Hugging Face's logo

mlfoundations-dev
/
llama3-1_8b_dcft-it

Text Generation
Transformers
Safetensors
llama
llama-factory
full
Generated from Trainer
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
gsmyrnis commited on
Commit
5a584db
•
1 Parent(s): 1e9279e

Training in progress, epoch 3

Browse files
Files changed (5) hide show
  1. model-00001-of-00004.safetensors +1 -1
  2. model-00002-of-00004.safetensors +1 -1
  3. model-00003-of-00004.safetensors +1 -1
  4. model-00004-of-00004.safetensors +1 -1
  5. trainer_log.jsonl +154 -0
model-00001-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:59f93c26509c5b454dda27fe5d23df2214389092cbe4ce80bafca4004ef2c032
3
  size 4976698672
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:377f140f768e06ba5524e662e502a1b956c6736006c40ca18a80287868dd8953
3
  size 4976698672
model-00002-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3797716d0136cd7ada7607cea5c3db3a0a132dc25f1e76fcbc117cd048dd00d4
3
  size 4999802720
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1ec30a25b2fb1771a572a266d781efcecb2b2434a97e515e027f58e11e20761
3
  size 4999802720
model-00003-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b55b519346fb6e34a8049742afb6d37742471dd21db718f4eca7261a9f1293d3
3
  size 4915916176
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9bce547e0e821fd6651b0a167d98a92e5fa53a6a43d6df85681cdb9922d57e7f
3
  size 4915916176
model-00004-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bc07798e43cbe66000b19eb49adbd3a755aa8ef65424fd35707de109e70a9ee0
3
  size 1168138808
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30841c3fa97577d9d8958f111aaaf23bcabe868094a38e08ab86c7e567fc370f
3
  size 1168138808
trainer_log.jsonl CHANGED
@@ -314,3 +314,157 @@
314
  {"current_steps": 3120, "total_steps": 4665, "loss": 0.6404, "learning_rate": 5e-06, "epoch": 2.0064308681672025, "percentage": 66.88, "elapsed_time": "5:36:34", "remaining_time": "2:46:40"}
315
  {"current_steps": 3130, "total_steps": 4665, "loss": 0.64, "learning_rate": 5e-06, "epoch": 2.012861736334405, "percentage": 67.1, "elapsed_time": "5:37:36", "remaining_time": "2:45:34"}
316
  {"current_steps": 3140, "total_steps": 4665, "loss": 0.6359, "learning_rate": 5e-06, "epoch": 2.0192926045016075, "percentage": 67.31, "elapsed_time": "5:38:39", "remaining_time": "2:44:28"}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
314
  {"current_steps": 3120, "total_steps": 4665, "loss": 0.6404, "learning_rate": 5e-06, "epoch": 2.0064308681672025, "percentage": 66.88, "elapsed_time": "5:36:34", "remaining_time": "2:46:40"}
315
  {"current_steps": 3130, "total_steps": 4665, "loss": 0.64, "learning_rate": 5e-06, "epoch": 2.012861736334405, "percentage": 67.1, "elapsed_time": "5:37:36", "remaining_time": "2:45:34"}
316
  {"current_steps": 3140, "total_steps": 4665, "loss": 0.6359, "learning_rate": 5e-06, "epoch": 2.0192926045016075, "percentage": 67.31, "elapsed_time": "5:38:39", "remaining_time": "2:44:28"}
317
+ {"current_steps": 3150, "total_steps": 4665, "loss": 0.6381, "learning_rate": 5e-06, "epoch": 2.0257234726688105, "percentage": 67.52, "elapsed_time": "5:39:41", "remaining_time": "2:43:22"}
318
+ {"current_steps": 3160, "total_steps": 4665, "loss": 0.6372, "learning_rate": 5e-06, "epoch": 2.032154340836013, "percentage": 67.74, "elapsed_time": "5:40:43", "remaining_time": "2:42:16"}
319
+ {"current_steps": 3170, "total_steps": 4665, "loss": 0.6394, "learning_rate": 5e-06, "epoch": 2.0385852090032155, "percentage": 67.95, "elapsed_time": "5:41:46", "remaining_time": "2:41:10"}
320
+ {"current_steps": 3180, "total_steps": 4665, "loss": 0.6429, "learning_rate": 5e-06, "epoch": 2.045016077170418, "percentage": 68.17, "elapsed_time": "5:42:48", "remaining_time": "2:40:05"}
321
+ {"current_steps": 3190, "total_steps": 4665, "loss": 0.6352, "learning_rate": 5e-06, "epoch": 2.0514469453376205, "percentage": 68.38, "elapsed_time": "5:43:50", "remaining_time": "2:38:59"}
322
+ {"current_steps": 3200, "total_steps": 4665, "loss": 0.6394, "learning_rate": 5e-06, "epoch": 2.057877813504823, "percentage": 68.6, "elapsed_time": "5:44:53", "remaining_time": "2:37:53"}
323
+ {"current_steps": 3210, "total_steps": 4665, "loss": 0.6367, "learning_rate": 5e-06, "epoch": 2.0643086816720255, "percentage": 68.81, "elapsed_time": "5:45:55", "remaining_time": "2:36:47"}
324
+ {"current_steps": 3220, "total_steps": 4665, "loss": 0.6364, "learning_rate": 5e-06, "epoch": 2.0707395498392285, "percentage": 69.02, "elapsed_time": "5:46:57", "remaining_time": "2:35:42"}
325
+ {"current_steps": 3230, "total_steps": 4665, "loss": 0.6441, "learning_rate": 5e-06, "epoch": 2.077170418006431, "percentage": 69.24, "elapsed_time": "5:48:00", "remaining_time": "2:34:36"}
326
+ {"current_steps": 3240, "total_steps": 4665, "loss": 0.6364, "learning_rate": 5e-06, "epoch": 2.0836012861736335, "percentage": 69.45, "elapsed_time": "5:49:02", "remaining_time": "2:33:30"}
327
+ {"current_steps": 3250, "total_steps": 4665, "loss": 0.6334, "learning_rate": 5e-06, "epoch": 2.090032154340836, "percentage": 69.67, "elapsed_time": "5:50:04", "remaining_time": "2:32:25"}
328
+ {"current_steps": 3260, "total_steps": 4665, "loss": 0.6395, "learning_rate": 5e-06, "epoch": 2.0964630225080385, "percentage": 69.88, "elapsed_time": "5:51:07", "remaining_time": "2:31:19"}
329
+ {"current_steps": 3270, "total_steps": 4665, "loss": 0.6444, "learning_rate": 5e-06, "epoch": 2.102893890675241, "percentage": 70.1, "elapsed_time": "5:52:09", "remaining_time": "2:30:13"}
330
+ {"current_steps": 3280, "total_steps": 4665, "loss": 0.6384, "learning_rate": 5e-06, "epoch": 2.1093247588424435, "percentage": 70.31, "elapsed_time": "5:53:11", "remaining_time": "2:29:08"}
331
+ {"current_steps": 3290, "total_steps": 4665, "loss": 0.638, "learning_rate": 5e-06, "epoch": 2.1157556270096465, "percentage": 70.53, "elapsed_time": "5:54:14", "remaining_time": "2:28:02"}
332
+ {"current_steps": 3300, "total_steps": 4665, "loss": 0.6466, "learning_rate": 5e-06, "epoch": 2.122186495176849, "percentage": 70.74, "elapsed_time": "5:55:16", "remaining_time": "2:26:57"}
333
+ {"current_steps": 3310, "total_steps": 4665, "loss": 0.6297, "learning_rate": 5e-06, "epoch": 2.1286173633440515, "percentage": 70.95, "elapsed_time": "5:56:18", "remaining_time": "2:25:51"}
334
+ {"current_steps": 3320, "total_steps": 4665, "loss": 0.6408, "learning_rate": 5e-06, "epoch": 2.135048231511254, "percentage": 71.17, "elapsed_time": "5:57:21", "remaining_time": "2:24:46"}
335
+ {"current_steps": 3330, "total_steps": 4665, "loss": 0.6416, "learning_rate": 5e-06, "epoch": 2.1414790996784565, "percentage": 71.38, "elapsed_time": "5:58:23", "remaining_time": "2:23:40"}
336
+ {"current_steps": 3340, "total_steps": 4665, "loss": 0.636, "learning_rate": 5e-06, "epoch": 2.147909967845659, "percentage": 71.6, "elapsed_time": "5:59:25", "remaining_time": "2:22:35"}
337
+ {"current_steps": 3350, "total_steps": 4665, "loss": 0.6395, "learning_rate": 5e-06, "epoch": 2.154340836012862, "percentage": 71.81, "elapsed_time": "6:00:28", "remaining_time": "2:21:29"}
338
+ {"current_steps": 3360, "total_steps": 4665, "loss": 0.6233, "learning_rate": 5e-06, "epoch": 2.1607717041800645, "percentage": 72.03, "elapsed_time": "6:01:30", "remaining_time": "2:20:24"}
339
+ {"current_steps": 3370, "total_steps": 4665, "loss": 0.6417, "learning_rate": 5e-06, "epoch": 2.167202572347267, "percentage": 72.24, "elapsed_time": "6:02:33", "remaining_time": "2:19:19"}
340
+ {"current_steps": 3380, "total_steps": 4665, "loss": 0.6418, "learning_rate": 5e-06, "epoch": 2.1736334405144695, "percentage": 72.45, "elapsed_time": "6:03:35", "remaining_time": "2:18:13"}
341
+ {"current_steps": 3390, "total_steps": 4665, "loss": 0.6369, "learning_rate": 5e-06, "epoch": 2.180064308681672, "percentage": 72.67, "elapsed_time": "6:04:37", "remaining_time": "2:17:08"}
342
+ {"current_steps": 3400, "total_steps": 4665, "loss": 0.6427, "learning_rate": 5e-06, "epoch": 2.1864951768488745, "percentage": 72.88, "elapsed_time": "6:05:40", "remaining_time": "2:16:03"}
343
+ {"current_steps": 3410, "total_steps": 4665, "loss": 0.6456, "learning_rate": 5e-06, "epoch": 2.192926045016077, "percentage": 73.1, "elapsed_time": "6:06:42", "remaining_time": "2:14:57"}
344
+ {"current_steps": 3420, "total_steps": 4665, "loss": 0.6435, "learning_rate": 5e-06, "epoch": 2.19935691318328, "percentage": 73.31, "elapsed_time": "6:07:44", "remaining_time": "2:13:52"}
345
+ {"current_steps": 3430, "total_steps": 4665, "loss": 0.6383, "learning_rate": 5e-06, "epoch": 2.2057877813504825, "percentage": 73.53, "elapsed_time": "6:08:47", "remaining_time": "2:12:47"}
346
+ {"current_steps": 3440, "total_steps": 4665, "loss": 0.634, "learning_rate": 5e-06, "epoch": 2.212218649517685, "percentage": 73.74, "elapsed_time": "6:09:49", "remaining_time": "2:11:41"}
347
+ {"current_steps": 3450, "total_steps": 4665, "loss": 0.6383, "learning_rate": 5e-06, "epoch": 2.2186495176848875, "percentage": 73.95, "elapsed_time": "6:10:51", "remaining_time": "2:10:36"}
348
+ {"current_steps": 3460, "total_steps": 4665, "loss": 0.6413, "learning_rate": 5e-06, "epoch": 2.22508038585209, "percentage": 74.17, "elapsed_time": "6:11:54", "remaining_time": "2:09:31"}
349
+ {"current_steps": 3470, "total_steps": 4665, "loss": 0.6348, "learning_rate": 5e-06, "epoch": 2.2315112540192925, "percentage": 74.38, "elapsed_time": "6:12:56", "remaining_time": "2:08:26"}
350
+ {"current_steps": 3480, "total_steps": 4665, "loss": 0.6489, "learning_rate": 5e-06, "epoch": 2.237942122186495, "percentage": 74.6, "elapsed_time": "6:13:58", "remaining_time": "2:07:20"}
351
+ {"current_steps": 3490, "total_steps": 4665, "loss": 0.6451, "learning_rate": 5e-06, "epoch": 2.244372990353698, "percentage": 74.81, "elapsed_time": "6:15:01", "remaining_time": "2:06:15"}
352
+ {"current_steps": 3500, "total_steps": 4665, "loss": 0.6355, "learning_rate": 5e-06, "epoch": 2.2508038585209005, "percentage": 75.03, "elapsed_time": "6:16:03", "remaining_time": "2:05:10"}
353
+ {"current_steps": 3510, "total_steps": 4665, "loss": 0.6374, "learning_rate": 5e-06, "epoch": 2.257234726688103, "percentage": 75.24, "elapsed_time": "6:17:05", "remaining_time": "2:04:05"}
354
+ {"current_steps": 3520, "total_steps": 4665, "loss": 0.6504, "learning_rate": 5e-06, "epoch": 2.2636655948553055, "percentage": 75.46, "elapsed_time": "6:18:08", "remaining_time": "2:03:00"}
355
+ {"current_steps": 3530, "total_steps": 4665, "loss": 0.6338, "learning_rate": 5e-06, "epoch": 2.270096463022508, "percentage": 75.67, "elapsed_time": "6:19:10", "remaining_time": "2:01:55"}
356
+ {"current_steps": 3540, "total_steps": 4665, "loss": 0.634, "learning_rate": 5e-06, "epoch": 2.2765273311897105, "percentage": 75.88, "elapsed_time": "6:20:13", "remaining_time": "2:00:49"}
357
+ {"current_steps": 3550, "total_steps": 4665, "loss": 0.6382, "learning_rate": 5e-06, "epoch": 2.282958199356913, "percentage": 76.1, "elapsed_time": "6:21:15", "remaining_time": "1:59:44"}
358
+ {"current_steps": 3560, "total_steps": 4665, "loss": 0.6351, "learning_rate": 5e-06, "epoch": 2.289389067524116, "percentage": 76.31, "elapsed_time": "6:22:17", "remaining_time": "1:58:39"}
359
+ {"current_steps": 3570, "total_steps": 4665, "loss": 0.6406, "learning_rate": 5e-06, "epoch": 2.2958199356913185, "percentage": 76.53, "elapsed_time": "6:23:20", "remaining_time": "1:57:34"}
360
+ {"current_steps": 3580, "total_steps": 4665, "loss": 0.6373, "learning_rate": 5e-06, "epoch": 2.302250803858521, "percentage": 76.74, "elapsed_time": "6:24:22", "remaining_time": "1:56:29"}
361
+ {"current_steps": 3590, "total_steps": 4665, "loss": 0.637, "learning_rate": 5e-06, "epoch": 2.3086816720257235, "percentage": 76.96, "elapsed_time": "6:25:25", "remaining_time": "1:55:24"}
362
+ {"current_steps": 3600, "total_steps": 4665, "loss": 0.6429, "learning_rate": 5e-06, "epoch": 2.315112540192926, "percentage": 77.17, "elapsed_time": "6:26:27", "remaining_time": "1:54:19"}
363
+ {"current_steps": 3610, "total_steps": 4665, "loss": 0.6411, "learning_rate": 5e-06, "epoch": 2.3215434083601285, "percentage": 77.38, "elapsed_time": "6:27:30", "remaining_time": "1:53:14"}
364
+ {"current_steps": 3620, "total_steps": 4665, "loss": 0.6404, "learning_rate": 5e-06, "epoch": 2.327974276527331, "percentage": 77.6, "elapsed_time": "6:28:32", "remaining_time": "1:52:09"}
365
+ {"current_steps": 3630, "total_steps": 4665, "loss": 0.6409, "learning_rate": 5e-06, "epoch": 2.334405144694534, "percentage": 77.81, "elapsed_time": "6:29:34", "remaining_time": "1:51:04"}
366
+ {"current_steps": 3640, "total_steps": 4665, "loss": 0.6397, "learning_rate": 5e-06, "epoch": 2.3408360128617365, "percentage": 78.03, "elapsed_time": "6:30:37", "remaining_time": "1:49:59"}
367
+ {"current_steps": 3650, "total_steps": 4665, "loss": 0.6437, "learning_rate": 5e-06, "epoch": 2.347266881028939, "percentage": 78.24, "elapsed_time": "6:31:39", "remaining_time": "1:48:54"}
368
+ {"current_steps": 3660, "total_steps": 4665, "loss": 0.6473, "learning_rate": 5e-06, "epoch": 2.3536977491961415, "percentage": 78.46, "elapsed_time": "6:32:42", "remaining_time": "1:47:49"}
369
+ {"current_steps": 3670, "total_steps": 4665, "loss": 0.6337, "learning_rate": 5e-06, "epoch": 2.360128617363344, "percentage": 78.67, "elapsed_time": "6:33:44", "remaining_time": "1:46:45"}
370
+ {"current_steps": 3680, "total_steps": 4665, "loss": 0.6393, "learning_rate": 5e-06, "epoch": 2.3665594855305465, "percentage": 78.89, "elapsed_time": "6:34:47", "remaining_time": "1:45:40"}
371
+ {"current_steps": 3690, "total_steps": 4665, "loss": 0.6354, "learning_rate": 5e-06, "epoch": 2.372990353697749, "percentage": 79.1, "elapsed_time": "6:35:49", "remaining_time": "1:44:35"}
372
+ {"current_steps": 3700, "total_steps": 4665, "loss": 0.6403, "learning_rate": 5e-06, "epoch": 2.379421221864952, "percentage": 79.31, "elapsed_time": "6:36:51", "remaining_time": "1:43:30"}
373
+ {"current_steps": 3710, "total_steps": 4665, "loss": 0.6377, "learning_rate": 5e-06, "epoch": 2.3858520900321545, "percentage": 79.53, "elapsed_time": "6:37:54", "remaining_time": "1:42:25"}
374
+ {"current_steps": 3720, "total_steps": 4665, "loss": 0.6391, "learning_rate": 5e-06, "epoch": 2.392282958199357, "percentage": 79.74, "elapsed_time": "6:38:56", "remaining_time": "1:41:20"}
375
+ {"current_steps": 3730, "total_steps": 4665, "loss": 0.6415, "learning_rate": 5e-06, "epoch": 2.3987138263665595, "percentage": 79.96, "elapsed_time": "6:39:59", "remaining_time": "1:40:15"}
376
+ {"current_steps": 3740, "total_steps": 4665, "loss": 0.6348, "learning_rate": 5e-06, "epoch": 2.405144694533762, "percentage": 80.17, "elapsed_time": "6:41:01", "remaining_time": "1:39:11"}
377
+ {"current_steps": 3750, "total_steps": 4665, "loss": 0.6415, "learning_rate": 5e-06, "epoch": 2.4115755627009645, "percentage": 80.39, "elapsed_time": "6:42:03", "remaining_time": "1:38:06"}
378
+ {"current_steps": 3760, "total_steps": 4665, "loss": 0.6377, "learning_rate": 5e-06, "epoch": 2.418006430868167, "percentage": 80.6, "elapsed_time": "6:43:06", "remaining_time": "1:37:01"}
379
+ {"current_steps": 3770, "total_steps": 4665, "loss": 0.6435, "learning_rate": 5e-06, "epoch": 2.42443729903537, "percentage": 80.81, "elapsed_time": "6:44:08", "remaining_time": "1:35:56"}
380
+ {"current_steps": 3780, "total_steps": 4665, "loss": 0.6401, "learning_rate": 5e-06, "epoch": 2.4308681672025725, "percentage": 81.03, "elapsed_time": "6:45:10", "remaining_time": "1:34:51"}
381
+ {"current_steps": 3790, "total_steps": 4665, "loss": 0.6454, "learning_rate": 5e-06, "epoch": 2.437299035369775, "percentage": 81.24, "elapsed_time": "6:46:13", "remaining_time": "1:33:47"}
382
+ {"current_steps": 3800, "total_steps": 4665, "loss": 0.6484, "learning_rate": 5e-06, "epoch": 2.4437299035369775, "percentage": 81.46, "elapsed_time": "6:47:15", "remaining_time": "1:32:42"}
383
+ {"current_steps": 3810, "total_steps": 4665, "loss": 0.6431, "learning_rate": 5e-06, "epoch": 2.45016077170418, "percentage": 81.67, "elapsed_time": "6:48:18", "remaining_time": "1:31:37"}
384
+ {"current_steps": 3820, "total_steps": 4665, "loss": 0.6371, "learning_rate": 5e-06, "epoch": 2.4565916398713825, "percentage": 81.89, "elapsed_time": "6:49:20", "remaining_time": "1:30:32"}
385
+ {"current_steps": 3830, "total_steps": 4665, "loss": 0.6398, "learning_rate": 5e-06, "epoch": 2.463022508038585, "percentage": 82.1, "elapsed_time": "6:50:22", "remaining_time": "1:29:28"}
386
+ {"current_steps": 3840, "total_steps": 4665, "loss": 0.6419, "learning_rate": 5e-06, "epoch": 2.469453376205788, "percentage": 82.32, "elapsed_time": "6:51:25", "remaining_time": "1:28:23"}
387
+ {"current_steps": 3850, "total_steps": 4665, "loss": 0.6379, "learning_rate": 5e-06, "epoch": 2.4758842443729905, "percentage": 82.53, "elapsed_time": "6:52:27", "remaining_time": "1:27:18"}
388
+ {"current_steps": 3860, "total_steps": 4665, "loss": 0.6439, "learning_rate": 5e-06, "epoch": 2.482315112540193, "percentage": 82.74, "elapsed_time": "6:53:29", "remaining_time": "1:26:14"}
389
+ {"current_steps": 3870, "total_steps": 4665, "loss": 0.6446, "learning_rate": 5e-06, "epoch": 2.4887459807073955, "percentage": 82.96, "elapsed_time": "6:54:32", "remaining_time": "1:25:09"}
390
+ {"current_steps": 3880, "total_steps": 4665, "loss": 0.6439, "learning_rate": 5e-06, "epoch": 2.495176848874598, "percentage": 83.17, "elapsed_time": "6:55:34", "remaining_time": "1:24:04"}
391
+ {"current_steps": 3890, "total_steps": 4665, "loss": 0.6479, "learning_rate": 5e-06, "epoch": 2.5016077170418005, "percentage": 83.39, "elapsed_time": "6:56:36", "remaining_time": "1:23:00"}
392
+ {"current_steps": 3900, "total_steps": 4665, "loss": 0.6448, "learning_rate": 5e-06, "epoch": 2.508038585209003, "percentage": 83.6, "elapsed_time": "6:57:39", "remaining_time": "1:21:55"}
393
+ {"current_steps": 3910, "total_steps": 4665, "loss": 0.6439, "learning_rate": 5e-06, "epoch": 2.514469453376206, "percentage": 83.82, "elapsed_time": "6:58:41", "remaining_time": "1:20:50"}
394
+ {"current_steps": 3920, "total_steps": 4665, "loss": 0.6369, "learning_rate": 5e-06, "epoch": 2.5209003215434085, "percentage": 84.03, "elapsed_time": "6:59:43", "remaining_time": "1:19:46"}
395
+ {"current_steps": 3930, "total_steps": 4665, "loss": 0.6446, "learning_rate": 5e-06, "epoch": 2.527331189710611, "percentage": 84.24, "elapsed_time": "7:00:45", "remaining_time": "1:18:41"}
396
+ {"current_steps": 3940, "total_steps": 4665, "loss": 0.6384, "learning_rate": 5e-06, "epoch": 2.5337620578778135, "percentage": 84.46, "elapsed_time": "7:01:48", "remaining_time": "1:17:36"}
397
+ {"current_steps": 3950, "total_steps": 4665, "loss": 0.6408, "learning_rate": 5e-06, "epoch": 2.540192926045016, "percentage": 84.67, "elapsed_time": "7:02:50", "remaining_time": "1:16:32"}
398
+ {"current_steps": 3960, "total_steps": 4665, "loss": 0.6303, "learning_rate": 5e-06, "epoch": 2.5466237942122185, "percentage": 84.89, "elapsed_time": "7:03:52", "remaining_time": "1:15:27"}
399
+ {"current_steps": 3970, "total_steps": 4665, "loss": 0.651, "learning_rate": 5e-06, "epoch": 2.553054662379421, "percentage": 85.1, "elapsed_time": "7:04:55", "remaining_time": "1:14:23"}
400
+ {"current_steps": 3980, "total_steps": 4665, "loss": 0.643, "learning_rate": 5e-06, "epoch": 2.559485530546624, "percentage": 85.32, "elapsed_time": "7:05:57", "remaining_time": "1:13:18"}
401
+ {"current_steps": 3990, "total_steps": 4665, "loss": 0.6388, "learning_rate": 5e-06, "epoch": 2.5659163987138265, "percentage": 85.53, "elapsed_time": "7:06:59", "remaining_time": "1:12:14"}
402
+ {"current_steps": 4000, "total_steps": 4665, "loss": 0.6418, "learning_rate": 5e-06, "epoch": 2.572347266881029, "percentage": 85.74, "elapsed_time": "7:08:02", "remaining_time": "1:11:09"}
403
+ {"current_steps": 4010, "total_steps": 4665, "loss": 0.6334, "learning_rate": 5e-06, "epoch": 2.5787781350482315, "percentage": 85.96, "elapsed_time": "7:09:04", "remaining_time": "1:10:05"}
404
+ {"current_steps": 4020, "total_steps": 4665, "loss": 0.6398, "learning_rate": 5e-06, "epoch": 2.585209003215434, "percentage": 86.17, "elapsed_time": "7:10:06", "remaining_time": "1:09:00"}
405
+ {"current_steps": 4030, "total_steps": 4665, "loss": 0.6392, "learning_rate": 5e-06, "epoch": 2.5916398713826365, "percentage": 86.39, "elapsed_time": "7:11:09", "remaining_time": "1:07:56"}
406
+ {"current_steps": 4040, "total_steps": 4665, "loss": 0.643, "learning_rate": 5e-06, "epoch": 2.598070739549839, "percentage": 86.6, "elapsed_time": "7:12:11", "remaining_time": "1:06:51"}
407
+ {"current_steps": 4050, "total_steps": 4665, "loss": 0.6407, "learning_rate": 5e-06, "epoch": 2.604501607717042, "percentage": 86.82, "elapsed_time": "7:13:13", "remaining_time": "1:05:47"}
408
+ {"current_steps": 4060, "total_steps": 4665, "loss": 0.635, "learning_rate": 5e-06, "epoch": 2.6109324758842445, "percentage": 87.03, "elapsed_time": "7:14:16", "remaining_time": "1:04:42"}
409
+ {"current_steps": 4070, "total_steps": 4665, "loss": 0.6364, "learning_rate": 5e-06, "epoch": 2.617363344051447, "percentage": 87.25, "elapsed_time": "7:15:18", "remaining_time": "1:03:38"}
410
+ {"current_steps": 4080, "total_steps": 4665, "loss": 0.6411, "learning_rate": 5e-06, "epoch": 2.6237942122186495, "percentage": 87.46, "elapsed_time": "7:16:20", "remaining_time": "1:02:33"}
411
+ {"current_steps": 4090, "total_steps": 4665, "loss": 0.6382, "learning_rate": 5e-06, "epoch": 2.630225080385852, "percentage": 87.67, "elapsed_time": "7:17:23", "remaining_time": "1:01:29"}
412
+ {"current_steps": 4100, "total_steps": 4665, "loss": 0.6429, "learning_rate": 5e-06, "epoch": 2.6366559485530545, "percentage": 87.89, "elapsed_time": "7:18:25", "remaining_time": "1:00:25"}
413
+ {"current_steps": 4110, "total_steps": 4665, "loss": 0.6378, "learning_rate": 5e-06, "epoch": 2.643086816720257, "percentage": 88.1, "elapsed_time": "7:19:27", "remaining_time": "0:59:20"}
414
+ {"current_steps": 4120, "total_steps": 4665, "loss": 0.6444, "learning_rate": 5e-06, "epoch": 2.64951768488746, "percentage": 88.32, "elapsed_time": "7:20:30", "remaining_time": "0:58:16"}
415
+ {"current_steps": 4130, "total_steps": 4665, "loss": 0.6424, "learning_rate": 5e-06, "epoch": 2.6559485530546625, "percentage": 88.53, "elapsed_time": "7:21:32", "remaining_time": "0:57:11"}
416
+ {"current_steps": 4140, "total_steps": 4665, "loss": 0.6402, "learning_rate": 5e-06, "epoch": 2.662379421221865, "percentage": 88.75, "elapsed_time": "7:22:34", "remaining_time": "0:56:07"}
417
+ {"current_steps": 4150, "total_steps": 4665, "loss": 0.6402, "learning_rate": 5e-06, "epoch": 2.6688102893890675, "percentage": 88.96, "elapsed_time": "7:23:36", "remaining_time": "0:55:03"}
418
+ {"current_steps": 4160, "total_steps": 4665, "loss": 0.6475, "learning_rate": 5e-06, "epoch": 2.67524115755627, "percentage": 89.17, "elapsed_time": "7:24:39", "remaining_time": "0:53:58"}
419
+ {"current_steps": 4170, "total_steps": 4665, "loss": 0.6443, "learning_rate": 5e-06, "epoch": 2.6816720257234725, "percentage": 89.39, "elapsed_time": "7:25:41", "remaining_time": "0:52:54"}
420
+ {"current_steps": 4180, "total_steps": 4665, "loss": 0.6425, "learning_rate": 5e-06, "epoch": 2.688102893890675, "percentage": 89.6, "elapsed_time": "7:26:43", "remaining_time": "0:51:50"}
421
+ {"current_steps": 4190, "total_steps": 4665, "loss": 0.6385, "learning_rate": 5e-06, "epoch": 2.694533762057878, "percentage": 89.82, "elapsed_time": "7:27:46", "remaining_time": "0:50:45"}
422
+ {"current_steps": 4200, "total_steps": 4665, "loss": 0.6415, "learning_rate": 5e-06, "epoch": 2.7009646302250805, "percentage": 90.03, "elapsed_time": "7:28:48", "remaining_time": "0:49:41"}
423
+ {"current_steps": 4210, "total_steps": 4665, "loss": 0.6392, "learning_rate": 5e-06, "epoch": 2.707395498392283, "percentage": 90.25, "elapsed_time": "7:29:50", "remaining_time": "0:48:37"}
424
+ {"current_steps": 4220, "total_steps": 4665, "loss": 0.6365, "learning_rate": 5e-06, "epoch": 2.7138263665594855, "percentage": 90.46, "elapsed_time": "7:30:53", "remaining_time": "0:47:32"}
425
+ {"current_steps": 4230, "total_steps": 4665, "loss": 0.6365, "learning_rate": 5e-06, "epoch": 2.720257234726688, "percentage": 90.68, "elapsed_time": "7:31:55", "remaining_time": "0:46:28"}
426
+ {"current_steps": 4240, "total_steps": 4665, "loss": 0.6425, "learning_rate": 5e-06, "epoch": 2.7266881028938905, "percentage": 90.89, "elapsed_time": "7:32:57", "remaining_time": "0:45:24"}
427
+ {"current_steps": 4250, "total_steps": 4665, "loss": 0.6471, "learning_rate": 5e-06, "epoch": 2.733118971061093, "percentage": 91.1, "elapsed_time": "7:34:00", "remaining_time": "0:44:19"}
428
+ {"current_steps": 4260, "total_steps": 4665, "loss": 0.6356, "learning_rate": 5e-06, "epoch": 2.739549839228296, "percentage": 91.32, "elapsed_time": "7:35:02", "remaining_time": "0:43:15"}
429
+ {"current_steps": 4270, "total_steps": 4665, "loss": 0.647, "learning_rate": 5e-06, "epoch": 2.7459807073954985, "percentage": 91.53, "elapsed_time": "7:36:04", "remaining_time": "0:42:11"}
430
+ {"current_steps": 4280, "total_steps": 4665, "loss": 0.64, "learning_rate": 5e-06, "epoch": 2.752411575562701, "percentage": 91.75, "elapsed_time": "7:37:07", "remaining_time": "0:41:07"}
431
+ {"current_steps": 4290, "total_steps": 4665, "loss": 0.6439, "learning_rate": 5e-06, "epoch": 2.7588424437299035, "percentage": 91.96, "elapsed_time": "7:38:09", "remaining_time": "0:40:02"}
432
+ {"current_steps": 4300, "total_steps": 4665, "loss": 0.6355, "learning_rate": 5e-06, "epoch": 2.765273311897106, "percentage": 92.18, "elapsed_time": "7:39:11", "remaining_time": "0:38:58"}
433
+ {"current_steps": 4310, "total_steps": 4665, "loss": 0.641, "learning_rate": 5e-06, "epoch": 2.7717041800643085, "percentage": 92.39, "elapsed_time": "7:40:14", "remaining_time": "0:37:54"}
434
+ {"current_steps": 4320, "total_steps": 4665, "loss": 0.6479, "learning_rate": 5e-06, "epoch": 2.778135048231511, "percentage": 92.6, "elapsed_time": "7:41:16", "remaining_time": "0:36:50"}
435
+ {"current_steps": 4330, "total_steps": 4665, "loss": 0.6474, "learning_rate": 5e-06, "epoch": 2.784565916398714, "percentage": 92.82, "elapsed_time": "7:42:18", "remaining_time": "0:35:46"}
436
+ {"current_steps": 4340, "total_steps": 4665, "loss": 0.6428, "learning_rate": 5e-06, "epoch": 2.7909967845659165, "percentage": 93.03, "elapsed_time": "7:43:21", "remaining_time": "0:34:41"}
437
+ {"current_steps": 4350, "total_steps": 4665, "loss": 0.6441, "learning_rate": 5e-06, "epoch": 2.797427652733119, "percentage": 93.25, "elapsed_time": "7:44:23", "remaining_time": "0:33:37"}
438
+ {"current_steps": 4360, "total_steps": 4665, "loss": 0.6477, "learning_rate": 5e-06, "epoch": 2.8038585209003215, "percentage": 93.46, "elapsed_time": "7:45:26", "remaining_time": "0:32:33"}
439
+ {"current_steps": 4370, "total_steps": 4665, "loss": 0.6504, "learning_rate": 5e-06, "epoch": 2.810289389067524, "percentage": 93.68, "elapsed_time": "7:46:28", "remaining_time": "0:31:29"}
440
+ {"current_steps": 4380, "total_steps": 4665, "loss": 0.6523, "learning_rate": 5e-06, "epoch": 2.816720257234727, "percentage": 93.89, "elapsed_time": "7:47:30", "remaining_time": "0:30:25"}
441
+ {"current_steps": 4390, "total_steps": 4665, "loss": 0.6423, "learning_rate": 5e-06, "epoch": 2.823151125401929, "percentage": 94.11, "elapsed_time": "7:48:33", "remaining_time": "0:29:21"}
442
+ {"current_steps": 4400, "total_steps": 4665, "loss": 0.6483, "learning_rate": 5e-06, "epoch": 2.829581993569132, "percentage": 94.32, "elapsed_time": "7:49:35", "remaining_time": "0:28:16"}
443
+ {"current_steps": 4410, "total_steps": 4665, "loss": 0.6426, "learning_rate": 5e-06, "epoch": 2.8360128617363345, "percentage": 94.53, "elapsed_time": "7:50:37", "remaining_time": "0:27:12"}
444
+ {"current_steps": 4420, "total_steps": 4665, "loss": 0.6383, "learning_rate": 5e-06, "epoch": 2.842443729903537, "percentage": 94.75, "elapsed_time": "7:51:40", "remaining_time": "0:26:08"}
445
+ {"current_steps": 4430, "total_steps": 4665, "loss": 0.6378, "learning_rate": 5e-06, "epoch": 2.8488745980707395, "percentage": 94.96, "elapsed_time": "7:52:42", "remaining_time": "0:25:04"}
446
+ {"current_steps": 4440, "total_steps": 4665, "loss": 0.6445, "learning_rate": 5e-06, "epoch": 2.855305466237942, "percentage": 95.18, "elapsed_time": "7:53:44", "remaining_time": "0:24:00"}
447
+ {"current_steps": 4450, "total_steps": 4665, "loss": 0.6429, "learning_rate": 5e-06, "epoch": 2.861736334405145, "percentage": 95.39, "elapsed_time": "7:54:47", "remaining_time": "0:22:56"}
448
+ {"current_steps": 4460, "total_steps": 4665, "loss": 0.6403, "learning_rate": 5e-06, "epoch": 2.868167202572347, "percentage": 95.61, "elapsed_time": "7:55:49", "remaining_time": "0:21:52"}
449
+ {"current_steps": 4470, "total_steps": 4665, "loss": 0.6423, "learning_rate": 5e-06, "epoch": 2.87459807073955, "percentage": 95.82, "elapsed_time": "7:56:51", "remaining_time": "0:20:48"}
450
+ {"current_steps": 4480, "total_steps": 4665, "loss": 0.6379, "learning_rate": 5e-06, "epoch": 2.8810289389067525, "percentage": 96.03, "elapsed_time": "7:57:54", "remaining_time": "0:19:44"}
451
+ {"current_steps": 4490, "total_steps": 4665, "loss": 0.6419, "learning_rate": 5e-06, "epoch": 2.887459807073955, "percentage": 96.25, "elapsed_time": "7:58:56", "remaining_time": "0:18:40"}
452
+ {"current_steps": 4500, "total_steps": 4665, "loss": 0.6361, "learning_rate": 5e-06, "epoch": 2.8938906752411575, "percentage": 96.46, "elapsed_time": "7:59:58", "remaining_time": "0:17:35"}
453
+ {"current_steps": 4510, "total_steps": 4665, "loss": 0.6386, "learning_rate": 5e-06, "epoch": 2.90032154340836, "percentage": 96.68, "elapsed_time": "8:01:01", "remaining_time": "0:16:31"}
454
+ {"current_steps": 4520, "total_steps": 4665, "loss": 0.6423, "learning_rate": 5e-06, "epoch": 2.906752411575563, "percentage": 96.89, "elapsed_time": "8:02:03", "remaining_time": "0:15:27"}
455
+ {"current_steps": 4530, "total_steps": 4665, "loss": 0.6422, "learning_rate": 5e-06, "epoch": 2.913183279742765, "percentage": 97.11, "elapsed_time": "8:03:05", "remaining_time": "0:14:23"}
456
+ {"current_steps": 4540, "total_steps": 4665, "loss": 0.6333, "learning_rate": 5e-06, "epoch": 2.919614147909968, "percentage": 97.32, "elapsed_time": "8:04:08", "remaining_time": "0:13:19"}
457
+ {"current_steps": 4550, "total_steps": 4665, "loss": 0.6419, "learning_rate": 5e-06, "epoch": 2.9260450160771705, "percentage": 97.53, "elapsed_time": "8:05:10", "remaining_time": "0:12:15"}
458
+ {"current_steps": 4560, "total_steps": 4665, "loss": 0.6393, "learning_rate": 5e-06, "epoch": 2.932475884244373, "percentage": 97.75, "elapsed_time": "8:06:12", "remaining_time": "0:11:11"}
459
+ {"current_steps": 4570, "total_steps": 4665, "loss": 0.6367, "learning_rate": 5e-06, "epoch": 2.9389067524115755, "percentage": 97.96, "elapsed_time": "8:07:15", "remaining_time": "0:10:07"}
460
+ {"current_steps": 4580, "total_steps": 4665, "loss": 0.6402, "learning_rate": 5e-06, "epoch": 2.945337620578778, "percentage": 98.18, "elapsed_time": "8:08:17", "remaining_time": "0:09:03"}
461
+ {"current_steps": 4590, "total_steps": 4665, "loss": 0.6372, "learning_rate": 5e-06, "epoch": 2.951768488745981, "percentage": 98.39, "elapsed_time": "8:09:19", "remaining_time": "0:07:59"}
462
+ {"current_steps": 4600, "total_steps": 4665, "loss": 0.6479, "learning_rate": 5e-06, "epoch": 2.958199356913183, "percentage": 98.61, "elapsed_time": "8:10:22", "remaining_time": "0:06:55"}
463
+ {"current_steps": 4610, "total_steps": 4665, "loss": 0.6391, "learning_rate": 5e-06, "epoch": 2.964630225080386, "percentage": 98.82, "elapsed_time": "8:11:24", "remaining_time": "0:05:51"}
464
+ {"current_steps": 4620, "total_steps": 4665, "loss": 0.643, "learning_rate": 5e-06, "epoch": 2.9710610932475885, "percentage": 99.04, "elapsed_time": "8:12:26", "remaining_time": "0:04:47"}
465
+ {"current_steps": 4630, "total_steps": 4665, "loss": 0.6369, "learning_rate": 5e-06, "epoch": 2.977491961414791, "percentage": 99.25, "elapsed_time": "8:13:29", "remaining_time": "0:03:43"}
466
+ {"current_steps": 4640, "total_steps": 4665, "loss": 0.6402, "learning_rate": 5e-06, "epoch": 2.9839228295819935, "percentage": 99.46, "elapsed_time": "8:14:31", "remaining_time": "0:02:39"}
467
+ {"current_steps": 4650, "total_steps": 4665, "loss": 0.6315, "learning_rate": 5e-06, "epoch": 2.990353697749196, "percentage": 99.68, "elapsed_time": "8:15:33", "remaining_time": "0:01:35"}
468
+ {"current_steps": 4660, "total_steps": 4665, "loss": 0.6421, "learning_rate": 5e-06, "epoch": 2.996784565916399, "percentage": 99.89, "elapsed_time": "8:16:36", "remaining_time": "0:00:31"}
469
+ {"current_steps": 4665, "total_steps": 4665, "eval_loss": 0.7122252583503723, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "8:23:17", "remaining_time": "0:00:00"}
470
+ {"current_steps": 4665, "total_steps": 4665, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "8:26:38", "remaining_time": "0:00:00"}