Training in progress, epoch 2

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dbbcdacea2e06a56e6af78c56a48cab17beff03c5a2a8dbe0d6a428f4a81f1f1
+oid sha256:4c0e2fd3bbbadcdf37892d384eb6e8f1ba11e46565ec61a118de5ab10054caa6
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1819d4a263e0b584854c253083657a96a8cef0920f93b6150455d96374f24a9e
+oid sha256:47012e74fafb5b856ae679cb38d61e5f2598159c33963bca286edeb370d34323
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:349ed342de23d53018e7ac529de65eecd60a362e8157a605ba07d0a2e8493fb3
+oid sha256:fa4b22f38357e7563bf835ecdd8009f0de3e289e37573e86c264617afb1918f7
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:676f8bea0e39053f01bc76cdcfc10b20be11489fba4c30018d9b7e01acade428
+oid sha256:2109fbaef76135fadd58e4a5a0690ea572491c657319eb4add1ca9684468f594
 size 1089994880

trainer_log.jsonl

@@ -68,3 +68,37 @@
 {"current_steps": 670, "total_steps": 1023, "loss": 0.5984, "learning_rate": 1.4607936393836542e-06, "epoch": 1.9615947329919532, "percentage": 65.49, "elapsed_time": "9:53:25", "remaining_time": "5:12:39"}
 {"current_steps": 680, "total_steps": 1023, "loss": 0.6129, "learning_rate": 1.387784194967904e-06, "epoch": 1.9908558888076078, "percentage": 66.47, "elapsed_time": "10:02:12", "remaining_time": "5:03:45"}
 {"current_steps": 682, "total_steps": 1023, "eval_loss": 0.6351735591888428, "epoch": 1.9967081199707388, "percentage": 66.67, "elapsed_time": "10:10:21", "remaining_time": "5:05:10"}
+{"current_steps": 690, "total_steps": 1023, "loss": 0.633, "learning_rate": 1.3159389099885194e-06, "epoch": 2.0212143379663496, "percentage": 67.45, "elapsed_time": "10:17:52", "remaining_time": "4:58:11"}
+{"current_steps": 700, "total_steps": 1023, "loss": 0.5844, "learning_rate": 1.2453329851089787e-06, "epoch": 2.050475493782004, "percentage": 68.43, "elapsed_time": "10:26:38", "remaining_time": "4:49:09"}
+{"current_steps": 710, "total_steps": 1023, "loss": 0.5866, "learning_rate": 1.1760403237510255e-06, "epoch": 2.0797366495976592, "percentage": 69.4, "elapsed_time": "10:35:24", "remaining_time": "4:40:07"}
+{"current_steps": 720, "total_steps": 1023, "loss": 0.5876, "learning_rate": 1.108133454739757e-06, "epoch": 2.108997805413314, "percentage": 70.38, "elapsed_time": "10:44:11", "remaining_time": "4:31:05"}
+{"current_steps": 730, "total_steps": 1023, "loss": 0.5892, "learning_rate": 1.0416834563874873e-06, "epoch": 2.1382589612289684, "percentage": 71.36, "elapsed_time": "10:52:57", "remaining_time": "4:22:04"}
+{"current_steps": 740, "total_steps": 1023, "loss": 0.5939, "learning_rate": 9.767598820958859e-07, "epoch": 2.1675201170446234, "percentage": 72.34, "elapsed_time": "11:01:42", "remaining_time": "4:13:03"}
+{"current_steps": 750, "total_steps": 1023, "loss": 0.5927, "learning_rate": 9.134306875542273e-07, "epoch": 2.196781272860278, "percentage": 73.31, "elapsed_time": "11:10:28", "remaining_time": "4:04:03"}
+{"current_steps": 760, "total_steps": 1023, "loss": 0.5898, "learning_rate": 8.517621596099837e-07, "epoch": 2.2260424286759326, "percentage": 74.29, "elapsed_time": "11:19:13", "remaining_time": "3:55:02"}
+{"current_steps": 770, "total_steps": 1023, "loss": 0.588, "learning_rate": 7.918188468861892e-07, "epoch": 2.255303584491587, "percentage": 75.27, "elapsed_time": "11:28:00", "remaining_time": "3:46:03"}
+{"current_steps": 780, "total_steps": 1023, "loss": 0.5875, "learning_rate": 7.336634922182184e-07, "epoch": 2.284564740307242, "percentage": 76.25, "elapsed_time": "11:36:46", "remaining_time": "3:37:04"}
+{"current_steps": 790, "total_steps": 1023, "loss": 0.5854, "learning_rate": 6.773569669806765e-07, "epoch": 2.313825896122897, "percentage": 77.22, "elapsed_time": "11:45:31", "remaining_time": "3:28:05"}
+{"current_steps": 800, "total_steps": 1023, "loss": 0.5931, "learning_rate": 6.22958207373166e-07, "epoch": 2.3430870519385514, "percentage": 78.2, "elapsed_time": "11:54:17", "remaining_time": "3:19:06"}
+{"current_steps": 810, "total_steps": 1023, "loss": 0.5881, "learning_rate": 5.705241527315924e-07, "epoch": 2.3723482077542064, "percentage": 79.18, "elapsed_time": "12:03:03", "remaining_time": "3:10:08"}
+{"current_steps": 820, "total_steps": 1023, "loss": 0.5943, "learning_rate": 5.201096859296056e-07, "epoch": 2.401609363569861, "percentage": 80.16, "elapsed_time": "12:11:48", "remaining_time": "3:01:10"}
+{"current_steps": 830, "total_steps": 1023, "loss": 0.5866, "learning_rate": 4.717675759325363e-07, "epoch": 2.4308705193855156, "percentage": 81.13, "elapsed_time": "12:20:34", "remaining_time": "2:52:12"}
+{"current_steps": 840, "total_steps": 1023, "loss": 0.5869, "learning_rate": 4.255484225639739e-07, "epoch": 2.4601316752011706, "percentage": 82.11, "elapsed_time": "12:29:19", "remaining_time": "2:43:14"}
+{"current_steps": 850, "total_steps": 1023, "loss": 0.5886, "learning_rate": 3.8150060354278373e-07, "epoch": 2.489392831016825, "percentage": 83.09, "elapsed_time": "12:38:05", "remaining_time": "2:34:17"}
+{"current_steps": 860, "total_steps": 1023, "loss": 0.5818, "learning_rate": 3.396702238460167e-07, "epoch": 2.5186539868324798, "percentage": 84.07, "elapsed_time": "12:46:51", "remaining_time": "2:25:20"}
+{"current_steps": 870, "total_steps": 1023, "loss": 0.5842, "learning_rate": 3.001010674506932e-07, "epoch": 2.547915142648135, "percentage": 85.04, "elapsed_time": "12:55:38", "remaining_time": "2:16:24"}
+{"current_steps": 880, "total_steps": 1023, "loss": 0.5882, "learning_rate": 2.628345515049954e-07, "epoch": 2.5771762984637894, "percentage": 86.02, "elapsed_time": "13:04:24", "remaining_time": "2:07:27"}
+{"current_steps": 890, "total_steps": 1023, "loss": 0.5875, "learning_rate": 2.2790968297681792e-07, "epoch": 2.606437454279444, "percentage": 87.0, "elapsed_time": "13:13:10", "remaining_time": "1:58:31"}
+{"current_steps": 900, "total_steps": 1023, "loss": 0.5921, "learning_rate": 1.9536301782506373e-07, "epoch": 2.6356986100950985, "percentage": 87.98, "elapsed_time": "13:21:55", "remaining_time": "1:49:35"}
+{"current_steps": 910, "total_steps": 1023, "loss": 0.5924, "learning_rate": 1.652286227364211e-07, "epoch": 2.6649597659107536, "percentage": 88.95, "elapsed_time": "13:30:40", "remaining_time": "1:40:40"}
+{"current_steps": 920, "total_steps": 1023, "loss": 0.5864, "learning_rate": 1.375380394676623e-07, "epoch": 2.694220921726408, "percentage": 89.93, "elapsed_time": "13:39:26", "remaining_time": "1:31:44"}
+{"current_steps": 930, "total_steps": 1023, "loss": 0.5898, "learning_rate": 1.1232025183080141e-07, "epoch": 2.723482077542063, "percentage": 90.91, "elapsed_time": "13:48:12", "remaining_time": "1:22:49"}
+{"current_steps": 940, "total_steps": 1023, "loss": 0.5888, "learning_rate": 8.960165535565168e-08, "epoch": 2.7527432333577178, "percentage": 91.89, "elapsed_time": "13:56:57", "remaining_time": "1:13:54"}
+{"current_steps": 950, "total_steps": 1023, "loss": 0.5875, "learning_rate": 6.940602966155341e-08, "epoch": 2.7820043891733723, "percentage": 92.86, "elapsed_time": "14:05:42", "remaining_time": "1:04:59"}
+{"current_steps": 960, "total_steps": 1023, "loss": 0.5914, "learning_rate": 5.175451356717737e-08, "epoch": 2.811265544989027, "percentage": 93.84, "elapsed_time": "14:14:27", "remaining_time": "0:56:04"}
+{"current_steps": 970, "total_steps": 1023, "loss": 0.5871, "learning_rate": 3.6665582964462986e-08, "epoch": 2.840526700804682, "percentage": 94.82, "elapsed_time": "14:23:14", "remaining_time": "0:47:09"}
+{"current_steps": 980, "total_steps": 1023, "loss": 0.5869, "learning_rate": 2.41550314798511e-08, "epoch": 2.8697878566203365, "percentage": 95.8, "elapsed_time": "14:32:00", "remaining_time": "0:38:15"}
+{"current_steps": 990, "total_steps": 1023, "loss": 0.5796, "learning_rate": 1.4235953943048942e-08, "epoch": 2.899049012435991, "percentage": 96.77, "elapsed_time": "14:40:47", "remaining_time": "0:29:21"}
+{"current_steps": 1000, "total_steps": 1023, "loss": 0.5876, "learning_rate": 6.918732680635809e-09, "epoch": 2.928310168251646, "percentage": 97.75, "elapsed_time": "14:49:32", "remaining_time": "0:20:27"}
+{"current_steps": 1010, "total_steps": 1023, "loss": 0.5789, "learning_rate": 2.21102664885231e-09, "epoch": 2.9575713240673007, "percentage": 98.73, "elapsed_time": "14:58:18", "remaining_time": "0:11:33"}
+{"current_steps": 1020, "total_steps": 1023, "loss": 0.591, "learning_rate": 1.177634169488573e-10, "epoch": 2.9868324798829553, "percentage": 99.71, "elapsed_time": "15:07:04", "remaining_time": "0:02:40"}