percentage": 66.67, "elapsed_time": "22:20:13", "remaining_time": "11:10:06"}
+{"current_steps": 890, "total_steps": 1329, "loss": 0.5767, "learning_rate": 1.7393026941362917e-06, "epoch": 2.008462623413258, "percentage": 66.97, "elapsed_time": "22:27:13", "remaining_time": "11:04:31"}
+{"current_steps": 900, "total_steps": 1329, "loss": 0.5017, "learning_rate": 1.6996830427892235e-06, "epoch": 2.0310296191819464, "percentage": 67.72, "elapsed_time": "22:42:06", "remaining_time": "10:49:16"}
+{"current_steps": 910, "total_steps": 1329, "loss": 0.5043, "learning_rate": 1.6600633914421555e-06, "epoch": 2.0535966149506346, "percentage": 68.47, "elapsed_time": "22:57:01", "remaining_time": "10:34:01"}
+{"current_steps": 920, "total_steps": 1329, "loss": 0.4963, "learning_rate": 1.6204437400950873e-06, "epoch": 2.076163610719323, "percentage": 69.22, "elapsed_time": "23:11:53", "remaining_time": "10:18:47"}
+{"current_steps": 930, "total_steps": 1329, "loss": 0.5033, "learning_rate": 1.5808240887480191e-06, "epoch": 2.098730606488011, "percentage": 69.98, "elapsed_time": "23:26:45", "remaining_time": "10:03:32"}
+{"current_steps": 940, "total_steps": 1329, "loss": 0.507, "learning_rate": 1.541204437400951e-06, "epoch": 2.1212976022566994, "percentage": 70.73, "elapsed_time": "23:41:39", "remaining_time": "9:48:19"}
+{"current_steps": 950, "total_steps": 1329, "loss": 0.5081, "learning_rate": 1.501584786053883e-06, "epoch": 2.143864598025388, "percentage": 71.48, "elapsed_time": "23:56:30", "remaining_time": "9:33:05"}
+{"current_steps": 960, "total_steps": 1329, "loss": 0.5042, "learning_rate": 1.4619651347068148e-06, "epoch": 2.1664315937940763, "percentage": 72.23, "elapsed_time": "1 day, 0:11:23", "remaining_time": "9:17:52"}
+{"current_steps": 970, "total_steps": 1329, "loss": 0.5009, "learning_rate": 1.4223454833597466e-06, "epoch": 2.1889985895627646, "percentage": 72.99, "elapsed_time": "1 day, 0:26:15", "remaining_time": "9:02:40"}
+{"current_steps": 980, "total_steps": 1329, "loss": 0.4953, "learning_rate": 1.3827258320126786e-06, "epoch": 2.211565585331453, "percentage": 73.74, "elapsed_time": "1 day, 0:41:07", "remaining_time": "8:47:27"}
+{"current_steps": 990, "total_steps": 1329, "loss": 0.4948, "learning_rate": 1.3431061806656104e-06, "epoch": 2.234132581100141, "percentage": 74.49, "elapsed_time": "1 day, 0:55:56", "remaining_time": "8:32:14"}
+{"current_steps": 1000, "total_steps": 1329, "loss": 0.5043, "learning_rate": 1.303486529318542e-06, "epoch": 2.2566995768688294, "percentage": 75.24, "elapsed_time": "1 day, 1:10:47", "remaining_time": "8:17:03"}
+{"current_steps": 1010, "total_steps": 1329, "loss": 0.51, "learning_rate": 1.2638668779714738e-06, "epoch": 2.2792665726375176, "percentage": 76.0, "elapsed_time": "1 day, 1:25:40", "remaining_time": "8:01:52"}
+{"current_steps": 1020, "total_steps": 1329, "loss": 0.5038, "learning_rate": 1.2242472266244059e-06, "epoch": 2.301833568406206, "percentage": 76.75, "elapsed_time": "1 day, 1:40:34", "remaining_time": "7:46:42"}
+{"current_steps": 1030, "total_steps": 1329, "loss": 0.5079, "learning_rate": 1.1846275752773377e-06, "epoch": 2.324400564174894, "percentage": 77.5, "elapsed_time": "1 day, 1:55:27", "remaining_time": "7:31:32"}
+{"current_steps": 1040, "total_steps": 1329, "loss": 0.505, "learning_rate": 1.1450079239302695e-06, "epoch": 2.3469675599435824, "percentage": 78.25, "elapsed_time": "1 day, 2:10:19", "remaining_time": "7:16:22"}
+{"current_steps": 1050, "total_steps": 1329, "loss": 0.5022, "learning_rate": 1.1053882725832013e-06, "epoch": 2.3695345557122707, "percentage": 79.01, "elapsed_time": "1 day, 2:25:11", "remaining_time": "7:01:12"}
+{"current_steps": 1060, "total_steps": 1329, "loss": 0.5014, "learning_rate": 1.0657686212361331e-06, "epoch": 2.392101551480959, "percentage": 79.76, "elapsed_time": "1 day, 2:40:01", "remaining_time": "6:46:02"}
+{"current_steps": 1070, "total_steps": 1329, "loss": 0.5016, "learning_rate": 1.0261489698890651e-06, "epoch": 2.414668547249647, "percentage": 80.51, "elapsed_time": "1 day, 2:54:54", "remaining_time": "6:30:53"}
+{"current_steps": 1080, "total_steps": 1329, "loss": 0.5053, "learning_rate": 9.86529318541997e-07, "epoch": 2.4372355430183354, "percentage": 81.26, "elapsed_time": "1 day, 3:09:43", "remaining_time": "6:15:44"}
+{"current_steps": 1090, "total_steps": 1329, "loss": 0.5032, "learning_rate": 9.469096671949288e-07, "epoch": 2.459802538787024, "percentage": 82.02, "elapsed_time": "1 day, 3:24:37", "remaining_time": "6:00:36"}
+{"current_steps": 1100, "total_steps": 1329, "loss": 0.5058, "learning_rate": 9.072900158478607e-07, "epoch": 2.4823695345557124, "percentage": 82.77, "elapsed_time": "1 day, 3:39:30", "remaining_time": "5:45:28"}
+{"current_steps": 1110, "total_steps": 1329, "loss": 0.5088, "learning_rate": 8.676703645007925e-07, "epoch": 2.5049365303244007, "percentage": 83.52, "elapsed_time": "1 day, 3:54:23", "remaining_time": "5:30:21"}
+{"current_steps": 1120, "total_steps": 1329, "loss": 0.4985, "learning_rate": 8.280507131537244e-07, "epoch": 2.527503526093089, "percentage": 84.27, "elapsed_time": "1 day, 4:09:14", "remaining_time": "5:15:13"}
+{"current_steps": 1130, "total_steps": 1329, "loss": 0.5035, "learning_rate": 7.884310618066561e-07, "epoch": 2.550070521861777, "percentage": 85.03, "elapsed_time": "1 day, 4:24:05", "remaining_time": "5:00:06"}
+{"current_steps": 1140, "total_steps": 1329, "loss": 0.5015, "learning_rate": 7.488114104595879e-07, "epoch": 2.5726375176304654, "percentage": 85.78, "elapsed_time": "1 day, 4:38:58", "remaining_time": "4:44:59"}
+{"current_steps": 1150, "total_steps": 1329, "loss": 0.5019, "learning_rate": 7.091917591125198e-07, "epoch": 2.5952045133991537, "percentage": 86.53, "elapsed_time": "1 day, 4:53:51", "remaining_time": "4:29:52"}
+{"current_steps": 1160, "total_steps": 1329, "loss": 0.5054, "learning_rate": 6.695721077654517e-07, "epoch": 2.617771509167842, "percentage": 87.28, "elapsed_time": "1 day, 5:08:43", "remaining_time": "4:14:46"}
+{"current_steps": 1170, "total_steps": 1329, "loss": 0.4938, "learning_rate": 6.299524564183836e-07, "epoch": 2.64033850493653, "percentage": 88.04, "elapsed_time": "1 day, 5:23:33", "remaining_time": "3:59:39"}
+{"current_steps": 1180, "total_steps": 1329, "loss": 0.5049, "learning_rate": 5.903328050713154e-07, "epoch": 2.6629055007052185, "percentage": 88.79, "elapsed_time": "1 day, 5:38:25", "remaining_time": "3:44:33"}
+{"current_steps": 1190, "total_steps": 1329, "loss": 0.5054, "learning_rate": 5.507131537242473e-07, "epoch": 2.685472496473907, "percentage": 89.54, "elapsed_time": "1 day, 5:53:18", "remaining_time": "3:29:28"}
+{"current_steps": 1200, "total_steps": 1329, "loss": 0.5036, "learning_rate": 5.110935023771791e-07, "epoch": 2.7080394922425954, "percentage": 90.29, "elapsed_time": "1 day, 6:08:11", "remaining_time": "3:14:22"}
+{"current_steps": 1210, "total_steps": 1329, "loss": 0.5043, "learning_rate": 4.71473851030111e-07, "epoch": 2.7306064880112837, "percentage": 91.05, "elapsed_time": "1 day, 6:23:02", "remaining_time": "2:59:17"}
+{"current_steps": 1220, "total_steps": 1329, "loss": 0.5006, "learning_rate": 4.3185419968304285e-07, "epoch": 2.753173483779972, "percentage": 91.8, "elapsed_time": "1 day, 6:37:55", "remaining_time": "2:44:12"}
+{"current_steps": 1230, "total_steps": 1329, "loss": 0.5075, "learning_rate": 3.922345483359747e-07, "epoch": 2.77574047954866, "percentage": 92.55, "elapsed_time": "1 day, 6:52:46", "remaining_time": "2:29:07"}
+{"current_steps": 1240, "total_steps": 1329, "loss": 0.5057, "learning_rate": 3.5261489698890647e-07, "epoch": 2.7983074753173485, "percentage": 93.3, "elapsed_time": "1 day, 7:07:36", "remaining_time": "2:14:02"}
+{"current_steps": 1250, "total_steps": 1329, "loss": 0.5053, "learning_rate": 3.129952456418384e-07, "epoch": 2.8208744710860367, "percentage": 94.06, "elapsed_time": "1 day, 7:22:29", "remaining_time": "1:58:58"}
+{"current_steps": 1260, "total_steps": 1329, "loss": 0.5069, "learning_rate": 2.7337559429477025e-07, "epoch": 2.843441466854725, "percentage": 94.81, "elapsed_time": "1 day, 7:37:21", "remaining_time": "1:43:54"}
+{"current_steps": 1270, "total_steps": 1329, "loss": 0.504, "learning_rate": 2.3375594294770207e-07, "epoch": 2.8660084626234132, "percentage": 95.56, "elapsed_time": "1 day, 7:52:13", "remaining_time": "1:28:50"}
+{"current_steps": 1280, "total_steps": 1329, "loss": 0.5014, "learning_rate": 1.9413629160063393e-07, "epoch": 2.8885754583921015, "percentage": 96.31, "elapsed_time": "1 day, 8:07:06", "remaining_time": "1:13:46"}
+{"current_steps": 1290, "total_steps": 1329, "loss": 0.5053, "learning_rate": 1.5451664025356577e-07, "epoch": 2.9111424541607898, "percentage": 97.07, "elapsed_time": "1 day, 8:22:00", "remaining_time": "0:58:42"}
+{"current_steps": 1300, "total_steps": 1329, "loss": 0.5003, "learning_rate": 1.1489698890649763e-07, "epoch": 2.933709449929478, "percentage": 97.82, "elapsed_time": "1 day, 8:36:52", "remaining_time": "0:43:39"}
+{"current_steps": 1310, "total_steps": 1329, "loss": 0.5043, "learning_rate": 7.527733755942949e-08, "epoch": 2.9562764456981663, "percentage": 98.57, "elapsed_time": "1 day, 8:51:45", "remaining_time": "0:28:35"}
+{"current_steps": 1320, "total_steps": 1329, "loss": 0.4998, "learning_rate": 3.565768621236134e-08, "epoch": 2.9788434414668545, "percentage": 99.32, "elapsed_time": "1 day, 9:06:35", "remaining_time": "0:13:32"}