root commited on
Commit
20c9c66
·
1 Parent(s): fb9f937

update full

Browse files
Files changed (45) hide show
  1. frozen/checkpoint-10000/model.safetensors +2 -2
  2. frozen/checkpoint-10000/scheduler.pt +1 -1
  3. frozen/checkpoint-10000/trainer_state.json +481 -481
  4. frozen/checkpoint-10720/model.safetensors +2 -2
  5. frozen/checkpoint-10720/scheduler.pt +1 -1
  6. frozen/checkpoint-10720/trainer_state.json +523 -523
  7. frozen/checkpoint-9000/model.safetensors +2 -2
  8. frozen/checkpoint-9000/scheduler.pt +1 -1
  9. frozen/checkpoint-9000/trainer_state.json +430 -430
  10. full/checkpoint-10000/chat_template.jinja +1 -0
  11. full/checkpoint-10000/config.json +29 -0
  12. full/checkpoint-10000/generation_config.json +7 -0
  13. full/checkpoint-10000/model.safetensors +3 -0
  14. full/checkpoint-10000/scheduler.pt +3 -0
  15. full/checkpoint-10000/special_tokens_map.json +24 -0
  16. full/checkpoint-10000/tokenizer.json +0 -0
  17. full/checkpoint-10000/tokenizer.model +3 -0
  18. full/checkpoint-10000/tokenizer_config.json +42 -0
  19. full/checkpoint-10000/trainer_state.json +1434 -0
  20. full/checkpoint-10000/training_args.bin +3 -0
  21. full/checkpoint-10000/zero_to_fp32.py +760 -0
  22. full/checkpoint-10720/chat_template.jinja +1 -0
  23. full/checkpoint-10720/config.json +29 -0
  24. full/checkpoint-10720/generation_config.json +7 -0
  25. full/checkpoint-10720/model.safetensors +3 -0
  26. full/checkpoint-10720/scheduler.pt +3 -0
  27. full/checkpoint-10720/special_tokens_map.json +24 -0
  28. full/checkpoint-10720/tokenizer.json +0 -0
  29. full/checkpoint-10720/tokenizer.model +3 -0
  30. full/checkpoint-10720/tokenizer_config.json +42 -0
  31. full/checkpoint-10720/trainer_state.json +1532 -0
  32. full/checkpoint-10720/training_args.bin +3 -0
  33. full/checkpoint-10720/zero_to_fp32.py +760 -0
  34. full/checkpoint-9000/chat_template.jinja +1 -0
  35. full/checkpoint-9000/config.json +29 -0
  36. full/checkpoint-9000/generation_config.json +7 -0
  37. full/checkpoint-9000/model.safetensors +3 -0
  38. full/checkpoint-9000/scheduler.pt +3 -0
  39. full/checkpoint-9000/special_tokens_map.json +24 -0
  40. full/checkpoint-9000/tokenizer.json +0 -0
  41. full/checkpoint-9000/tokenizer.model +3 -0
  42. full/checkpoint-9000/tokenizer_config.json +42 -0
  43. full/checkpoint-9000/trainer_state.json +1294 -0
  44. full/checkpoint-9000/training_args.bin +3 -0
  45. full/checkpoint-9000/zero_to_fp32.py +760 -0
frozen/checkpoint-10000/model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:30d10fc2f65b25f104d2151e5fb1152390037c7f98e820e1a7917ec2e08ebe0e
3
- size 1067471056
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9e24c4e1c89b4e46a94a7ac30cd33217a5956a2c269b6c85a88824120b7a559
3
+ size 800608432
frozen/checkpoint-10000/scheduler.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ace32f2a668fac0e33c3139e3324a1b087ed7c7d914c1e38edef3c3ce92aa339
3
  size 1465
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb68357ef47649a4fbfa03b4aef53b36545a5650f7adfe410545ad8d5e8447ee
3
  size 1465
frozen/checkpoint-10000/trainer_state.json CHANGED
@@ -11,1402 +11,1402 @@
11
  "log_history": [
12
  {
13
  "epoch": 0.0046641791044776115,
14
- "grad_norm": 1.907758116722107,
15
  "learning_rate": 4.9000000000000005e-06,
16
- "loss": 10.0793,
17
  "step": 50
18
  },
19
  {
20
  "epoch": 0.009328358208955223,
21
- "grad_norm": 1.399185061454773,
22
  "learning_rate": 9.900000000000002e-06,
23
- "loss": 8.9287,
24
  "step": 100
25
  },
26
  {
27
  "epoch": 0.013992537313432836,
28
- "grad_norm": 1.0101089477539062,
29
  "learning_rate": 1.49e-05,
30
- "loss": 8.3288,
31
  "step": 150
32
  },
33
  {
34
  "epoch": 0.018656716417910446,
35
- "grad_norm": 0.7379341125488281,
36
  "learning_rate": 1.9900000000000003e-05,
37
- "loss": 7.6969,
38
  "step": 200
39
  },
40
  {
41
  "epoch": 0.02332089552238806,
42
- "grad_norm": 0.9656306505203247,
43
  "learning_rate": 2.4900000000000002e-05,
44
- "loss": 7.2698,
45
  "step": 250
46
  },
47
  {
48
  "epoch": 0.027985074626865673,
49
- "grad_norm": 1.3790279626846313,
50
  "learning_rate": 2.9900000000000002e-05,
51
- "loss": 6.9685,
52
  "step": 300
53
  },
54
  {
55
  "epoch": 0.03264925373134328,
56
- "grad_norm": 0.6374709606170654,
57
  "learning_rate": 3.49e-05,
58
- "loss": 6.6752,
59
  "step": 350
60
  },
61
  {
62
  "epoch": 0.03731343283582089,
63
- "grad_norm": 1.5095465183258057,
64
  "learning_rate": 3.99e-05,
65
- "loss": 6.3805,
66
  "step": 400
67
  },
68
  {
69
  "epoch": 0.04197761194029851,
70
- "grad_norm": 1.394410490989685,
71
  "learning_rate": 4.49e-05,
72
- "loss": 6.1667,
73
  "step": 450
74
  },
75
  {
76
  "epoch": 0.04664179104477612,
77
- "grad_norm": 1.7006834745407104,
78
  "learning_rate": 4.99e-05,
79
- "loss": 5.9861,
80
  "step": 500
81
  },
82
  {
83
  "epoch": 0.051305970149253734,
84
- "grad_norm": 0.9934776425361633,
85
  "learning_rate": 5.4900000000000006e-05,
86
- "loss": 5.8166,
87
  "step": 550
88
  },
89
  {
90
  "epoch": 0.055970149253731345,
91
- "grad_norm": 2.0550477504730225,
92
  "learning_rate": 5.99e-05,
93
- "loss": 5.6646,
94
  "step": 600
95
  },
96
  {
97
  "epoch": 0.06063432835820896,
98
- "grad_norm": 2.2504913806915283,
99
  "learning_rate": 6.49e-05,
100
- "loss": 5.5353,
101
  "step": 650
102
  },
103
  {
104
  "epoch": 0.06529850746268656,
105
- "grad_norm": 2.468212604522705,
106
  "learning_rate": 6.99e-05,
107
- "loss": 5.3982,
108
  "step": 700
109
  },
110
  {
111
  "epoch": 0.06996268656716417,
112
- "grad_norm": 2.4010939598083496,
113
  "learning_rate": 7.49e-05,
114
- "loss": 5.2855,
115
  "step": 750
116
  },
117
  {
118
  "epoch": 0.07462686567164178,
119
- "grad_norm": 1.4586199522018433,
120
  "learning_rate": 7.99e-05,
121
- "loss": 5.1901,
122
  "step": 800
123
  },
124
  {
125
  "epoch": 0.07929104477611941,
126
- "grad_norm": 1.2821402549743652,
127
  "learning_rate": 8.49e-05,
128
- "loss": 5.0973,
129
  "step": 850
130
  },
131
  {
132
  "epoch": 0.08395522388059702,
133
- "grad_norm": 2.1037180423736572,
134
  "learning_rate": 8.99e-05,
135
- "loss": 5.0062,
136
  "step": 900
137
  },
138
  {
139
  "epoch": 0.08861940298507463,
140
- "grad_norm": 1.370751976966858,
141
  "learning_rate": 9.49e-05,
142
- "loss": 4.9392,
143
  "step": 950
144
  },
145
  {
146
  "epoch": 0.09328358208955224,
147
- "grad_norm": 1.276108741760254,
148
  "learning_rate": 9.99e-05,
149
- "loss": 4.848,
150
  "step": 1000
151
  },
152
  {
153
  "epoch": 0.09794776119402986,
154
- "grad_norm": 1.1077498197555542,
155
  "learning_rate": 0.0001049,
156
- "loss": 4.7863,
157
  "step": 1050
158
  },
159
  {
160
  "epoch": 0.10261194029850747,
161
- "grad_norm": 1.0212163925170898,
162
  "learning_rate": 0.0001099,
163
- "loss": 4.7102,
164
  "step": 1100
165
  },
166
  {
167
  "epoch": 0.10727611940298508,
168
- "grad_norm": 1.1480289697647095,
169
  "learning_rate": 0.0001149,
170
- "loss": 4.6551,
171
  "step": 1150
172
  },
173
  {
174
  "epoch": 0.11194029850746269,
175
- "grad_norm": 1.3106518983840942,
176
  "learning_rate": 0.00011990000000000001,
177
- "loss": 4.5627,
178
  "step": 1200
179
  },
180
  {
181
  "epoch": 0.1166044776119403,
182
- "grad_norm": 1.4304360151290894,
183
  "learning_rate": 0.00012490000000000003,
184
- "loss": 4.4823,
185
  "step": 1250
186
  },
187
  {
188
  "epoch": 0.12126865671641791,
189
- "grad_norm": 0.883263885974884,
190
  "learning_rate": 0.00012989999999999999,
191
- "loss": 4.4091,
192
  "step": 1300
193
  },
194
  {
195
  "epoch": 0.1259328358208955,
196
- "grad_norm": 1.2335679531097412,
197
  "learning_rate": 0.0001349,
198
- "loss": 4.3284,
199
  "step": 1350
200
  },
201
  {
202
  "epoch": 0.13059701492537312,
203
- "grad_norm": 1.3587566614151,
204
  "learning_rate": 0.0001399,
205
- "loss": 4.2523,
206
  "step": 1400
207
  },
208
  {
209
  "epoch": 0.13526119402985073,
210
- "grad_norm": 1.4397273063659668,
211
  "learning_rate": 0.00014490000000000003,
212
- "loss": 4.1874,
213
  "step": 1450
214
  },
215
  {
216
  "epoch": 0.13992537313432835,
217
- "grad_norm": 0.9128643274307251,
218
  "learning_rate": 0.0001499,
219
- "loss": 4.1382,
220
  "step": 1500
221
  },
222
  {
223
  "epoch": 0.14458955223880596,
224
- "grad_norm": 1.1429646015167236,
225
  "learning_rate": 0.0001549,
226
- "loss": 4.0713,
227
  "step": 1550
228
  },
229
  {
230
  "epoch": 0.14925373134328357,
231
- "grad_norm": 0.7839804291725159,
232
  "learning_rate": 0.0001599,
233
- "loss": 4.0327,
234
  "step": 1600
235
  },
236
  {
237
  "epoch": 0.15391791044776118,
238
- "grad_norm": 0.5934556126594543,
239
  "learning_rate": 0.0001649,
240
- "loss": 4.0535,
241
  "step": 1650
242
  },
243
  {
244
  "epoch": 0.15858208955223882,
245
- "grad_norm": 0.6320812702178955,
246
  "learning_rate": 0.0001699,
247
- "loss": 3.9615,
248
  "step": 1700
249
  },
250
  {
251
  "epoch": 0.16324626865671643,
252
- "grad_norm": 0.6207342743873596,
253
  "learning_rate": 0.00017490000000000002,
254
- "loss": 3.9086,
255
  "step": 1750
256
  },
257
  {
258
  "epoch": 0.16791044776119404,
259
- "grad_norm": 0.8524695634841919,
260
  "learning_rate": 0.0001799,
261
- "loss": 3.8703,
262
  "step": 1800
263
  },
264
  {
265
  "epoch": 0.17257462686567165,
266
- "grad_norm": 0.9148792028427124,
267
  "learning_rate": 0.0001849,
268
- "loss": 3.8388,
269
  "step": 1850
270
  },
271
  {
272
  "epoch": 0.17723880597014927,
273
- "grad_norm": 0.7455419301986694,
274
  "learning_rate": 0.0001899,
275
- "loss": 3.8031,
276
  "step": 1900
277
  },
278
  {
279
  "epoch": 0.18190298507462688,
280
- "grad_norm": 0.9777323603630066,
281
  "learning_rate": 0.00019490000000000002,
282
- "loss": 3.781,
283
  "step": 1950
284
  },
285
  {
286
  "epoch": 0.1865671641791045,
287
- "grad_norm": 0.6018832921981812,
288
  "learning_rate": 0.0001999,
289
- "loss": 3.7421,
290
  "step": 2000
291
  },
292
  {
293
  "epoch": 0.1912313432835821,
294
- "grad_norm": 0.7007433772087097,
295
  "learning_rate": 0.00019887614678899083,
296
- "loss": 3.7148,
297
  "step": 2050
298
  },
299
  {
300
  "epoch": 0.1958955223880597,
301
- "grad_norm": 0.7864815592765808,
302
  "learning_rate": 0.00019772935779816515,
303
- "loss": 3.6907,
304
  "step": 2100
305
  },
306
  {
307
  "epoch": 0.20055970149253732,
308
- "grad_norm": 0.6575994491577148,
309
  "learning_rate": 0.00019658256880733944,
310
- "loss": 3.6504,
311
  "step": 2150
312
  },
313
  {
314
  "epoch": 0.20522388059701493,
315
- "grad_norm": 0.524401843547821,
316
  "learning_rate": 0.0001954357798165138,
317
- "loss": 3.6333,
318
  "step": 2200
319
  },
320
  {
321
  "epoch": 0.20988805970149255,
322
- "grad_norm": 0.8242985010147095,
323
  "learning_rate": 0.00019428899082568808,
324
- "loss": 3.6084,
325
  "step": 2250
326
  },
327
  {
328
  "epoch": 0.21455223880597016,
329
- "grad_norm": 0.5965232849121094,
330
  "learning_rate": 0.0001931422018348624,
331
- "loss": 3.5914,
332
  "step": 2300
333
  },
334
  {
335
  "epoch": 0.21921641791044777,
336
- "grad_norm": 0.5473845601081848,
337
  "learning_rate": 0.00019199541284403671,
338
- "loss": 3.572,
339
  "step": 2350
340
  },
341
  {
342
  "epoch": 0.22388059701492538,
343
- "grad_norm": 0.7030166983604431,
344
  "learning_rate": 0.00019084862385321103,
345
- "loss": 3.5548,
346
  "step": 2400
347
  },
348
  {
349
  "epoch": 0.228544776119403,
350
- "grad_norm": 0.5844017267227173,
351
- "learning_rate": 0.00018972477064220186,
352
- "loss": 3.5326,
353
  "step": 2450
354
  },
355
  {
356
  "epoch": 0.2332089552238806,
357
- "grad_norm": 0.4722403883934021,
358
- "learning_rate": 0.00018857798165137615,
359
- "loss": 3.5113,
360
  "step": 2500
361
  },
362
  {
363
  "epoch": 0.23787313432835822,
364
- "grad_norm": 0.4691997468471527,
365
- "learning_rate": 0.00018743119266055047,
366
- "loss": 3.4985,
367
  "step": 2550
368
  },
369
  {
370
  "epoch": 0.24253731343283583,
371
- "grad_norm": 0.5742476582527161,
372
- "learning_rate": 0.00018628440366972478,
373
- "loss": 3.4689,
374
  "step": 2600
375
  },
376
  {
377
  "epoch": 0.24720149253731344,
378
- "grad_norm": 0.4867892563343048,
379
- "learning_rate": 0.0001851376146788991,
380
- "loss": 3.4737,
381
  "step": 2650
382
  },
383
  {
384
  "epoch": 0.251865671641791,
385
- "grad_norm": 0.5876694321632385,
386
- "learning_rate": 0.0001839908256880734,
387
- "loss": 3.4516,
388
  "step": 2700
389
  },
390
  {
391
  "epoch": 0.25652985074626866,
392
- "grad_norm": 0.4691877067089081,
393
- "learning_rate": 0.0001828440366972477,
394
- "loss": 3.4411,
395
  "step": 2750
396
  },
397
  {
398
  "epoch": 0.26119402985074625,
399
- "grad_norm": 0.497503399848938,
400
- "learning_rate": 0.00018169724770642203,
401
- "loss": 3.4311,
402
  "step": 2800
403
  },
404
  {
405
  "epoch": 0.2658582089552239,
406
- "grad_norm": 0.4610927700996399,
407
- "learning_rate": 0.00018055045871559635,
408
- "loss": 3.4122,
409
  "step": 2850
410
  },
411
  {
412
  "epoch": 0.27052238805970147,
413
- "grad_norm": 0.5312972664833069,
414
- "learning_rate": 0.00017940366972477064,
415
- "loss": 3.4004,
416
  "step": 2900
417
  },
418
  {
419
  "epoch": 0.2751865671641791,
420
- "grad_norm": 0.5256180167198181,
421
- "learning_rate": 0.00017825688073394496,
422
- "loss": 3.3834,
423
  "step": 2950
424
  },
425
  {
426
  "epoch": 0.2798507462686567,
427
- "grad_norm": 0.5108594298362732,
428
- "learning_rate": 0.00017711009174311928,
429
- "loss": 3.379,
430
  "step": 3000
431
  },
432
  {
433
  "epoch": 0.28451492537313433,
434
- "grad_norm": 0.4818528890609741,
435
- "learning_rate": 0.00017596330275229357,
436
- "loss": 3.3645,
437
  "step": 3050
438
  },
439
  {
440
  "epoch": 0.2891791044776119,
441
- "grad_norm": 0.4077577590942383,
442
- "learning_rate": 0.0001748165137614679,
443
- "loss": 3.3659,
444
  "step": 3100
445
  },
446
  {
447
  "epoch": 0.29384328358208955,
448
- "grad_norm": 0.4791859984397888,
449
- "learning_rate": 0.0001736697247706422,
450
- "loss": 3.3427,
451
  "step": 3150
452
  },
453
  {
454
  "epoch": 0.29850746268656714,
455
- "grad_norm": 0.6330239176750183,
456
- "learning_rate": 0.00017252293577981652,
457
- "loss": 3.3484,
458
  "step": 3200
459
  },
460
  {
461
  "epoch": 0.3031716417910448,
462
- "grad_norm": 0.4511379897594452,
463
- "learning_rate": 0.00017137614678899084,
464
- "loss": 3.3302,
465
  "step": 3250
466
  },
467
  {
468
  "epoch": 0.30783582089552236,
469
- "grad_norm": 0.4571825861930847,
470
- "learning_rate": 0.00017022935779816516,
471
- "loss": 3.3144,
472
  "step": 3300
473
  },
474
  {
475
  "epoch": 0.3125,
476
- "grad_norm": 0.44557932019233704,
477
- "learning_rate": 0.00016908256880733945,
478
- "loss": 3.3071,
479
  "step": 3350
480
  },
481
  {
482
  "epoch": 0.31716417910447764,
483
- "grad_norm": 0.42434263229370117,
484
  "learning_rate": 0.00016793577981651377,
485
- "loss": 3.2986,
486
  "step": 3400
487
  },
488
  {
489
  "epoch": 0.3218283582089552,
490
- "grad_norm": 0.41813352704048157,
491
  "learning_rate": 0.0001667889908256881,
492
- "loss": 3.2902,
493
  "step": 3450
494
  },
495
  {
496
  "epoch": 0.32649253731343286,
497
- "grad_norm": 0.4084812104701996,
498
  "learning_rate": 0.00016564220183486238,
499
- "loss": 3.283,
500
  "step": 3500
501
  },
502
  {
503
  "epoch": 0.33115671641791045,
504
- "grad_norm": 0.38210785388946533,
505
  "learning_rate": 0.00016449541284403672,
506
- "loss": 3.2769,
507
  "step": 3550
508
  },
509
  {
510
  "epoch": 0.3358208955223881,
511
- "grad_norm": 0.5375980734825134,
512
  "learning_rate": 0.00016334862385321102,
513
- "loss": 3.2697,
514
  "step": 3600
515
  },
516
  {
517
  "epoch": 0.34048507462686567,
518
- "grad_norm": 0.38781264424324036,
519
  "learning_rate": 0.00016220183486238533,
520
- "loss": 3.2556,
521
  "step": 3650
522
  },
523
  {
524
  "epoch": 0.3451492537313433,
525
- "grad_norm": 0.41560590267181396,
526
  "learning_rate": 0.00016105504587155965,
527
- "loss": 3.2565,
528
  "step": 3700
529
  },
530
  {
531
  "epoch": 0.3498134328358209,
532
- "grad_norm": 0.4119664132595062,
533
  "learning_rate": 0.00015990825688073394,
534
- "loss": 3.2386,
535
  "step": 3750
536
  },
537
  {
538
  "epoch": 0.35447761194029853,
539
- "grad_norm": 0.4190433919429779,
540
  "learning_rate": 0.00015876146788990826,
541
- "loss": 3.2345,
542
  "step": 3800
543
  },
544
  {
545
  "epoch": 0.3591417910447761,
546
- "grad_norm": 0.37028583884239197,
547
  "learning_rate": 0.00015761467889908258,
548
- "loss": 3.226,
549
  "step": 3850
550
  },
551
  {
552
  "epoch": 0.36380597014925375,
553
- "grad_norm": 0.3971349000930786,
554
  "learning_rate": 0.0001564678899082569,
555
- "loss": 3.2213,
556
  "step": 3900
557
  },
558
  {
559
  "epoch": 0.36847014925373134,
560
- "grad_norm": 0.34641316533088684,
561
  "learning_rate": 0.0001553211009174312,
562
- "loss": 3.2144,
563
  "step": 3950
564
  },
565
  {
566
  "epoch": 0.373134328358209,
567
- "grad_norm": 0.40990573167800903,
568
  "learning_rate": 0.0001541743119266055,
569
- "loss": 3.2141,
570
  "step": 4000
571
  },
572
  {
573
  "epoch": 0.37779850746268656,
574
- "grad_norm": 0.35828983783721924,
575
  "learning_rate": 0.00015302752293577983,
576
- "loss": 3.2054,
577
  "step": 4050
578
  },
579
  {
580
  "epoch": 0.3824626865671642,
581
- "grad_norm": 0.37572401762008667,
582
  "learning_rate": 0.00015188073394495414,
583
- "loss": 3.1943,
584
  "step": 4100
585
  },
586
  {
587
  "epoch": 0.3871268656716418,
588
- "grad_norm": 0.3795240521430969,
589
  "learning_rate": 0.00015073394495412844,
590
- "loss": 3.1971,
591
  "step": 4150
592
  },
593
  {
594
  "epoch": 0.3917910447761194,
595
- "grad_norm": 0.3884343206882477,
596
  "learning_rate": 0.00014958715596330275,
597
- "loss": 3.1877,
598
  "step": 4200
599
  },
600
  {
601
  "epoch": 0.396455223880597,
602
- "grad_norm": 0.374661922454834,
603
  "learning_rate": 0.00014844036697247707,
604
- "loss": 3.174,
605
  "step": 4250
606
  },
607
  {
608
  "epoch": 0.40111940298507465,
609
- "grad_norm": 0.4330432415008545,
610
  "learning_rate": 0.00014729357798165136,
611
- "loss": 3.1787,
612
  "step": 4300
613
  },
614
  {
615
  "epoch": 0.40578358208955223,
616
- "grad_norm": 0.40475285053253174,
617
  "learning_rate": 0.0001461467889908257,
618
- "loss": 3.1781,
619
  "step": 4350
620
  },
621
  {
622
  "epoch": 0.41044776119402987,
623
- "grad_norm": 0.36028817296028137,
624
  "learning_rate": 0.000145,
625
- "loss": 3.1653,
626
  "step": 4400
627
  },
628
  {
629
  "epoch": 0.41511194029850745,
630
- "grad_norm": 0.37242797017097473,
631
  "learning_rate": 0.00014385321100917432,
632
- "loss": 3.1608,
633
  "step": 4450
634
  },
635
  {
636
  "epoch": 0.4197761194029851,
637
- "grad_norm": 0.42880979180336,
638
  "learning_rate": 0.00014270642201834864,
639
- "loss": 3.1656,
640
  "step": 4500
641
  },
642
  {
643
  "epoch": 0.4244402985074627,
644
- "grad_norm": 0.3562730848789215,
645
  "learning_rate": 0.00014155963302752296,
646
- "loss": 3.1409,
647
  "step": 4550
648
  },
649
  {
650
  "epoch": 0.4291044776119403,
651
- "grad_norm": 0.3420323133468628,
652
  "learning_rate": 0.00014041284403669725,
653
- "loss": 3.1502,
654
  "step": 4600
655
  },
656
  {
657
  "epoch": 0.4337686567164179,
658
- "grad_norm": 0.33985841274261475,
659
  "learning_rate": 0.00013926605504587157,
660
- "loss": 3.1437,
661
  "step": 4650
662
  },
663
  {
664
  "epoch": 0.43843283582089554,
665
- "grad_norm": 0.35252872109413147,
666
  "learning_rate": 0.00013811926605504588,
667
- "loss": 3.1375,
668
  "step": 4700
669
  },
670
  {
671
  "epoch": 0.4430970149253731,
672
- "grad_norm": 0.33299019932746887,
673
  "learning_rate": 0.00013697247706422018,
674
- "loss": 3.1207,
675
  "step": 4750
676
  },
677
  {
678
  "epoch": 0.44776119402985076,
679
- "grad_norm": 0.3340663015842438,
680
  "learning_rate": 0.00013582568807339452,
681
- "loss": 3.1285,
682
  "step": 4800
683
  },
684
  {
685
  "epoch": 0.45242537313432835,
686
- "grad_norm": 0.3459039330482483,
687
  "learning_rate": 0.00013470183486238532,
688
- "loss": 3.1187,
689
  "step": 4850
690
  },
691
  {
692
  "epoch": 0.457089552238806,
693
- "grad_norm": 0.34190812706947327,
694
  "learning_rate": 0.00013355504587155963,
695
- "loss": 3.1128,
696
  "step": 4900
697
  },
698
  {
699
  "epoch": 0.46175373134328357,
700
- "grad_norm": 0.38482001423835754,
701
  "learning_rate": 0.00013240825688073395,
702
- "loss": 3.1081,
703
  "step": 4950
704
  },
705
  {
706
  "epoch": 0.4664179104477612,
707
- "grad_norm": 0.3969718813896179,
708
  "learning_rate": 0.00013126146788990827,
709
- "loss": 3.111,
710
  "step": 5000
711
  },
712
  {
713
  "epoch": 0.4710820895522388,
714
- "grad_norm": 0.30978360772132874,
715
  "learning_rate": 0.0001301146788990826,
716
- "loss": 3.106,
717
  "step": 5050
718
  },
719
  {
720
  "epoch": 0.47574626865671643,
721
- "grad_norm": 0.32029977440834045,
722
  "learning_rate": 0.00012896788990825688,
723
- "loss": 3.1095,
724
  "step": 5100
725
  },
726
  {
727
  "epoch": 0.480410447761194,
728
- "grad_norm": 0.36698678135871887,
729
  "learning_rate": 0.0001278440366972477,
730
- "loss": 3.1017,
731
  "step": 5150
732
  },
733
  {
734
  "epoch": 0.48507462686567165,
735
- "grad_norm": 0.3710266947746277,
736
  "learning_rate": 0.00012669724770642202,
737
- "loss": 3.0863,
738
  "step": 5200
739
  },
740
  {
741
  "epoch": 0.48973880597014924,
742
- "grad_norm": 0.33469706773757935,
743
  "learning_rate": 0.00012555045871559634,
744
- "loss": 3.0949,
745
  "step": 5250
746
  },
747
  {
748
  "epoch": 0.4944029850746269,
749
- "grad_norm": 0.33493635058403015,
750
  "learning_rate": 0.00012440366972477063,
751
- "loss": 3.0794,
752
  "step": 5300
753
  },
754
  {
755
  "epoch": 0.49906716417910446,
756
- "grad_norm": 0.3489319682121277,
757
  "learning_rate": 0.00012325688073394498,
758
- "loss": 3.0752,
759
  "step": 5350
760
  },
761
  {
762
  "epoch": 0.503731343283582,
763
- "grad_norm": 0.32940617203712463,
764
  "learning_rate": 0.00012211009174311927,
765
- "loss": 3.0911,
766
  "step": 5400
767
  },
768
  {
769
  "epoch": 0.5083955223880597,
770
- "grad_norm": 0.3039519190788269,
771
  "learning_rate": 0.0001209633027522936,
772
- "loss": 3.0749,
773
  "step": 5450
774
  },
775
  {
776
  "epoch": 0.5130597014925373,
777
- "grad_norm": 0.3196834325790405,
778
  "learning_rate": 0.00011981651376146789,
779
- "loss": 3.0685,
780
  "step": 5500
781
  },
782
  {
783
  "epoch": 0.5177238805970149,
784
- "grad_norm": 0.35171428322792053,
785
  "learning_rate": 0.0001186697247706422,
786
- "loss": 3.0684,
787
  "step": 5550
788
  },
789
  {
790
  "epoch": 0.5223880597014925,
791
- "grad_norm": 0.3927328884601593,
792
  "learning_rate": 0.00011752293577981653,
793
- "loss": 3.0676,
794
  "step": 5600
795
  },
796
  {
797
  "epoch": 0.5270522388059702,
798
- "grad_norm": 0.32428842782974243,
799
  "learning_rate": 0.00011637614678899082,
800
- "loss": 3.0635,
801
  "step": 5650
802
  },
803
  {
804
  "epoch": 0.5317164179104478,
805
- "grad_norm": 0.3297617435455322,
806
  "learning_rate": 0.00011522935779816515,
807
- "loss": 3.0525,
808
  "step": 5700
809
  },
810
  {
811
  "epoch": 0.5363805970149254,
812
- "grad_norm": 0.3118542432785034,
813
  "learning_rate": 0.00011408256880733946,
814
- "loss": 3.0566,
815
  "step": 5750
816
  },
817
  {
818
  "epoch": 0.5410447761194029,
819
- "grad_norm": 0.29949086904525757,
820
  "learning_rate": 0.00011293577981651377,
821
- "loss": 3.0454,
822
  "step": 5800
823
  },
824
  {
825
  "epoch": 0.5457089552238806,
826
- "grad_norm": 0.3567812740802765,
827
  "learning_rate": 0.00011178899082568808,
828
- "loss": 3.0462,
829
  "step": 5850
830
  },
831
  {
832
  "epoch": 0.5503731343283582,
833
- "grad_norm": 0.3179719150066376,
834
  "learning_rate": 0.00011064220183486238,
835
- "loss": 3.0409,
836
  "step": 5900
837
  },
838
  {
839
  "epoch": 0.5550373134328358,
840
- "grad_norm": 0.32883220911026,
841
  "learning_rate": 0.0001094954128440367,
842
- "loss": 3.0527,
843
  "step": 5950
844
  },
845
  {
846
  "epoch": 0.5597014925373134,
847
- "grad_norm": 0.31516268849372864,
848
  "learning_rate": 0.000108348623853211,
849
- "loss": 3.0359,
850
  "step": 6000
851
  },
852
  {
853
  "epoch": 0.5643656716417911,
854
- "grad_norm": 0.298753559589386,
855
  "learning_rate": 0.00010720183486238533,
856
- "loss": 3.0343,
857
  "step": 6050
858
  },
859
  {
860
  "epoch": 0.5690298507462687,
861
- "grad_norm": 0.32886001467704773,
862
  "learning_rate": 0.00010605504587155963,
863
- "loss": 3.0229,
864
  "step": 6100
865
  },
866
  {
867
  "epoch": 0.5736940298507462,
868
- "grad_norm": 0.3526987135410309,
869
  "learning_rate": 0.00010490825688073396,
870
- "loss": 3.0227,
871
  "step": 6150
872
  },
873
  {
874
  "epoch": 0.5783582089552238,
875
- "grad_norm": 0.31915777921676636,
876
  "learning_rate": 0.00010376146788990825,
877
- "loss": 3.0311,
878
  "step": 6200
879
  },
880
  {
881
  "epoch": 0.5830223880597015,
882
- "grad_norm": 0.31342729926109314,
883
  "learning_rate": 0.00010261467889908258,
884
- "loss": 3.0257,
885
  "step": 6250
886
  },
887
  {
888
  "epoch": 0.5876865671641791,
889
- "grad_norm": 0.3711305856704712,
890
  "learning_rate": 0.00010146788990825689,
891
- "loss": 3.0125,
892
  "step": 6300
893
  },
894
  {
895
  "epoch": 0.5923507462686567,
896
- "grad_norm": 0.32861173152923584,
897
  "learning_rate": 0.00010032110091743118,
898
- "loss": 3.0235,
899
  "step": 6350
900
  },
901
  {
902
  "epoch": 0.5970149253731343,
903
- "grad_norm": 0.30850905179977417,
904
  "learning_rate": 9.917431192660551e-05,
905
- "loss": 3.0171,
906
  "step": 6400
907
  },
908
  {
909
  "epoch": 0.601679104477612,
910
- "grad_norm": 0.31126561760902405,
911
  "learning_rate": 9.802752293577983e-05,
912
- "loss": 3.0177,
913
  "step": 6450
914
  },
915
  {
916
  "epoch": 0.6063432835820896,
917
- "grad_norm": 0.3373335897922516,
918
  "learning_rate": 9.688073394495414e-05,
919
- "loss": 3.0138,
920
  "step": 6500
921
  },
922
  {
923
  "epoch": 0.6110074626865671,
924
- "grad_norm": 0.32974258065223694,
925
- "learning_rate": 9.575688073394496e-05,
926
- "loss": 3.0092,
927
  "step": 6550
928
  },
929
  {
930
  "epoch": 0.6156716417910447,
931
- "grad_norm": 0.29914727807044983,
932
- "learning_rate": 9.463302752293578e-05,
933
- "loss": 3.0075,
934
  "step": 6600
935
  },
936
  {
937
  "epoch": 0.6203358208955224,
938
- "grad_norm": 0.3006197214126587,
939
- "learning_rate": 9.34862385321101e-05,
940
- "loss": 3.015,
941
  "step": 6650
942
  },
943
  {
944
  "epoch": 0.625,
945
- "grad_norm": 0.2997278869152069,
946
- "learning_rate": 9.23394495412844e-05,
947
- "loss": 3.0037,
948
  "step": 6700
949
  },
950
  {
951
  "epoch": 0.6296641791044776,
952
- "grad_norm": 0.31719309091567993,
953
- "learning_rate": 9.119266055045872e-05,
954
- "loss": 2.9969,
955
  "step": 6750
956
  },
957
  {
958
  "epoch": 0.6343283582089553,
959
- "grad_norm": 0.3004010319709778,
960
- "learning_rate": 9.004587155963303e-05,
961
- "loss": 2.9909,
962
  "step": 6800
963
  },
964
  {
965
  "epoch": 0.6389925373134329,
966
- "grad_norm": 0.3123314082622528,
967
- "learning_rate": 8.889908256880735e-05,
968
- "loss": 2.9853,
969
  "step": 6850
970
  },
971
  {
972
  "epoch": 0.6436567164179104,
973
- "grad_norm": 0.334837406873703,
974
- "learning_rate": 8.775229357798166e-05,
975
- "loss": 2.9923,
976
  "step": 6900
977
  },
978
  {
979
  "epoch": 0.648320895522388,
980
- "grad_norm": 0.3055480420589447,
981
- "learning_rate": 8.660550458715597e-05,
982
- "loss": 2.9897,
983
  "step": 6950
984
  },
985
  {
986
  "epoch": 0.6529850746268657,
987
- "grad_norm": 0.33151915669441223,
988
- "learning_rate": 8.545871559633027e-05,
989
- "loss": 2.9843,
990
  "step": 7000
991
  },
992
  {
993
  "epoch": 0.6576492537313433,
994
- "grad_norm": 0.32511621713638306,
995
- "learning_rate": 8.431192660550459e-05,
996
- "loss": 2.9797,
997
  "step": 7050
998
  },
999
  {
1000
  "epoch": 0.6623134328358209,
1001
- "grad_norm": 0.29789265990257263,
1002
- "learning_rate": 8.31651376146789e-05,
1003
- "loss": 2.9789,
1004
  "step": 7100
1005
  },
1006
  {
1007
  "epoch": 0.6669776119402985,
1008
- "grad_norm": 0.3272823393344879,
1009
- "learning_rate": 8.201834862385321e-05,
1010
- "loss": 2.971,
1011
  "step": 7150
1012
  },
1013
  {
1014
  "epoch": 0.6716417910447762,
1015
- "grad_norm": 0.2983226478099823,
1016
- "learning_rate": 8.087155963302753e-05,
1017
- "loss": 2.9818,
1018
  "step": 7200
1019
  },
1020
  {
1021
  "epoch": 0.6763059701492538,
1022
- "grad_norm": 0.30532923340797424,
1023
- "learning_rate": 7.972477064220184e-05,
1024
- "loss": 2.9739,
1025
  "step": 7250
1026
  },
1027
  {
1028
  "epoch": 0.6809701492537313,
1029
- "grad_norm": 0.3046092092990875,
1030
- "learning_rate": 7.857798165137616e-05,
1031
- "loss": 2.9575,
1032
  "step": 7300
1033
  },
1034
  {
1035
  "epoch": 0.6856343283582089,
1036
- "grad_norm": 0.3133400082588196,
1037
- "learning_rate": 7.743119266055046e-05,
1038
- "loss": 2.9794,
1039
  "step": 7350
1040
  },
1041
  {
1042
  "epoch": 0.6902985074626866,
1043
- "grad_norm": 0.3090495765209198,
1044
- "learning_rate": 7.628440366972477e-05,
1045
- "loss": 2.9632,
1046
  "step": 7400
1047
  },
1048
  {
1049
  "epoch": 0.6949626865671642,
1050
- "grad_norm": 0.2925854027271271,
1051
- "learning_rate": 7.513761467889908e-05,
1052
- "loss": 2.9649,
1053
  "step": 7450
1054
  },
1055
  {
1056
  "epoch": 0.6996268656716418,
1057
- "grad_norm": 0.30464622378349304,
1058
- "learning_rate": 7.399082568807339e-05,
1059
- "loss": 2.9606,
1060
  "step": 7500
1061
  },
1062
  {
1063
  "epoch": 0.7042910447761194,
1064
- "grad_norm": 0.29478827118873596,
1065
- "learning_rate": 7.284403669724771e-05,
1066
- "loss": 2.9574,
1067
  "step": 7550
1068
  },
1069
  {
1070
  "epoch": 0.7089552238805971,
1071
- "grad_norm": 0.30368906259536743,
1072
- "learning_rate": 7.169724770642203e-05,
1073
- "loss": 2.9583,
1074
  "step": 7600
1075
  },
1076
  {
1077
  "epoch": 0.7136194029850746,
1078
- "grad_norm": 0.2974824905395508,
1079
- "learning_rate": 7.055045871559633e-05,
1080
- "loss": 2.9566,
1081
  "step": 7650
1082
  },
1083
  {
1084
  "epoch": 0.7182835820895522,
1085
- "grad_norm": 0.3006385564804077,
1086
- "learning_rate": 6.940366972477065e-05,
1087
- "loss": 2.953,
1088
  "step": 7700
1089
  },
1090
  {
1091
  "epoch": 0.7229477611940298,
1092
- "grad_norm": 0.2787078022956848,
1093
- "learning_rate": 6.825688073394497e-05,
1094
- "loss": 2.9466,
1095
  "step": 7750
1096
  },
1097
  {
1098
  "epoch": 0.7276119402985075,
1099
- "grad_norm": 0.2966342568397522,
1100
- "learning_rate": 6.711009174311927e-05,
1101
- "loss": 2.9446,
1102
  "step": 7800
1103
  },
1104
  {
1105
  "epoch": 0.7322761194029851,
1106
- "grad_norm": 0.32728853821754456,
1107
- "learning_rate": 6.600917431192661e-05,
1108
- "loss": 2.9486,
1109
  "step": 7850
1110
  },
1111
  {
1112
  "epoch": 0.7369402985074627,
1113
- "grad_norm": 0.325846403837204,
1114
- "learning_rate": 6.486238532110092e-05,
1115
- "loss": 2.9456,
1116
  "step": 7900
1117
  },
1118
  {
1119
  "epoch": 0.7416044776119403,
1120
- "grad_norm": 0.2933451235294342,
1121
- "learning_rate": 6.371559633027524e-05,
1122
- "loss": 2.9404,
1123
  "step": 7950
1124
  },
1125
  {
1126
  "epoch": 0.746268656716418,
1127
- "grad_norm": 0.3031325042247772,
1128
- "learning_rate": 6.256880733944954e-05,
1129
- "loss": 2.9486,
1130
  "step": 8000
1131
  },
1132
  {
1133
  "epoch": 0.7509328358208955,
1134
- "grad_norm": 0.30206266045570374,
1135
- "learning_rate": 6.142201834862386e-05,
1136
- "loss": 2.9364,
1137
  "step": 8050
1138
  },
1139
  {
1140
  "epoch": 0.7555970149253731,
1141
- "grad_norm": 0.29706859588623047,
1142
- "learning_rate": 6.027522935779817e-05,
1143
- "loss": 2.9483,
1144
  "step": 8100
1145
  },
1146
  {
1147
  "epoch": 0.7602611940298507,
1148
- "grad_norm": 0.2909660339355469,
1149
- "learning_rate": 5.912844036697248e-05,
1150
- "loss": 2.9405,
1151
  "step": 8150
1152
  },
1153
  {
1154
  "epoch": 0.7649253731343284,
1155
- "grad_norm": 0.3136695623397827,
1156
- "learning_rate": 5.798165137614679e-05,
1157
- "loss": 2.9366,
1158
  "step": 8200
1159
  },
1160
  {
1161
  "epoch": 0.769589552238806,
1162
- "grad_norm": 0.29548221826553345,
1163
- "learning_rate": 5.683486238532111e-05,
1164
- "loss": 2.9305,
1165
  "step": 8250
1166
  },
1167
  {
1168
  "epoch": 0.7742537313432836,
1169
- "grad_norm": 0.2777513265609741,
1170
- "learning_rate": 5.568807339449541e-05,
1171
- "loss": 2.9377,
1172
  "step": 8300
1173
  },
1174
  {
1175
  "epoch": 0.7789179104477612,
1176
- "grad_norm": 0.28673887252807617,
1177
- "learning_rate": 5.454128440366972e-05,
1178
- "loss": 2.9342,
1179
  "step": 8350
1180
  },
1181
  {
1182
  "epoch": 0.7835820895522388,
1183
- "grad_norm": 0.2911510169506073,
1184
- "learning_rate": 5.339449541284404e-05,
1185
- "loss": 2.938,
1186
  "step": 8400
1187
  },
1188
  {
1189
  "epoch": 0.7882462686567164,
1190
- "grad_norm": 0.3052498698234558,
1191
- "learning_rate": 5.224770642201835e-05,
1192
- "loss": 2.9235,
1193
  "step": 8450
1194
  },
1195
  {
1196
  "epoch": 0.792910447761194,
1197
- "grad_norm": 0.30528631806373596,
1198
- "learning_rate": 5.110091743119266e-05,
1199
- "loss": 2.9393,
1200
  "step": 8500
1201
  },
1202
  {
1203
  "epoch": 0.7975746268656716,
1204
- "grad_norm": 0.27922648191452026,
1205
- "learning_rate": 4.9954128440366974e-05,
1206
- "loss": 2.9153,
1207
  "step": 8550
1208
  },
1209
  {
1210
  "epoch": 0.8022388059701493,
1211
- "grad_norm": 0.2799355685710907,
1212
- "learning_rate": 4.880733944954129e-05,
1213
- "loss": 2.9329,
1214
  "step": 8600
1215
  },
1216
  {
1217
  "epoch": 0.8069029850746269,
1218
- "grad_norm": 0.3089926242828369,
1219
- "learning_rate": 4.76605504587156e-05,
1220
- "loss": 2.9203,
1221
  "step": 8650
1222
  },
1223
  {
1224
  "epoch": 0.8115671641791045,
1225
- "grad_norm": 0.3086063265800476,
1226
- "learning_rate": 4.651376146788991e-05,
1227
- "loss": 2.9223,
1228
  "step": 8700
1229
  },
1230
  {
1231
  "epoch": 0.816231343283582,
1232
- "grad_norm": 0.3019801676273346,
1233
- "learning_rate": 4.536697247706422e-05,
1234
- "loss": 2.9159,
1235
  "step": 8750
1236
  },
1237
  {
1238
  "epoch": 0.8208955223880597,
1239
- "grad_norm": 0.2975527048110962,
1240
- "learning_rate": 4.422018348623854e-05,
1241
- "loss": 2.9215,
1242
  "step": 8800
1243
  },
1244
  {
1245
  "epoch": 0.8255597014925373,
1246
- "grad_norm": 0.30920466780662537,
1247
- "learning_rate": 4.3073394495412844e-05,
1248
- "loss": 2.9213,
1249
  "step": 8850
1250
  },
1251
  {
1252
  "epoch": 0.8302238805970149,
1253
- "grad_norm": 0.2723771929740906,
1254
- "learning_rate": 4.1926605504587155e-05,
1255
- "loss": 2.9119,
1256
  "step": 8900
1257
  },
1258
  {
1259
  "epoch": 0.8348880597014925,
1260
- "grad_norm": 0.28407421708106995,
1261
- "learning_rate": 4.0779816513761474e-05,
1262
- "loss": 2.9101,
1263
  "step": 8950
1264
  },
1265
  {
1266
  "epoch": 0.8395522388059702,
1267
- "grad_norm": 0.2752555012702942,
1268
- "learning_rate": 3.9633027522935785e-05,
1269
- "loss": 2.9087,
1270
  "step": 9000
1271
  },
1272
  {
1273
  "epoch": 0.8442164179104478,
1274
- "grad_norm": 0.30377471446990967,
1275
- "learning_rate": 3.848623853211009e-05,
1276
- "loss": 2.9003,
1277
  "step": 9050
1278
  },
1279
  {
1280
  "epoch": 0.8488805970149254,
1281
- "grad_norm": 0.29200997948646545,
1282
- "learning_rate": 3.73394495412844e-05,
1283
- "loss": 2.9064,
1284
  "step": 9100
1285
  },
1286
  {
1287
  "epoch": 0.8535447761194029,
1288
- "grad_norm": 0.2830120623111725,
1289
- "learning_rate": 3.619266055045872e-05,
1290
- "loss": 2.9085,
1291
  "step": 9150
1292
  },
1293
  {
1294
  "epoch": 0.8582089552238806,
1295
- "grad_norm": 0.32428160309791565,
1296
  "learning_rate": 3.504587155963303e-05,
1297
- "loss": 2.9116,
1298
  "step": 9200
1299
  },
1300
  {
1301
  "epoch": 0.8628731343283582,
1302
- "grad_norm": 0.2852170169353485,
1303
  "learning_rate": 3.3899082568807337e-05,
1304
- "loss": 2.9119,
1305
  "step": 9250
1306
  },
1307
  {
1308
  "epoch": 0.8675373134328358,
1309
- "grad_norm": 0.2776082754135132,
1310
  "learning_rate": 3.2752293577981655e-05,
1311
- "loss": 2.9004,
1312
  "step": 9300
1313
  },
1314
  {
1315
  "epoch": 0.8722014925373134,
1316
- "grad_norm": 0.2881547212600708,
1317
  "learning_rate": 3.1605504587155967e-05,
1318
- "loss": 2.9024,
1319
  "step": 9350
1320
  },
1321
  {
1322
  "epoch": 0.8768656716417911,
1323
- "grad_norm": 0.2810397148132324,
1324
  "learning_rate": 3.0458715596330278e-05,
1325
- "loss": 2.9094,
1326
  "step": 9400
1327
  },
1328
  {
1329
  "epoch": 0.8815298507462687,
1330
- "grad_norm": 0.2975148558616638,
1331
  "learning_rate": 2.9311926605504586e-05,
1332
- "loss": 2.8971,
1333
  "step": 9450
1334
  },
1335
  {
1336
  "epoch": 0.8861940298507462,
1337
- "grad_norm": 0.29773736000061035,
1338
  "learning_rate": 2.8165137614678898e-05,
1339
- "loss": 2.8943,
1340
  "step": 9500
1341
  },
1342
  {
1343
  "epoch": 0.8908582089552238,
1344
- "grad_norm": 0.2762524485588074,
1345
  "learning_rate": 2.7018348623853213e-05,
1346
- "loss": 2.8899,
1347
  "step": 9550
1348
  },
1349
  {
1350
  "epoch": 0.8955223880597015,
1351
- "grad_norm": 0.2947123348712921,
1352
  "learning_rate": 2.5871559633027525e-05,
1353
- "loss": 2.8965,
1354
  "step": 9600
1355
  },
1356
  {
1357
  "epoch": 0.9001865671641791,
1358
- "grad_norm": 0.27293965220451355,
1359
- "learning_rate": 2.4724770642201836e-05,
1360
- "loss": 2.8869,
1361
  "step": 9650
1362
  },
1363
  {
1364
  "epoch": 0.9048507462686567,
1365
- "grad_norm": 0.2870621681213379,
1366
- "learning_rate": 2.3577981651376148e-05,
1367
- "loss": 2.8885,
1368
  "step": 9700
1369
  },
1370
  {
1371
  "epoch": 0.9095149253731343,
1372
- "grad_norm": 0.2940194606781006,
1373
- "learning_rate": 2.243119266055046e-05,
1374
- "loss": 2.8933,
1375
  "step": 9750
1376
  },
1377
  {
1378
  "epoch": 0.914179104477612,
1379
- "grad_norm": 0.270706444978714,
1380
- "learning_rate": 2.1284403669724774e-05,
1381
- "loss": 2.8889,
1382
  "step": 9800
1383
  },
1384
  {
1385
  "epoch": 0.9188432835820896,
1386
- "grad_norm": 0.27721747756004333,
1387
- "learning_rate": 2.0137614678899082e-05,
1388
- "loss": 2.8893,
1389
  "step": 9850
1390
  },
1391
  {
1392
  "epoch": 0.9235074626865671,
1393
- "grad_norm": 0.2803277373313904,
1394
- "learning_rate": 1.8990825688073397e-05,
1395
- "loss": 2.8868,
1396
  "step": 9900
1397
  },
1398
  {
1399
  "epoch": 0.9281716417910447,
1400
- "grad_norm": 0.2814282476902008,
1401
- "learning_rate": 1.7844036697247706e-05,
1402
- "loss": 2.8906,
1403
  "step": 9950
1404
  },
1405
  {
1406
  "epoch": 0.9328358208955224,
1407
- "grad_norm": 0.2730121612548828,
1408
- "learning_rate": 1.669724770642202e-05,
1409
- "loss": 2.8886,
1410
  "step": 10000
1411
  }
1412
  ],
@@ -1427,7 +1427,7 @@
1427
  "attributes": {}
1428
  }
1429
  },
1430
- "total_flos": 2124398876164096.0,
1431
  "train_batch_size": 64,
1432
  "trial_name": null,
1433
  "trial_params": null
 
11
  "log_history": [
12
  {
13
  "epoch": 0.0046641791044776115,
14
+ "grad_norm": 2.434795618057251,
15
  "learning_rate": 4.9000000000000005e-06,
16
+ "loss": 10.2918,
17
  "step": 50
18
  },
19
  {
20
  "epoch": 0.009328358208955223,
21
+ "grad_norm": 1.3679819107055664,
22
  "learning_rate": 9.900000000000002e-06,
23
+ "loss": 9.0755,
24
  "step": 100
25
  },
26
  {
27
  "epoch": 0.013992537313432836,
28
+ "grad_norm": 0.9769303798675537,
29
  "learning_rate": 1.49e-05,
30
+ "loss": 8.3639,
31
  "step": 150
32
  },
33
  {
34
  "epoch": 0.018656716417910446,
35
+ "grad_norm": 0.5544369220733643,
36
  "learning_rate": 1.9900000000000003e-05,
37
+ "loss": 7.8229,
38
  "step": 200
39
  },
40
  {
41
  "epoch": 0.02332089552238806,
42
+ "grad_norm": 0.32632946968078613,
43
  "learning_rate": 2.4900000000000002e-05,
44
+ "loss": 7.4323,
45
  "step": 250
46
  },
47
  {
48
  "epoch": 0.027985074626865673,
49
+ "grad_norm": 0.34820762276649475,
50
  "learning_rate": 2.9900000000000002e-05,
51
+ "loss": 7.1956,
52
  "step": 300
53
  },
54
  {
55
  "epoch": 0.03264925373134328,
56
+ "grad_norm": 0.5745918154716492,
57
  "learning_rate": 3.49e-05,
58
+ "loss": 6.9913,
59
  "step": 350
60
  },
61
  {
62
  "epoch": 0.03731343283582089,
63
+ "grad_norm": 0.7313559651374817,
64
  "learning_rate": 3.99e-05,
65
+ "loss": 6.7761,
66
  "step": 400
67
  },
68
  {
69
  "epoch": 0.04197761194029851,
70
+ "grad_norm": 0.5769322514533997,
71
  "learning_rate": 4.49e-05,
72
+ "loss": 6.598,
73
  "step": 450
74
  },
75
  {
76
  "epoch": 0.04664179104477612,
77
+ "grad_norm": 0.5906850099563599,
78
  "learning_rate": 4.99e-05,
79
+ "loss": 6.4384,
80
  "step": 500
81
  },
82
  {
83
  "epoch": 0.051305970149253734,
84
+ "grad_norm": 1.5271503925323486,
85
  "learning_rate": 5.4900000000000006e-05,
86
+ "loss": 6.3052,
87
  "step": 550
88
  },
89
  {
90
  "epoch": 0.055970149253731345,
91
+ "grad_norm": 0.980595052242279,
92
  "learning_rate": 5.99e-05,
93
+ "loss": 6.1786,
94
  "step": 600
95
  },
96
  {
97
  "epoch": 0.06063432835820896,
98
+ "grad_norm": 0.8249331712722778,
99
  "learning_rate": 6.49e-05,
100
+ "loss": 6.0769,
101
  "step": 650
102
  },
103
  {
104
  "epoch": 0.06529850746268656,
105
+ "grad_norm": 1.6615272760391235,
106
  "learning_rate": 6.99e-05,
107
+ "loss": 5.9624,
108
  "step": 700
109
  },
110
  {
111
  "epoch": 0.06996268656716417,
112
+ "grad_norm": 1.4796066284179688,
113
  "learning_rate": 7.49e-05,
114
+ "loss": 5.8661,
115
  "step": 750
116
  },
117
  {
118
  "epoch": 0.07462686567164178,
119
+ "grad_norm": 1.050801157951355,
120
  "learning_rate": 7.99e-05,
121
+ "loss": 5.7773,
122
  "step": 800
123
  },
124
  {
125
  "epoch": 0.07929104477611941,
126
+ "grad_norm": 1.1273882389068604,
127
  "learning_rate": 8.49e-05,
128
+ "loss": 5.6794,
129
  "step": 850
130
  },
131
  {
132
  "epoch": 0.08395522388059702,
133
+ "grad_norm": 1.3447613716125488,
134
  "learning_rate": 8.99e-05,
135
+ "loss": 5.5883,
136
  "step": 900
137
  },
138
  {
139
  "epoch": 0.08861940298507463,
140
+ "grad_norm": 0.8979071378707886,
141
  "learning_rate": 9.49e-05,
142
+ "loss": 5.5152,
143
  "step": 950
144
  },
145
  {
146
  "epoch": 0.09328358208955224,
147
+ "grad_norm": 1.7317239046096802,
148
  "learning_rate": 9.99e-05,
149
+ "loss": 5.4185,
150
  "step": 1000
151
  },
152
  {
153
  "epoch": 0.09794776119402986,
154
+ "grad_norm": 1.6568114757537842,
155
  "learning_rate": 0.0001049,
156
+ "loss": 5.3531,
157
  "step": 1050
158
  },
159
  {
160
  "epoch": 0.10261194029850747,
161
+ "grad_norm": 1.260619044303894,
162
  "learning_rate": 0.0001099,
163
+ "loss": 5.2675,
164
  "step": 1100
165
  },
166
  {
167
  "epoch": 0.10727611940298508,
168
+ "grad_norm": 1.6347819566726685,
169
  "learning_rate": 0.0001149,
170
+ "loss": 5.1987,
171
  "step": 1150
172
  },
173
  {
174
  "epoch": 0.11194029850746269,
175
+ "grad_norm": 1.6571054458618164,
176
  "learning_rate": 0.00011990000000000001,
177
+ "loss": 5.1358,
178
  "step": 1200
179
  },
180
  {
181
  "epoch": 0.1166044776119403,
182
+ "grad_norm": 1.2196329832077026,
183
  "learning_rate": 0.00012490000000000003,
184
+ "loss": 5.0755,
185
  "step": 1250
186
  },
187
  {
188
  "epoch": 0.12126865671641791,
189
+ "grad_norm": 1.2077012062072754,
190
  "learning_rate": 0.00012989999999999999,
191
+ "loss": 5.0243,
192
  "step": 1300
193
  },
194
  {
195
  "epoch": 0.1259328358208955,
196
+ "grad_norm": 1.3209941387176514,
197
  "learning_rate": 0.0001349,
198
+ "loss": 4.9678,
199
  "step": 1350
200
  },
201
  {
202
  "epoch": 0.13059701492537312,
203
+ "grad_norm": 0.8279642462730408,
204
  "learning_rate": 0.0001399,
205
+ "loss": 4.9159,
206
  "step": 1400
207
  },
208
  {
209
  "epoch": 0.13526119402985073,
210
+ "grad_norm": 0.9633798599243164,
211
  "learning_rate": 0.00014490000000000003,
212
+ "loss": 4.8596,
213
  "step": 1450
214
  },
215
  {
216
  "epoch": 0.13992537313432835,
217
+ "grad_norm": 0.987281084060669,
218
  "learning_rate": 0.0001499,
219
+ "loss": 4.8162,
220
  "step": 1500
221
  },
222
  {
223
  "epoch": 0.14458955223880596,
224
+ "grad_norm": 1.094568133354187,
225
  "learning_rate": 0.0001549,
226
+ "loss": 4.7598,
227
  "step": 1550
228
  },
229
  {
230
  "epoch": 0.14925373134328357,
231
+ "grad_norm": 0.8916253447532654,
232
  "learning_rate": 0.0001599,
233
+ "loss": 4.7184,
234
  "step": 1600
235
  },
236
  {
237
  "epoch": 0.15391791044776118,
238
+ "grad_norm": 0.8389304876327515,
239
  "learning_rate": 0.0001649,
240
+ "loss": 4.6699,
241
  "step": 1650
242
  },
243
  {
244
  "epoch": 0.15858208955223882,
245
+ "grad_norm": 1.113956093788147,
246
  "learning_rate": 0.0001699,
247
+ "loss": 4.6177,
248
  "step": 1700
249
  },
250
  {
251
  "epoch": 0.16324626865671643,
252
+ "grad_norm": 0.9380860924720764,
253
  "learning_rate": 0.00017490000000000002,
254
+ "loss": 4.5619,
255
  "step": 1750
256
  },
257
  {
258
  "epoch": 0.16791044776119404,
259
+ "grad_norm": 1.2293964624404907,
260
  "learning_rate": 0.0001799,
261
+ "loss": 4.5084,
262
  "step": 1800
263
  },
264
  {
265
  "epoch": 0.17257462686567165,
266
+ "grad_norm": 1.1242707967758179,
267
  "learning_rate": 0.0001849,
268
+ "loss": 4.4553,
269
  "step": 1850
270
  },
271
  {
272
  "epoch": 0.17723880597014927,
273
+ "grad_norm": 0.7840976715087891,
274
  "learning_rate": 0.0001899,
275
+ "loss": 4.4074,
276
  "step": 1900
277
  },
278
  {
279
  "epoch": 0.18190298507462688,
280
+ "grad_norm": 0.7234581112861633,
281
  "learning_rate": 0.00019490000000000002,
282
+ "loss": 4.3652,
283
  "step": 1950
284
  },
285
  {
286
  "epoch": 0.1865671641791045,
287
+ "grad_norm": 0.8629528880119324,
288
  "learning_rate": 0.0001999,
289
+ "loss": 4.317,
290
  "step": 2000
291
  },
292
  {
293
  "epoch": 0.1912313432835821,
294
+ "grad_norm": 0.8042444586753845,
295
  "learning_rate": 0.00019887614678899083,
296
+ "loss": 4.2762,
297
  "step": 2050
298
  },
299
  {
300
  "epoch": 0.1958955223880597,
301
+ "grad_norm": 0.80702143907547,
302
  "learning_rate": 0.00019772935779816515,
303
+ "loss": 4.2375,
304
  "step": 2100
305
  },
306
  {
307
  "epoch": 0.20055970149253732,
308
+ "grad_norm": 0.6999443769454956,
309
  "learning_rate": 0.00019658256880733944,
310
+ "loss": 4.1876,
311
  "step": 2150
312
  },
313
  {
314
  "epoch": 0.20522388059701493,
315
+ "grad_norm": 0.7686917781829834,
316
  "learning_rate": 0.0001954357798165138,
317
+ "loss": 4.1574,
318
  "step": 2200
319
  },
320
  {
321
  "epoch": 0.20988805970149255,
322
+ "grad_norm": 0.7840579152107239,
323
  "learning_rate": 0.00019428899082568808,
324
+ "loss": 4.1217,
325
  "step": 2250
326
  },
327
  {
328
  "epoch": 0.21455223880597016,
329
+ "grad_norm": 0.8510863184928894,
330
  "learning_rate": 0.0001931422018348624,
331
+ "loss": 4.0979,
332
  "step": 2300
333
  },
334
  {
335
  "epoch": 0.21921641791044777,
336
+ "grad_norm": 0.679388165473938,
337
  "learning_rate": 0.00019199541284403671,
338
+ "loss": 4.0661,
339
  "step": 2350
340
  },
341
  {
342
  "epoch": 0.22388059701492538,
343
+ "grad_norm": 0.5541307330131531,
344
  "learning_rate": 0.00019084862385321103,
345
+ "loss": 4.0407,
346
  "step": 2400
347
  },
348
  {
349
  "epoch": 0.228544776119403,
350
+ "grad_norm": 0.6176748871803284,
351
+ "learning_rate": 0.00018970183486238532,
352
+ "loss": 4.0098,
353
  "step": 2450
354
  },
355
  {
356
  "epoch": 0.2332089552238806,
357
+ "grad_norm": 0.6339316368103027,
358
+ "learning_rate": 0.00018855504587155964,
359
+ "loss": 3.9825,
360
  "step": 2500
361
  },
362
  {
363
  "epoch": 0.23787313432835822,
364
+ "grad_norm": 0.6165521740913391,
365
+ "learning_rate": 0.00018740825688073396,
366
+ "loss": 3.9626,
367
  "step": 2550
368
  },
369
  {
370
  "epoch": 0.24253731343283583,
371
+ "grad_norm": 0.6364430785179138,
372
+ "learning_rate": 0.00018626146788990825,
373
+ "loss": 3.9288,
374
  "step": 2600
375
  },
376
  {
377
  "epoch": 0.24720149253731344,
378
+ "grad_norm": 0.5691106915473938,
379
+ "learning_rate": 0.00018511467889908257,
380
+ "loss": 3.9221,
381
  "step": 2650
382
  },
383
  {
384
  "epoch": 0.251865671641791,
385
+ "grad_norm": 0.5992695689201355,
386
+ "learning_rate": 0.0001839678899082569,
387
+ "loss": 3.8961,
388
  "step": 2700
389
  },
390
  {
391
  "epoch": 0.25652985074626866,
392
+ "grad_norm": 0.5932413935661316,
393
+ "learning_rate": 0.0001828211009174312,
394
+ "loss": 3.8808,
395
  "step": 2750
396
  },
397
  {
398
  "epoch": 0.26119402985074625,
399
+ "grad_norm": 0.6363717913627625,
400
+ "learning_rate": 0.0001816743119266055,
401
+ "loss": 3.8652,
402
  "step": 2800
403
  },
404
  {
405
  "epoch": 0.2658582089552239,
406
+ "grad_norm": 0.5456887483596802,
407
+ "learning_rate": 0.00018052752293577984,
408
+ "loss": 3.842,
409
  "step": 2850
410
  },
411
  {
412
  "epoch": 0.27052238805970147,
413
+ "grad_norm": 0.5930595397949219,
414
+ "learning_rate": 0.00017938073394495414,
415
+ "loss": 3.8244,
416
  "step": 2900
417
  },
418
  {
419
  "epoch": 0.2751865671641791,
420
+ "grad_norm": 0.5566945672035217,
421
+ "learning_rate": 0.00017823394495412843,
422
+ "loss": 3.804,
423
  "step": 2950
424
  },
425
  {
426
  "epoch": 0.2798507462686567,
427
+ "grad_norm": 0.6539527773857117,
428
+ "learning_rate": 0.00017708715596330277,
429
+ "loss": 3.7985,
430
  "step": 3000
431
  },
432
  {
433
  "epoch": 0.28451492537313433,
434
+ "grad_norm": 0.5921605229377747,
435
+ "learning_rate": 0.00017594036697247706,
436
+ "loss": 3.7834,
437
  "step": 3050
438
  },
439
  {
440
  "epoch": 0.2891791044776119,
441
+ "grad_norm": 0.5476294159889221,
442
+ "learning_rate": 0.00017479357798165138,
443
+ "loss": 3.7735,
444
  "step": 3100
445
  },
446
  {
447
  "epoch": 0.29384328358208955,
448
+ "grad_norm": 0.5736291408538818,
449
+ "learning_rate": 0.0001736467889908257,
450
+ "loss": 3.7535,
451
  "step": 3150
452
  },
453
  {
454
  "epoch": 0.29850746268656714,
455
+ "grad_norm": 0.5571457743644714,
456
+ "learning_rate": 0.00017250000000000002,
457
+ "loss": 3.7517,
458
  "step": 3200
459
  },
460
  {
461
  "epoch": 0.3031716417910448,
462
+ "grad_norm": 0.5658199787139893,
463
+ "learning_rate": 0.0001713532110091743,
464
+ "loss": 3.7308,
465
  "step": 3250
466
  },
467
  {
468
  "epoch": 0.30783582089552236,
469
+ "grad_norm": 0.5400105714797974,
470
+ "learning_rate": 0.00017020642201834863,
471
+ "loss": 3.713,
472
  "step": 3300
473
  },
474
  {
475
  "epoch": 0.3125,
476
+ "grad_norm": 0.5364307761192322,
477
+ "learning_rate": 0.00016905963302752295,
478
+ "loss": 3.7018,
479
  "step": 3350
480
  },
481
  {
482
  "epoch": 0.31716417910447764,
483
+ "grad_norm": 0.5792276859283447,
484
  "learning_rate": 0.00016793577981651377,
485
+ "loss": 3.6918,
486
  "step": 3400
487
  },
488
  {
489
  "epoch": 0.3218283582089552,
490
+ "grad_norm": 0.44588422775268555,
491
  "learning_rate": 0.0001667889908256881,
492
+ "loss": 3.6813,
493
  "step": 3450
494
  },
495
  {
496
  "epoch": 0.32649253731343286,
497
+ "grad_norm": 0.5175060033798218,
498
  "learning_rate": 0.00016564220183486238,
499
+ "loss": 3.6705,
500
  "step": 3500
501
  },
502
  {
503
  "epoch": 0.33115671641791045,
504
+ "grad_norm": 0.5310373306274414,
505
  "learning_rate": 0.00016449541284403672,
506
+ "loss": 3.6645,
507
  "step": 3550
508
  },
509
  {
510
  "epoch": 0.3358208955223881,
511
+ "grad_norm": 0.5658640265464783,
512
  "learning_rate": 0.00016334862385321102,
513
+ "loss": 3.6542,
514
  "step": 3600
515
  },
516
  {
517
  "epoch": 0.34048507462686567,
518
+ "grad_norm": 0.5425497889518738,
519
  "learning_rate": 0.00016220183486238533,
520
+ "loss": 3.6375,
521
  "step": 3650
522
  },
523
  {
524
  "epoch": 0.3451492537313433,
525
+ "grad_norm": 0.4865582585334778,
526
  "learning_rate": 0.00016105504587155965,
527
+ "loss": 3.6384,
528
  "step": 3700
529
  },
530
  {
531
  "epoch": 0.3498134328358209,
532
+ "grad_norm": 0.4987630248069763,
533
  "learning_rate": 0.00015990825688073394,
534
+ "loss": 3.6156,
535
  "step": 3750
536
  },
537
  {
538
  "epoch": 0.35447761194029853,
539
+ "grad_norm": 0.5758020877838135,
540
  "learning_rate": 0.00015876146788990826,
541
+ "loss": 3.6123,
542
  "step": 3800
543
  },
544
  {
545
  "epoch": 0.3591417910447761,
546
+ "grad_norm": 0.5264896154403687,
547
  "learning_rate": 0.00015761467889908258,
548
+ "loss": 3.6019,
549
  "step": 3850
550
  },
551
  {
552
  "epoch": 0.36380597014925375,
553
+ "grad_norm": 0.4835953414440155,
554
  "learning_rate": 0.0001564678899082569,
555
+ "loss": 3.5959,
556
  "step": 3900
557
  },
558
  {
559
  "epoch": 0.36847014925373134,
560
+ "grad_norm": 0.4341369569301605,
561
  "learning_rate": 0.0001553211009174312,
562
+ "loss": 3.5873,
563
  "step": 3950
564
  },
565
  {
566
  "epoch": 0.373134328358209,
567
+ "grad_norm": 0.47072407603263855,
568
  "learning_rate": 0.0001541743119266055,
569
+ "loss": 3.5852,
570
  "step": 4000
571
  },
572
  {
573
  "epoch": 0.37779850746268656,
574
+ "grad_norm": 0.46912139654159546,
575
  "learning_rate": 0.00015302752293577983,
576
+ "loss": 3.575,
577
  "step": 4050
578
  },
579
  {
580
  "epoch": 0.3824626865671642,
581
+ "grad_norm": 0.5946963429450989,
582
  "learning_rate": 0.00015188073394495414,
583
+ "loss": 3.5639,
584
  "step": 4100
585
  },
586
  {
587
  "epoch": 0.3871268656716418,
588
+ "grad_norm": 0.4828120470046997,
589
  "learning_rate": 0.00015073394495412844,
590
+ "loss": 3.5659,
591
  "step": 4150
592
  },
593
  {
594
  "epoch": 0.3917910447761194,
595
+ "grad_norm": 0.48630213737487793,
596
  "learning_rate": 0.00014958715596330275,
597
+ "loss": 3.5538,
598
  "step": 4200
599
  },
600
  {
601
  "epoch": 0.396455223880597,
602
+ "grad_norm": 0.4966210126876831,
603
  "learning_rate": 0.00014844036697247707,
604
+ "loss": 3.5394,
605
  "step": 4250
606
  },
607
  {
608
  "epoch": 0.40111940298507465,
609
+ "grad_norm": 0.8680199384689331,
610
  "learning_rate": 0.00014729357798165136,
611
+ "loss": 3.5434,
612
  "step": 4300
613
  },
614
  {
615
  "epoch": 0.40578358208955223,
616
+ "grad_norm": 0.4786328971385956,
617
  "learning_rate": 0.0001461467889908257,
618
+ "loss": 3.5429,
619
  "step": 4350
620
  },
621
  {
622
  "epoch": 0.41044776119402987,
623
+ "grad_norm": 0.4968096911907196,
624
  "learning_rate": 0.000145,
625
+ "loss": 3.5282,
626
  "step": 4400
627
  },
628
  {
629
  "epoch": 0.41511194029850745,
630
+ "grad_norm": 0.5361109375953674,
631
  "learning_rate": 0.00014385321100917432,
632
+ "loss": 3.5219,
633
  "step": 4450
634
  },
635
  {
636
  "epoch": 0.4197761194029851,
637
+ "grad_norm": 0.4819442629814148,
638
  "learning_rate": 0.00014270642201834864,
639
+ "loss": 3.5247,
640
  "step": 4500
641
  },
642
  {
643
  "epoch": 0.4244402985074627,
644
+ "grad_norm": 0.43222498893737793,
645
  "learning_rate": 0.00014155963302752296,
646
+ "loss": 3.5003,
647
  "step": 4550
648
  },
649
  {
650
  "epoch": 0.4291044776119403,
651
+ "grad_norm": 0.40276777744293213,
652
  "learning_rate": 0.00014041284403669725,
653
+ "loss": 3.5076,
654
  "step": 4600
655
  },
656
  {
657
  "epoch": 0.4337686567164179,
658
+ "grad_norm": 0.4238736033439636,
659
  "learning_rate": 0.00013926605504587157,
660
+ "loss": 3.5017,
661
  "step": 4650
662
  },
663
  {
664
  "epoch": 0.43843283582089554,
665
+ "grad_norm": 0.4680284559726715,
666
  "learning_rate": 0.00013811926605504588,
667
+ "loss": 3.4945,
668
  "step": 4700
669
  },
670
  {
671
  "epoch": 0.4430970149253731,
672
+ "grad_norm": 0.42782142758369446,
673
  "learning_rate": 0.00013697247706422018,
674
+ "loss": 3.4769,
675
  "step": 4750
676
  },
677
  {
678
  "epoch": 0.44776119402985076,
679
+ "grad_norm": 0.48704081773757935,
680
  "learning_rate": 0.00013582568807339452,
681
+ "loss": 3.4835,
682
  "step": 4800
683
  },
684
  {
685
  "epoch": 0.45242537313432835,
686
+ "grad_norm": 0.4254228472709656,
687
  "learning_rate": 0.00013470183486238532,
688
+ "loss": 3.4746,
689
  "step": 4850
690
  },
691
  {
692
  "epoch": 0.457089552238806,
693
+ "grad_norm": 0.46868908405303955,
694
  "learning_rate": 0.00013355504587155963,
695
+ "loss": 3.4669,
696
  "step": 4900
697
  },
698
  {
699
  "epoch": 0.46175373134328357,
700
+ "grad_norm": 0.39554563164711,
701
  "learning_rate": 0.00013240825688073395,
702
+ "loss": 3.4612,
703
  "step": 4950
704
  },
705
  {
706
  "epoch": 0.4664179104477612,
707
+ "grad_norm": 0.5025322437286377,
708
  "learning_rate": 0.00013126146788990827,
709
+ "loss": 3.4626,
710
  "step": 5000
711
  },
712
  {
713
  "epoch": 0.4710820895522388,
714
+ "grad_norm": 0.45474347472190857,
715
  "learning_rate": 0.0001301146788990826,
716
+ "loss": 3.4573,
717
  "step": 5050
718
  },
719
  {
720
  "epoch": 0.47574626865671643,
721
+ "grad_norm": 0.45531395077705383,
722
  "learning_rate": 0.00012896788990825688,
723
+ "loss": 3.4622,
724
  "step": 5100
725
  },
726
  {
727
  "epoch": 0.480410447761194,
728
+ "grad_norm": 0.5105506181716919,
729
  "learning_rate": 0.0001278440366972477,
730
+ "loss": 3.4522,
731
  "step": 5150
732
  },
733
  {
734
  "epoch": 0.48507462686567165,
735
+ "grad_norm": 0.41265401244163513,
736
  "learning_rate": 0.00012669724770642202,
737
+ "loss": 3.4378,
738
  "step": 5200
739
  },
740
  {
741
  "epoch": 0.48973880597014924,
742
+ "grad_norm": 0.4541892409324646,
743
  "learning_rate": 0.00012555045871559634,
744
+ "loss": 3.4452,
745
  "step": 5250
746
  },
747
  {
748
  "epoch": 0.4944029850746269,
749
+ "grad_norm": 0.4062674641609192,
750
  "learning_rate": 0.00012440366972477063,
751
+ "loss": 3.4299,
752
  "step": 5300
753
  },
754
  {
755
  "epoch": 0.49906716417910446,
756
+ "grad_norm": 0.39839744567871094,
757
  "learning_rate": 0.00012325688073394498,
758
+ "loss": 3.4231,
759
  "step": 5350
760
  },
761
  {
762
  "epoch": 0.503731343283582,
763
+ "grad_norm": 0.4296305775642395,
764
  "learning_rate": 0.00012211009174311927,
765
+ "loss": 3.4399,
766
  "step": 5400
767
  },
768
  {
769
  "epoch": 0.5083955223880597,
770
+ "grad_norm": 0.40859147906303406,
771
  "learning_rate": 0.0001209633027522936,
772
+ "loss": 3.4226,
773
  "step": 5450
774
  },
775
  {
776
  "epoch": 0.5130597014925373,
777
+ "grad_norm": 0.38735997676849365,
778
  "learning_rate": 0.00011981651376146789,
779
+ "loss": 3.4168,
780
  "step": 5500
781
  },
782
  {
783
  "epoch": 0.5177238805970149,
784
+ "grad_norm": 0.5080564022064209,
785
  "learning_rate": 0.0001186697247706422,
786
+ "loss": 3.4154,
787
  "step": 5550
788
  },
789
  {
790
  "epoch": 0.5223880597014925,
791
+ "grad_norm": 0.42273515462875366,
792
  "learning_rate": 0.00011752293577981653,
793
+ "loss": 3.4155,
794
  "step": 5600
795
  },
796
  {
797
  "epoch": 0.5270522388059702,
798
+ "grad_norm": 0.41893094778060913,
799
  "learning_rate": 0.00011637614678899082,
800
+ "loss": 3.4087,
801
  "step": 5650
802
  },
803
  {
804
  "epoch": 0.5317164179104478,
805
+ "grad_norm": 0.4470965564250946,
806
  "learning_rate": 0.00011522935779816515,
807
+ "loss": 3.3979,
808
  "step": 5700
809
  },
810
  {
811
  "epoch": 0.5363805970149254,
812
+ "grad_norm": 0.40868085622787476,
813
  "learning_rate": 0.00011408256880733946,
814
+ "loss": 3.4007,
815
  "step": 5750
816
  },
817
  {
818
  "epoch": 0.5410447761194029,
819
+ "grad_norm": 0.42737168073654175,
820
  "learning_rate": 0.00011293577981651377,
821
+ "loss": 3.3904,
822
  "step": 5800
823
  },
824
  {
825
  "epoch": 0.5457089552238806,
826
+ "grad_norm": 0.5149052143096924,
827
  "learning_rate": 0.00011178899082568808,
828
+ "loss": 3.3914,
829
  "step": 5850
830
  },
831
  {
832
  "epoch": 0.5503731343283582,
833
+ "grad_norm": 0.4031182825565338,
834
  "learning_rate": 0.00011064220183486238,
835
+ "loss": 3.3845,
836
  "step": 5900
837
  },
838
  {
839
  "epoch": 0.5550373134328358,
840
+ "grad_norm": 0.41353049874305725,
841
  "learning_rate": 0.0001094954128440367,
842
+ "loss": 3.3976,
843
  "step": 5950
844
  },
845
  {
846
  "epoch": 0.5597014925373134,
847
+ "grad_norm": 0.412335604429245,
848
  "learning_rate": 0.000108348623853211,
849
+ "loss": 3.3782,
850
  "step": 6000
851
  },
852
  {
853
  "epoch": 0.5643656716417911,
854
+ "grad_norm": 0.3940559923648834,
855
  "learning_rate": 0.00010720183486238533,
856
+ "loss": 3.3755,
857
  "step": 6050
858
  },
859
  {
860
  "epoch": 0.5690298507462687,
861
+ "grad_norm": 0.3801034092903137,
862
  "learning_rate": 0.00010605504587155963,
863
+ "loss": 3.3668,
864
  "step": 6100
865
  },
866
  {
867
  "epoch": 0.5736940298507462,
868
+ "grad_norm": 0.5560407638549805,
869
  "learning_rate": 0.00010490825688073396,
870
+ "loss": 3.3648,
871
  "step": 6150
872
  },
873
  {
874
  "epoch": 0.5783582089552238,
875
+ "grad_norm": 0.4365626275539398,
876
  "learning_rate": 0.00010376146788990825,
877
+ "loss": 3.374,
878
  "step": 6200
879
  },
880
  {
881
  "epoch": 0.5830223880597015,
882
+ "grad_norm": 0.39503213763237,
883
  "learning_rate": 0.00010261467889908258,
884
+ "loss": 3.3679,
885
  "step": 6250
886
  },
887
  {
888
  "epoch": 0.5876865671641791,
889
+ "grad_norm": 0.41522470116615295,
890
  "learning_rate": 0.00010146788990825689,
891
+ "loss": 3.3532,
892
  "step": 6300
893
  },
894
  {
895
  "epoch": 0.5923507462686567,
896
+ "grad_norm": 0.3890146315097809,
897
  "learning_rate": 0.00010032110091743118,
898
+ "loss": 3.3667,
899
  "step": 6350
900
  },
901
  {
902
  "epoch": 0.5970149253731343,
903
+ "grad_norm": 0.5168399810791016,
904
  "learning_rate": 9.917431192660551e-05,
905
+ "loss": 3.358,
906
  "step": 6400
907
  },
908
  {
909
  "epoch": 0.601679104477612,
910
+ "grad_norm": 0.3710983395576477,
911
  "learning_rate": 9.802752293577983e-05,
912
+ "loss": 3.3585,
913
  "step": 6450
914
  },
915
  {
916
  "epoch": 0.6063432835820896,
917
+ "grad_norm": 0.4420715570449829,
918
  "learning_rate": 9.688073394495414e-05,
919
+ "loss": 3.3539,
920
  "step": 6500
921
  },
922
  {
923
  "epoch": 0.6110074626865671,
924
+ "grad_norm": 0.3815698027610779,
925
+ "learning_rate": 9.573394495412844e-05,
926
+ "loss": 3.3484,
927
  "step": 6550
928
  },
929
  {
930
  "epoch": 0.6156716417910447,
931
+ "grad_norm": 0.38033589720726013,
932
+ "learning_rate": 9.458715596330276e-05,
933
+ "loss": 3.3471,
934
  "step": 6600
935
  },
936
  {
937
  "epoch": 0.6203358208955224,
938
+ "grad_norm": 0.35299167037010193,
939
+ "learning_rate": 9.344036697247706e-05,
940
+ "loss": 3.3555,
941
  "step": 6650
942
  },
943
  {
944
  "epoch": 0.625,
945
+ "grad_norm": 0.3831453025341034,
946
+ "learning_rate": 9.229357798165138e-05,
947
+ "loss": 3.343,
948
  "step": 6700
949
  },
950
  {
951
  "epoch": 0.6296641791044776,
952
+ "grad_norm": 0.38805127143859863,
953
+ "learning_rate": 9.114678899082569e-05,
954
+ "loss": 3.3362,
955
  "step": 6750
956
  },
957
  {
958
  "epoch": 0.6343283582089553,
959
+ "grad_norm": 0.360837459564209,
960
+ "learning_rate": 9e-05,
961
+ "loss": 3.329,
962
  "step": 6800
963
  },
964
  {
965
  "epoch": 0.6389925373134329,
966
+ "grad_norm": 0.48795634508132935,
967
+ "learning_rate": 8.885321100917432e-05,
968
+ "loss": 3.3244,
969
  "step": 6850
970
  },
971
  {
972
  "epoch": 0.6436567164179104,
973
+ "grad_norm": 0.3907620310783386,
974
+ "learning_rate": 8.770642201834863e-05,
975
+ "loss": 3.3326,
976
  "step": 6900
977
  },
978
  {
979
  "epoch": 0.648320895522388,
980
+ "grad_norm": 0.4346226751804352,
981
+ "learning_rate": 8.655963302752293e-05,
982
+ "loss": 3.3291,
983
  "step": 6950
984
  },
985
  {
986
  "epoch": 0.6529850746268657,
987
+ "grad_norm": 0.4240095019340515,
988
+ "learning_rate": 8.541284403669725e-05,
989
+ "loss": 3.3236,
990
  "step": 7000
991
  },
992
  {
993
  "epoch": 0.6576492537313433,
994
+ "grad_norm": 0.4035656452178955,
995
+ "learning_rate": 8.426605504587156e-05,
996
+ "loss": 3.3188,
997
  "step": 7050
998
  },
999
  {
1000
  "epoch": 0.6623134328358209,
1001
+ "grad_norm": 0.44728022813796997,
1002
+ "learning_rate": 8.311926605504588e-05,
1003
+ "loss": 3.3179,
1004
  "step": 7100
1005
  },
1006
  {
1007
  "epoch": 0.6669776119402985,
1008
+ "grad_norm": 0.3958663046360016,
1009
+ "learning_rate": 8.19724770642202e-05,
1010
+ "loss": 3.3081,
1011
  "step": 7150
1012
  },
1013
  {
1014
  "epoch": 0.6716417910447762,
1015
+ "grad_norm": 0.4779495894908905,
1016
+ "learning_rate": 8.08256880733945e-05,
1017
+ "loss": 3.3194,
1018
  "step": 7200
1019
  },
1020
  {
1021
  "epoch": 0.6763059701492538,
1022
+ "grad_norm": 0.3885568082332611,
1023
+ "learning_rate": 7.967889908256882e-05,
1024
+ "loss": 3.3123,
1025
  "step": 7250
1026
  },
1027
  {
1028
  "epoch": 0.6809701492537313,
1029
+ "grad_norm": 0.46929308772087097,
1030
+ "learning_rate": 7.853211009174312e-05,
1031
+ "loss": 3.2946,
1032
  "step": 7300
1033
  },
1034
  {
1035
  "epoch": 0.6856343283582089,
1036
+ "grad_norm": 0.3649180829524994,
1037
+ "learning_rate": 7.738532110091743e-05,
1038
+ "loss": 3.3172,
1039
  "step": 7350
1040
  },
1041
  {
1042
  "epoch": 0.6902985074626866,
1043
+ "grad_norm": 0.3710353672504425,
1044
+ "learning_rate": 7.623853211009174e-05,
1045
+ "loss": 3.3026,
1046
  "step": 7400
1047
  },
1048
  {
1049
  "epoch": 0.6949626865671642,
1050
+ "grad_norm": 0.3968711197376251,
1051
+ "learning_rate": 7.509174311926605e-05,
1052
+ "loss": 3.3026,
1053
  "step": 7450
1054
  },
1055
  {
1056
  "epoch": 0.6996268656716418,
1057
+ "grad_norm": 0.4033445715904236,
1058
+ "learning_rate": 7.394495412844037e-05,
1059
+ "loss": 3.2982,
1060
  "step": 7500
1061
  },
1062
  {
1063
  "epoch": 0.7042910447761194,
1064
+ "grad_norm": 0.4648160934448242,
1065
+ "learning_rate": 7.279816513761469e-05,
1066
+ "loss": 3.2959,
1067
  "step": 7550
1068
  },
1069
  {
1070
  "epoch": 0.7089552238805971,
1071
+ "grad_norm": 0.399700790643692,
1072
+ "learning_rate": 7.165137614678899e-05,
1073
+ "loss": 3.2955,
1074
  "step": 7600
1075
  },
1076
  {
1077
  "epoch": 0.7136194029850746,
1078
+ "grad_norm": 0.43972837924957275,
1079
+ "learning_rate": 7.050458715596331e-05,
1080
+ "loss": 3.2947,
1081
  "step": 7650
1082
  },
1083
  {
1084
  "epoch": 0.7182835820895522,
1085
+ "grad_norm": 0.3754163086414337,
1086
+ "learning_rate": 6.935779816513763e-05,
1087
+ "loss": 3.2901,
1088
  "step": 7700
1089
  },
1090
  {
1091
  "epoch": 0.7229477611940298,
1092
+ "grad_norm": 0.38772958517074585,
1093
+ "learning_rate": 6.821100917431193e-05,
1094
+ "loss": 3.2843,
1095
  "step": 7750
1096
  },
1097
  {
1098
  "epoch": 0.7276119402985075,
1099
+ "grad_norm": 0.3898366689682007,
1100
+ "learning_rate": 6.706422018348624e-05,
1101
+ "loss": 3.2818,
1102
  "step": 7800
1103
  },
1104
  {
1105
  "epoch": 0.7322761194029851,
1106
+ "grad_norm": 0.3694920539855957,
1107
+ "learning_rate": 6.591743119266056e-05,
1108
+ "loss": 3.285,
1109
  "step": 7850
1110
  },
1111
  {
1112
  "epoch": 0.7369402985074627,
1113
+ "grad_norm": 0.3696017861366272,
1114
+ "learning_rate": 6.477064220183486e-05,
1115
+ "loss": 3.282,
1116
  "step": 7900
1117
  },
1118
  {
1119
  "epoch": 0.7416044776119403,
1120
+ "grad_norm": 0.4159897565841675,
1121
+ "learning_rate": 6.362385321100918e-05,
1122
+ "loss": 3.2778,
1123
  "step": 7950
1124
  },
1125
  {
1126
  "epoch": 0.746268656716418,
1127
+ "grad_norm": 0.3910117447376251,
1128
+ "learning_rate": 6.247706422018348e-05,
1129
+ "loss": 3.2852,
1130
  "step": 8000
1131
  },
1132
  {
1133
  "epoch": 0.7509328358208955,
1134
+ "grad_norm": 0.44550907611846924,
1135
+ "learning_rate": 6.13302752293578e-05,
1136
+ "loss": 3.2731,
1137
  "step": 8050
1138
  },
1139
  {
1140
  "epoch": 0.7555970149253731,
1141
+ "grad_norm": 0.4173608422279358,
1142
+ "learning_rate": 6.0183486238532114e-05,
1143
+ "loss": 3.2865,
1144
  "step": 8100
1145
  },
1146
  {
1147
  "epoch": 0.7602611940298507,
1148
+ "grad_norm": 0.3633171021938324,
1149
+ "learning_rate": 5.903669724770643e-05,
1150
+ "loss": 3.2783,
1151
  "step": 8150
1152
  },
1153
  {
1154
  "epoch": 0.7649253731343284,
1155
+ "grad_norm": 0.42995285987854004,
1156
+ "learning_rate": 5.793577981651377e-05,
1157
+ "loss": 3.2732,
1158
  "step": 8200
1159
  },
1160
  {
1161
  "epoch": 0.769589552238806,
1162
+ "grad_norm": 0.3819643557071686,
1163
+ "learning_rate": 5.678899082568807e-05,
1164
+ "loss": 3.2677,
1165
  "step": 8250
1166
  },
1167
  {
1168
  "epoch": 0.7742537313432836,
1169
+ "grad_norm": 0.3480379581451416,
1170
+ "learning_rate": 5.564220183486238e-05,
1171
+ "loss": 3.2749,
1172
  "step": 8300
1173
  },
1174
  {
1175
  "epoch": 0.7789179104477612,
1176
+ "grad_norm": 0.3577636480331421,
1177
+ "learning_rate": 5.44954128440367e-05,
1178
+ "loss": 3.2725,
1179
  "step": 8350
1180
  },
1181
  {
1182
  "epoch": 0.7835820895522388,
1183
+ "grad_norm": 0.36848926544189453,
1184
+ "learning_rate": 5.334862385321101e-05,
1185
+ "loss": 3.2744,
1186
  "step": 8400
1187
  },
1188
  {
1189
  "epoch": 0.7882462686567164,
1190
+ "grad_norm": 0.35368892550468445,
1191
+ "learning_rate": 5.220183486238532e-05,
1192
+ "loss": 3.2616,
1193
  "step": 8450
1194
  },
1195
  {
1196
  "epoch": 0.792910447761194,
1197
+ "grad_norm": 0.41831615567207336,
1198
+ "learning_rate": 5.1055045871559635e-05,
1199
+ "loss": 3.277,
1200
  "step": 8500
1201
  },
1202
  {
1203
  "epoch": 0.7975746268656716,
1204
+ "grad_norm": 0.41986221075057983,
1205
+ "learning_rate": 4.990825688073395e-05,
1206
+ "loss": 3.2522,
1207
  "step": 8550
1208
  },
1209
  {
1210
  "epoch": 0.8022388059701493,
1211
+ "grad_norm": 0.3512188494205475,
1212
+ "learning_rate": 4.876146788990826e-05,
1213
+ "loss": 3.2709,
1214
  "step": 8600
1215
  },
1216
  {
1217
  "epoch": 0.8069029850746269,
1218
+ "grad_norm": 0.3639850318431854,
1219
+ "learning_rate": 4.761467889908257e-05,
1220
+ "loss": 3.2571,
1221
  "step": 8650
1222
  },
1223
  {
1224
  "epoch": 0.8115671641791045,
1225
+ "grad_norm": 0.4546779692173004,
1226
+ "learning_rate": 4.646788990825688e-05,
1227
+ "loss": 3.2599,
1228
  "step": 8700
1229
  },
1230
  {
1231
  "epoch": 0.816231343283582,
1232
+ "grad_norm": 0.35850730538368225,
1233
+ "learning_rate": 4.53211009174312e-05,
1234
+ "loss": 3.2531,
1235
  "step": 8750
1236
  },
1237
  {
1238
  "epoch": 0.8208955223880597,
1239
+ "grad_norm": 0.3843632936477661,
1240
+ "learning_rate": 4.4174311926605504e-05,
1241
+ "loss": 3.2598,
1242
  "step": 8800
1243
  },
1244
  {
1245
  "epoch": 0.8255597014925373,
1246
+ "grad_norm": 0.3926890790462494,
1247
+ "learning_rate": 4.3027522935779816e-05,
1248
+ "loss": 3.2579,
1249
  "step": 8850
1250
  },
1251
  {
1252
  "epoch": 0.8302238805970149,
1253
+ "grad_norm": 0.36206337809562683,
1254
+ "learning_rate": 4.1880733944954134e-05,
1255
+ "loss": 3.2503,
1256
  "step": 8900
1257
  },
1258
  {
1259
  "epoch": 0.8348880597014925,
1260
+ "grad_norm": 0.3471512198448181,
1261
+ "learning_rate": 4.0733944954128446e-05,
1262
+ "loss": 3.2478,
1263
  "step": 8950
1264
  },
1265
  {
1266
  "epoch": 0.8395522388059702,
1267
+ "grad_norm": 0.3711581230163574,
1268
+ "learning_rate": 3.958715596330275e-05,
1269
+ "loss": 3.2461,
1270
  "step": 9000
1271
  },
1272
  {
1273
  "epoch": 0.8442164179104478,
1274
+ "grad_norm": 0.3600054979324341,
1275
+ "learning_rate": 3.844036697247706e-05,
1276
+ "loss": 3.2375,
1277
  "step": 9050
1278
  },
1279
  {
1280
  "epoch": 0.8488805970149254,
1281
+ "grad_norm": 0.3525969088077545,
1282
+ "learning_rate": 3.729357798165138e-05,
1283
+ "loss": 3.2446,
1284
  "step": 9100
1285
  },
1286
  {
1287
  "epoch": 0.8535447761194029,
1288
+ "grad_norm": 0.3669526278972626,
1289
+ "learning_rate": 3.614678899082569e-05,
1290
+ "loss": 3.2459,
1291
  "step": 9150
1292
  },
1293
  {
1294
  "epoch": 0.8582089552238806,
1295
+ "grad_norm": 0.3348044753074646,
1296
  "learning_rate": 3.504587155963303e-05,
1297
+ "loss": 3.2498,
1298
  "step": 9200
1299
  },
1300
  {
1301
  "epoch": 0.8628731343283582,
1302
+ "grad_norm": 0.3626518249511719,
1303
  "learning_rate": 3.3899082568807337e-05,
1304
+ "loss": 3.2501,
1305
  "step": 9250
1306
  },
1307
  {
1308
  "epoch": 0.8675373134328358,
1309
+ "grad_norm": 0.377811998128891,
1310
  "learning_rate": 3.2752293577981655e-05,
1311
+ "loss": 3.2385,
1312
  "step": 9300
1313
  },
1314
  {
1315
  "epoch": 0.8722014925373134,
1316
+ "grad_norm": 0.3583707809448242,
1317
  "learning_rate": 3.1605504587155967e-05,
1318
+ "loss": 3.2411,
1319
  "step": 9350
1320
  },
1321
  {
1322
  "epoch": 0.8768656716417911,
1323
+ "grad_norm": 0.35831955075263977,
1324
  "learning_rate": 3.0458715596330278e-05,
1325
+ "loss": 3.2486,
1326
  "step": 9400
1327
  },
1328
  {
1329
  "epoch": 0.8815298507462687,
1330
+ "grad_norm": 0.3578301668167114,
1331
  "learning_rate": 2.9311926605504586e-05,
1332
+ "loss": 3.2352,
1333
  "step": 9450
1334
  },
1335
  {
1336
  "epoch": 0.8861940298507462,
1337
+ "grad_norm": 0.3495340645313263,
1338
  "learning_rate": 2.8165137614678898e-05,
1339
+ "loss": 3.2336,
1340
  "step": 9500
1341
  },
1342
  {
1343
  "epoch": 0.8908582089552238,
1344
+ "grad_norm": 0.34632202982902527,
1345
  "learning_rate": 2.7018348623853213e-05,
1346
+ "loss": 3.2288,
1347
  "step": 9550
1348
  },
1349
  {
1350
  "epoch": 0.8955223880597015,
1351
+ "grad_norm": 0.3287692070007324,
1352
  "learning_rate": 2.5871559633027525e-05,
1353
+ "loss": 3.2349,
1354
  "step": 9600
1355
  },
1356
  {
1357
  "epoch": 0.9001865671641791,
1358
+ "grad_norm": 0.329765260219574,
1359
+ "learning_rate": 2.474770642201835e-05,
1360
+ "loss": 3.2271,
1361
  "step": 9650
1362
  },
1363
  {
1364
  "epoch": 0.9048507462686567,
1365
+ "grad_norm": 0.3470270037651062,
1366
+ "learning_rate": 2.3600917431192662e-05,
1367
+ "loss": 3.2279,
1368
  "step": 9700
1369
  },
1370
  {
1371
  "epoch": 0.9095149253731343,
1372
+ "grad_norm": 0.34217342734336853,
1373
+ "learning_rate": 2.2454128440366973e-05,
1374
+ "loss": 3.2329,
1375
  "step": 9750
1376
  },
1377
  {
1378
  "epoch": 0.914179104477612,
1379
+ "grad_norm": 0.33399471640586853,
1380
+ "learning_rate": 2.1307339449541285e-05,
1381
+ "loss": 3.2284,
1382
  "step": 9800
1383
  },
1384
  {
1385
  "epoch": 0.9188432835820896,
1386
+ "grad_norm": 0.3363756835460663,
1387
+ "learning_rate": 2.0160550458715597e-05,
1388
+ "loss": 3.2302,
1389
  "step": 9850
1390
  },
1391
  {
1392
  "epoch": 0.9235074626865671,
1393
+ "grad_norm": 0.3226417303085327,
1394
+ "learning_rate": 1.9013761467889908e-05,
1395
+ "loss": 3.226,
1396
  "step": 9900
1397
  },
1398
  {
1399
  "epoch": 0.9281716417910447,
1400
+ "grad_norm": 0.33726227283477783,
1401
+ "learning_rate": 1.786697247706422e-05,
1402
+ "loss": 3.23,
1403
  "step": 9950
1404
  },
1405
  {
1406
  "epoch": 0.9328358208955224,
1407
+ "grad_norm": 0.321735143661499,
1408
+ "learning_rate": 1.6720183486238535e-05,
1409
+ "loss": 3.2278,
1410
  "step": 10000
1411
  }
1412
  ],
 
1427
  "attributes": {}
1428
  }
1429
  },
1430
+ "total_flos": 2124187846049792.0,
1431
  "train_batch_size": 64,
1432
  "trial_name": null,
1433
  "trial_params": null
frozen/checkpoint-10720/model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:801f27268aa61bcbd737f0ac81310a163e93f798aa2731904a74ce142d761898
3
- size 1067471056
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a311c3477778d4618305c6a250f22332b3347e78451e29bd06e1d4af932bb2a
3
+ size 800608432
frozen/checkpoint-10720/scheduler.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:27636c48b899230a9cd8b7e63ecb4b9bd567dd54336eed166836cff4fa52e881
3
  size 1465
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a264713324385af4418bfbf723e4ed00b6014d8f19715bd066960210d5e0f2d1
3
  size 1465
frozen/checkpoint-10720/trainer_state.json CHANGED
@@ -11,1500 +11,1500 @@
11
  "log_history": [
12
  {
13
  "epoch": 0.0046641791044776115,
14
- "grad_norm": 1.907758116722107,
15
  "learning_rate": 4.9000000000000005e-06,
16
- "loss": 10.0793,
17
  "step": 50
18
  },
19
  {
20
  "epoch": 0.009328358208955223,
21
- "grad_norm": 1.399185061454773,
22
  "learning_rate": 9.900000000000002e-06,
23
- "loss": 8.9287,
24
  "step": 100
25
  },
26
  {
27
  "epoch": 0.013992537313432836,
28
- "grad_norm": 1.0101089477539062,
29
  "learning_rate": 1.49e-05,
30
- "loss": 8.3288,
31
  "step": 150
32
  },
33
  {
34
  "epoch": 0.018656716417910446,
35
- "grad_norm": 0.7379341125488281,
36
  "learning_rate": 1.9900000000000003e-05,
37
- "loss": 7.6969,
38
  "step": 200
39
  },
40
  {
41
  "epoch": 0.02332089552238806,
42
- "grad_norm": 0.9656306505203247,
43
  "learning_rate": 2.4900000000000002e-05,
44
- "loss": 7.2698,
45
  "step": 250
46
  },
47
  {
48
  "epoch": 0.027985074626865673,
49
- "grad_norm": 1.3790279626846313,
50
  "learning_rate": 2.9900000000000002e-05,
51
- "loss": 6.9685,
52
  "step": 300
53
  },
54
  {
55
  "epoch": 0.03264925373134328,
56
- "grad_norm": 0.6374709606170654,
57
  "learning_rate": 3.49e-05,
58
- "loss": 6.6752,
59
  "step": 350
60
  },
61
  {
62
  "epoch": 0.03731343283582089,
63
- "grad_norm": 1.5095465183258057,
64
  "learning_rate": 3.99e-05,
65
- "loss": 6.3805,
66
  "step": 400
67
  },
68
  {
69
  "epoch": 0.04197761194029851,
70
- "grad_norm": 1.394410490989685,
71
  "learning_rate": 4.49e-05,
72
- "loss": 6.1667,
73
  "step": 450
74
  },
75
  {
76
  "epoch": 0.04664179104477612,
77
- "grad_norm": 1.7006834745407104,
78
  "learning_rate": 4.99e-05,
79
- "loss": 5.9861,
80
  "step": 500
81
  },
82
  {
83
  "epoch": 0.051305970149253734,
84
- "grad_norm": 0.9934776425361633,
85
  "learning_rate": 5.4900000000000006e-05,
86
- "loss": 5.8166,
87
  "step": 550
88
  },
89
  {
90
  "epoch": 0.055970149253731345,
91
- "grad_norm": 2.0550477504730225,
92
  "learning_rate": 5.99e-05,
93
- "loss": 5.6646,
94
  "step": 600
95
  },
96
  {
97
  "epoch": 0.06063432835820896,
98
- "grad_norm": 2.2504913806915283,
99
  "learning_rate": 6.49e-05,
100
- "loss": 5.5353,
101
  "step": 650
102
  },
103
  {
104
  "epoch": 0.06529850746268656,
105
- "grad_norm": 2.468212604522705,
106
  "learning_rate": 6.99e-05,
107
- "loss": 5.3982,
108
  "step": 700
109
  },
110
  {
111
  "epoch": 0.06996268656716417,
112
- "grad_norm": 2.4010939598083496,
113
  "learning_rate": 7.49e-05,
114
- "loss": 5.2855,
115
  "step": 750
116
  },
117
  {
118
  "epoch": 0.07462686567164178,
119
- "grad_norm": 1.4586199522018433,
120
  "learning_rate": 7.99e-05,
121
- "loss": 5.1901,
122
  "step": 800
123
  },
124
  {
125
  "epoch": 0.07929104477611941,
126
- "grad_norm": 1.2821402549743652,
127
  "learning_rate": 8.49e-05,
128
- "loss": 5.0973,
129
  "step": 850
130
  },
131
  {
132
  "epoch": 0.08395522388059702,
133
- "grad_norm": 2.1037180423736572,
134
  "learning_rate": 8.99e-05,
135
- "loss": 5.0062,
136
  "step": 900
137
  },
138
  {
139
  "epoch": 0.08861940298507463,
140
- "grad_norm": 1.370751976966858,
141
  "learning_rate": 9.49e-05,
142
- "loss": 4.9392,
143
  "step": 950
144
  },
145
  {
146
  "epoch": 0.09328358208955224,
147
- "grad_norm": 1.276108741760254,
148
  "learning_rate": 9.99e-05,
149
- "loss": 4.848,
150
  "step": 1000
151
  },
152
  {
153
  "epoch": 0.09794776119402986,
154
- "grad_norm": 1.1077498197555542,
155
  "learning_rate": 0.0001049,
156
- "loss": 4.7863,
157
  "step": 1050
158
  },
159
  {
160
  "epoch": 0.10261194029850747,
161
- "grad_norm": 1.0212163925170898,
162
  "learning_rate": 0.0001099,
163
- "loss": 4.7102,
164
  "step": 1100
165
  },
166
  {
167
  "epoch": 0.10727611940298508,
168
- "grad_norm": 1.1480289697647095,
169
  "learning_rate": 0.0001149,
170
- "loss": 4.6551,
171
  "step": 1150
172
  },
173
  {
174
  "epoch": 0.11194029850746269,
175
- "grad_norm": 1.3106518983840942,
176
  "learning_rate": 0.00011990000000000001,
177
- "loss": 4.5627,
178
  "step": 1200
179
  },
180
  {
181
  "epoch": 0.1166044776119403,
182
- "grad_norm": 1.4304360151290894,
183
  "learning_rate": 0.00012490000000000003,
184
- "loss": 4.4823,
185
  "step": 1250
186
  },
187
  {
188
  "epoch": 0.12126865671641791,
189
- "grad_norm": 0.883263885974884,
190
  "learning_rate": 0.00012989999999999999,
191
- "loss": 4.4091,
192
  "step": 1300
193
  },
194
  {
195
  "epoch": 0.1259328358208955,
196
- "grad_norm": 1.2335679531097412,
197
  "learning_rate": 0.0001349,
198
- "loss": 4.3284,
199
  "step": 1350
200
  },
201
  {
202
  "epoch": 0.13059701492537312,
203
- "grad_norm": 1.3587566614151,
204
  "learning_rate": 0.0001399,
205
- "loss": 4.2523,
206
  "step": 1400
207
  },
208
  {
209
  "epoch": 0.13526119402985073,
210
- "grad_norm": 1.4397273063659668,
211
  "learning_rate": 0.00014490000000000003,
212
- "loss": 4.1874,
213
  "step": 1450
214
  },
215
  {
216
  "epoch": 0.13992537313432835,
217
- "grad_norm": 0.9128643274307251,
218
  "learning_rate": 0.0001499,
219
- "loss": 4.1382,
220
  "step": 1500
221
  },
222
  {
223
  "epoch": 0.14458955223880596,
224
- "grad_norm": 1.1429646015167236,
225
  "learning_rate": 0.0001549,
226
- "loss": 4.0713,
227
  "step": 1550
228
  },
229
  {
230
  "epoch": 0.14925373134328357,
231
- "grad_norm": 0.7839804291725159,
232
  "learning_rate": 0.0001599,
233
- "loss": 4.0327,
234
  "step": 1600
235
  },
236
  {
237
  "epoch": 0.15391791044776118,
238
- "grad_norm": 0.5934556126594543,
239
  "learning_rate": 0.0001649,
240
- "loss": 4.0535,
241
  "step": 1650
242
  },
243
  {
244
  "epoch": 0.15858208955223882,
245
- "grad_norm": 0.6320812702178955,
246
  "learning_rate": 0.0001699,
247
- "loss": 3.9615,
248
  "step": 1700
249
  },
250
  {
251
  "epoch": 0.16324626865671643,
252
- "grad_norm": 0.6207342743873596,
253
  "learning_rate": 0.00017490000000000002,
254
- "loss": 3.9086,
255
  "step": 1750
256
  },
257
  {
258
  "epoch": 0.16791044776119404,
259
- "grad_norm": 0.8524695634841919,
260
  "learning_rate": 0.0001799,
261
- "loss": 3.8703,
262
  "step": 1800
263
  },
264
  {
265
  "epoch": 0.17257462686567165,
266
- "grad_norm": 0.9148792028427124,
267
  "learning_rate": 0.0001849,
268
- "loss": 3.8388,
269
  "step": 1850
270
  },
271
  {
272
  "epoch": 0.17723880597014927,
273
- "grad_norm": 0.7455419301986694,
274
  "learning_rate": 0.0001899,
275
- "loss": 3.8031,
276
  "step": 1900
277
  },
278
  {
279
  "epoch": 0.18190298507462688,
280
- "grad_norm": 0.9777323603630066,
281
  "learning_rate": 0.00019490000000000002,
282
- "loss": 3.781,
283
  "step": 1950
284
  },
285
  {
286
  "epoch": 0.1865671641791045,
287
- "grad_norm": 0.6018832921981812,
288
  "learning_rate": 0.0001999,
289
- "loss": 3.7421,
290
  "step": 2000
291
  },
292
  {
293
  "epoch": 0.1912313432835821,
294
- "grad_norm": 0.7007433772087097,
295
  "learning_rate": 0.00019887614678899083,
296
- "loss": 3.7148,
297
  "step": 2050
298
  },
299
  {
300
  "epoch": 0.1958955223880597,
301
- "grad_norm": 0.7864815592765808,
302
  "learning_rate": 0.00019772935779816515,
303
- "loss": 3.6907,
304
  "step": 2100
305
  },
306
  {
307
  "epoch": 0.20055970149253732,
308
- "grad_norm": 0.6575994491577148,
309
  "learning_rate": 0.00019658256880733944,
310
- "loss": 3.6504,
311
  "step": 2150
312
  },
313
  {
314
  "epoch": 0.20522388059701493,
315
- "grad_norm": 0.524401843547821,
316
  "learning_rate": 0.0001954357798165138,
317
- "loss": 3.6333,
318
  "step": 2200
319
  },
320
  {
321
  "epoch": 0.20988805970149255,
322
- "grad_norm": 0.8242985010147095,
323
  "learning_rate": 0.00019428899082568808,
324
- "loss": 3.6084,
325
  "step": 2250
326
  },
327
  {
328
  "epoch": 0.21455223880597016,
329
- "grad_norm": 0.5965232849121094,
330
  "learning_rate": 0.0001931422018348624,
331
- "loss": 3.5914,
332
  "step": 2300
333
  },
334
  {
335
  "epoch": 0.21921641791044777,
336
- "grad_norm": 0.5473845601081848,
337
  "learning_rate": 0.00019199541284403671,
338
- "loss": 3.572,
339
  "step": 2350
340
  },
341
  {
342
  "epoch": 0.22388059701492538,
343
- "grad_norm": 0.7030166983604431,
344
  "learning_rate": 0.00019084862385321103,
345
- "loss": 3.5548,
346
  "step": 2400
347
  },
348
  {
349
  "epoch": 0.228544776119403,
350
- "grad_norm": 0.5844017267227173,
351
- "learning_rate": 0.00018972477064220186,
352
- "loss": 3.5326,
353
  "step": 2450
354
  },
355
  {
356
  "epoch": 0.2332089552238806,
357
- "grad_norm": 0.4722403883934021,
358
- "learning_rate": 0.00018857798165137615,
359
- "loss": 3.5113,
360
  "step": 2500
361
  },
362
  {
363
  "epoch": 0.23787313432835822,
364
- "grad_norm": 0.4691997468471527,
365
- "learning_rate": 0.00018743119266055047,
366
- "loss": 3.4985,
367
  "step": 2550
368
  },
369
  {
370
  "epoch": 0.24253731343283583,
371
- "grad_norm": 0.5742476582527161,
372
- "learning_rate": 0.00018628440366972478,
373
- "loss": 3.4689,
374
  "step": 2600
375
  },
376
  {
377
  "epoch": 0.24720149253731344,
378
- "grad_norm": 0.4867892563343048,
379
- "learning_rate": 0.0001851376146788991,
380
- "loss": 3.4737,
381
  "step": 2650
382
  },
383
  {
384
  "epoch": 0.251865671641791,
385
- "grad_norm": 0.5876694321632385,
386
- "learning_rate": 0.0001839908256880734,
387
- "loss": 3.4516,
388
  "step": 2700
389
  },
390
  {
391
  "epoch": 0.25652985074626866,
392
- "grad_norm": 0.4691877067089081,
393
- "learning_rate": 0.0001828440366972477,
394
- "loss": 3.4411,
395
  "step": 2750
396
  },
397
  {
398
  "epoch": 0.26119402985074625,
399
- "grad_norm": 0.497503399848938,
400
- "learning_rate": 0.00018169724770642203,
401
- "loss": 3.4311,
402
  "step": 2800
403
  },
404
  {
405
  "epoch": 0.2658582089552239,
406
- "grad_norm": 0.4610927700996399,
407
- "learning_rate": 0.00018055045871559635,
408
- "loss": 3.4122,
409
  "step": 2850
410
  },
411
  {
412
  "epoch": 0.27052238805970147,
413
- "grad_norm": 0.5312972664833069,
414
- "learning_rate": 0.00017940366972477064,
415
- "loss": 3.4004,
416
  "step": 2900
417
  },
418
  {
419
  "epoch": 0.2751865671641791,
420
- "grad_norm": 0.5256180167198181,
421
- "learning_rate": 0.00017825688073394496,
422
- "loss": 3.3834,
423
  "step": 2950
424
  },
425
  {
426
  "epoch": 0.2798507462686567,
427
- "grad_norm": 0.5108594298362732,
428
- "learning_rate": 0.00017711009174311928,
429
- "loss": 3.379,
430
  "step": 3000
431
  },
432
  {
433
  "epoch": 0.28451492537313433,
434
- "grad_norm": 0.4818528890609741,
435
- "learning_rate": 0.00017596330275229357,
436
- "loss": 3.3645,
437
  "step": 3050
438
  },
439
  {
440
  "epoch": 0.2891791044776119,
441
- "grad_norm": 0.4077577590942383,
442
- "learning_rate": 0.0001748165137614679,
443
- "loss": 3.3659,
444
  "step": 3100
445
  },
446
  {
447
  "epoch": 0.29384328358208955,
448
- "grad_norm": 0.4791859984397888,
449
- "learning_rate": 0.0001736697247706422,
450
- "loss": 3.3427,
451
  "step": 3150
452
  },
453
  {
454
  "epoch": 0.29850746268656714,
455
- "grad_norm": 0.6330239176750183,
456
- "learning_rate": 0.00017252293577981652,
457
- "loss": 3.3484,
458
  "step": 3200
459
  },
460
  {
461
  "epoch": 0.3031716417910448,
462
- "grad_norm": 0.4511379897594452,
463
- "learning_rate": 0.00017137614678899084,
464
- "loss": 3.3302,
465
  "step": 3250
466
  },
467
  {
468
  "epoch": 0.30783582089552236,
469
- "grad_norm": 0.4571825861930847,
470
- "learning_rate": 0.00017022935779816516,
471
- "loss": 3.3144,
472
  "step": 3300
473
  },
474
  {
475
  "epoch": 0.3125,
476
- "grad_norm": 0.44557932019233704,
477
- "learning_rate": 0.00016908256880733945,
478
- "loss": 3.3071,
479
  "step": 3350
480
  },
481
  {
482
  "epoch": 0.31716417910447764,
483
- "grad_norm": 0.42434263229370117,
484
  "learning_rate": 0.00016793577981651377,
485
- "loss": 3.2986,
486
  "step": 3400
487
  },
488
  {
489
  "epoch": 0.3218283582089552,
490
- "grad_norm": 0.41813352704048157,
491
  "learning_rate": 0.0001667889908256881,
492
- "loss": 3.2902,
493
  "step": 3450
494
  },
495
  {
496
  "epoch": 0.32649253731343286,
497
- "grad_norm": 0.4084812104701996,
498
  "learning_rate": 0.00016564220183486238,
499
- "loss": 3.283,
500
  "step": 3500
501
  },
502
  {
503
  "epoch": 0.33115671641791045,
504
- "grad_norm": 0.38210785388946533,
505
  "learning_rate": 0.00016449541284403672,
506
- "loss": 3.2769,
507
  "step": 3550
508
  },
509
  {
510
  "epoch": 0.3358208955223881,
511
- "grad_norm": 0.5375980734825134,
512
  "learning_rate": 0.00016334862385321102,
513
- "loss": 3.2697,
514
  "step": 3600
515
  },
516
  {
517
  "epoch": 0.34048507462686567,
518
- "grad_norm": 0.38781264424324036,
519
  "learning_rate": 0.00016220183486238533,
520
- "loss": 3.2556,
521
  "step": 3650
522
  },
523
  {
524
  "epoch": 0.3451492537313433,
525
- "grad_norm": 0.41560590267181396,
526
  "learning_rate": 0.00016105504587155965,
527
- "loss": 3.2565,
528
  "step": 3700
529
  },
530
  {
531
  "epoch": 0.3498134328358209,
532
- "grad_norm": 0.4119664132595062,
533
  "learning_rate": 0.00015990825688073394,
534
- "loss": 3.2386,
535
  "step": 3750
536
  },
537
  {
538
  "epoch": 0.35447761194029853,
539
- "grad_norm": 0.4190433919429779,
540
  "learning_rate": 0.00015876146788990826,
541
- "loss": 3.2345,
542
  "step": 3800
543
  },
544
  {
545
  "epoch": 0.3591417910447761,
546
- "grad_norm": 0.37028583884239197,
547
  "learning_rate": 0.00015761467889908258,
548
- "loss": 3.226,
549
  "step": 3850
550
  },
551
  {
552
  "epoch": 0.36380597014925375,
553
- "grad_norm": 0.3971349000930786,
554
  "learning_rate": 0.0001564678899082569,
555
- "loss": 3.2213,
556
  "step": 3900
557
  },
558
  {
559
  "epoch": 0.36847014925373134,
560
- "grad_norm": 0.34641316533088684,
561
  "learning_rate": 0.0001553211009174312,
562
- "loss": 3.2144,
563
  "step": 3950
564
  },
565
  {
566
  "epoch": 0.373134328358209,
567
- "grad_norm": 0.40990573167800903,
568
  "learning_rate": 0.0001541743119266055,
569
- "loss": 3.2141,
570
  "step": 4000
571
  },
572
  {
573
  "epoch": 0.37779850746268656,
574
- "grad_norm": 0.35828983783721924,
575
  "learning_rate": 0.00015302752293577983,
576
- "loss": 3.2054,
577
  "step": 4050
578
  },
579
  {
580
  "epoch": 0.3824626865671642,
581
- "grad_norm": 0.37572401762008667,
582
  "learning_rate": 0.00015188073394495414,
583
- "loss": 3.1943,
584
  "step": 4100
585
  },
586
  {
587
  "epoch": 0.3871268656716418,
588
- "grad_norm": 0.3795240521430969,
589
  "learning_rate": 0.00015073394495412844,
590
- "loss": 3.1971,
591
  "step": 4150
592
  },
593
  {
594
  "epoch": 0.3917910447761194,
595
- "grad_norm": 0.3884343206882477,
596
  "learning_rate": 0.00014958715596330275,
597
- "loss": 3.1877,
598
  "step": 4200
599
  },
600
  {
601
  "epoch": 0.396455223880597,
602
- "grad_norm": 0.374661922454834,
603
  "learning_rate": 0.00014844036697247707,
604
- "loss": 3.174,
605
  "step": 4250
606
  },
607
  {
608
  "epoch": 0.40111940298507465,
609
- "grad_norm": 0.4330432415008545,
610
  "learning_rate": 0.00014729357798165136,
611
- "loss": 3.1787,
612
  "step": 4300
613
  },
614
  {
615
  "epoch": 0.40578358208955223,
616
- "grad_norm": 0.40475285053253174,
617
  "learning_rate": 0.0001461467889908257,
618
- "loss": 3.1781,
619
  "step": 4350
620
  },
621
  {
622
  "epoch": 0.41044776119402987,
623
- "grad_norm": 0.36028817296028137,
624
  "learning_rate": 0.000145,
625
- "loss": 3.1653,
626
  "step": 4400
627
  },
628
  {
629
  "epoch": 0.41511194029850745,
630
- "grad_norm": 0.37242797017097473,
631
  "learning_rate": 0.00014385321100917432,
632
- "loss": 3.1608,
633
  "step": 4450
634
  },
635
  {
636
  "epoch": 0.4197761194029851,
637
- "grad_norm": 0.42880979180336,
638
  "learning_rate": 0.00014270642201834864,
639
- "loss": 3.1656,
640
  "step": 4500
641
  },
642
  {
643
  "epoch": 0.4244402985074627,
644
- "grad_norm": 0.3562730848789215,
645
  "learning_rate": 0.00014155963302752296,
646
- "loss": 3.1409,
647
  "step": 4550
648
  },
649
  {
650
  "epoch": 0.4291044776119403,
651
- "grad_norm": 0.3420323133468628,
652
  "learning_rate": 0.00014041284403669725,
653
- "loss": 3.1502,
654
  "step": 4600
655
  },
656
  {
657
  "epoch": 0.4337686567164179,
658
- "grad_norm": 0.33985841274261475,
659
  "learning_rate": 0.00013926605504587157,
660
- "loss": 3.1437,
661
  "step": 4650
662
  },
663
  {
664
  "epoch": 0.43843283582089554,
665
- "grad_norm": 0.35252872109413147,
666
  "learning_rate": 0.00013811926605504588,
667
- "loss": 3.1375,
668
  "step": 4700
669
  },
670
  {
671
  "epoch": 0.4430970149253731,
672
- "grad_norm": 0.33299019932746887,
673
  "learning_rate": 0.00013697247706422018,
674
- "loss": 3.1207,
675
  "step": 4750
676
  },
677
  {
678
  "epoch": 0.44776119402985076,
679
- "grad_norm": 0.3340663015842438,
680
  "learning_rate": 0.00013582568807339452,
681
- "loss": 3.1285,
682
  "step": 4800
683
  },
684
  {
685
  "epoch": 0.45242537313432835,
686
- "grad_norm": 0.3459039330482483,
687
  "learning_rate": 0.00013470183486238532,
688
- "loss": 3.1187,
689
  "step": 4850
690
  },
691
  {
692
  "epoch": 0.457089552238806,
693
- "grad_norm": 0.34190812706947327,
694
  "learning_rate": 0.00013355504587155963,
695
- "loss": 3.1128,
696
  "step": 4900
697
  },
698
  {
699
  "epoch": 0.46175373134328357,
700
- "grad_norm": 0.38482001423835754,
701
  "learning_rate": 0.00013240825688073395,
702
- "loss": 3.1081,
703
  "step": 4950
704
  },
705
  {
706
  "epoch": 0.4664179104477612,
707
- "grad_norm": 0.3969718813896179,
708
  "learning_rate": 0.00013126146788990827,
709
- "loss": 3.111,
710
  "step": 5000
711
  },
712
  {
713
  "epoch": 0.4710820895522388,
714
- "grad_norm": 0.30978360772132874,
715
  "learning_rate": 0.0001301146788990826,
716
- "loss": 3.106,
717
  "step": 5050
718
  },
719
  {
720
  "epoch": 0.47574626865671643,
721
- "grad_norm": 0.32029977440834045,
722
  "learning_rate": 0.00012896788990825688,
723
- "loss": 3.1095,
724
  "step": 5100
725
  },
726
  {
727
  "epoch": 0.480410447761194,
728
- "grad_norm": 0.36698678135871887,
729
  "learning_rate": 0.0001278440366972477,
730
- "loss": 3.1017,
731
  "step": 5150
732
  },
733
  {
734
  "epoch": 0.48507462686567165,
735
- "grad_norm": 0.3710266947746277,
736
  "learning_rate": 0.00012669724770642202,
737
- "loss": 3.0863,
738
  "step": 5200
739
  },
740
  {
741
  "epoch": 0.48973880597014924,
742
- "grad_norm": 0.33469706773757935,
743
  "learning_rate": 0.00012555045871559634,
744
- "loss": 3.0949,
745
  "step": 5250
746
  },
747
  {
748
  "epoch": 0.4944029850746269,
749
- "grad_norm": 0.33493635058403015,
750
  "learning_rate": 0.00012440366972477063,
751
- "loss": 3.0794,
752
  "step": 5300
753
  },
754
  {
755
  "epoch": 0.49906716417910446,
756
- "grad_norm": 0.3489319682121277,
757
  "learning_rate": 0.00012325688073394498,
758
- "loss": 3.0752,
759
  "step": 5350
760
  },
761
  {
762
  "epoch": 0.503731343283582,
763
- "grad_norm": 0.32940617203712463,
764
  "learning_rate": 0.00012211009174311927,
765
- "loss": 3.0911,
766
  "step": 5400
767
  },
768
  {
769
  "epoch": 0.5083955223880597,
770
- "grad_norm": 0.3039519190788269,
771
  "learning_rate": 0.0001209633027522936,
772
- "loss": 3.0749,
773
  "step": 5450
774
  },
775
  {
776
  "epoch": 0.5130597014925373,
777
- "grad_norm": 0.3196834325790405,
778
  "learning_rate": 0.00011981651376146789,
779
- "loss": 3.0685,
780
  "step": 5500
781
  },
782
  {
783
  "epoch": 0.5177238805970149,
784
- "grad_norm": 0.35171428322792053,
785
  "learning_rate": 0.0001186697247706422,
786
- "loss": 3.0684,
787
  "step": 5550
788
  },
789
  {
790
  "epoch": 0.5223880597014925,
791
- "grad_norm": 0.3927328884601593,
792
  "learning_rate": 0.00011752293577981653,
793
- "loss": 3.0676,
794
  "step": 5600
795
  },
796
  {
797
  "epoch": 0.5270522388059702,
798
- "grad_norm": 0.32428842782974243,
799
  "learning_rate": 0.00011637614678899082,
800
- "loss": 3.0635,
801
  "step": 5650
802
  },
803
  {
804
  "epoch": 0.5317164179104478,
805
- "grad_norm": 0.3297617435455322,
806
  "learning_rate": 0.00011522935779816515,
807
- "loss": 3.0525,
808
  "step": 5700
809
  },
810
  {
811
  "epoch": 0.5363805970149254,
812
- "grad_norm": 0.3118542432785034,
813
  "learning_rate": 0.00011408256880733946,
814
- "loss": 3.0566,
815
  "step": 5750
816
  },
817
  {
818
  "epoch": 0.5410447761194029,
819
- "grad_norm": 0.29949086904525757,
820
  "learning_rate": 0.00011293577981651377,
821
- "loss": 3.0454,
822
  "step": 5800
823
  },
824
  {
825
  "epoch": 0.5457089552238806,
826
- "grad_norm": 0.3567812740802765,
827
  "learning_rate": 0.00011178899082568808,
828
- "loss": 3.0462,
829
  "step": 5850
830
  },
831
  {
832
  "epoch": 0.5503731343283582,
833
- "grad_norm": 0.3179719150066376,
834
  "learning_rate": 0.00011064220183486238,
835
- "loss": 3.0409,
836
  "step": 5900
837
  },
838
  {
839
  "epoch": 0.5550373134328358,
840
- "grad_norm": 0.32883220911026,
841
  "learning_rate": 0.0001094954128440367,
842
- "loss": 3.0527,
843
  "step": 5950
844
  },
845
  {
846
  "epoch": 0.5597014925373134,
847
- "grad_norm": 0.31516268849372864,
848
  "learning_rate": 0.000108348623853211,
849
- "loss": 3.0359,
850
  "step": 6000
851
  },
852
  {
853
  "epoch": 0.5643656716417911,
854
- "grad_norm": 0.298753559589386,
855
  "learning_rate": 0.00010720183486238533,
856
- "loss": 3.0343,
857
  "step": 6050
858
  },
859
  {
860
  "epoch": 0.5690298507462687,
861
- "grad_norm": 0.32886001467704773,
862
  "learning_rate": 0.00010605504587155963,
863
- "loss": 3.0229,
864
  "step": 6100
865
  },
866
  {
867
  "epoch": 0.5736940298507462,
868
- "grad_norm": 0.3526987135410309,
869
  "learning_rate": 0.00010490825688073396,
870
- "loss": 3.0227,
871
  "step": 6150
872
  },
873
  {
874
  "epoch": 0.5783582089552238,
875
- "grad_norm": 0.31915777921676636,
876
  "learning_rate": 0.00010376146788990825,
877
- "loss": 3.0311,
878
  "step": 6200
879
  },
880
  {
881
  "epoch": 0.5830223880597015,
882
- "grad_norm": 0.31342729926109314,
883
  "learning_rate": 0.00010261467889908258,
884
- "loss": 3.0257,
885
  "step": 6250
886
  },
887
  {
888
  "epoch": 0.5876865671641791,
889
- "grad_norm": 0.3711305856704712,
890
  "learning_rate": 0.00010146788990825689,
891
- "loss": 3.0125,
892
  "step": 6300
893
  },
894
  {
895
  "epoch": 0.5923507462686567,
896
- "grad_norm": 0.32861173152923584,
897
  "learning_rate": 0.00010032110091743118,
898
- "loss": 3.0235,
899
  "step": 6350
900
  },
901
  {
902
  "epoch": 0.5970149253731343,
903
- "grad_norm": 0.30850905179977417,
904
  "learning_rate": 9.917431192660551e-05,
905
- "loss": 3.0171,
906
  "step": 6400
907
  },
908
  {
909
  "epoch": 0.601679104477612,
910
- "grad_norm": 0.31126561760902405,
911
  "learning_rate": 9.802752293577983e-05,
912
- "loss": 3.0177,
913
  "step": 6450
914
  },
915
  {
916
  "epoch": 0.6063432835820896,
917
- "grad_norm": 0.3373335897922516,
918
  "learning_rate": 9.688073394495414e-05,
919
- "loss": 3.0138,
920
  "step": 6500
921
  },
922
  {
923
  "epoch": 0.6110074626865671,
924
- "grad_norm": 0.32974258065223694,
925
- "learning_rate": 9.575688073394496e-05,
926
- "loss": 3.0092,
927
  "step": 6550
928
  },
929
  {
930
  "epoch": 0.6156716417910447,
931
- "grad_norm": 0.29914727807044983,
932
- "learning_rate": 9.463302752293578e-05,
933
- "loss": 3.0075,
934
  "step": 6600
935
  },
936
  {
937
  "epoch": 0.6203358208955224,
938
- "grad_norm": 0.3006197214126587,
939
- "learning_rate": 9.34862385321101e-05,
940
- "loss": 3.015,
941
  "step": 6650
942
  },
943
  {
944
  "epoch": 0.625,
945
- "grad_norm": 0.2997278869152069,
946
- "learning_rate": 9.23394495412844e-05,
947
- "loss": 3.0037,
948
  "step": 6700
949
  },
950
  {
951
  "epoch": 0.6296641791044776,
952
- "grad_norm": 0.31719309091567993,
953
- "learning_rate": 9.119266055045872e-05,
954
- "loss": 2.9969,
955
  "step": 6750
956
  },
957
  {
958
  "epoch": 0.6343283582089553,
959
- "grad_norm": 0.3004010319709778,
960
- "learning_rate": 9.004587155963303e-05,
961
- "loss": 2.9909,
962
  "step": 6800
963
  },
964
  {
965
  "epoch": 0.6389925373134329,
966
- "grad_norm": 0.3123314082622528,
967
- "learning_rate": 8.889908256880735e-05,
968
- "loss": 2.9853,
969
  "step": 6850
970
  },
971
  {
972
  "epoch": 0.6436567164179104,
973
- "grad_norm": 0.334837406873703,
974
- "learning_rate": 8.775229357798166e-05,
975
- "loss": 2.9923,
976
  "step": 6900
977
  },
978
  {
979
  "epoch": 0.648320895522388,
980
- "grad_norm": 0.3055480420589447,
981
- "learning_rate": 8.660550458715597e-05,
982
- "loss": 2.9897,
983
  "step": 6950
984
  },
985
  {
986
  "epoch": 0.6529850746268657,
987
- "grad_norm": 0.33151915669441223,
988
- "learning_rate": 8.545871559633027e-05,
989
- "loss": 2.9843,
990
  "step": 7000
991
  },
992
  {
993
  "epoch": 0.6576492537313433,
994
- "grad_norm": 0.32511621713638306,
995
- "learning_rate": 8.431192660550459e-05,
996
- "loss": 2.9797,
997
  "step": 7050
998
  },
999
  {
1000
  "epoch": 0.6623134328358209,
1001
- "grad_norm": 0.29789265990257263,
1002
- "learning_rate": 8.31651376146789e-05,
1003
- "loss": 2.9789,
1004
  "step": 7100
1005
  },
1006
  {
1007
  "epoch": 0.6669776119402985,
1008
- "grad_norm": 0.3272823393344879,
1009
- "learning_rate": 8.201834862385321e-05,
1010
- "loss": 2.971,
1011
  "step": 7150
1012
  },
1013
  {
1014
  "epoch": 0.6716417910447762,
1015
- "grad_norm": 0.2983226478099823,
1016
- "learning_rate": 8.087155963302753e-05,
1017
- "loss": 2.9818,
1018
  "step": 7200
1019
  },
1020
  {
1021
  "epoch": 0.6763059701492538,
1022
- "grad_norm": 0.30532923340797424,
1023
- "learning_rate": 7.972477064220184e-05,
1024
- "loss": 2.9739,
1025
  "step": 7250
1026
  },
1027
  {
1028
  "epoch": 0.6809701492537313,
1029
- "grad_norm": 0.3046092092990875,
1030
- "learning_rate": 7.857798165137616e-05,
1031
- "loss": 2.9575,
1032
  "step": 7300
1033
  },
1034
  {
1035
  "epoch": 0.6856343283582089,
1036
- "grad_norm": 0.3133400082588196,
1037
- "learning_rate": 7.743119266055046e-05,
1038
- "loss": 2.9794,
1039
  "step": 7350
1040
  },
1041
  {
1042
  "epoch": 0.6902985074626866,
1043
- "grad_norm": 0.3090495765209198,
1044
- "learning_rate": 7.628440366972477e-05,
1045
- "loss": 2.9632,
1046
  "step": 7400
1047
  },
1048
  {
1049
  "epoch": 0.6949626865671642,
1050
- "grad_norm": 0.2925854027271271,
1051
- "learning_rate": 7.513761467889908e-05,
1052
- "loss": 2.9649,
1053
  "step": 7450
1054
  },
1055
  {
1056
  "epoch": 0.6996268656716418,
1057
- "grad_norm": 0.30464622378349304,
1058
- "learning_rate": 7.399082568807339e-05,
1059
- "loss": 2.9606,
1060
  "step": 7500
1061
  },
1062
  {
1063
  "epoch": 0.7042910447761194,
1064
- "grad_norm": 0.29478827118873596,
1065
- "learning_rate": 7.284403669724771e-05,
1066
- "loss": 2.9574,
1067
  "step": 7550
1068
  },
1069
  {
1070
  "epoch": 0.7089552238805971,
1071
- "grad_norm": 0.30368906259536743,
1072
- "learning_rate": 7.169724770642203e-05,
1073
- "loss": 2.9583,
1074
  "step": 7600
1075
  },
1076
  {
1077
  "epoch": 0.7136194029850746,
1078
- "grad_norm": 0.2974824905395508,
1079
- "learning_rate": 7.055045871559633e-05,
1080
- "loss": 2.9566,
1081
  "step": 7650
1082
  },
1083
  {
1084
  "epoch": 0.7182835820895522,
1085
- "grad_norm": 0.3006385564804077,
1086
- "learning_rate": 6.940366972477065e-05,
1087
- "loss": 2.953,
1088
  "step": 7700
1089
  },
1090
  {
1091
  "epoch": 0.7229477611940298,
1092
- "grad_norm": 0.2787078022956848,
1093
- "learning_rate": 6.825688073394497e-05,
1094
- "loss": 2.9466,
1095
  "step": 7750
1096
  },
1097
  {
1098
  "epoch": 0.7276119402985075,
1099
- "grad_norm": 0.2966342568397522,
1100
- "learning_rate": 6.711009174311927e-05,
1101
- "loss": 2.9446,
1102
  "step": 7800
1103
  },
1104
  {
1105
  "epoch": 0.7322761194029851,
1106
- "grad_norm": 0.32728853821754456,
1107
- "learning_rate": 6.600917431192661e-05,
1108
- "loss": 2.9486,
1109
  "step": 7850
1110
  },
1111
  {
1112
  "epoch": 0.7369402985074627,
1113
- "grad_norm": 0.325846403837204,
1114
- "learning_rate": 6.486238532110092e-05,
1115
- "loss": 2.9456,
1116
  "step": 7900
1117
  },
1118
  {
1119
  "epoch": 0.7416044776119403,
1120
- "grad_norm": 0.2933451235294342,
1121
- "learning_rate": 6.371559633027524e-05,
1122
- "loss": 2.9404,
1123
  "step": 7950
1124
  },
1125
  {
1126
  "epoch": 0.746268656716418,
1127
- "grad_norm": 0.3031325042247772,
1128
- "learning_rate": 6.256880733944954e-05,
1129
- "loss": 2.9486,
1130
  "step": 8000
1131
  },
1132
  {
1133
  "epoch": 0.7509328358208955,
1134
- "grad_norm": 0.30206266045570374,
1135
- "learning_rate": 6.142201834862386e-05,
1136
- "loss": 2.9364,
1137
  "step": 8050
1138
  },
1139
  {
1140
  "epoch": 0.7555970149253731,
1141
- "grad_norm": 0.29706859588623047,
1142
- "learning_rate": 6.027522935779817e-05,
1143
- "loss": 2.9483,
1144
  "step": 8100
1145
  },
1146
  {
1147
  "epoch": 0.7602611940298507,
1148
- "grad_norm": 0.2909660339355469,
1149
- "learning_rate": 5.912844036697248e-05,
1150
- "loss": 2.9405,
1151
  "step": 8150
1152
  },
1153
  {
1154
  "epoch": 0.7649253731343284,
1155
- "grad_norm": 0.3136695623397827,
1156
- "learning_rate": 5.798165137614679e-05,
1157
- "loss": 2.9366,
1158
  "step": 8200
1159
  },
1160
  {
1161
  "epoch": 0.769589552238806,
1162
- "grad_norm": 0.29548221826553345,
1163
- "learning_rate": 5.683486238532111e-05,
1164
- "loss": 2.9305,
1165
  "step": 8250
1166
  },
1167
  {
1168
  "epoch": 0.7742537313432836,
1169
- "grad_norm": 0.2777513265609741,
1170
- "learning_rate": 5.568807339449541e-05,
1171
- "loss": 2.9377,
1172
  "step": 8300
1173
  },
1174
  {
1175
  "epoch": 0.7789179104477612,
1176
- "grad_norm": 0.28673887252807617,
1177
- "learning_rate": 5.454128440366972e-05,
1178
- "loss": 2.9342,
1179
  "step": 8350
1180
  },
1181
  {
1182
  "epoch": 0.7835820895522388,
1183
- "grad_norm": 0.2911510169506073,
1184
- "learning_rate": 5.339449541284404e-05,
1185
- "loss": 2.938,
1186
  "step": 8400
1187
  },
1188
  {
1189
  "epoch": 0.7882462686567164,
1190
- "grad_norm": 0.3052498698234558,
1191
- "learning_rate": 5.224770642201835e-05,
1192
- "loss": 2.9235,
1193
  "step": 8450
1194
  },
1195
  {
1196
  "epoch": 0.792910447761194,
1197
- "grad_norm": 0.30528631806373596,
1198
- "learning_rate": 5.110091743119266e-05,
1199
- "loss": 2.9393,
1200
  "step": 8500
1201
  },
1202
  {
1203
  "epoch": 0.7975746268656716,
1204
- "grad_norm": 0.27922648191452026,
1205
- "learning_rate": 4.9954128440366974e-05,
1206
- "loss": 2.9153,
1207
  "step": 8550
1208
  },
1209
  {
1210
  "epoch": 0.8022388059701493,
1211
- "grad_norm": 0.2799355685710907,
1212
- "learning_rate": 4.880733944954129e-05,
1213
- "loss": 2.9329,
1214
  "step": 8600
1215
  },
1216
  {
1217
  "epoch": 0.8069029850746269,
1218
- "grad_norm": 0.3089926242828369,
1219
- "learning_rate": 4.76605504587156e-05,
1220
- "loss": 2.9203,
1221
  "step": 8650
1222
  },
1223
  {
1224
  "epoch": 0.8115671641791045,
1225
- "grad_norm": 0.3086063265800476,
1226
- "learning_rate": 4.651376146788991e-05,
1227
- "loss": 2.9223,
1228
  "step": 8700
1229
  },
1230
  {
1231
  "epoch": 0.816231343283582,
1232
- "grad_norm": 0.3019801676273346,
1233
- "learning_rate": 4.536697247706422e-05,
1234
- "loss": 2.9159,
1235
  "step": 8750
1236
  },
1237
  {
1238
  "epoch": 0.8208955223880597,
1239
- "grad_norm": 0.2975527048110962,
1240
- "learning_rate": 4.422018348623854e-05,
1241
- "loss": 2.9215,
1242
  "step": 8800
1243
  },
1244
  {
1245
  "epoch": 0.8255597014925373,
1246
- "grad_norm": 0.30920466780662537,
1247
- "learning_rate": 4.3073394495412844e-05,
1248
- "loss": 2.9213,
1249
  "step": 8850
1250
  },
1251
  {
1252
  "epoch": 0.8302238805970149,
1253
- "grad_norm": 0.2723771929740906,
1254
- "learning_rate": 4.1926605504587155e-05,
1255
- "loss": 2.9119,
1256
  "step": 8900
1257
  },
1258
  {
1259
  "epoch": 0.8348880597014925,
1260
- "grad_norm": 0.28407421708106995,
1261
- "learning_rate": 4.0779816513761474e-05,
1262
- "loss": 2.9101,
1263
  "step": 8950
1264
  },
1265
  {
1266
  "epoch": 0.8395522388059702,
1267
- "grad_norm": 0.2752555012702942,
1268
- "learning_rate": 3.9633027522935785e-05,
1269
- "loss": 2.9087,
1270
  "step": 9000
1271
  },
1272
  {
1273
  "epoch": 0.8442164179104478,
1274
- "grad_norm": 0.30377471446990967,
1275
- "learning_rate": 3.848623853211009e-05,
1276
- "loss": 2.9003,
1277
  "step": 9050
1278
  },
1279
  {
1280
  "epoch": 0.8488805970149254,
1281
- "grad_norm": 0.29200997948646545,
1282
- "learning_rate": 3.73394495412844e-05,
1283
- "loss": 2.9064,
1284
  "step": 9100
1285
  },
1286
  {
1287
  "epoch": 0.8535447761194029,
1288
- "grad_norm": 0.2830120623111725,
1289
- "learning_rate": 3.619266055045872e-05,
1290
- "loss": 2.9085,
1291
  "step": 9150
1292
  },
1293
  {
1294
  "epoch": 0.8582089552238806,
1295
- "grad_norm": 0.32428160309791565,
1296
  "learning_rate": 3.504587155963303e-05,
1297
- "loss": 2.9116,
1298
  "step": 9200
1299
  },
1300
  {
1301
  "epoch": 0.8628731343283582,
1302
- "grad_norm": 0.2852170169353485,
1303
  "learning_rate": 3.3899082568807337e-05,
1304
- "loss": 2.9119,
1305
  "step": 9250
1306
  },
1307
  {
1308
  "epoch": 0.8675373134328358,
1309
- "grad_norm": 0.2776082754135132,
1310
  "learning_rate": 3.2752293577981655e-05,
1311
- "loss": 2.9004,
1312
  "step": 9300
1313
  },
1314
  {
1315
  "epoch": 0.8722014925373134,
1316
- "grad_norm": 0.2881547212600708,
1317
  "learning_rate": 3.1605504587155967e-05,
1318
- "loss": 2.9024,
1319
  "step": 9350
1320
  },
1321
  {
1322
  "epoch": 0.8768656716417911,
1323
- "grad_norm": 0.2810397148132324,
1324
  "learning_rate": 3.0458715596330278e-05,
1325
- "loss": 2.9094,
1326
  "step": 9400
1327
  },
1328
  {
1329
  "epoch": 0.8815298507462687,
1330
- "grad_norm": 0.2975148558616638,
1331
  "learning_rate": 2.9311926605504586e-05,
1332
- "loss": 2.8971,
1333
  "step": 9450
1334
  },
1335
  {
1336
  "epoch": 0.8861940298507462,
1337
- "grad_norm": 0.29773736000061035,
1338
  "learning_rate": 2.8165137614678898e-05,
1339
- "loss": 2.8943,
1340
  "step": 9500
1341
  },
1342
  {
1343
  "epoch": 0.8908582089552238,
1344
- "grad_norm": 0.2762524485588074,
1345
  "learning_rate": 2.7018348623853213e-05,
1346
- "loss": 2.8899,
1347
  "step": 9550
1348
  },
1349
  {
1350
  "epoch": 0.8955223880597015,
1351
- "grad_norm": 0.2947123348712921,
1352
  "learning_rate": 2.5871559633027525e-05,
1353
- "loss": 2.8965,
1354
  "step": 9600
1355
  },
1356
  {
1357
  "epoch": 0.9001865671641791,
1358
- "grad_norm": 0.27293965220451355,
1359
- "learning_rate": 2.4724770642201836e-05,
1360
- "loss": 2.8869,
1361
  "step": 9650
1362
  },
1363
  {
1364
  "epoch": 0.9048507462686567,
1365
- "grad_norm": 0.2870621681213379,
1366
- "learning_rate": 2.3577981651376148e-05,
1367
- "loss": 2.8885,
1368
  "step": 9700
1369
  },
1370
  {
1371
  "epoch": 0.9095149253731343,
1372
- "grad_norm": 0.2940194606781006,
1373
- "learning_rate": 2.243119266055046e-05,
1374
- "loss": 2.8933,
1375
  "step": 9750
1376
  },
1377
  {
1378
  "epoch": 0.914179104477612,
1379
- "grad_norm": 0.270706444978714,
1380
- "learning_rate": 2.1284403669724774e-05,
1381
- "loss": 2.8889,
1382
  "step": 9800
1383
  },
1384
  {
1385
  "epoch": 0.9188432835820896,
1386
- "grad_norm": 0.27721747756004333,
1387
- "learning_rate": 2.0137614678899082e-05,
1388
- "loss": 2.8893,
1389
  "step": 9850
1390
  },
1391
  {
1392
  "epoch": 0.9235074626865671,
1393
- "grad_norm": 0.2803277373313904,
1394
- "learning_rate": 1.8990825688073397e-05,
1395
- "loss": 2.8868,
1396
  "step": 9900
1397
  },
1398
  {
1399
  "epoch": 0.9281716417910447,
1400
- "grad_norm": 0.2814282476902008,
1401
- "learning_rate": 1.7844036697247706e-05,
1402
- "loss": 2.8906,
1403
  "step": 9950
1404
  },
1405
  {
1406
  "epoch": 0.9328358208955224,
1407
- "grad_norm": 0.2730121612548828,
1408
- "learning_rate": 1.669724770642202e-05,
1409
- "loss": 2.8886,
1410
  "step": 10000
1411
  },
1412
  {
1413
  "epoch": 0.9375,
1414
- "grad_norm": 0.2759721577167511,
1415
- "learning_rate": 1.555045871559633e-05,
1416
- "loss": 2.8906,
1417
  "step": 10050
1418
  },
1419
  {
1420
  "epoch": 0.9421641791044776,
1421
- "grad_norm": 0.28318414092063904,
1422
- "learning_rate": 1.4403669724770644e-05,
1423
- "loss": 2.8834,
1424
  "step": 10100
1425
  },
1426
  {
1427
  "epoch": 0.9468283582089553,
1428
- "grad_norm": 0.27038970589637756,
1429
- "learning_rate": 1.3256880733944954e-05,
1430
- "loss": 2.8866,
1431
  "step": 10150
1432
  },
1433
  {
1434
  "epoch": 0.9514925373134329,
1435
- "grad_norm": 0.29715093970298767,
1436
- "learning_rate": 1.2110091743119267e-05,
1437
- "loss": 2.8749,
1438
  "step": 10200
1439
  },
1440
  {
1441
  "epoch": 0.9561567164179104,
1442
- "grad_norm": 0.27278682589530945,
1443
- "learning_rate": 1.0986238532110093e-05,
1444
- "loss": 2.8771,
1445
  "step": 10250
1446
  },
1447
  {
1448
  "epoch": 0.960820895522388,
1449
- "grad_norm": 0.26732149720191956,
1450
- "learning_rate": 9.839449541284404e-06,
1451
- "loss": 2.8857,
1452
  "step": 10300
1453
  },
1454
  {
1455
  "epoch": 0.9654850746268657,
1456
- "grad_norm": 0.2564278244972229,
1457
- "learning_rate": 8.692660550458716e-06,
1458
- "loss": 2.8801,
1459
  "step": 10350
1460
  },
1461
  {
1462
  "epoch": 0.9701492537313433,
1463
- "grad_norm": 0.2738514244556427,
1464
- "learning_rate": 7.545871559633027e-06,
1465
- "loss": 2.8833,
1466
  "step": 10400
1467
  },
1468
  {
1469
  "epoch": 0.9748134328358209,
1470
- "grad_norm": 0.26929736137390137,
1471
- "learning_rate": 6.399082568807339e-06,
1472
- "loss": 2.8706,
1473
  "step": 10450
1474
  },
1475
  {
1476
  "epoch": 0.9794776119402985,
1477
- "grad_norm": 0.2783481180667877,
1478
- "learning_rate": 5.252293577981652e-06,
1479
- "loss": 2.8758,
1480
  "step": 10500
1481
  },
1482
  {
1483
  "epoch": 0.9841417910447762,
1484
- "grad_norm": 0.2605401575565338,
1485
- "learning_rate": 4.105504587155964e-06,
1486
- "loss": 2.8718,
1487
  "step": 10550
1488
  },
1489
  {
1490
  "epoch": 0.9888059701492538,
1491
- "grad_norm": 0.2604130506515503,
1492
- "learning_rate": 2.9587155963302755e-06,
1493
- "loss": 2.87,
1494
  "step": 10600
1495
  },
1496
  {
1497
  "epoch": 0.9934701492537313,
1498
- "grad_norm": 0.259562224149704,
1499
- "learning_rate": 1.8119266055045873e-06,
1500
- "loss": 2.8781,
1501
  "step": 10650
1502
  },
1503
  {
1504
  "epoch": 0.9981343283582089,
1505
- "grad_norm": 0.2627047300338745,
1506
- "learning_rate": 6.651376146788992e-07,
1507
- "loss": 2.8782,
1508
  "step": 10700
1509
  }
1510
  ],
@@ -1525,7 +1525,7 @@
1525
  "attributes": {}
1526
  }
1527
  },
1528
- "total_flos": 2277239225843712.0,
1529
  "train_batch_size": 64,
1530
  "trial_name": null,
1531
  "trial_params": null
 
11
  "log_history": [
12
  {
13
  "epoch": 0.0046641791044776115,
14
+ "grad_norm": 2.434795618057251,
15
  "learning_rate": 4.9000000000000005e-06,
16
+ "loss": 10.2918,
17
  "step": 50
18
  },
19
  {
20
  "epoch": 0.009328358208955223,
21
+ "grad_norm": 1.3679819107055664,
22
  "learning_rate": 9.900000000000002e-06,
23
+ "loss": 9.0755,
24
  "step": 100
25
  },
26
  {
27
  "epoch": 0.013992537313432836,
28
+ "grad_norm": 0.9769303798675537,
29
  "learning_rate": 1.49e-05,
30
+ "loss": 8.3639,
31
  "step": 150
32
  },
33
  {
34
  "epoch": 0.018656716417910446,
35
+ "grad_norm": 0.5544369220733643,
36
  "learning_rate": 1.9900000000000003e-05,
37
+ "loss": 7.8229,
38
  "step": 200
39
  },
40
  {
41
  "epoch": 0.02332089552238806,
42
+ "grad_norm": 0.32632946968078613,
43
  "learning_rate": 2.4900000000000002e-05,
44
+ "loss": 7.4323,
45
  "step": 250
46
  },
47
  {
48
  "epoch": 0.027985074626865673,
49
+ "grad_norm": 0.34820762276649475,
50
  "learning_rate": 2.9900000000000002e-05,
51
+ "loss": 7.1956,
52
  "step": 300
53
  },
54
  {
55
  "epoch": 0.03264925373134328,
56
+ "grad_norm": 0.5745918154716492,
57
  "learning_rate": 3.49e-05,
58
+ "loss": 6.9913,
59
  "step": 350
60
  },
61
  {
62
  "epoch": 0.03731343283582089,
63
+ "grad_norm": 0.7313559651374817,
64
  "learning_rate": 3.99e-05,
65
+ "loss": 6.7761,
66
  "step": 400
67
  },
68
  {
69
  "epoch": 0.04197761194029851,
70
+ "grad_norm": 0.5769322514533997,
71
  "learning_rate": 4.49e-05,
72
+ "loss": 6.598,
73
  "step": 450
74
  },
75
  {
76
  "epoch": 0.04664179104477612,
77
+ "grad_norm": 0.5906850099563599,
78
  "learning_rate": 4.99e-05,
79
+ "loss": 6.4384,
80
  "step": 500
81
  },
82
  {
83
  "epoch": 0.051305970149253734,
84
+ "grad_norm": 1.5271503925323486,
85
  "learning_rate": 5.4900000000000006e-05,
86
+ "loss": 6.3052,
87
  "step": 550
88
  },
89
  {
90
  "epoch": 0.055970149253731345,
91
+ "grad_norm": 0.980595052242279,
92
  "learning_rate": 5.99e-05,
93
+ "loss": 6.1786,
94
  "step": 600
95
  },
96
  {
97
  "epoch": 0.06063432835820896,
98
+ "grad_norm": 0.8249331712722778,
99
  "learning_rate": 6.49e-05,
100
+ "loss": 6.0769,
101
  "step": 650
102
  },
103
  {
104
  "epoch": 0.06529850746268656,
105
+ "grad_norm": 1.6615272760391235,
106
  "learning_rate": 6.99e-05,
107
+ "loss": 5.9624,
108
  "step": 700
109
  },
110
  {
111
  "epoch": 0.06996268656716417,
112
+ "grad_norm": 1.4796066284179688,
113
  "learning_rate": 7.49e-05,
114
+ "loss": 5.8661,
115
  "step": 750
116
  },
117
  {
118
  "epoch": 0.07462686567164178,
119
+ "grad_norm": 1.050801157951355,
120
  "learning_rate": 7.99e-05,
121
+ "loss": 5.7773,
122
  "step": 800
123
  },
124
  {
125
  "epoch": 0.07929104477611941,
126
+ "grad_norm": 1.1273882389068604,
127
  "learning_rate": 8.49e-05,
128
+ "loss": 5.6794,
129
  "step": 850
130
  },
131
  {
132
  "epoch": 0.08395522388059702,
133
+ "grad_norm": 1.3447613716125488,
134
  "learning_rate": 8.99e-05,
135
+ "loss": 5.5883,
136
  "step": 900
137
  },
138
  {
139
  "epoch": 0.08861940298507463,
140
+ "grad_norm": 0.8979071378707886,
141
  "learning_rate": 9.49e-05,
142
+ "loss": 5.5152,
143
  "step": 950
144
  },
145
  {
146
  "epoch": 0.09328358208955224,
147
+ "grad_norm": 1.7317239046096802,
148
  "learning_rate": 9.99e-05,
149
+ "loss": 5.4185,
150
  "step": 1000
151
  },
152
  {
153
  "epoch": 0.09794776119402986,
154
+ "grad_norm": 1.6568114757537842,
155
  "learning_rate": 0.0001049,
156
+ "loss": 5.3531,
157
  "step": 1050
158
  },
159
  {
160
  "epoch": 0.10261194029850747,
161
+ "grad_norm": 1.260619044303894,
162
  "learning_rate": 0.0001099,
163
+ "loss": 5.2675,
164
  "step": 1100
165
  },
166
  {
167
  "epoch": 0.10727611940298508,
168
+ "grad_norm": 1.6347819566726685,
169
  "learning_rate": 0.0001149,
170
+ "loss": 5.1987,
171
  "step": 1150
172
  },
173
  {
174
  "epoch": 0.11194029850746269,
175
+ "grad_norm": 1.6571054458618164,
176
  "learning_rate": 0.00011990000000000001,
177
+ "loss": 5.1358,
178
  "step": 1200
179
  },
180
  {
181
  "epoch": 0.1166044776119403,
182
+ "grad_norm": 1.2196329832077026,
183
  "learning_rate": 0.00012490000000000003,
184
+ "loss": 5.0755,
185
  "step": 1250
186
  },
187
  {
188
  "epoch": 0.12126865671641791,
189
+ "grad_norm": 1.2077012062072754,
190
  "learning_rate": 0.00012989999999999999,
191
+ "loss": 5.0243,
192
  "step": 1300
193
  },
194
  {
195
  "epoch": 0.1259328358208955,
196
+ "grad_norm": 1.3209941387176514,
197
  "learning_rate": 0.0001349,
198
+ "loss": 4.9678,
199
  "step": 1350
200
  },
201
  {
202
  "epoch": 0.13059701492537312,
203
+ "grad_norm": 0.8279642462730408,
204
  "learning_rate": 0.0001399,
205
+ "loss": 4.9159,
206
  "step": 1400
207
  },
208
  {
209
  "epoch": 0.13526119402985073,
210
+ "grad_norm": 0.9633798599243164,
211
  "learning_rate": 0.00014490000000000003,
212
+ "loss": 4.8596,
213
  "step": 1450
214
  },
215
  {
216
  "epoch": 0.13992537313432835,
217
+ "grad_norm": 0.987281084060669,
218
  "learning_rate": 0.0001499,
219
+ "loss": 4.8162,
220
  "step": 1500
221
  },
222
  {
223
  "epoch": 0.14458955223880596,
224
+ "grad_norm": 1.094568133354187,
225
  "learning_rate": 0.0001549,
226
+ "loss": 4.7598,
227
  "step": 1550
228
  },
229
  {
230
  "epoch": 0.14925373134328357,
231
+ "grad_norm": 0.8916253447532654,
232
  "learning_rate": 0.0001599,
233
+ "loss": 4.7184,
234
  "step": 1600
235
  },
236
  {
237
  "epoch": 0.15391791044776118,
238
+ "grad_norm": 0.8389304876327515,
239
  "learning_rate": 0.0001649,
240
+ "loss": 4.6699,
241
  "step": 1650
242
  },
243
  {
244
  "epoch": 0.15858208955223882,
245
+ "grad_norm": 1.113956093788147,
246
  "learning_rate": 0.0001699,
247
+ "loss": 4.6177,
248
  "step": 1700
249
  },
250
  {
251
  "epoch": 0.16324626865671643,
252
+ "grad_norm": 0.9380860924720764,
253
  "learning_rate": 0.00017490000000000002,
254
+ "loss": 4.5619,
255
  "step": 1750
256
  },
257
  {
258
  "epoch": 0.16791044776119404,
259
+ "grad_norm": 1.2293964624404907,
260
  "learning_rate": 0.0001799,
261
+ "loss": 4.5084,
262
  "step": 1800
263
  },
264
  {
265
  "epoch": 0.17257462686567165,
266
+ "grad_norm": 1.1242707967758179,
267
  "learning_rate": 0.0001849,
268
+ "loss": 4.4553,
269
  "step": 1850
270
  },
271
  {
272
  "epoch": 0.17723880597014927,
273
+ "grad_norm": 0.7840976715087891,
274
  "learning_rate": 0.0001899,
275
+ "loss": 4.4074,
276
  "step": 1900
277
  },
278
  {
279
  "epoch": 0.18190298507462688,
280
+ "grad_norm": 0.7234581112861633,
281
  "learning_rate": 0.00019490000000000002,
282
+ "loss": 4.3652,
283
  "step": 1950
284
  },
285
  {
286
  "epoch": 0.1865671641791045,
287
+ "grad_norm": 0.8629528880119324,
288
  "learning_rate": 0.0001999,
289
+ "loss": 4.317,
290
  "step": 2000
291
  },
292
  {
293
  "epoch": 0.1912313432835821,
294
+ "grad_norm": 0.8042444586753845,
295
  "learning_rate": 0.00019887614678899083,
296
+ "loss": 4.2762,
297
  "step": 2050
298
  },
299
  {
300
  "epoch": 0.1958955223880597,
301
+ "grad_norm": 0.80702143907547,
302
  "learning_rate": 0.00019772935779816515,
303
+ "loss": 4.2375,
304
  "step": 2100
305
  },
306
  {
307
  "epoch": 0.20055970149253732,
308
+ "grad_norm": 0.6999443769454956,
309
  "learning_rate": 0.00019658256880733944,
310
+ "loss": 4.1876,
311
  "step": 2150
312
  },
313
  {
314
  "epoch": 0.20522388059701493,
315
+ "grad_norm": 0.7686917781829834,
316
  "learning_rate": 0.0001954357798165138,
317
+ "loss": 4.1574,
318
  "step": 2200
319
  },
320
  {
321
  "epoch": 0.20988805970149255,
322
+ "grad_norm": 0.7840579152107239,
323
  "learning_rate": 0.00019428899082568808,
324
+ "loss": 4.1217,
325
  "step": 2250
326
  },
327
  {
328
  "epoch": 0.21455223880597016,
329
+ "grad_norm": 0.8510863184928894,
330
  "learning_rate": 0.0001931422018348624,
331
+ "loss": 4.0979,
332
  "step": 2300
333
  },
334
  {
335
  "epoch": 0.21921641791044777,
336
+ "grad_norm": 0.679388165473938,
337
  "learning_rate": 0.00019199541284403671,
338
+ "loss": 4.0661,
339
  "step": 2350
340
  },
341
  {
342
  "epoch": 0.22388059701492538,
343
+ "grad_norm": 0.5541307330131531,
344
  "learning_rate": 0.00019084862385321103,
345
+ "loss": 4.0407,
346
  "step": 2400
347
  },
348
  {
349
  "epoch": 0.228544776119403,
350
+ "grad_norm": 0.6176748871803284,
351
+ "learning_rate": 0.00018970183486238532,
352
+ "loss": 4.0098,
353
  "step": 2450
354
  },
355
  {
356
  "epoch": 0.2332089552238806,
357
+ "grad_norm": 0.6339316368103027,
358
+ "learning_rate": 0.00018855504587155964,
359
+ "loss": 3.9825,
360
  "step": 2500
361
  },
362
  {
363
  "epoch": 0.23787313432835822,
364
+ "grad_norm": 0.6165521740913391,
365
+ "learning_rate": 0.00018740825688073396,
366
+ "loss": 3.9626,
367
  "step": 2550
368
  },
369
  {
370
  "epoch": 0.24253731343283583,
371
+ "grad_norm": 0.6364430785179138,
372
+ "learning_rate": 0.00018626146788990825,
373
+ "loss": 3.9288,
374
  "step": 2600
375
  },
376
  {
377
  "epoch": 0.24720149253731344,
378
+ "grad_norm": 0.5691106915473938,
379
+ "learning_rate": 0.00018511467889908257,
380
+ "loss": 3.9221,
381
  "step": 2650
382
  },
383
  {
384
  "epoch": 0.251865671641791,
385
+ "grad_norm": 0.5992695689201355,
386
+ "learning_rate": 0.0001839678899082569,
387
+ "loss": 3.8961,
388
  "step": 2700
389
  },
390
  {
391
  "epoch": 0.25652985074626866,
392
+ "grad_norm": 0.5932413935661316,
393
+ "learning_rate": 0.0001828211009174312,
394
+ "loss": 3.8808,
395
  "step": 2750
396
  },
397
  {
398
  "epoch": 0.26119402985074625,
399
+ "grad_norm": 0.6363717913627625,
400
+ "learning_rate": 0.0001816743119266055,
401
+ "loss": 3.8652,
402
  "step": 2800
403
  },
404
  {
405
  "epoch": 0.2658582089552239,
406
+ "grad_norm": 0.5456887483596802,
407
+ "learning_rate": 0.00018052752293577984,
408
+ "loss": 3.842,
409
  "step": 2850
410
  },
411
  {
412
  "epoch": 0.27052238805970147,
413
+ "grad_norm": 0.5930595397949219,
414
+ "learning_rate": 0.00017938073394495414,
415
+ "loss": 3.8244,
416
  "step": 2900
417
  },
418
  {
419
  "epoch": 0.2751865671641791,
420
+ "grad_norm": 0.5566945672035217,
421
+ "learning_rate": 0.00017823394495412843,
422
+ "loss": 3.804,
423
  "step": 2950
424
  },
425
  {
426
  "epoch": 0.2798507462686567,
427
+ "grad_norm": 0.6539527773857117,
428
+ "learning_rate": 0.00017708715596330277,
429
+ "loss": 3.7985,
430
  "step": 3000
431
  },
432
  {
433
  "epoch": 0.28451492537313433,
434
+ "grad_norm": 0.5921605229377747,
435
+ "learning_rate": 0.00017594036697247706,
436
+ "loss": 3.7834,
437
  "step": 3050
438
  },
439
  {
440
  "epoch": 0.2891791044776119,
441
+ "grad_norm": 0.5476294159889221,
442
+ "learning_rate": 0.00017479357798165138,
443
+ "loss": 3.7735,
444
  "step": 3100
445
  },
446
  {
447
  "epoch": 0.29384328358208955,
448
+ "grad_norm": 0.5736291408538818,
449
+ "learning_rate": 0.0001736467889908257,
450
+ "loss": 3.7535,
451
  "step": 3150
452
  },
453
  {
454
  "epoch": 0.29850746268656714,
455
+ "grad_norm": 0.5571457743644714,
456
+ "learning_rate": 0.00017250000000000002,
457
+ "loss": 3.7517,
458
  "step": 3200
459
  },
460
  {
461
  "epoch": 0.3031716417910448,
462
+ "grad_norm": 0.5658199787139893,
463
+ "learning_rate": 0.0001713532110091743,
464
+ "loss": 3.7308,
465
  "step": 3250
466
  },
467
  {
468
  "epoch": 0.30783582089552236,
469
+ "grad_norm": 0.5400105714797974,
470
+ "learning_rate": 0.00017020642201834863,
471
+ "loss": 3.713,
472
  "step": 3300
473
  },
474
  {
475
  "epoch": 0.3125,
476
+ "grad_norm": 0.5364307761192322,
477
+ "learning_rate": 0.00016905963302752295,
478
+ "loss": 3.7018,
479
  "step": 3350
480
  },
481
  {
482
  "epoch": 0.31716417910447764,
483
+ "grad_norm": 0.5792276859283447,
484
  "learning_rate": 0.00016793577981651377,
485
+ "loss": 3.6918,
486
  "step": 3400
487
  },
488
  {
489
  "epoch": 0.3218283582089552,
490
+ "grad_norm": 0.44588422775268555,
491
  "learning_rate": 0.0001667889908256881,
492
+ "loss": 3.6813,
493
  "step": 3450
494
  },
495
  {
496
  "epoch": 0.32649253731343286,
497
+ "grad_norm": 0.5175060033798218,
498
  "learning_rate": 0.00016564220183486238,
499
+ "loss": 3.6705,
500
  "step": 3500
501
  },
502
  {
503
  "epoch": 0.33115671641791045,
504
+ "grad_norm": 0.5310373306274414,
505
  "learning_rate": 0.00016449541284403672,
506
+ "loss": 3.6645,
507
  "step": 3550
508
  },
509
  {
510
  "epoch": 0.3358208955223881,
511
+ "grad_norm": 0.5658640265464783,
512
  "learning_rate": 0.00016334862385321102,
513
+ "loss": 3.6542,
514
  "step": 3600
515
  },
516
  {
517
  "epoch": 0.34048507462686567,
518
+ "grad_norm": 0.5425497889518738,
519
  "learning_rate": 0.00016220183486238533,
520
+ "loss": 3.6375,
521
  "step": 3650
522
  },
523
  {
524
  "epoch": 0.3451492537313433,
525
+ "grad_norm": 0.4865582585334778,
526
  "learning_rate": 0.00016105504587155965,
527
+ "loss": 3.6384,
528
  "step": 3700
529
  },
530
  {
531
  "epoch": 0.3498134328358209,
532
+ "grad_norm": 0.4987630248069763,
533
  "learning_rate": 0.00015990825688073394,
534
+ "loss": 3.6156,
535
  "step": 3750
536
  },
537
  {
538
  "epoch": 0.35447761194029853,
539
+ "grad_norm": 0.5758020877838135,
540
  "learning_rate": 0.00015876146788990826,
541
+ "loss": 3.6123,
542
  "step": 3800
543
  },
544
  {
545
  "epoch": 0.3591417910447761,
546
+ "grad_norm": 0.5264896154403687,
547
  "learning_rate": 0.00015761467889908258,
548
+ "loss": 3.6019,
549
  "step": 3850
550
  },
551
  {
552
  "epoch": 0.36380597014925375,
553
+ "grad_norm": 0.4835953414440155,
554
  "learning_rate": 0.0001564678899082569,
555
+ "loss": 3.5959,
556
  "step": 3900
557
  },
558
  {
559
  "epoch": 0.36847014925373134,
560
+ "grad_norm": 0.4341369569301605,
561
  "learning_rate": 0.0001553211009174312,
562
+ "loss": 3.5873,
563
  "step": 3950
564
  },
565
  {
566
  "epoch": 0.373134328358209,
567
+ "grad_norm": 0.47072407603263855,
568
  "learning_rate": 0.0001541743119266055,
569
+ "loss": 3.5852,
570
  "step": 4000
571
  },
572
  {
573
  "epoch": 0.37779850746268656,
574
+ "grad_norm": 0.46912139654159546,
575
  "learning_rate": 0.00015302752293577983,
576
+ "loss": 3.575,
577
  "step": 4050
578
  },
579
  {
580
  "epoch": 0.3824626865671642,
581
+ "grad_norm": 0.5946963429450989,
582
  "learning_rate": 0.00015188073394495414,
583
+ "loss": 3.5639,
584
  "step": 4100
585
  },
586
  {
587
  "epoch": 0.3871268656716418,
588
+ "grad_norm": 0.4828120470046997,
589
  "learning_rate": 0.00015073394495412844,
590
+ "loss": 3.5659,
591
  "step": 4150
592
  },
593
  {
594
  "epoch": 0.3917910447761194,
595
+ "grad_norm": 0.48630213737487793,
596
  "learning_rate": 0.00014958715596330275,
597
+ "loss": 3.5538,
598
  "step": 4200
599
  },
600
  {
601
  "epoch": 0.396455223880597,
602
+ "grad_norm": 0.4966210126876831,
603
  "learning_rate": 0.00014844036697247707,
604
+ "loss": 3.5394,
605
  "step": 4250
606
  },
607
  {
608
  "epoch": 0.40111940298507465,
609
+ "grad_norm": 0.8680199384689331,
610
  "learning_rate": 0.00014729357798165136,
611
+ "loss": 3.5434,
612
  "step": 4300
613
  },
614
  {
615
  "epoch": 0.40578358208955223,
616
+ "grad_norm": 0.4786328971385956,
617
  "learning_rate": 0.0001461467889908257,
618
+ "loss": 3.5429,
619
  "step": 4350
620
  },
621
  {
622
  "epoch": 0.41044776119402987,
623
+ "grad_norm": 0.4968096911907196,
624
  "learning_rate": 0.000145,
625
+ "loss": 3.5282,
626
  "step": 4400
627
  },
628
  {
629
  "epoch": 0.41511194029850745,
630
+ "grad_norm": 0.5361109375953674,
631
  "learning_rate": 0.00014385321100917432,
632
+ "loss": 3.5219,
633
  "step": 4450
634
  },
635
  {
636
  "epoch": 0.4197761194029851,
637
+ "grad_norm": 0.4819442629814148,
638
  "learning_rate": 0.00014270642201834864,
639
+ "loss": 3.5247,
640
  "step": 4500
641
  },
642
  {
643
  "epoch": 0.4244402985074627,
644
+ "grad_norm": 0.43222498893737793,
645
  "learning_rate": 0.00014155963302752296,
646
+ "loss": 3.5003,
647
  "step": 4550
648
  },
649
  {
650
  "epoch": 0.4291044776119403,
651
+ "grad_norm": 0.40276777744293213,
652
  "learning_rate": 0.00014041284403669725,
653
+ "loss": 3.5076,
654
  "step": 4600
655
  },
656
  {
657
  "epoch": 0.4337686567164179,
658
+ "grad_norm": 0.4238736033439636,
659
  "learning_rate": 0.00013926605504587157,
660
+ "loss": 3.5017,
661
  "step": 4650
662
  },
663
  {
664
  "epoch": 0.43843283582089554,
665
+ "grad_norm": 0.4680284559726715,
666
  "learning_rate": 0.00013811926605504588,
667
+ "loss": 3.4945,
668
  "step": 4700
669
  },
670
  {
671
  "epoch": 0.4430970149253731,
672
+ "grad_norm": 0.42782142758369446,
673
  "learning_rate": 0.00013697247706422018,
674
+ "loss": 3.4769,
675
  "step": 4750
676
  },
677
  {
678
  "epoch": 0.44776119402985076,
679
+ "grad_norm": 0.48704081773757935,
680
  "learning_rate": 0.00013582568807339452,
681
+ "loss": 3.4835,
682
  "step": 4800
683
  },
684
  {
685
  "epoch": 0.45242537313432835,
686
+ "grad_norm": 0.4254228472709656,
687
  "learning_rate": 0.00013470183486238532,
688
+ "loss": 3.4746,
689
  "step": 4850
690
  },
691
  {
692
  "epoch": 0.457089552238806,
693
+ "grad_norm": 0.46868908405303955,
694
  "learning_rate": 0.00013355504587155963,
695
+ "loss": 3.4669,
696
  "step": 4900
697
  },
698
  {
699
  "epoch": 0.46175373134328357,
700
+ "grad_norm": 0.39554563164711,
701
  "learning_rate": 0.00013240825688073395,
702
+ "loss": 3.4612,
703
  "step": 4950
704
  },
705
  {
706
  "epoch": 0.4664179104477612,
707
+ "grad_norm": 0.5025322437286377,
708
  "learning_rate": 0.00013126146788990827,
709
+ "loss": 3.4626,
710
  "step": 5000
711
  },
712
  {
713
  "epoch": 0.4710820895522388,
714
+ "grad_norm": 0.45474347472190857,
715
  "learning_rate": 0.0001301146788990826,
716
+ "loss": 3.4573,
717
  "step": 5050
718
  },
719
  {
720
  "epoch": 0.47574626865671643,
721
+ "grad_norm": 0.45531395077705383,
722
  "learning_rate": 0.00012896788990825688,
723
+ "loss": 3.4622,
724
  "step": 5100
725
  },
726
  {
727
  "epoch": 0.480410447761194,
728
+ "grad_norm": 0.5105506181716919,
729
  "learning_rate": 0.0001278440366972477,
730
+ "loss": 3.4522,
731
  "step": 5150
732
  },
733
  {
734
  "epoch": 0.48507462686567165,
735
+ "grad_norm": 0.41265401244163513,
736
  "learning_rate": 0.00012669724770642202,
737
+ "loss": 3.4378,
738
  "step": 5200
739
  },
740
  {
741
  "epoch": 0.48973880597014924,
742
+ "grad_norm": 0.4541892409324646,
743
  "learning_rate": 0.00012555045871559634,
744
+ "loss": 3.4452,
745
  "step": 5250
746
  },
747
  {
748
  "epoch": 0.4944029850746269,
749
+ "grad_norm": 0.4062674641609192,
750
  "learning_rate": 0.00012440366972477063,
751
+ "loss": 3.4299,
752
  "step": 5300
753
  },
754
  {
755
  "epoch": 0.49906716417910446,
756
+ "grad_norm": 0.39839744567871094,
757
  "learning_rate": 0.00012325688073394498,
758
+ "loss": 3.4231,
759
  "step": 5350
760
  },
761
  {
762
  "epoch": 0.503731343283582,
763
+ "grad_norm": 0.4296305775642395,
764
  "learning_rate": 0.00012211009174311927,
765
+ "loss": 3.4399,
766
  "step": 5400
767
  },
768
  {
769
  "epoch": 0.5083955223880597,
770
+ "grad_norm": 0.40859147906303406,
771
  "learning_rate": 0.0001209633027522936,
772
+ "loss": 3.4226,
773
  "step": 5450
774
  },
775
  {
776
  "epoch": 0.5130597014925373,
777
+ "grad_norm": 0.38735997676849365,
778
  "learning_rate": 0.00011981651376146789,
779
+ "loss": 3.4168,
780
  "step": 5500
781
  },
782
  {
783
  "epoch": 0.5177238805970149,
784
+ "grad_norm": 0.5080564022064209,
785
  "learning_rate": 0.0001186697247706422,
786
+ "loss": 3.4154,
787
  "step": 5550
788
  },
789
  {
790
  "epoch": 0.5223880597014925,
791
+ "grad_norm": 0.42273515462875366,
792
  "learning_rate": 0.00011752293577981653,
793
+ "loss": 3.4155,
794
  "step": 5600
795
  },
796
  {
797
  "epoch": 0.5270522388059702,
798
+ "grad_norm": 0.41893094778060913,
799
  "learning_rate": 0.00011637614678899082,
800
+ "loss": 3.4087,
801
  "step": 5650
802
  },
803
  {
804
  "epoch": 0.5317164179104478,
805
+ "grad_norm": 0.4470965564250946,
806
  "learning_rate": 0.00011522935779816515,
807
+ "loss": 3.3979,
808
  "step": 5700
809
  },
810
  {
811
  "epoch": 0.5363805970149254,
812
+ "grad_norm": 0.40868085622787476,
813
  "learning_rate": 0.00011408256880733946,
814
+ "loss": 3.4007,
815
  "step": 5750
816
  },
817
  {
818
  "epoch": 0.5410447761194029,
819
+ "grad_norm": 0.42737168073654175,
820
  "learning_rate": 0.00011293577981651377,
821
+ "loss": 3.3904,
822
  "step": 5800
823
  },
824
  {
825
  "epoch": 0.5457089552238806,
826
+ "grad_norm": 0.5149052143096924,
827
  "learning_rate": 0.00011178899082568808,
828
+ "loss": 3.3914,
829
  "step": 5850
830
  },
831
  {
832
  "epoch": 0.5503731343283582,
833
+ "grad_norm": 0.4031182825565338,
834
  "learning_rate": 0.00011064220183486238,
835
+ "loss": 3.3845,
836
  "step": 5900
837
  },
838
  {
839
  "epoch": 0.5550373134328358,
840
+ "grad_norm": 0.41353049874305725,
841
  "learning_rate": 0.0001094954128440367,
842
+ "loss": 3.3976,
843
  "step": 5950
844
  },
845
  {
846
  "epoch": 0.5597014925373134,
847
+ "grad_norm": 0.412335604429245,
848
  "learning_rate": 0.000108348623853211,
849
+ "loss": 3.3782,
850
  "step": 6000
851
  },
852
  {
853
  "epoch": 0.5643656716417911,
854
+ "grad_norm": 0.3940559923648834,
855
  "learning_rate": 0.00010720183486238533,
856
+ "loss": 3.3755,
857
  "step": 6050
858
  },
859
  {
860
  "epoch": 0.5690298507462687,
861
+ "grad_norm": 0.3801034092903137,
862
  "learning_rate": 0.00010605504587155963,
863
+ "loss": 3.3668,
864
  "step": 6100
865
  },
866
  {
867
  "epoch": 0.5736940298507462,
868
+ "grad_norm": 0.5560407638549805,
869
  "learning_rate": 0.00010490825688073396,
870
+ "loss": 3.3648,
871
  "step": 6150
872
  },
873
  {
874
  "epoch": 0.5783582089552238,
875
+ "grad_norm": 0.4365626275539398,
876
  "learning_rate": 0.00010376146788990825,
877
+ "loss": 3.374,
878
  "step": 6200
879
  },
880
  {
881
  "epoch": 0.5830223880597015,
882
+ "grad_norm": 0.39503213763237,
883
  "learning_rate": 0.00010261467889908258,
884
+ "loss": 3.3679,
885
  "step": 6250
886
  },
887
  {
888
  "epoch": 0.5876865671641791,
889
+ "grad_norm": 0.41522470116615295,
890
  "learning_rate": 0.00010146788990825689,
891
+ "loss": 3.3532,
892
  "step": 6300
893
  },
894
  {
895
  "epoch": 0.5923507462686567,
896
+ "grad_norm": 0.3890146315097809,
897
  "learning_rate": 0.00010032110091743118,
898
+ "loss": 3.3667,
899
  "step": 6350
900
  },
901
  {
902
  "epoch": 0.5970149253731343,
903
+ "grad_norm": 0.5168399810791016,
904
  "learning_rate": 9.917431192660551e-05,
905
+ "loss": 3.358,
906
  "step": 6400
907
  },
908
  {
909
  "epoch": 0.601679104477612,
910
+ "grad_norm": 0.3710983395576477,
911
  "learning_rate": 9.802752293577983e-05,
912
+ "loss": 3.3585,
913
  "step": 6450
914
  },
915
  {
916
  "epoch": 0.6063432835820896,
917
+ "grad_norm": 0.4420715570449829,
918
  "learning_rate": 9.688073394495414e-05,
919
+ "loss": 3.3539,
920
  "step": 6500
921
  },
922
  {
923
  "epoch": 0.6110074626865671,
924
+ "grad_norm": 0.3815698027610779,
925
+ "learning_rate": 9.573394495412844e-05,
926
+ "loss": 3.3484,
927
  "step": 6550
928
  },
929
  {
930
  "epoch": 0.6156716417910447,
931
+ "grad_norm": 0.38033589720726013,
932
+ "learning_rate": 9.458715596330276e-05,
933
+ "loss": 3.3471,
934
  "step": 6600
935
  },
936
  {
937
  "epoch": 0.6203358208955224,
938
+ "grad_norm": 0.35299167037010193,
939
+ "learning_rate": 9.344036697247706e-05,
940
+ "loss": 3.3555,
941
  "step": 6650
942
  },
943
  {
944
  "epoch": 0.625,
945
+ "grad_norm": 0.3831453025341034,
946
+ "learning_rate": 9.229357798165138e-05,
947
+ "loss": 3.343,
948
  "step": 6700
949
  },
950
  {
951
  "epoch": 0.6296641791044776,
952
+ "grad_norm": 0.38805127143859863,
953
+ "learning_rate": 9.114678899082569e-05,
954
+ "loss": 3.3362,
955
  "step": 6750
956
  },
957
  {
958
  "epoch": 0.6343283582089553,
959
+ "grad_norm": 0.360837459564209,
960
+ "learning_rate": 9e-05,
961
+ "loss": 3.329,
962
  "step": 6800
963
  },
964
  {
965
  "epoch": 0.6389925373134329,
966
+ "grad_norm": 0.48795634508132935,
967
+ "learning_rate": 8.885321100917432e-05,
968
+ "loss": 3.3244,
969
  "step": 6850
970
  },
971
  {
972
  "epoch": 0.6436567164179104,
973
+ "grad_norm": 0.3907620310783386,
974
+ "learning_rate": 8.770642201834863e-05,
975
+ "loss": 3.3326,
976
  "step": 6900
977
  },
978
  {
979
  "epoch": 0.648320895522388,
980
+ "grad_norm": 0.4346226751804352,
981
+ "learning_rate": 8.655963302752293e-05,
982
+ "loss": 3.3291,
983
  "step": 6950
984
  },
985
  {
986
  "epoch": 0.6529850746268657,
987
+ "grad_norm": 0.4240095019340515,
988
+ "learning_rate": 8.541284403669725e-05,
989
+ "loss": 3.3236,
990
  "step": 7000
991
  },
992
  {
993
  "epoch": 0.6576492537313433,
994
+ "grad_norm": 0.4035656452178955,
995
+ "learning_rate": 8.426605504587156e-05,
996
+ "loss": 3.3188,
997
  "step": 7050
998
  },
999
  {
1000
  "epoch": 0.6623134328358209,
1001
+ "grad_norm": 0.44728022813796997,
1002
+ "learning_rate": 8.311926605504588e-05,
1003
+ "loss": 3.3179,
1004
  "step": 7100
1005
  },
1006
  {
1007
  "epoch": 0.6669776119402985,
1008
+ "grad_norm": 0.3958663046360016,
1009
+ "learning_rate": 8.19724770642202e-05,
1010
+ "loss": 3.3081,
1011
  "step": 7150
1012
  },
1013
  {
1014
  "epoch": 0.6716417910447762,
1015
+ "grad_norm": 0.4779495894908905,
1016
+ "learning_rate": 8.08256880733945e-05,
1017
+ "loss": 3.3194,
1018
  "step": 7200
1019
  },
1020
  {
1021
  "epoch": 0.6763059701492538,
1022
+ "grad_norm": 0.3885568082332611,
1023
+ "learning_rate": 7.967889908256882e-05,
1024
+ "loss": 3.3123,
1025
  "step": 7250
1026
  },
1027
  {
1028
  "epoch": 0.6809701492537313,
1029
+ "grad_norm": 0.46929308772087097,
1030
+ "learning_rate": 7.853211009174312e-05,
1031
+ "loss": 3.2946,
1032
  "step": 7300
1033
  },
1034
  {
1035
  "epoch": 0.6856343283582089,
1036
+ "grad_norm": 0.3649180829524994,
1037
+ "learning_rate": 7.738532110091743e-05,
1038
+ "loss": 3.3172,
1039
  "step": 7350
1040
  },
1041
  {
1042
  "epoch": 0.6902985074626866,
1043
+ "grad_norm": 0.3710353672504425,
1044
+ "learning_rate": 7.623853211009174e-05,
1045
+ "loss": 3.3026,
1046
  "step": 7400
1047
  },
1048
  {
1049
  "epoch": 0.6949626865671642,
1050
+ "grad_norm": 0.3968711197376251,
1051
+ "learning_rate": 7.509174311926605e-05,
1052
+ "loss": 3.3026,
1053
  "step": 7450
1054
  },
1055
  {
1056
  "epoch": 0.6996268656716418,
1057
+ "grad_norm": 0.4033445715904236,
1058
+ "learning_rate": 7.394495412844037e-05,
1059
+ "loss": 3.2982,
1060
  "step": 7500
1061
  },
1062
  {
1063
  "epoch": 0.7042910447761194,
1064
+ "grad_norm": 0.4648160934448242,
1065
+ "learning_rate": 7.279816513761469e-05,
1066
+ "loss": 3.2959,
1067
  "step": 7550
1068
  },
1069
  {
1070
  "epoch": 0.7089552238805971,
1071
+ "grad_norm": 0.399700790643692,
1072
+ "learning_rate": 7.165137614678899e-05,
1073
+ "loss": 3.2955,
1074
  "step": 7600
1075
  },
1076
  {
1077
  "epoch": 0.7136194029850746,
1078
+ "grad_norm": 0.43972837924957275,
1079
+ "learning_rate": 7.050458715596331e-05,
1080
+ "loss": 3.2947,
1081
  "step": 7650
1082
  },
1083
  {
1084
  "epoch": 0.7182835820895522,
1085
+ "grad_norm": 0.3754163086414337,
1086
+ "learning_rate": 6.935779816513763e-05,
1087
+ "loss": 3.2901,
1088
  "step": 7700
1089
  },
1090
  {
1091
  "epoch": 0.7229477611940298,
1092
+ "grad_norm": 0.38772958517074585,
1093
+ "learning_rate": 6.821100917431193e-05,
1094
+ "loss": 3.2843,
1095
  "step": 7750
1096
  },
1097
  {
1098
  "epoch": 0.7276119402985075,
1099
+ "grad_norm": 0.3898366689682007,
1100
+ "learning_rate": 6.706422018348624e-05,
1101
+ "loss": 3.2818,
1102
  "step": 7800
1103
  },
1104
  {
1105
  "epoch": 0.7322761194029851,
1106
+ "grad_norm": 0.3694920539855957,
1107
+ "learning_rate": 6.591743119266056e-05,
1108
+ "loss": 3.285,
1109
  "step": 7850
1110
  },
1111
  {
1112
  "epoch": 0.7369402985074627,
1113
+ "grad_norm": 0.3696017861366272,
1114
+ "learning_rate": 6.477064220183486e-05,
1115
+ "loss": 3.282,
1116
  "step": 7900
1117
  },
1118
  {
1119
  "epoch": 0.7416044776119403,
1120
+ "grad_norm": 0.4159897565841675,
1121
+ "learning_rate": 6.362385321100918e-05,
1122
+ "loss": 3.2778,
1123
  "step": 7950
1124
  },
1125
  {
1126
  "epoch": 0.746268656716418,
1127
+ "grad_norm": 0.3910117447376251,
1128
+ "learning_rate": 6.247706422018348e-05,
1129
+ "loss": 3.2852,
1130
  "step": 8000
1131
  },
1132
  {
1133
  "epoch": 0.7509328358208955,
1134
+ "grad_norm": 0.44550907611846924,
1135
+ "learning_rate": 6.13302752293578e-05,
1136
+ "loss": 3.2731,
1137
  "step": 8050
1138
  },
1139
  {
1140
  "epoch": 0.7555970149253731,
1141
+ "grad_norm": 0.4173608422279358,
1142
+ "learning_rate": 6.0183486238532114e-05,
1143
+ "loss": 3.2865,
1144
  "step": 8100
1145
  },
1146
  {
1147
  "epoch": 0.7602611940298507,
1148
+ "grad_norm": 0.3633171021938324,
1149
+ "learning_rate": 5.903669724770643e-05,
1150
+ "loss": 3.2783,
1151
  "step": 8150
1152
  },
1153
  {
1154
  "epoch": 0.7649253731343284,
1155
+ "grad_norm": 0.42995285987854004,
1156
+ "learning_rate": 5.793577981651377e-05,
1157
+ "loss": 3.2732,
1158
  "step": 8200
1159
  },
1160
  {
1161
  "epoch": 0.769589552238806,
1162
+ "grad_norm": 0.3819643557071686,
1163
+ "learning_rate": 5.678899082568807e-05,
1164
+ "loss": 3.2677,
1165
  "step": 8250
1166
  },
1167
  {
1168
  "epoch": 0.7742537313432836,
1169
+ "grad_norm": 0.3480379581451416,
1170
+ "learning_rate": 5.564220183486238e-05,
1171
+ "loss": 3.2749,
1172
  "step": 8300
1173
  },
1174
  {
1175
  "epoch": 0.7789179104477612,
1176
+ "grad_norm": 0.3577636480331421,
1177
+ "learning_rate": 5.44954128440367e-05,
1178
+ "loss": 3.2725,
1179
  "step": 8350
1180
  },
1181
  {
1182
  "epoch": 0.7835820895522388,
1183
+ "grad_norm": 0.36848926544189453,
1184
+ "learning_rate": 5.334862385321101e-05,
1185
+ "loss": 3.2744,
1186
  "step": 8400
1187
  },
1188
  {
1189
  "epoch": 0.7882462686567164,
1190
+ "grad_norm": 0.35368892550468445,
1191
+ "learning_rate": 5.220183486238532e-05,
1192
+ "loss": 3.2616,
1193
  "step": 8450
1194
  },
1195
  {
1196
  "epoch": 0.792910447761194,
1197
+ "grad_norm": 0.41831615567207336,
1198
+ "learning_rate": 5.1055045871559635e-05,
1199
+ "loss": 3.277,
1200
  "step": 8500
1201
  },
1202
  {
1203
  "epoch": 0.7975746268656716,
1204
+ "grad_norm": 0.41986221075057983,
1205
+ "learning_rate": 4.990825688073395e-05,
1206
+ "loss": 3.2522,
1207
  "step": 8550
1208
  },
1209
  {
1210
  "epoch": 0.8022388059701493,
1211
+ "grad_norm": 0.3512188494205475,
1212
+ "learning_rate": 4.876146788990826e-05,
1213
+ "loss": 3.2709,
1214
  "step": 8600
1215
  },
1216
  {
1217
  "epoch": 0.8069029850746269,
1218
+ "grad_norm": 0.3639850318431854,
1219
+ "learning_rate": 4.761467889908257e-05,
1220
+ "loss": 3.2571,
1221
  "step": 8650
1222
  },
1223
  {
1224
  "epoch": 0.8115671641791045,
1225
+ "grad_norm": 0.4546779692173004,
1226
+ "learning_rate": 4.646788990825688e-05,
1227
+ "loss": 3.2599,
1228
  "step": 8700
1229
  },
1230
  {
1231
  "epoch": 0.816231343283582,
1232
+ "grad_norm": 0.35850730538368225,
1233
+ "learning_rate": 4.53211009174312e-05,
1234
+ "loss": 3.2531,
1235
  "step": 8750
1236
  },
1237
  {
1238
  "epoch": 0.8208955223880597,
1239
+ "grad_norm": 0.3843632936477661,
1240
+ "learning_rate": 4.4174311926605504e-05,
1241
+ "loss": 3.2598,
1242
  "step": 8800
1243
  },
1244
  {
1245
  "epoch": 0.8255597014925373,
1246
+ "grad_norm": 0.3926890790462494,
1247
+ "learning_rate": 4.3027522935779816e-05,
1248
+ "loss": 3.2579,
1249
  "step": 8850
1250
  },
1251
  {
1252
  "epoch": 0.8302238805970149,
1253
+ "grad_norm": 0.36206337809562683,
1254
+ "learning_rate": 4.1880733944954134e-05,
1255
+ "loss": 3.2503,
1256
  "step": 8900
1257
  },
1258
  {
1259
  "epoch": 0.8348880597014925,
1260
+ "grad_norm": 0.3471512198448181,
1261
+ "learning_rate": 4.0733944954128446e-05,
1262
+ "loss": 3.2478,
1263
  "step": 8950
1264
  },
1265
  {
1266
  "epoch": 0.8395522388059702,
1267
+ "grad_norm": 0.3711581230163574,
1268
+ "learning_rate": 3.958715596330275e-05,
1269
+ "loss": 3.2461,
1270
  "step": 9000
1271
  },
1272
  {
1273
  "epoch": 0.8442164179104478,
1274
+ "grad_norm": 0.3600054979324341,
1275
+ "learning_rate": 3.844036697247706e-05,
1276
+ "loss": 3.2375,
1277
  "step": 9050
1278
  },
1279
  {
1280
  "epoch": 0.8488805970149254,
1281
+ "grad_norm": 0.3525969088077545,
1282
+ "learning_rate": 3.729357798165138e-05,
1283
+ "loss": 3.2446,
1284
  "step": 9100
1285
  },
1286
  {
1287
  "epoch": 0.8535447761194029,
1288
+ "grad_norm": 0.3669526278972626,
1289
+ "learning_rate": 3.614678899082569e-05,
1290
+ "loss": 3.2459,
1291
  "step": 9150
1292
  },
1293
  {
1294
  "epoch": 0.8582089552238806,
1295
+ "grad_norm": 0.3348044753074646,
1296
  "learning_rate": 3.504587155963303e-05,
1297
+ "loss": 3.2498,
1298
  "step": 9200
1299
  },
1300
  {
1301
  "epoch": 0.8628731343283582,
1302
+ "grad_norm": 0.3626518249511719,
1303
  "learning_rate": 3.3899082568807337e-05,
1304
+ "loss": 3.2501,
1305
  "step": 9250
1306
  },
1307
  {
1308
  "epoch": 0.8675373134328358,
1309
+ "grad_norm": 0.377811998128891,
1310
  "learning_rate": 3.2752293577981655e-05,
1311
+ "loss": 3.2385,
1312
  "step": 9300
1313
  },
1314
  {
1315
  "epoch": 0.8722014925373134,
1316
+ "grad_norm": 0.3583707809448242,
1317
  "learning_rate": 3.1605504587155967e-05,
1318
+ "loss": 3.2411,
1319
  "step": 9350
1320
  },
1321
  {
1322
  "epoch": 0.8768656716417911,
1323
+ "grad_norm": 0.35831955075263977,
1324
  "learning_rate": 3.0458715596330278e-05,
1325
+ "loss": 3.2486,
1326
  "step": 9400
1327
  },
1328
  {
1329
  "epoch": 0.8815298507462687,
1330
+ "grad_norm": 0.3578301668167114,
1331
  "learning_rate": 2.9311926605504586e-05,
1332
+ "loss": 3.2352,
1333
  "step": 9450
1334
  },
1335
  {
1336
  "epoch": 0.8861940298507462,
1337
+ "grad_norm": 0.3495340645313263,
1338
  "learning_rate": 2.8165137614678898e-05,
1339
+ "loss": 3.2336,
1340
  "step": 9500
1341
  },
1342
  {
1343
  "epoch": 0.8908582089552238,
1344
+ "grad_norm": 0.34632202982902527,
1345
  "learning_rate": 2.7018348623853213e-05,
1346
+ "loss": 3.2288,
1347
  "step": 9550
1348
  },
1349
  {
1350
  "epoch": 0.8955223880597015,
1351
+ "grad_norm": 0.3287692070007324,
1352
  "learning_rate": 2.5871559633027525e-05,
1353
+ "loss": 3.2349,
1354
  "step": 9600
1355
  },
1356
  {
1357
  "epoch": 0.9001865671641791,
1358
+ "grad_norm": 0.329765260219574,
1359
+ "learning_rate": 2.474770642201835e-05,
1360
+ "loss": 3.2271,
1361
  "step": 9650
1362
  },
1363
  {
1364
  "epoch": 0.9048507462686567,
1365
+ "grad_norm": 0.3470270037651062,
1366
+ "learning_rate": 2.3600917431192662e-05,
1367
+ "loss": 3.2279,
1368
  "step": 9700
1369
  },
1370
  {
1371
  "epoch": 0.9095149253731343,
1372
+ "grad_norm": 0.34217342734336853,
1373
+ "learning_rate": 2.2454128440366973e-05,
1374
+ "loss": 3.2329,
1375
  "step": 9750
1376
  },
1377
  {
1378
  "epoch": 0.914179104477612,
1379
+ "grad_norm": 0.33399471640586853,
1380
+ "learning_rate": 2.1307339449541285e-05,
1381
+ "loss": 3.2284,
1382
  "step": 9800
1383
  },
1384
  {
1385
  "epoch": 0.9188432835820896,
1386
+ "grad_norm": 0.3363756835460663,
1387
+ "learning_rate": 2.0160550458715597e-05,
1388
+ "loss": 3.2302,
1389
  "step": 9850
1390
  },
1391
  {
1392
  "epoch": 0.9235074626865671,
1393
+ "grad_norm": 0.3226417303085327,
1394
+ "learning_rate": 1.9013761467889908e-05,
1395
+ "loss": 3.226,
1396
  "step": 9900
1397
  },
1398
  {
1399
  "epoch": 0.9281716417910447,
1400
+ "grad_norm": 0.33726227283477783,
1401
+ "learning_rate": 1.786697247706422e-05,
1402
+ "loss": 3.23,
1403
  "step": 9950
1404
  },
1405
  {
1406
  "epoch": 0.9328358208955224,
1407
+ "grad_norm": 0.321735143661499,
1408
+ "learning_rate": 1.6720183486238535e-05,
1409
+ "loss": 3.2278,
1410
  "step": 10000
1411
  },
1412
  {
1413
  "epoch": 0.9375,
1414
+ "grad_norm": 0.3822382986545563,
1415
+ "learning_rate": 1.5573394495412843e-05,
1416
+ "loss": 3.2295,
1417
  "step": 10050
1418
  },
1419
  {
1420
  "epoch": 0.9421641791044776,
1421
+ "grad_norm": 0.3251081109046936,
1422
+ "learning_rate": 1.4426605504587156e-05,
1423
+ "loss": 3.224,
1424
  "step": 10100
1425
  },
1426
  {
1427
  "epoch": 0.9468283582089553,
1428
+ "grad_norm": 0.31546786427497864,
1429
+ "learning_rate": 1.3279816513761468e-05,
1430
+ "loss": 3.2254,
1431
  "step": 10150
1432
  },
1433
  {
1434
  "epoch": 0.9514925373134329,
1435
+ "grad_norm": 0.3238883912563324,
1436
+ "learning_rate": 1.2133027522935781e-05,
1437
+ "loss": 3.2152,
1438
  "step": 10200
1439
  },
1440
  {
1441
  "epoch": 0.9561567164179104,
1442
+ "grad_norm": 0.32408469915390015,
1443
+ "learning_rate": 1.1009174311926607e-05,
1444
+ "loss": 3.2176,
1445
  "step": 10250
1446
  },
1447
  {
1448
  "epoch": 0.960820895522388,
1449
+ "grad_norm": 0.3206406533718109,
1450
+ "learning_rate": 9.862385321100918e-06,
1451
+ "loss": 3.2268,
1452
  "step": 10300
1453
  },
1454
  {
1455
  "epoch": 0.9654850746268657,
1456
+ "grad_norm": 0.317878782749176,
1457
+ "learning_rate": 8.71559633027523e-06,
1458
+ "loss": 3.2213,
1459
  "step": 10350
1460
  },
1461
  {
1462
  "epoch": 0.9701492537313433,
1463
+ "grad_norm": 0.3347484767436981,
1464
+ "learning_rate": 7.568807339449542e-06,
1465
+ "loss": 3.2236,
1466
  "step": 10400
1467
  },
1468
  {
1469
  "epoch": 0.9748134328358209,
1470
+ "grad_norm": 0.3140642046928406,
1471
+ "learning_rate": 6.422018348623854e-06,
1472
+ "loss": 3.2108,
1473
  "step": 10450
1474
  },
1475
  {
1476
  "epoch": 0.9794776119402985,
1477
+ "grad_norm": 0.31455010175704956,
1478
+ "learning_rate": 5.2752293577981655e-06,
1479
+ "loss": 3.217,
1480
  "step": 10500
1481
  },
1482
  {
1483
  "epoch": 0.9841417910447762,
1484
+ "grad_norm": 0.30823615193367004,
1485
+ "learning_rate": 4.128440366972477e-06,
1486
+ "loss": 3.213,
1487
  "step": 10550
1488
  },
1489
  {
1490
  "epoch": 0.9888059701492538,
1491
+ "grad_norm": 0.28951770067214966,
1492
+ "learning_rate": 2.981651376146789e-06,
1493
+ "loss": 3.2116,
1494
  "step": 10600
1495
  },
1496
  {
1497
  "epoch": 0.9934701492537313,
1498
+ "grad_norm": 0.29542475938796997,
1499
+ "learning_rate": 1.8348623853211011e-06,
1500
+ "loss": 3.2197,
1501
  "step": 10650
1502
  },
1503
  {
1504
  "epoch": 0.9981343283582089,
1505
+ "grad_norm": 0.29293331503868103,
1506
+ "learning_rate": 6.880733944954129e-07,
1507
+ "loss": 3.219,
1508
  "step": 10700
1509
  }
1510
  ],
 
1525
  "attributes": {}
1526
  }
1527
  },
1528
+ "total_flos": 2277028195729408.0,
1529
  "train_batch_size": 64,
1530
  "trial_name": null,
1531
  "trial_params": null
frozen/checkpoint-9000/model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fb3b84d468444da94f4aca832d8968c5168101a33c276e95372b4c6b1c2203ad
3
- size 1067471056
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ba68a4c8af979d53d897204bb16c9ab5bd41c443b433835e3dd60d203142122
3
+ size 800608432
frozen/checkpoint-9000/scheduler.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cae9caf33f74c42ea1c1debccd7a6aeeb6572a94f1f322fe88d1c2a83ac3e3e7
3
  size 1465
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ca87206bfccfb745d29684b6420eb2f5e59bda9ff5f222514b2806a71a78480
3
  size 1465
frozen/checkpoint-9000/trainer_state.json CHANGED
@@ -11,1262 +11,1262 @@
11
  "log_history": [
12
  {
13
  "epoch": 0.0046641791044776115,
14
- "grad_norm": 1.907758116722107,
15
  "learning_rate": 4.9000000000000005e-06,
16
- "loss": 10.0793,
17
  "step": 50
18
  },
19
  {
20
  "epoch": 0.009328358208955223,
21
- "grad_norm": 1.399185061454773,
22
  "learning_rate": 9.900000000000002e-06,
23
- "loss": 8.9287,
24
  "step": 100
25
  },
26
  {
27
  "epoch": 0.013992537313432836,
28
- "grad_norm": 1.0101089477539062,
29
  "learning_rate": 1.49e-05,
30
- "loss": 8.3288,
31
  "step": 150
32
  },
33
  {
34
  "epoch": 0.018656716417910446,
35
- "grad_norm": 0.7379341125488281,
36
  "learning_rate": 1.9900000000000003e-05,
37
- "loss": 7.6969,
38
  "step": 200
39
  },
40
  {
41
  "epoch": 0.02332089552238806,
42
- "grad_norm": 0.9656306505203247,
43
  "learning_rate": 2.4900000000000002e-05,
44
- "loss": 7.2698,
45
  "step": 250
46
  },
47
  {
48
  "epoch": 0.027985074626865673,
49
- "grad_norm": 1.3790279626846313,
50
  "learning_rate": 2.9900000000000002e-05,
51
- "loss": 6.9685,
52
  "step": 300
53
  },
54
  {
55
  "epoch": 0.03264925373134328,
56
- "grad_norm": 0.6374709606170654,
57
  "learning_rate": 3.49e-05,
58
- "loss": 6.6752,
59
  "step": 350
60
  },
61
  {
62
  "epoch": 0.03731343283582089,
63
- "grad_norm": 1.5095465183258057,
64
  "learning_rate": 3.99e-05,
65
- "loss": 6.3805,
66
  "step": 400
67
  },
68
  {
69
  "epoch": 0.04197761194029851,
70
- "grad_norm": 1.394410490989685,
71
  "learning_rate": 4.49e-05,
72
- "loss": 6.1667,
73
  "step": 450
74
  },
75
  {
76
  "epoch": 0.04664179104477612,
77
- "grad_norm": 1.7006834745407104,
78
  "learning_rate": 4.99e-05,
79
- "loss": 5.9861,
80
  "step": 500
81
  },
82
  {
83
  "epoch": 0.051305970149253734,
84
- "grad_norm": 0.9934776425361633,
85
  "learning_rate": 5.4900000000000006e-05,
86
- "loss": 5.8166,
87
  "step": 550
88
  },
89
  {
90
  "epoch": 0.055970149253731345,
91
- "grad_norm": 2.0550477504730225,
92
  "learning_rate": 5.99e-05,
93
- "loss": 5.6646,
94
  "step": 600
95
  },
96
  {
97
  "epoch": 0.06063432835820896,
98
- "grad_norm": 2.2504913806915283,
99
  "learning_rate": 6.49e-05,
100
- "loss": 5.5353,
101
  "step": 650
102
  },
103
  {
104
  "epoch": 0.06529850746268656,
105
- "grad_norm": 2.468212604522705,
106
  "learning_rate": 6.99e-05,
107
- "loss": 5.3982,
108
  "step": 700
109
  },
110
  {
111
  "epoch": 0.06996268656716417,
112
- "grad_norm": 2.4010939598083496,
113
  "learning_rate": 7.49e-05,
114
- "loss": 5.2855,
115
  "step": 750
116
  },
117
  {
118
  "epoch": 0.07462686567164178,
119
- "grad_norm": 1.4586199522018433,
120
  "learning_rate": 7.99e-05,
121
- "loss": 5.1901,
122
  "step": 800
123
  },
124
  {
125
  "epoch": 0.07929104477611941,
126
- "grad_norm": 1.2821402549743652,
127
  "learning_rate": 8.49e-05,
128
- "loss": 5.0973,
129
  "step": 850
130
  },
131
  {
132
  "epoch": 0.08395522388059702,
133
- "grad_norm": 2.1037180423736572,
134
  "learning_rate": 8.99e-05,
135
- "loss": 5.0062,
136
  "step": 900
137
  },
138
  {
139
  "epoch": 0.08861940298507463,
140
- "grad_norm": 1.370751976966858,
141
  "learning_rate": 9.49e-05,
142
- "loss": 4.9392,
143
  "step": 950
144
  },
145
  {
146
  "epoch": 0.09328358208955224,
147
- "grad_norm": 1.276108741760254,
148
  "learning_rate": 9.99e-05,
149
- "loss": 4.848,
150
  "step": 1000
151
  },
152
  {
153
  "epoch": 0.09794776119402986,
154
- "grad_norm": 1.1077498197555542,
155
  "learning_rate": 0.0001049,
156
- "loss": 4.7863,
157
  "step": 1050
158
  },
159
  {
160
  "epoch": 0.10261194029850747,
161
- "grad_norm": 1.0212163925170898,
162
  "learning_rate": 0.0001099,
163
- "loss": 4.7102,
164
  "step": 1100
165
  },
166
  {
167
  "epoch": 0.10727611940298508,
168
- "grad_norm": 1.1480289697647095,
169
  "learning_rate": 0.0001149,
170
- "loss": 4.6551,
171
  "step": 1150
172
  },
173
  {
174
  "epoch": 0.11194029850746269,
175
- "grad_norm": 1.3106518983840942,
176
  "learning_rate": 0.00011990000000000001,
177
- "loss": 4.5627,
178
  "step": 1200
179
  },
180
  {
181
  "epoch": 0.1166044776119403,
182
- "grad_norm": 1.4304360151290894,
183
  "learning_rate": 0.00012490000000000003,
184
- "loss": 4.4823,
185
  "step": 1250
186
  },
187
  {
188
  "epoch": 0.12126865671641791,
189
- "grad_norm": 0.883263885974884,
190
  "learning_rate": 0.00012989999999999999,
191
- "loss": 4.4091,
192
  "step": 1300
193
  },
194
  {
195
  "epoch": 0.1259328358208955,
196
- "grad_norm": 1.2335679531097412,
197
  "learning_rate": 0.0001349,
198
- "loss": 4.3284,
199
  "step": 1350
200
  },
201
  {
202
  "epoch": 0.13059701492537312,
203
- "grad_norm": 1.3587566614151,
204
  "learning_rate": 0.0001399,
205
- "loss": 4.2523,
206
  "step": 1400
207
  },
208
  {
209
  "epoch": 0.13526119402985073,
210
- "grad_norm": 1.4397273063659668,
211
  "learning_rate": 0.00014490000000000003,
212
- "loss": 4.1874,
213
  "step": 1450
214
  },
215
  {
216
  "epoch": 0.13992537313432835,
217
- "grad_norm": 0.9128643274307251,
218
  "learning_rate": 0.0001499,
219
- "loss": 4.1382,
220
  "step": 1500
221
  },
222
  {
223
  "epoch": 0.14458955223880596,
224
- "grad_norm": 1.1429646015167236,
225
  "learning_rate": 0.0001549,
226
- "loss": 4.0713,
227
  "step": 1550
228
  },
229
  {
230
  "epoch": 0.14925373134328357,
231
- "grad_norm": 0.7839804291725159,
232
  "learning_rate": 0.0001599,
233
- "loss": 4.0327,
234
  "step": 1600
235
  },
236
  {
237
  "epoch": 0.15391791044776118,
238
- "grad_norm": 0.5934556126594543,
239
  "learning_rate": 0.0001649,
240
- "loss": 4.0535,
241
  "step": 1650
242
  },
243
  {
244
  "epoch": 0.15858208955223882,
245
- "grad_norm": 0.6320812702178955,
246
  "learning_rate": 0.0001699,
247
- "loss": 3.9615,
248
  "step": 1700
249
  },
250
  {
251
  "epoch": 0.16324626865671643,
252
- "grad_norm": 0.6207342743873596,
253
  "learning_rate": 0.00017490000000000002,
254
- "loss": 3.9086,
255
  "step": 1750
256
  },
257
  {
258
  "epoch": 0.16791044776119404,
259
- "grad_norm": 0.8524695634841919,
260
  "learning_rate": 0.0001799,
261
- "loss": 3.8703,
262
  "step": 1800
263
  },
264
  {
265
  "epoch": 0.17257462686567165,
266
- "grad_norm": 0.9148792028427124,
267
  "learning_rate": 0.0001849,
268
- "loss": 3.8388,
269
  "step": 1850
270
  },
271
  {
272
  "epoch": 0.17723880597014927,
273
- "grad_norm": 0.7455419301986694,
274
  "learning_rate": 0.0001899,
275
- "loss": 3.8031,
276
  "step": 1900
277
  },
278
  {
279
  "epoch": 0.18190298507462688,
280
- "grad_norm": 0.9777323603630066,
281
  "learning_rate": 0.00019490000000000002,
282
- "loss": 3.781,
283
  "step": 1950
284
  },
285
  {
286
  "epoch": 0.1865671641791045,
287
- "grad_norm": 0.6018832921981812,
288
  "learning_rate": 0.0001999,
289
- "loss": 3.7421,
290
  "step": 2000
291
  },
292
  {
293
  "epoch": 0.1912313432835821,
294
- "grad_norm": 0.7007433772087097,
295
  "learning_rate": 0.00019887614678899083,
296
- "loss": 3.7148,
297
  "step": 2050
298
  },
299
  {
300
  "epoch": 0.1958955223880597,
301
- "grad_norm": 0.7864815592765808,
302
  "learning_rate": 0.00019772935779816515,
303
- "loss": 3.6907,
304
  "step": 2100
305
  },
306
  {
307
  "epoch": 0.20055970149253732,
308
- "grad_norm": 0.6575994491577148,
309
  "learning_rate": 0.00019658256880733944,
310
- "loss": 3.6504,
311
  "step": 2150
312
  },
313
  {
314
  "epoch": 0.20522388059701493,
315
- "grad_norm": 0.524401843547821,
316
  "learning_rate": 0.0001954357798165138,
317
- "loss": 3.6333,
318
  "step": 2200
319
  },
320
  {
321
  "epoch": 0.20988805970149255,
322
- "grad_norm": 0.8242985010147095,
323
  "learning_rate": 0.00019428899082568808,
324
- "loss": 3.6084,
325
  "step": 2250
326
  },
327
  {
328
  "epoch": 0.21455223880597016,
329
- "grad_norm": 0.5965232849121094,
330
  "learning_rate": 0.0001931422018348624,
331
- "loss": 3.5914,
332
  "step": 2300
333
  },
334
  {
335
  "epoch": 0.21921641791044777,
336
- "grad_norm": 0.5473845601081848,
337
  "learning_rate": 0.00019199541284403671,
338
- "loss": 3.572,
339
  "step": 2350
340
  },
341
  {
342
  "epoch": 0.22388059701492538,
343
- "grad_norm": 0.7030166983604431,
344
  "learning_rate": 0.00019084862385321103,
345
- "loss": 3.5548,
346
  "step": 2400
347
  },
348
  {
349
  "epoch": 0.228544776119403,
350
- "grad_norm": 0.5844017267227173,
351
- "learning_rate": 0.00018972477064220186,
352
- "loss": 3.5326,
353
  "step": 2450
354
  },
355
  {
356
  "epoch": 0.2332089552238806,
357
- "grad_norm": 0.4722403883934021,
358
- "learning_rate": 0.00018857798165137615,
359
- "loss": 3.5113,
360
  "step": 2500
361
  },
362
  {
363
  "epoch": 0.23787313432835822,
364
- "grad_norm": 0.4691997468471527,
365
- "learning_rate": 0.00018743119266055047,
366
- "loss": 3.4985,
367
  "step": 2550
368
  },
369
  {
370
  "epoch": 0.24253731343283583,
371
- "grad_norm": 0.5742476582527161,
372
- "learning_rate": 0.00018628440366972478,
373
- "loss": 3.4689,
374
  "step": 2600
375
  },
376
  {
377
  "epoch": 0.24720149253731344,
378
- "grad_norm": 0.4867892563343048,
379
- "learning_rate": 0.0001851376146788991,
380
- "loss": 3.4737,
381
  "step": 2650
382
  },
383
  {
384
  "epoch": 0.251865671641791,
385
- "grad_norm": 0.5876694321632385,
386
- "learning_rate": 0.0001839908256880734,
387
- "loss": 3.4516,
388
  "step": 2700
389
  },
390
  {
391
  "epoch": 0.25652985074626866,
392
- "grad_norm": 0.4691877067089081,
393
- "learning_rate": 0.0001828440366972477,
394
- "loss": 3.4411,
395
  "step": 2750
396
  },
397
  {
398
  "epoch": 0.26119402985074625,
399
- "grad_norm": 0.497503399848938,
400
- "learning_rate": 0.00018169724770642203,
401
- "loss": 3.4311,
402
  "step": 2800
403
  },
404
  {
405
  "epoch": 0.2658582089552239,
406
- "grad_norm": 0.4610927700996399,
407
- "learning_rate": 0.00018055045871559635,
408
- "loss": 3.4122,
409
  "step": 2850
410
  },
411
  {
412
  "epoch": 0.27052238805970147,
413
- "grad_norm": 0.5312972664833069,
414
- "learning_rate": 0.00017940366972477064,
415
- "loss": 3.4004,
416
  "step": 2900
417
  },
418
  {
419
  "epoch": 0.2751865671641791,
420
- "grad_norm": 0.5256180167198181,
421
- "learning_rate": 0.00017825688073394496,
422
- "loss": 3.3834,
423
  "step": 2950
424
  },
425
  {
426
  "epoch": 0.2798507462686567,
427
- "grad_norm": 0.5108594298362732,
428
- "learning_rate": 0.00017711009174311928,
429
- "loss": 3.379,
430
  "step": 3000
431
  },
432
  {
433
  "epoch": 0.28451492537313433,
434
- "grad_norm": 0.4818528890609741,
435
- "learning_rate": 0.00017596330275229357,
436
- "loss": 3.3645,
437
  "step": 3050
438
  },
439
  {
440
  "epoch": 0.2891791044776119,
441
- "grad_norm": 0.4077577590942383,
442
- "learning_rate": 0.0001748165137614679,
443
- "loss": 3.3659,
444
  "step": 3100
445
  },
446
  {
447
  "epoch": 0.29384328358208955,
448
- "grad_norm": 0.4791859984397888,
449
- "learning_rate": 0.0001736697247706422,
450
- "loss": 3.3427,
451
  "step": 3150
452
  },
453
  {
454
  "epoch": 0.29850746268656714,
455
- "grad_norm": 0.6330239176750183,
456
- "learning_rate": 0.00017252293577981652,
457
- "loss": 3.3484,
458
  "step": 3200
459
  },
460
  {
461
  "epoch": 0.3031716417910448,
462
- "grad_norm": 0.4511379897594452,
463
- "learning_rate": 0.00017137614678899084,
464
- "loss": 3.3302,
465
  "step": 3250
466
  },
467
  {
468
  "epoch": 0.30783582089552236,
469
- "grad_norm": 0.4571825861930847,
470
- "learning_rate": 0.00017022935779816516,
471
- "loss": 3.3144,
472
  "step": 3300
473
  },
474
  {
475
  "epoch": 0.3125,
476
- "grad_norm": 0.44557932019233704,
477
- "learning_rate": 0.00016908256880733945,
478
- "loss": 3.3071,
479
  "step": 3350
480
  },
481
  {
482
  "epoch": 0.31716417910447764,
483
- "grad_norm": 0.42434263229370117,
484
  "learning_rate": 0.00016793577981651377,
485
- "loss": 3.2986,
486
  "step": 3400
487
  },
488
  {
489
  "epoch": 0.3218283582089552,
490
- "grad_norm": 0.41813352704048157,
491
  "learning_rate": 0.0001667889908256881,
492
- "loss": 3.2902,
493
  "step": 3450
494
  },
495
  {
496
  "epoch": 0.32649253731343286,
497
- "grad_norm": 0.4084812104701996,
498
  "learning_rate": 0.00016564220183486238,
499
- "loss": 3.283,
500
  "step": 3500
501
  },
502
  {
503
  "epoch": 0.33115671641791045,
504
- "grad_norm": 0.38210785388946533,
505
  "learning_rate": 0.00016449541284403672,
506
- "loss": 3.2769,
507
  "step": 3550
508
  },
509
  {
510
  "epoch": 0.3358208955223881,
511
- "grad_norm": 0.5375980734825134,
512
  "learning_rate": 0.00016334862385321102,
513
- "loss": 3.2697,
514
  "step": 3600
515
  },
516
  {
517
  "epoch": 0.34048507462686567,
518
- "grad_norm": 0.38781264424324036,
519
  "learning_rate": 0.00016220183486238533,
520
- "loss": 3.2556,
521
  "step": 3650
522
  },
523
  {
524
  "epoch": 0.3451492537313433,
525
- "grad_norm": 0.41560590267181396,
526
  "learning_rate": 0.00016105504587155965,
527
- "loss": 3.2565,
528
  "step": 3700
529
  },
530
  {
531
  "epoch": 0.3498134328358209,
532
- "grad_norm": 0.4119664132595062,
533
  "learning_rate": 0.00015990825688073394,
534
- "loss": 3.2386,
535
  "step": 3750
536
  },
537
  {
538
  "epoch": 0.35447761194029853,
539
- "grad_norm": 0.4190433919429779,
540
  "learning_rate": 0.00015876146788990826,
541
- "loss": 3.2345,
542
  "step": 3800
543
  },
544
  {
545
  "epoch": 0.3591417910447761,
546
- "grad_norm": 0.37028583884239197,
547
  "learning_rate": 0.00015761467889908258,
548
- "loss": 3.226,
549
  "step": 3850
550
  },
551
  {
552
  "epoch": 0.36380597014925375,
553
- "grad_norm": 0.3971349000930786,
554
  "learning_rate": 0.0001564678899082569,
555
- "loss": 3.2213,
556
  "step": 3900
557
  },
558
  {
559
  "epoch": 0.36847014925373134,
560
- "grad_norm": 0.34641316533088684,
561
  "learning_rate": 0.0001553211009174312,
562
- "loss": 3.2144,
563
  "step": 3950
564
  },
565
  {
566
  "epoch": 0.373134328358209,
567
- "grad_norm": 0.40990573167800903,
568
  "learning_rate": 0.0001541743119266055,
569
- "loss": 3.2141,
570
  "step": 4000
571
  },
572
  {
573
  "epoch": 0.37779850746268656,
574
- "grad_norm": 0.35828983783721924,
575
  "learning_rate": 0.00015302752293577983,
576
- "loss": 3.2054,
577
  "step": 4050
578
  },
579
  {
580
  "epoch": 0.3824626865671642,
581
- "grad_norm": 0.37572401762008667,
582
  "learning_rate": 0.00015188073394495414,
583
- "loss": 3.1943,
584
  "step": 4100
585
  },
586
  {
587
  "epoch": 0.3871268656716418,
588
- "grad_norm": 0.3795240521430969,
589
  "learning_rate": 0.00015073394495412844,
590
- "loss": 3.1971,
591
  "step": 4150
592
  },
593
  {
594
  "epoch": 0.3917910447761194,
595
- "grad_norm": 0.3884343206882477,
596
  "learning_rate": 0.00014958715596330275,
597
- "loss": 3.1877,
598
  "step": 4200
599
  },
600
  {
601
  "epoch": 0.396455223880597,
602
- "grad_norm": 0.374661922454834,
603
  "learning_rate": 0.00014844036697247707,
604
- "loss": 3.174,
605
  "step": 4250
606
  },
607
  {
608
  "epoch": 0.40111940298507465,
609
- "grad_norm": 0.4330432415008545,
610
  "learning_rate": 0.00014729357798165136,
611
- "loss": 3.1787,
612
  "step": 4300
613
  },
614
  {
615
  "epoch": 0.40578358208955223,
616
- "grad_norm": 0.40475285053253174,
617
  "learning_rate": 0.0001461467889908257,
618
- "loss": 3.1781,
619
  "step": 4350
620
  },
621
  {
622
  "epoch": 0.41044776119402987,
623
- "grad_norm": 0.36028817296028137,
624
  "learning_rate": 0.000145,
625
- "loss": 3.1653,
626
  "step": 4400
627
  },
628
  {
629
  "epoch": 0.41511194029850745,
630
- "grad_norm": 0.37242797017097473,
631
  "learning_rate": 0.00014385321100917432,
632
- "loss": 3.1608,
633
  "step": 4450
634
  },
635
  {
636
  "epoch": 0.4197761194029851,
637
- "grad_norm": 0.42880979180336,
638
  "learning_rate": 0.00014270642201834864,
639
- "loss": 3.1656,
640
  "step": 4500
641
  },
642
  {
643
  "epoch": 0.4244402985074627,
644
- "grad_norm": 0.3562730848789215,
645
  "learning_rate": 0.00014155963302752296,
646
- "loss": 3.1409,
647
  "step": 4550
648
  },
649
  {
650
  "epoch": 0.4291044776119403,
651
- "grad_norm": 0.3420323133468628,
652
  "learning_rate": 0.00014041284403669725,
653
- "loss": 3.1502,
654
  "step": 4600
655
  },
656
  {
657
  "epoch": 0.4337686567164179,
658
- "grad_norm": 0.33985841274261475,
659
  "learning_rate": 0.00013926605504587157,
660
- "loss": 3.1437,
661
  "step": 4650
662
  },
663
  {
664
  "epoch": 0.43843283582089554,
665
- "grad_norm": 0.35252872109413147,
666
  "learning_rate": 0.00013811926605504588,
667
- "loss": 3.1375,
668
  "step": 4700
669
  },
670
  {
671
  "epoch": 0.4430970149253731,
672
- "grad_norm": 0.33299019932746887,
673
  "learning_rate": 0.00013697247706422018,
674
- "loss": 3.1207,
675
  "step": 4750
676
  },
677
  {
678
  "epoch": 0.44776119402985076,
679
- "grad_norm": 0.3340663015842438,
680
  "learning_rate": 0.00013582568807339452,
681
- "loss": 3.1285,
682
  "step": 4800
683
  },
684
  {
685
  "epoch": 0.45242537313432835,
686
- "grad_norm": 0.3459039330482483,
687
  "learning_rate": 0.00013470183486238532,
688
- "loss": 3.1187,
689
  "step": 4850
690
  },
691
  {
692
  "epoch": 0.457089552238806,
693
- "grad_norm": 0.34190812706947327,
694
  "learning_rate": 0.00013355504587155963,
695
- "loss": 3.1128,
696
  "step": 4900
697
  },
698
  {
699
  "epoch": 0.46175373134328357,
700
- "grad_norm": 0.38482001423835754,
701
  "learning_rate": 0.00013240825688073395,
702
- "loss": 3.1081,
703
  "step": 4950
704
  },
705
  {
706
  "epoch": 0.4664179104477612,
707
- "grad_norm": 0.3969718813896179,
708
  "learning_rate": 0.00013126146788990827,
709
- "loss": 3.111,
710
  "step": 5000
711
  },
712
  {
713
  "epoch": 0.4710820895522388,
714
- "grad_norm": 0.30978360772132874,
715
  "learning_rate": 0.0001301146788990826,
716
- "loss": 3.106,
717
  "step": 5050
718
  },
719
  {
720
  "epoch": 0.47574626865671643,
721
- "grad_norm": 0.32029977440834045,
722
  "learning_rate": 0.00012896788990825688,
723
- "loss": 3.1095,
724
  "step": 5100
725
  },
726
  {
727
  "epoch": 0.480410447761194,
728
- "grad_norm": 0.36698678135871887,
729
  "learning_rate": 0.0001278440366972477,
730
- "loss": 3.1017,
731
  "step": 5150
732
  },
733
  {
734
  "epoch": 0.48507462686567165,
735
- "grad_norm": 0.3710266947746277,
736
  "learning_rate": 0.00012669724770642202,
737
- "loss": 3.0863,
738
  "step": 5200
739
  },
740
  {
741
  "epoch": 0.48973880597014924,
742
- "grad_norm": 0.33469706773757935,
743
  "learning_rate": 0.00012555045871559634,
744
- "loss": 3.0949,
745
  "step": 5250
746
  },
747
  {
748
  "epoch": 0.4944029850746269,
749
- "grad_norm": 0.33493635058403015,
750
  "learning_rate": 0.00012440366972477063,
751
- "loss": 3.0794,
752
  "step": 5300
753
  },
754
  {
755
  "epoch": 0.49906716417910446,
756
- "grad_norm": 0.3489319682121277,
757
  "learning_rate": 0.00012325688073394498,
758
- "loss": 3.0752,
759
  "step": 5350
760
  },
761
  {
762
  "epoch": 0.503731343283582,
763
- "grad_norm": 0.32940617203712463,
764
  "learning_rate": 0.00012211009174311927,
765
- "loss": 3.0911,
766
  "step": 5400
767
  },
768
  {
769
  "epoch": 0.5083955223880597,
770
- "grad_norm": 0.3039519190788269,
771
  "learning_rate": 0.0001209633027522936,
772
- "loss": 3.0749,
773
  "step": 5450
774
  },
775
  {
776
  "epoch": 0.5130597014925373,
777
- "grad_norm": 0.3196834325790405,
778
  "learning_rate": 0.00011981651376146789,
779
- "loss": 3.0685,
780
  "step": 5500
781
  },
782
  {
783
  "epoch": 0.5177238805970149,
784
- "grad_norm": 0.35171428322792053,
785
  "learning_rate": 0.0001186697247706422,
786
- "loss": 3.0684,
787
  "step": 5550
788
  },
789
  {
790
  "epoch": 0.5223880597014925,
791
- "grad_norm": 0.3927328884601593,
792
  "learning_rate": 0.00011752293577981653,
793
- "loss": 3.0676,
794
  "step": 5600
795
  },
796
  {
797
  "epoch": 0.5270522388059702,
798
- "grad_norm": 0.32428842782974243,
799
  "learning_rate": 0.00011637614678899082,
800
- "loss": 3.0635,
801
  "step": 5650
802
  },
803
  {
804
  "epoch": 0.5317164179104478,
805
- "grad_norm": 0.3297617435455322,
806
  "learning_rate": 0.00011522935779816515,
807
- "loss": 3.0525,
808
  "step": 5700
809
  },
810
  {
811
  "epoch": 0.5363805970149254,
812
- "grad_norm": 0.3118542432785034,
813
  "learning_rate": 0.00011408256880733946,
814
- "loss": 3.0566,
815
  "step": 5750
816
  },
817
  {
818
  "epoch": 0.5410447761194029,
819
- "grad_norm": 0.29949086904525757,
820
  "learning_rate": 0.00011293577981651377,
821
- "loss": 3.0454,
822
  "step": 5800
823
  },
824
  {
825
  "epoch": 0.5457089552238806,
826
- "grad_norm": 0.3567812740802765,
827
  "learning_rate": 0.00011178899082568808,
828
- "loss": 3.0462,
829
  "step": 5850
830
  },
831
  {
832
  "epoch": 0.5503731343283582,
833
- "grad_norm": 0.3179719150066376,
834
  "learning_rate": 0.00011064220183486238,
835
- "loss": 3.0409,
836
  "step": 5900
837
  },
838
  {
839
  "epoch": 0.5550373134328358,
840
- "grad_norm": 0.32883220911026,
841
  "learning_rate": 0.0001094954128440367,
842
- "loss": 3.0527,
843
  "step": 5950
844
  },
845
  {
846
  "epoch": 0.5597014925373134,
847
- "grad_norm": 0.31516268849372864,
848
  "learning_rate": 0.000108348623853211,
849
- "loss": 3.0359,
850
  "step": 6000
851
  },
852
  {
853
  "epoch": 0.5643656716417911,
854
- "grad_norm": 0.298753559589386,
855
  "learning_rate": 0.00010720183486238533,
856
- "loss": 3.0343,
857
  "step": 6050
858
  },
859
  {
860
  "epoch": 0.5690298507462687,
861
- "grad_norm": 0.32886001467704773,
862
  "learning_rate": 0.00010605504587155963,
863
- "loss": 3.0229,
864
  "step": 6100
865
  },
866
  {
867
  "epoch": 0.5736940298507462,
868
- "grad_norm": 0.3526987135410309,
869
  "learning_rate": 0.00010490825688073396,
870
- "loss": 3.0227,
871
  "step": 6150
872
  },
873
  {
874
  "epoch": 0.5783582089552238,
875
- "grad_norm": 0.31915777921676636,
876
  "learning_rate": 0.00010376146788990825,
877
- "loss": 3.0311,
878
  "step": 6200
879
  },
880
  {
881
  "epoch": 0.5830223880597015,
882
- "grad_norm": 0.31342729926109314,
883
  "learning_rate": 0.00010261467889908258,
884
- "loss": 3.0257,
885
  "step": 6250
886
  },
887
  {
888
  "epoch": 0.5876865671641791,
889
- "grad_norm": 0.3711305856704712,
890
  "learning_rate": 0.00010146788990825689,
891
- "loss": 3.0125,
892
  "step": 6300
893
  },
894
  {
895
  "epoch": 0.5923507462686567,
896
- "grad_norm": 0.32861173152923584,
897
  "learning_rate": 0.00010032110091743118,
898
- "loss": 3.0235,
899
  "step": 6350
900
  },
901
  {
902
  "epoch": 0.5970149253731343,
903
- "grad_norm": 0.30850905179977417,
904
  "learning_rate": 9.917431192660551e-05,
905
- "loss": 3.0171,
906
  "step": 6400
907
  },
908
  {
909
  "epoch": 0.601679104477612,
910
- "grad_norm": 0.31126561760902405,
911
  "learning_rate": 9.802752293577983e-05,
912
- "loss": 3.0177,
913
  "step": 6450
914
  },
915
  {
916
  "epoch": 0.6063432835820896,
917
- "grad_norm": 0.3373335897922516,
918
  "learning_rate": 9.688073394495414e-05,
919
- "loss": 3.0138,
920
  "step": 6500
921
  },
922
  {
923
  "epoch": 0.6110074626865671,
924
- "grad_norm": 0.32974258065223694,
925
- "learning_rate": 9.575688073394496e-05,
926
- "loss": 3.0092,
927
  "step": 6550
928
  },
929
  {
930
  "epoch": 0.6156716417910447,
931
- "grad_norm": 0.29914727807044983,
932
- "learning_rate": 9.463302752293578e-05,
933
- "loss": 3.0075,
934
  "step": 6600
935
  },
936
  {
937
  "epoch": 0.6203358208955224,
938
- "grad_norm": 0.3006197214126587,
939
- "learning_rate": 9.34862385321101e-05,
940
- "loss": 3.015,
941
  "step": 6650
942
  },
943
  {
944
  "epoch": 0.625,
945
- "grad_norm": 0.2997278869152069,
946
- "learning_rate": 9.23394495412844e-05,
947
- "loss": 3.0037,
948
  "step": 6700
949
  },
950
  {
951
  "epoch": 0.6296641791044776,
952
- "grad_norm": 0.31719309091567993,
953
- "learning_rate": 9.119266055045872e-05,
954
- "loss": 2.9969,
955
  "step": 6750
956
  },
957
  {
958
  "epoch": 0.6343283582089553,
959
- "grad_norm": 0.3004010319709778,
960
- "learning_rate": 9.004587155963303e-05,
961
- "loss": 2.9909,
962
  "step": 6800
963
  },
964
  {
965
  "epoch": 0.6389925373134329,
966
- "grad_norm": 0.3123314082622528,
967
- "learning_rate": 8.889908256880735e-05,
968
- "loss": 2.9853,
969
  "step": 6850
970
  },
971
  {
972
  "epoch": 0.6436567164179104,
973
- "grad_norm": 0.334837406873703,
974
- "learning_rate": 8.775229357798166e-05,
975
- "loss": 2.9923,
976
  "step": 6900
977
  },
978
  {
979
  "epoch": 0.648320895522388,
980
- "grad_norm": 0.3055480420589447,
981
- "learning_rate": 8.660550458715597e-05,
982
- "loss": 2.9897,
983
  "step": 6950
984
  },
985
  {
986
  "epoch": 0.6529850746268657,
987
- "grad_norm": 0.33151915669441223,
988
- "learning_rate": 8.545871559633027e-05,
989
- "loss": 2.9843,
990
  "step": 7000
991
  },
992
  {
993
  "epoch": 0.6576492537313433,
994
- "grad_norm": 0.32511621713638306,
995
- "learning_rate": 8.431192660550459e-05,
996
- "loss": 2.9797,
997
  "step": 7050
998
  },
999
  {
1000
  "epoch": 0.6623134328358209,
1001
- "grad_norm": 0.29789265990257263,
1002
- "learning_rate": 8.31651376146789e-05,
1003
- "loss": 2.9789,
1004
  "step": 7100
1005
  },
1006
  {
1007
  "epoch": 0.6669776119402985,
1008
- "grad_norm": 0.3272823393344879,
1009
- "learning_rate": 8.201834862385321e-05,
1010
- "loss": 2.971,
1011
  "step": 7150
1012
  },
1013
  {
1014
  "epoch": 0.6716417910447762,
1015
- "grad_norm": 0.2983226478099823,
1016
- "learning_rate": 8.087155963302753e-05,
1017
- "loss": 2.9818,
1018
  "step": 7200
1019
  },
1020
  {
1021
  "epoch": 0.6763059701492538,
1022
- "grad_norm": 0.30532923340797424,
1023
- "learning_rate": 7.972477064220184e-05,
1024
- "loss": 2.9739,
1025
  "step": 7250
1026
  },
1027
  {
1028
  "epoch": 0.6809701492537313,
1029
- "grad_norm": 0.3046092092990875,
1030
- "learning_rate": 7.857798165137616e-05,
1031
- "loss": 2.9575,
1032
  "step": 7300
1033
  },
1034
  {
1035
  "epoch": 0.6856343283582089,
1036
- "grad_norm": 0.3133400082588196,
1037
- "learning_rate": 7.743119266055046e-05,
1038
- "loss": 2.9794,
1039
  "step": 7350
1040
  },
1041
  {
1042
  "epoch": 0.6902985074626866,
1043
- "grad_norm": 0.3090495765209198,
1044
- "learning_rate": 7.628440366972477e-05,
1045
- "loss": 2.9632,
1046
  "step": 7400
1047
  },
1048
  {
1049
  "epoch": 0.6949626865671642,
1050
- "grad_norm": 0.2925854027271271,
1051
- "learning_rate": 7.513761467889908e-05,
1052
- "loss": 2.9649,
1053
  "step": 7450
1054
  },
1055
  {
1056
  "epoch": 0.6996268656716418,
1057
- "grad_norm": 0.30464622378349304,
1058
- "learning_rate": 7.399082568807339e-05,
1059
- "loss": 2.9606,
1060
  "step": 7500
1061
  },
1062
  {
1063
  "epoch": 0.7042910447761194,
1064
- "grad_norm": 0.29478827118873596,
1065
- "learning_rate": 7.284403669724771e-05,
1066
- "loss": 2.9574,
1067
  "step": 7550
1068
  },
1069
  {
1070
  "epoch": 0.7089552238805971,
1071
- "grad_norm": 0.30368906259536743,
1072
- "learning_rate": 7.169724770642203e-05,
1073
- "loss": 2.9583,
1074
  "step": 7600
1075
  },
1076
  {
1077
  "epoch": 0.7136194029850746,
1078
- "grad_norm": 0.2974824905395508,
1079
- "learning_rate": 7.055045871559633e-05,
1080
- "loss": 2.9566,
1081
  "step": 7650
1082
  },
1083
  {
1084
  "epoch": 0.7182835820895522,
1085
- "grad_norm": 0.3006385564804077,
1086
- "learning_rate": 6.940366972477065e-05,
1087
- "loss": 2.953,
1088
  "step": 7700
1089
  },
1090
  {
1091
  "epoch": 0.7229477611940298,
1092
- "grad_norm": 0.2787078022956848,
1093
- "learning_rate": 6.825688073394497e-05,
1094
- "loss": 2.9466,
1095
  "step": 7750
1096
  },
1097
  {
1098
  "epoch": 0.7276119402985075,
1099
- "grad_norm": 0.2966342568397522,
1100
- "learning_rate": 6.711009174311927e-05,
1101
- "loss": 2.9446,
1102
  "step": 7800
1103
  },
1104
  {
1105
  "epoch": 0.7322761194029851,
1106
- "grad_norm": 0.32728853821754456,
1107
- "learning_rate": 6.600917431192661e-05,
1108
- "loss": 2.9486,
1109
  "step": 7850
1110
  },
1111
  {
1112
  "epoch": 0.7369402985074627,
1113
- "grad_norm": 0.325846403837204,
1114
- "learning_rate": 6.486238532110092e-05,
1115
- "loss": 2.9456,
1116
  "step": 7900
1117
  },
1118
  {
1119
  "epoch": 0.7416044776119403,
1120
- "grad_norm": 0.2933451235294342,
1121
- "learning_rate": 6.371559633027524e-05,
1122
- "loss": 2.9404,
1123
  "step": 7950
1124
  },
1125
  {
1126
  "epoch": 0.746268656716418,
1127
- "grad_norm": 0.3031325042247772,
1128
- "learning_rate": 6.256880733944954e-05,
1129
- "loss": 2.9486,
1130
  "step": 8000
1131
  },
1132
  {
1133
  "epoch": 0.7509328358208955,
1134
- "grad_norm": 0.30206266045570374,
1135
- "learning_rate": 6.142201834862386e-05,
1136
- "loss": 2.9364,
1137
  "step": 8050
1138
  },
1139
  {
1140
  "epoch": 0.7555970149253731,
1141
- "grad_norm": 0.29706859588623047,
1142
- "learning_rate": 6.027522935779817e-05,
1143
- "loss": 2.9483,
1144
  "step": 8100
1145
  },
1146
  {
1147
  "epoch": 0.7602611940298507,
1148
- "grad_norm": 0.2909660339355469,
1149
- "learning_rate": 5.912844036697248e-05,
1150
- "loss": 2.9405,
1151
  "step": 8150
1152
  },
1153
  {
1154
  "epoch": 0.7649253731343284,
1155
- "grad_norm": 0.3136695623397827,
1156
- "learning_rate": 5.798165137614679e-05,
1157
- "loss": 2.9366,
1158
  "step": 8200
1159
  },
1160
  {
1161
  "epoch": 0.769589552238806,
1162
- "grad_norm": 0.29548221826553345,
1163
- "learning_rate": 5.683486238532111e-05,
1164
- "loss": 2.9305,
1165
  "step": 8250
1166
  },
1167
  {
1168
  "epoch": 0.7742537313432836,
1169
- "grad_norm": 0.2777513265609741,
1170
- "learning_rate": 5.568807339449541e-05,
1171
- "loss": 2.9377,
1172
  "step": 8300
1173
  },
1174
  {
1175
  "epoch": 0.7789179104477612,
1176
- "grad_norm": 0.28673887252807617,
1177
- "learning_rate": 5.454128440366972e-05,
1178
- "loss": 2.9342,
1179
  "step": 8350
1180
  },
1181
  {
1182
  "epoch": 0.7835820895522388,
1183
- "grad_norm": 0.2911510169506073,
1184
- "learning_rate": 5.339449541284404e-05,
1185
- "loss": 2.938,
1186
  "step": 8400
1187
  },
1188
  {
1189
  "epoch": 0.7882462686567164,
1190
- "grad_norm": 0.3052498698234558,
1191
- "learning_rate": 5.224770642201835e-05,
1192
- "loss": 2.9235,
1193
  "step": 8450
1194
  },
1195
  {
1196
  "epoch": 0.792910447761194,
1197
- "grad_norm": 0.30528631806373596,
1198
- "learning_rate": 5.110091743119266e-05,
1199
- "loss": 2.9393,
1200
  "step": 8500
1201
  },
1202
  {
1203
  "epoch": 0.7975746268656716,
1204
- "grad_norm": 0.27922648191452026,
1205
- "learning_rate": 4.9954128440366974e-05,
1206
- "loss": 2.9153,
1207
  "step": 8550
1208
  },
1209
  {
1210
  "epoch": 0.8022388059701493,
1211
- "grad_norm": 0.2799355685710907,
1212
- "learning_rate": 4.880733944954129e-05,
1213
- "loss": 2.9329,
1214
  "step": 8600
1215
  },
1216
  {
1217
  "epoch": 0.8069029850746269,
1218
- "grad_norm": 0.3089926242828369,
1219
- "learning_rate": 4.76605504587156e-05,
1220
- "loss": 2.9203,
1221
  "step": 8650
1222
  },
1223
  {
1224
  "epoch": 0.8115671641791045,
1225
- "grad_norm": 0.3086063265800476,
1226
- "learning_rate": 4.651376146788991e-05,
1227
- "loss": 2.9223,
1228
  "step": 8700
1229
  },
1230
  {
1231
  "epoch": 0.816231343283582,
1232
- "grad_norm": 0.3019801676273346,
1233
- "learning_rate": 4.536697247706422e-05,
1234
- "loss": 2.9159,
1235
  "step": 8750
1236
  },
1237
  {
1238
  "epoch": 0.8208955223880597,
1239
- "grad_norm": 0.2975527048110962,
1240
- "learning_rate": 4.422018348623854e-05,
1241
- "loss": 2.9215,
1242
  "step": 8800
1243
  },
1244
  {
1245
  "epoch": 0.8255597014925373,
1246
- "grad_norm": 0.30920466780662537,
1247
- "learning_rate": 4.3073394495412844e-05,
1248
- "loss": 2.9213,
1249
  "step": 8850
1250
  },
1251
  {
1252
  "epoch": 0.8302238805970149,
1253
- "grad_norm": 0.2723771929740906,
1254
- "learning_rate": 4.1926605504587155e-05,
1255
- "loss": 2.9119,
1256
  "step": 8900
1257
  },
1258
  {
1259
  "epoch": 0.8348880597014925,
1260
- "grad_norm": 0.28407421708106995,
1261
- "learning_rate": 4.0779816513761474e-05,
1262
- "loss": 2.9101,
1263
  "step": 8950
1264
  },
1265
  {
1266
  "epoch": 0.8395522388059702,
1267
- "grad_norm": 0.2752555012702942,
1268
- "learning_rate": 3.9633027522935785e-05,
1269
- "loss": 2.9087,
1270
  "step": 9000
1271
  }
1272
  ],
@@ -1287,7 +1287,7 @@
1287
  "attributes": {}
1288
  }
1289
  },
1290
- "total_flos": 1911800851333120.0,
1291
  "train_batch_size": 64,
1292
  "trial_name": null,
1293
  "trial_params": null
 
11
  "log_history": [
12
  {
13
  "epoch": 0.0046641791044776115,
14
+ "grad_norm": 2.434795618057251,
15
  "learning_rate": 4.9000000000000005e-06,
16
+ "loss": 10.2918,
17
  "step": 50
18
  },
19
  {
20
  "epoch": 0.009328358208955223,
21
+ "grad_norm": 1.3679819107055664,
22
  "learning_rate": 9.900000000000002e-06,
23
+ "loss": 9.0755,
24
  "step": 100
25
  },
26
  {
27
  "epoch": 0.013992537313432836,
28
+ "grad_norm": 0.9769303798675537,
29
  "learning_rate": 1.49e-05,
30
+ "loss": 8.3639,
31
  "step": 150
32
  },
33
  {
34
  "epoch": 0.018656716417910446,
35
+ "grad_norm": 0.5544369220733643,
36
  "learning_rate": 1.9900000000000003e-05,
37
+ "loss": 7.8229,
38
  "step": 200
39
  },
40
  {
41
  "epoch": 0.02332089552238806,
42
+ "grad_norm": 0.32632946968078613,
43
  "learning_rate": 2.4900000000000002e-05,
44
+ "loss": 7.4323,
45
  "step": 250
46
  },
47
  {
48
  "epoch": 0.027985074626865673,
49
+ "grad_norm": 0.34820762276649475,
50
  "learning_rate": 2.9900000000000002e-05,
51
+ "loss": 7.1956,
52
  "step": 300
53
  },
54
  {
55
  "epoch": 0.03264925373134328,
56
+ "grad_norm": 0.5745918154716492,
57
  "learning_rate": 3.49e-05,
58
+ "loss": 6.9913,
59
  "step": 350
60
  },
61
  {
62
  "epoch": 0.03731343283582089,
63
+ "grad_norm": 0.7313559651374817,
64
  "learning_rate": 3.99e-05,
65
+ "loss": 6.7761,
66
  "step": 400
67
  },
68
  {
69
  "epoch": 0.04197761194029851,
70
+ "grad_norm": 0.5769322514533997,
71
  "learning_rate": 4.49e-05,
72
+ "loss": 6.598,
73
  "step": 450
74
  },
75
  {
76
  "epoch": 0.04664179104477612,
77
+ "grad_norm": 0.5906850099563599,
78
  "learning_rate": 4.99e-05,
79
+ "loss": 6.4384,
80
  "step": 500
81
  },
82
  {
83
  "epoch": 0.051305970149253734,
84
+ "grad_norm": 1.5271503925323486,
85
  "learning_rate": 5.4900000000000006e-05,
86
+ "loss": 6.3052,
87
  "step": 550
88
  },
89
  {
90
  "epoch": 0.055970149253731345,
91
+ "grad_norm": 0.980595052242279,
92
  "learning_rate": 5.99e-05,
93
+ "loss": 6.1786,
94
  "step": 600
95
  },
96
  {
97
  "epoch": 0.06063432835820896,
98
+ "grad_norm": 0.8249331712722778,
99
  "learning_rate": 6.49e-05,
100
+ "loss": 6.0769,
101
  "step": 650
102
  },
103
  {
104
  "epoch": 0.06529850746268656,
105
+ "grad_norm": 1.6615272760391235,
106
  "learning_rate": 6.99e-05,
107
+ "loss": 5.9624,
108
  "step": 700
109
  },
110
  {
111
  "epoch": 0.06996268656716417,
112
+ "grad_norm": 1.4796066284179688,
113
  "learning_rate": 7.49e-05,
114
+ "loss": 5.8661,
115
  "step": 750
116
  },
117
  {
118
  "epoch": 0.07462686567164178,
119
+ "grad_norm": 1.050801157951355,
120
  "learning_rate": 7.99e-05,
121
+ "loss": 5.7773,
122
  "step": 800
123
  },
124
  {
125
  "epoch": 0.07929104477611941,
126
+ "grad_norm": 1.1273882389068604,
127
  "learning_rate": 8.49e-05,
128
+ "loss": 5.6794,
129
  "step": 850
130
  },
131
  {
132
  "epoch": 0.08395522388059702,
133
+ "grad_norm": 1.3447613716125488,
134
  "learning_rate": 8.99e-05,
135
+ "loss": 5.5883,
136
  "step": 900
137
  },
138
  {
139
  "epoch": 0.08861940298507463,
140
+ "grad_norm": 0.8979071378707886,
141
  "learning_rate": 9.49e-05,
142
+ "loss": 5.5152,
143
  "step": 950
144
  },
145
  {
146
  "epoch": 0.09328358208955224,
147
+ "grad_norm": 1.7317239046096802,
148
  "learning_rate": 9.99e-05,
149
+ "loss": 5.4185,
150
  "step": 1000
151
  },
152
  {
153
  "epoch": 0.09794776119402986,
154
+ "grad_norm": 1.6568114757537842,
155
  "learning_rate": 0.0001049,
156
+ "loss": 5.3531,
157
  "step": 1050
158
  },
159
  {
160
  "epoch": 0.10261194029850747,
161
+ "grad_norm": 1.260619044303894,
162
  "learning_rate": 0.0001099,
163
+ "loss": 5.2675,
164
  "step": 1100
165
  },
166
  {
167
  "epoch": 0.10727611940298508,
168
+ "grad_norm": 1.6347819566726685,
169
  "learning_rate": 0.0001149,
170
+ "loss": 5.1987,
171
  "step": 1150
172
  },
173
  {
174
  "epoch": 0.11194029850746269,
175
+ "grad_norm": 1.6571054458618164,
176
  "learning_rate": 0.00011990000000000001,
177
+ "loss": 5.1358,
178
  "step": 1200
179
  },
180
  {
181
  "epoch": 0.1166044776119403,
182
+ "grad_norm": 1.2196329832077026,
183
  "learning_rate": 0.00012490000000000003,
184
+ "loss": 5.0755,
185
  "step": 1250
186
  },
187
  {
188
  "epoch": 0.12126865671641791,
189
+ "grad_norm": 1.2077012062072754,
190
  "learning_rate": 0.00012989999999999999,
191
+ "loss": 5.0243,
192
  "step": 1300
193
  },
194
  {
195
  "epoch": 0.1259328358208955,
196
+ "grad_norm": 1.3209941387176514,
197
  "learning_rate": 0.0001349,
198
+ "loss": 4.9678,
199
  "step": 1350
200
  },
201
  {
202
  "epoch": 0.13059701492537312,
203
+ "grad_norm": 0.8279642462730408,
204
  "learning_rate": 0.0001399,
205
+ "loss": 4.9159,
206
  "step": 1400
207
  },
208
  {
209
  "epoch": 0.13526119402985073,
210
+ "grad_norm": 0.9633798599243164,
211
  "learning_rate": 0.00014490000000000003,
212
+ "loss": 4.8596,
213
  "step": 1450
214
  },
215
  {
216
  "epoch": 0.13992537313432835,
217
+ "grad_norm": 0.987281084060669,
218
  "learning_rate": 0.0001499,
219
+ "loss": 4.8162,
220
  "step": 1500
221
  },
222
  {
223
  "epoch": 0.14458955223880596,
224
+ "grad_norm": 1.094568133354187,
225
  "learning_rate": 0.0001549,
226
+ "loss": 4.7598,
227
  "step": 1550
228
  },
229
  {
230
  "epoch": 0.14925373134328357,
231
+ "grad_norm": 0.8916253447532654,
232
  "learning_rate": 0.0001599,
233
+ "loss": 4.7184,
234
  "step": 1600
235
  },
236
  {
237
  "epoch": 0.15391791044776118,
238
+ "grad_norm": 0.8389304876327515,
239
  "learning_rate": 0.0001649,
240
+ "loss": 4.6699,
241
  "step": 1650
242
  },
243
  {
244
  "epoch": 0.15858208955223882,
245
+ "grad_norm": 1.113956093788147,
246
  "learning_rate": 0.0001699,
247
+ "loss": 4.6177,
248
  "step": 1700
249
  },
250
  {
251
  "epoch": 0.16324626865671643,
252
+ "grad_norm": 0.9380860924720764,
253
  "learning_rate": 0.00017490000000000002,
254
+ "loss": 4.5619,
255
  "step": 1750
256
  },
257
  {
258
  "epoch": 0.16791044776119404,
259
+ "grad_norm": 1.2293964624404907,
260
  "learning_rate": 0.0001799,
261
+ "loss": 4.5084,
262
  "step": 1800
263
  },
264
  {
265
  "epoch": 0.17257462686567165,
266
+ "grad_norm": 1.1242707967758179,
267
  "learning_rate": 0.0001849,
268
+ "loss": 4.4553,
269
  "step": 1850
270
  },
271
  {
272
  "epoch": 0.17723880597014927,
273
+ "grad_norm": 0.7840976715087891,
274
  "learning_rate": 0.0001899,
275
+ "loss": 4.4074,
276
  "step": 1900
277
  },
278
  {
279
  "epoch": 0.18190298507462688,
280
+ "grad_norm": 0.7234581112861633,
281
  "learning_rate": 0.00019490000000000002,
282
+ "loss": 4.3652,
283
  "step": 1950
284
  },
285
  {
286
  "epoch": 0.1865671641791045,
287
+ "grad_norm": 0.8629528880119324,
288
  "learning_rate": 0.0001999,
289
+ "loss": 4.317,
290
  "step": 2000
291
  },
292
  {
293
  "epoch": 0.1912313432835821,
294
+ "grad_norm": 0.8042444586753845,
295
  "learning_rate": 0.00019887614678899083,
296
+ "loss": 4.2762,
297
  "step": 2050
298
  },
299
  {
300
  "epoch": 0.1958955223880597,
301
+ "grad_norm": 0.80702143907547,
302
  "learning_rate": 0.00019772935779816515,
303
+ "loss": 4.2375,
304
  "step": 2100
305
  },
306
  {
307
  "epoch": 0.20055970149253732,
308
+ "grad_norm": 0.6999443769454956,
309
  "learning_rate": 0.00019658256880733944,
310
+ "loss": 4.1876,
311
  "step": 2150
312
  },
313
  {
314
  "epoch": 0.20522388059701493,
315
+ "grad_norm": 0.7686917781829834,
316
  "learning_rate": 0.0001954357798165138,
317
+ "loss": 4.1574,
318
  "step": 2200
319
  },
320
  {
321
  "epoch": 0.20988805970149255,
322
+ "grad_norm": 0.7840579152107239,
323
  "learning_rate": 0.00019428899082568808,
324
+ "loss": 4.1217,
325
  "step": 2250
326
  },
327
  {
328
  "epoch": 0.21455223880597016,
329
+ "grad_norm": 0.8510863184928894,
330
  "learning_rate": 0.0001931422018348624,
331
+ "loss": 4.0979,
332
  "step": 2300
333
  },
334
  {
335
  "epoch": 0.21921641791044777,
336
+ "grad_norm": 0.679388165473938,
337
  "learning_rate": 0.00019199541284403671,
338
+ "loss": 4.0661,
339
  "step": 2350
340
  },
341
  {
342
  "epoch": 0.22388059701492538,
343
+ "grad_norm": 0.5541307330131531,
344
  "learning_rate": 0.00019084862385321103,
345
+ "loss": 4.0407,
346
  "step": 2400
347
  },
348
  {
349
  "epoch": 0.228544776119403,
350
+ "grad_norm": 0.6176748871803284,
351
+ "learning_rate": 0.00018970183486238532,
352
+ "loss": 4.0098,
353
  "step": 2450
354
  },
355
  {
356
  "epoch": 0.2332089552238806,
357
+ "grad_norm": 0.6339316368103027,
358
+ "learning_rate": 0.00018855504587155964,
359
+ "loss": 3.9825,
360
  "step": 2500
361
  },
362
  {
363
  "epoch": 0.23787313432835822,
364
+ "grad_norm": 0.6165521740913391,
365
+ "learning_rate": 0.00018740825688073396,
366
+ "loss": 3.9626,
367
  "step": 2550
368
  },
369
  {
370
  "epoch": 0.24253731343283583,
371
+ "grad_norm": 0.6364430785179138,
372
+ "learning_rate": 0.00018626146788990825,
373
+ "loss": 3.9288,
374
  "step": 2600
375
  },
376
  {
377
  "epoch": 0.24720149253731344,
378
+ "grad_norm": 0.5691106915473938,
379
+ "learning_rate": 0.00018511467889908257,
380
+ "loss": 3.9221,
381
  "step": 2650
382
  },
383
  {
384
  "epoch": 0.251865671641791,
385
+ "grad_norm": 0.5992695689201355,
386
+ "learning_rate": 0.0001839678899082569,
387
+ "loss": 3.8961,
388
  "step": 2700
389
  },
390
  {
391
  "epoch": 0.25652985074626866,
392
+ "grad_norm": 0.5932413935661316,
393
+ "learning_rate": 0.0001828211009174312,
394
+ "loss": 3.8808,
395
  "step": 2750
396
  },
397
  {
398
  "epoch": 0.26119402985074625,
399
+ "grad_norm": 0.6363717913627625,
400
+ "learning_rate": 0.0001816743119266055,
401
+ "loss": 3.8652,
402
  "step": 2800
403
  },
404
  {
405
  "epoch": 0.2658582089552239,
406
+ "grad_norm": 0.5456887483596802,
407
+ "learning_rate": 0.00018052752293577984,
408
+ "loss": 3.842,
409
  "step": 2850
410
  },
411
  {
412
  "epoch": 0.27052238805970147,
413
+ "grad_norm": 0.5930595397949219,
414
+ "learning_rate": 0.00017938073394495414,
415
+ "loss": 3.8244,
416
  "step": 2900
417
  },
418
  {
419
  "epoch": 0.2751865671641791,
420
+ "grad_norm": 0.5566945672035217,
421
+ "learning_rate": 0.00017823394495412843,
422
+ "loss": 3.804,
423
  "step": 2950
424
  },
425
  {
426
  "epoch": 0.2798507462686567,
427
+ "grad_norm": 0.6539527773857117,
428
+ "learning_rate": 0.00017708715596330277,
429
+ "loss": 3.7985,
430
  "step": 3000
431
  },
432
  {
433
  "epoch": 0.28451492537313433,
434
+ "grad_norm": 0.5921605229377747,
435
+ "learning_rate": 0.00017594036697247706,
436
+ "loss": 3.7834,
437
  "step": 3050
438
  },
439
  {
440
  "epoch": 0.2891791044776119,
441
+ "grad_norm": 0.5476294159889221,
442
+ "learning_rate": 0.00017479357798165138,
443
+ "loss": 3.7735,
444
  "step": 3100
445
  },
446
  {
447
  "epoch": 0.29384328358208955,
448
+ "grad_norm": 0.5736291408538818,
449
+ "learning_rate": 0.0001736467889908257,
450
+ "loss": 3.7535,
451
  "step": 3150
452
  },
453
  {
454
  "epoch": 0.29850746268656714,
455
+ "grad_norm": 0.5571457743644714,
456
+ "learning_rate": 0.00017250000000000002,
457
+ "loss": 3.7517,
458
  "step": 3200
459
  },
460
  {
461
  "epoch": 0.3031716417910448,
462
+ "grad_norm": 0.5658199787139893,
463
+ "learning_rate": 0.0001713532110091743,
464
+ "loss": 3.7308,
465
  "step": 3250
466
  },
467
  {
468
  "epoch": 0.30783582089552236,
469
+ "grad_norm": 0.5400105714797974,
470
+ "learning_rate": 0.00017020642201834863,
471
+ "loss": 3.713,
472
  "step": 3300
473
  },
474
  {
475
  "epoch": 0.3125,
476
+ "grad_norm": 0.5364307761192322,
477
+ "learning_rate": 0.00016905963302752295,
478
+ "loss": 3.7018,
479
  "step": 3350
480
  },
481
  {
482
  "epoch": 0.31716417910447764,
483
+ "grad_norm": 0.5792276859283447,
484
  "learning_rate": 0.00016793577981651377,
485
+ "loss": 3.6918,
486
  "step": 3400
487
  },
488
  {
489
  "epoch": 0.3218283582089552,
490
+ "grad_norm": 0.44588422775268555,
491
  "learning_rate": 0.0001667889908256881,
492
+ "loss": 3.6813,
493
  "step": 3450
494
  },
495
  {
496
  "epoch": 0.32649253731343286,
497
+ "grad_norm": 0.5175060033798218,
498
  "learning_rate": 0.00016564220183486238,
499
+ "loss": 3.6705,
500
  "step": 3500
501
  },
502
  {
503
  "epoch": 0.33115671641791045,
504
+ "grad_norm": 0.5310373306274414,
505
  "learning_rate": 0.00016449541284403672,
506
+ "loss": 3.6645,
507
  "step": 3550
508
  },
509
  {
510
  "epoch": 0.3358208955223881,
511
+ "grad_norm": 0.5658640265464783,
512
  "learning_rate": 0.00016334862385321102,
513
+ "loss": 3.6542,
514
  "step": 3600
515
  },
516
  {
517
  "epoch": 0.34048507462686567,
518
+ "grad_norm": 0.5425497889518738,
519
  "learning_rate": 0.00016220183486238533,
520
+ "loss": 3.6375,
521
  "step": 3650
522
  },
523
  {
524
  "epoch": 0.3451492537313433,
525
+ "grad_norm": 0.4865582585334778,
526
  "learning_rate": 0.00016105504587155965,
527
+ "loss": 3.6384,
528
  "step": 3700
529
  },
530
  {
531
  "epoch": 0.3498134328358209,
532
+ "grad_norm": 0.4987630248069763,
533
  "learning_rate": 0.00015990825688073394,
534
+ "loss": 3.6156,
535
  "step": 3750
536
  },
537
  {
538
  "epoch": 0.35447761194029853,
539
+ "grad_norm": 0.5758020877838135,
540
  "learning_rate": 0.00015876146788990826,
541
+ "loss": 3.6123,
542
  "step": 3800
543
  },
544
  {
545
  "epoch": 0.3591417910447761,
546
+ "grad_norm": 0.5264896154403687,
547
  "learning_rate": 0.00015761467889908258,
548
+ "loss": 3.6019,
549
  "step": 3850
550
  },
551
  {
552
  "epoch": 0.36380597014925375,
553
+ "grad_norm": 0.4835953414440155,
554
  "learning_rate": 0.0001564678899082569,
555
+ "loss": 3.5959,
556
  "step": 3900
557
  },
558
  {
559
  "epoch": 0.36847014925373134,
560
+ "grad_norm": 0.4341369569301605,
561
  "learning_rate": 0.0001553211009174312,
562
+ "loss": 3.5873,
563
  "step": 3950
564
  },
565
  {
566
  "epoch": 0.373134328358209,
567
+ "grad_norm": 0.47072407603263855,
568
  "learning_rate": 0.0001541743119266055,
569
+ "loss": 3.5852,
570
  "step": 4000
571
  },
572
  {
573
  "epoch": 0.37779850746268656,
574
+ "grad_norm": 0.46912139654159546,
575
  "learning_rate": 0.00015302752293577983,
576
+ "loss": 3.575,
577
  "step": 4050
578
  },
579
  {
580
  "epoch": 0.3824626865671642,
581
+ "grad_norm": 0.5946963429450989,
582
  "learning_rate": 0.00015188073394495414,
583
+ "loss": 3.5639,
584
  "step": 4100
585
  },
586
  {
587
  "epoch": 0.3871268656716418,
588
+ "grad_norm": 0.4828120470046997,
589
  "learning_rate": 0.00015073394495412844,
590
+ "loss": 3.5659,
591
  "step": 4150
592
  },
593
  {
594
  "epoch": 0.3917910447761194,
595
+ "grad_norm": 0.48630213737487793,
596
  "learning_rate": 0.00014958715596330275,
597
+ "loss": 3.5538,
598
  "step": 4200
599
  },
600
  {
601
  "epoch": 0.396455223880597,
602
+ "grad_norm": 0.4966210126876831,
603
  "learning_rate": 0.00014844036697247707,
604
+ "loss": 3.5394,
605
  "step": 4250
606
  },
607
  {
608
  "epoch": 0.40111940298507465,
609
+ "grad_norm": 0.8680199384689331,
610
  "learning_rate": 0.00014729357798165136,
611
+ "loss": 3.5434,
612
  "step": 4300
613
  },
614
  {
615
  "epoch": 0.40578358208955223,
616
+ "grad_norm": 0.4786328971385956,
617
  "learning_rate": 0.0001461467889908257,
618
+ "loss": 3.5429,
619
  "step": 4350
620
  },
621
  {
622
  "epoch": 0.41044776119402987,
623
+ "grad_norm": 0.4968096911907196,
624
  "learning_rate": 0.000145,
625
+ "loss": 3.5282,
626
  "step": 4400
627
  },
628
  {
629
  "epoch": 0.41511194029850745,
630
+ "grad_norm": 0.5361109375953674,
631
  "learning_rate": 0.00014385321100917432,
632
+ "loss": 3.5219,
633
  "step": 4450
634
  },
635
  {
636
  "epoch": 0.4197761194029851,
637
+ "grad_norm": 0.4819442629814148,
638
  "learning_rate": 0.00014270642201834864,
639
+ "loss": 3.5247,
640
  "step": 4500
641
  },
642
  {
643
  "epoch": 0.4244402985074627,
644
+ "grad_norm": 0.43222498893737793,
645
  "learning_rate": 0.00014155963302752296,
646
+ "loss": 3.5003,
647
  "step": 4550
648
  },
649
  {
650
  "epoch": 0.4291044776119403,
651
+ "grad_norm": 0.40276777744293213,
652
  "learning_rate": 0.00014041284403669725,
653
+ "loss": 3.5076,
654
  "step": 4600
655
  },
656
  {
657
  "epoch": 0.4337686567164179,
658
+ "grad_norm": 0.4238736033439636,
659
  "learning_rate": 0.00013926605504587157,
660
+ "loss": 3.5017,
661
  "step": 4650
662
  },
663
  {
664
  "epoch": 0.43843283582089554,
665
+ "grad_norm": 0.4680284559726715,
666
  "learning_rate": 0.00013811926605504588,
667
+ "loss": 3.4945,
668
  "step": 4700
669
  },
670
  {
671
  "epoch": 0.4430970149253731,
672
+ "grad_norm": 0.42782142758369446,
673
  "learning_rate": 0.00013697247706422018,
674
+ "loss": 3.4769,
675
  "step": 4750
676
  },
677
  {
678
  "epoch": 0.44776119402985076,
679
+ "grad_norm": 0.48704081773757935,
680
  "learning_rate": 0.00013582568807339452,
681
+ "loss": 3.4835,
682
  "step": 4800
683
  },
684
  {
685
  "epoch": 0.45242537313432835,
686
+ "grad_norm": 0.4254228472709656,
687
  "learning_rate": 0.00013470183486238532,
688
+ "loss": 3.4746,
689
  "step": 4850
690
  },
691
  {
692
  "epoch": 0.457089552238806,
693
+ "grad_norm": 0.46868908405303955,
694
  "learning_rate": 0.00013355504587155963,
695
+ "loss": 3.4669,
696
  "step": 4900
697
  },
698
  {
699
  "epoch": 0.46175373134328357,
700
+ "grad_norm": 0.39554563164711,
701
  "learning_rate": 0.00013240825688073395,
702
+ "loss": 3.4612,
703
  "step": 4950
704
  },
705
  {
706
  "epoch": 0.4664179104477612,
707
+ "grad_norm": 0.5025322437286377,
708
  "learning_rate": 0.00013126146788990827,
709
+ "loss": 3.4626,
710
  "step": 5000
711
  },
712
  {
713
  "epoch": 0.4710820895522388,
714
+ "grad_norm": 0.45474347472190857,
715
  "learning_rate": 0.0001301146788990826,
716
+ "loss": 3.4573,
717
  "step": 5050
718
  },
719
  {
720
  "epoch": 0.47574626865671643,
721
+ "grad_norm": 0.45531395077705383,
722
  "learning_rate": 0.00012896788990825688,
723
+ "loss": 3.4622,
724
  "step": 5100
725
  },
726
  {
727
  "epoch": 0.480410447761194,
728
+ "grad_norm": 0.5105506181716919,
729
  "learning_rate": 0.0001278440366972477,
730
+ "loss": 3.4522,
731
  "step": 5150
732
  },
733
  {
734
  "epoch": 0.48507462686567165,
735
+ "grad_norm": 0.41265401244163513,
736
  "learning_rate": 0.00012669724770642202,
737
+ "loss": 3.4378,
738
  "step": 5200
739
  },
740
  {
741
  "epoch": 0.48973880597014924,
742
+ "grad_norm": 0.4541892409324646,
743
  "learning_rate": 0.00012555045871559634,
744
+ "loss": 3.4452,
745
  "step": 5250
746
  },
747
  {
748
  "epoch": 0.4944029850746269,
749
+ "grad_norm": 0.4062674641609192,
750
  "learning_rate": 0.00012440366972477063,
751
+ "loss": 3.4299,
752
  "step": 5300
753
  },
754
  {
755
  "epoch": 0.49906716417910446,
756
+ "grad_norm": 0.39839744567871094,
757
  "learning_rate": 0.00012325688073394498,
758
+ "loss": 3.4231,
759
  "step": 5350
760
  },
761
  {
762
  "epoch": 0.503731343283582,
763
+ "grad_norm": 0.4296305775642395,
764
  "learning_rate": 0.00012211009174311927,
765
+ "loss": 3.4399,
766
  "step": 5400
767
  },
768
  {
769
  "epoch": 0.5083955223880597,
770
+ "grad_norm": 0.40859147906303406,
771
  "learning_rate": 0.0001209633027522936,
772
+ "loss": 3.4226,
773
  "step": 5450
774
  },
775
  {
776
  "epoch": 0.5130597014925373,
777
+ "grad_norm": 0.38735997676849365,
778
  "learning_rate": 0.00011981651376146789,
779
+ "loss": 3.4168,
780
  "step": 5500
781
  },
782
  {
783
  "epoch": 0.5177238805970149,
784
+ "grad_norm": 0.5080564022064209,
785
  "learning_rate": 0.0001186697247706422,
786
+ "loss": 3.4154,
787
  "step": 5550
788
  },
789
  {
790
  "epoch": 0.5223880597014925,
791
+ "grad_norm": 0.42273515462875366,
792
  "learning_rate": 0.00011752293577981653,
793
+ "loss": 3.4155,
794
  "step": 5600
795
  },
796
  {
797
  "epoch": 0.5270522388059702,
798
+ "grad_norm": 0.41893094778060913,
799
  "learning_rate": 0.00011637614678899082,
800
+ "loss": 3.4087,
801
  "step": 5650
802
  },
803
  {
804
  "epoch": 0.5317164179104478,
805
+ "grad_norm": 0.4470965564250946,
806
  "learning_rate": 0.00011522935779816515,
807
+ "loss": 3.3979,
808
  "step": 5700
809
  },
810
  {
811
  "epoch": 0.5363805970149254,
812
+ "grad_norm": 0.40868085622787476,
813
  "learning_rate": 0.00011408256880733946,
814
+ "loss": 3.4007,
815
  "step": 5750
816
  },
817
  {
818
  "epoch": 0.5410447761194029,
819
+ "grad_norm": 0.42737168073654175,
820
  "learning_rate": 0.00011293577981651377,
821
+ "loss": 3.3904,
822
  "step": 5800
823
  },
824
  {
825
  "epoch": 0.5457089552238806,
826
+ "grad_norm": 0.5149052143096924,
827
  "learning_rate": 0.00011178899082568808,
828
+ "loss": 3.3914,
829
  "step": 5850
830
  },
831
  {
832
  "epoch": 0.5503731343283582,
833
+ "grad_norm": 0.4031182825565338,
834
  "learning_rate": 0.00011064220183486238,
835
+ "loss": 3.3845,
836
  "step": 5900
837
  },
838
  {
839
  "epoch": 0.5550373134328358,
840
+ "grad_norm": 0.41353049874305725,
841
  "learning_rate": 0.0001094954128440367,
842
+ "loss": 3.3976,
843
  "step": 5950
844
  },
845
  {
846
  "epoch": 0.5597014925373134,
847
+ "grad_norm": 0.412335604429245,
848
  "learning_rate": 0.000108348623853211,
849
+ "loss": 3.3782,
850
  "step": 6000
851
  },
852
  {
853
  "epoch": 0.5643656716417911,
854
+ "grad_norm": 0.3940559923648834,
855
  "learning_rate": 0.00010720183486238533,
856
+ "loss": 3.3755,
857
  "step": 6050
858
  },
859
  {
860
  "epoch": 0.5690298507462687,
861
+ "grad_norm": 0.3801034092903137,
862
  "learning_rate": 0.00010605504587155963,
863
+ "loss": 3.3668,
864
  "step": 6100
865
  },
866
  {
867
  "epoch": 0.5736940298507462,
868
+ "grad_norm": 0.5560407638549805,
869
  "learning_rate": 0.00010490825688073396,
870
+ "loss": 3.3648,
871
  "step": 6150
872
  },
873
  {
874
  "epoch": 0.5783582089552238,
875
+ "grad_norm": 0.4365626275539398,
876
  "learning_rate": 0.00010376146788990825,
877
+ "loss": 3.374,
878
  "step": 6200
879
  },
880
  {
881
  "epoch": 0.5830223880597015,
882
+ "grad_norm": 0.39503213763237,
883
  "learning_rate": 0.00010261467889908258,
884
+ "loss": 3.3679,
885
  "step": 6250
886
  },
887
  {
888
  "epoch": 0.5876865671641791,
889
+ "grad_norm": 0.41522470116615295,
890
  "learning_rate": 0.00010146788990825689,
891
+ "loss": 3.3532,
892
  "step": 6300
893
  },
894
  {
895
  "epoch": 0.5923507462686567,
896
+ "grad_norm": 0.3890146315097809,
897
  "learning_rate": 0.00010032110091743118,
898
+ "loss": 3.3667,
899
  "step": 6350
900
  },
901
  {
902
  "epoch": 0.5970149253731343,
903
+ "grad_norm": 0.5168399810791016,
904
  "learning_rate": 9.917431192660551e-05,
905
+ "loss": 3.358,
906
  "step": 6400
907
  },
908
  {
909
  "epoch": 0.601679104477612,
910
+ "grad_norm": 0.3710983395576477,
911
  "learning_rate": 9.802752293577983e-05,
912
+ "loss": 3.3585,
913
  "step": 6450
914
  },
915
  {
916
  "epoch": 0.6063432835820896,
917
+ "grad_norm": 0.4420715570449829,
918
  "learning_rate": 9.688073394495414e-05,
919
+ "loss": 3.3539,
920
  "step": 6500
921
  },
922
  {
923
  "epoch": 0.6110074626865671,
924
+ "grad_norm": 0.3815698027610779,
925
+ "learning_rate": 9.573394495412844e-05,
926
+ "loss": 3.3484,
927
  "step": 6550
928
  },
929
  {
930
  "epoch": 0.6156716417910447,
931
+ "grad_norm": 0.38033589720726013,
932
+ "learning_rate": 9.458715596330276e-05,
933
+ "loss": 3.3471,
934
  "step": 6600
935
  },
936
  {
937
  "epoch": 0.6203358208955224,
938
+ "grad_norm": 0.35299167037010193,
939
+ "learning_rate": 9.344036697247706e-05,
940
+ "loss": 3.3555,
941
  "step": 6650
942
  },
943
  {
944
  "epoch": 0.625,
945
+ "grad_norm": 0.3831453025341034,
946
+ "learning_rate": 9.229357798165138e-05,
947
+ "loss": 3.343,
948
  "step": 6700
949
  },
950
  {
951
  "epoch": 0.6296641791044776,
952
+ "grad_norm": 0.38805127143859863,
953
+ "learning_rate": 9.114678899082569e-05,
954
+ "loss": 3.3362,
955
  "step": 6750
956
  },
957
  {
958
  "epoch": 0.6343283582089553,
959
+ "grad_norm": 0.360837459564209,
960
+ "learning_rate": 9e-05,
961
+ "loss": 3.329,
962
  "step": 6800
963
  },
964
  {
965
  "epoch": 0.6389925373134329,
966
+ "grad_norm": 0.48795634508132935,
967
+ "learning_rate": 8.885321100917432e-05,
968
+ "loss": 3.3244,
969
  "step": 6850
970
  },
971
  {
972
  "epoch": 0.6436567164179104,
973
+ "grad_norm": 0.3907620310783386,
974
+ "learning_rate": 8.770642201834863e-05,
975
+ "loss": 3.3326,
976
  "step": 6900
977
  },
978
  {
979
  "epoch": 0.648320895522388,
980
+ "grad_norm": 0.4346226751804352,
981
+ "learning_rate": 8.655963302752293e-05,
982
+ "loss": 3.3291,
983
  "step": 6950
984
  },
985
  {
986
  "epoch": 0.6529850746268657,
987
+ "grad_norm": 0.4240095019340515,
988
+ "learning_rate": 8.541284403669725e-05,
989
+ "loss": 3.3236,
990
  "step": 7000
991
  },
992
  {
993
  "epoch": 0.6576492537313433,
994
+ "grad_norm": 0.4035656452178955,
995
+ "learning_rate": 8.426605504587156e-05,
996
+ "loss": 3.3188,
997
  "step": 7050
998
  },
999
  {
1000
  "epoch": 0.6623134328358209,
1001
+ "grad_norm": 0.44728022813796997,
1002
+ "learning_rate": 8.311926605504588e-05,
1003
+ "loss": 3.3179,
1004
  "step": 7100
1005
  },
1006
  {
1007
  "epoch": 0.6669776119402985,
1008
+ "grad_norm": 0.3958663046360016,
1009
+ "learning_rate": 8.19724770642202e-05,
1010
+ "loss": 3.3081,
1011
  "step": 7150
1012
  },
1013
  {
1014
  "epoch": 0.6716417910447762,
1015
+ "grad_norm": 0.4779495894908905,
1016
+ "learning_rate": 8.08256880733945e-05,
1017
+ "loss": 3.3194,
1018
  "step": 7200
1019
  },
1020
  {
1021
  "epoch": 0.6763059701492538,
1022
+ "grad_norm": 0.3885568082332611,
1023
+ "learning_rate": 7.967889908256882e-05,
1024
+ "loss": 3.3123,
1025
  "step": 7250
1026
  },
1027
  {
1028
  "epoch": 0.6809701492537313,
1029
+ "grad_norm": 0.46929308772087097,
1030
+ "learning_rate": 7.853211009174312e-05,
1031
+ "loss": 3.2946,
1032
  "step": 7300
1033
  },
1034
  {
1035
  "epoch": 0.6856343283582089,
1036
+ "grad_norm": 0.3649180829524994,
1037
+ "learning_rate": 7.738532110091743e-05,
1038
+ "loss": 3.3172,
1039
  "step": 7350
1040
  },
1041
  {
1042
  "epoch": 0.6902985074626866,
1043
+ "grad_norm": 0.3710353672504425,
1044
+ "learning_rate": 7.623853211009174e-05,
1045
+ "loss": 3.3026,
1046
  "step": 7400
1047
  },
1048
  {
1049
  "epoch": 0.6949626865671642,
1050
+ "grad_norm": 0.3968711197376251,
1051
+ "learning_rate": 7.509174311926605e-05,
1052
+ "loss": 3.3026,
1053
  "step": 7450
1054
  },
1055
  {
1056
  "epoch": 0.6996268656716418,
1057
+ "grad_norm": 0.4033445715904236,
1058
+ "learning_rate": 7.394495412844037e-05,
1059
+ "loss": 3.2982,
1060
  "step": 7500
1061
  },
1062
  {
1063
  "epoch": 0.7042910447761194,
1064
+ "grad_norm": 0.4648160934448242,
1065
+ "learning_rate": 7.279816513761469e-05,
1066
+ "loss": 3.2959,
1067
  "step": 7550
1068
  },
1069
  {
1070
  "epoch": 0.7089552238805971,
1071
+ "grad_norm": 0.399700790643692,
1072
+ "learning_rate": 7.165137614678899e-05,
1073
+ "loss": 3.2955,
1074
  "step": 7600
1075
  },
1076
  {
1077
  "epoch": 0.7136194029850746,
1078
+ "grad_norm": 0.43972837924957275,
1079
+ "learning_rate": 7.050458715596331e-05,
1080
+ "loss": 3.2947,
1081
  "step": 7650
1082
  },
1083
  {
1084
  "epoch": 0.7182835820895522,
1085
+ "grad_norm": 0.3754163086414337,
1086
+ "learning_rate": 6.935779816513763e-05,
1087
+ "loss": 3.2901,
1088
  "step": 7700
1089
  },
1090
  {
1091
  "epoch": 0.7229477611940298,
1092
+ "grad_norm": 0.38772958517074585,
1093
+ "learning_rate": 6.821100917431193e-05,
1094
+ "loss": 3.2843,
1095
  "step": 7750
1096
  },
1097
  {
1098
  "epoch": 0.7276119402985075,
1099
+ "grad_norm": 0.3898366689682007,
1100
+ "learning_rate": 6.706422018348624e-05,
1101
+ "loss": 3.2818,
1102
  "step": 7800
1103
  },
1104
  {
1105
  "epoch": 0.7322761194029851,
1106
+ "grad_norm": 0.3694920539855957,
1107
+ "learning_rate": 6.591743119266056e-05,
1108
+ "loss": 3.285,
1109
  "step": 7850
1110
  },
1111
  {
1112
  "epoch": 0.7369402985074627,
1113
+ "grad_norm": 0.3696017861366272,
1114
+ "learning_rate": 6.477064220183486e-05,
1115
+ "loss": 3.282,
1116
  "step": 7900
1117
  },
1118
  {
1119
  "epoch": 0.7416044776119403,
1120
+ "grad_norm": 0.4159897565841675,
1121
+ "learning_rate": 6.362385321100918e-05,
1122
+ "loss": 3.2778,
1123
  "step": 7950
1124
  },
1125
  {
1126
  "epoch": 0.746268656716418,
1127
+ "grad_norm": 0.3910117447376251,
1128
+ "learning_rate": 6.247706422018348e-05,
1129
+ "loss": 3.2852,
1130
  "step": 8000
1131
  },
1132
  {
1133
  "epoch": 0.7509328358208955,
1134
+ "grad_norm": 0.44550907611846924,
1135
+ "learning_rate": 6.13302752293578e-05,
1136
+ "loss": 3.2731,
1137
  "step": 8050
1138
  },
1139
  {
1140
  "epoch": 0.7555970149253731,
1141
+ "grad_norm": 0.4173608422279358,
1142
+ "learning_rate": 6.0183486238532114e-05,
1143
+ "loss": 3.2865,
1144
  "step": 8100
1145
  },
1146
  {
1147
  "epoch": 0.7602611940298507,
1148
+ "grad_norm": 0.3633171021938324,
1149
+ "learning_rate": 5.903669724770643e-05,
1150
+ "loss": 3.2783,
1151
  "step": 8150
1152
  },
1153
  {
1154
  "epoch": 0.7649253731343284,
1155
+ "grad_norm": 0.42995285987854004,
1156
+ "learning_rate": 5.793577981651377e-05,
1157
+ "loss": 3.2732,
1158
  "step": 8200
1159
  },
1160
  {
1161
  "epoch": 0.769589552238806,
1162
+ "grad_norm": 0.3819643557071686,
1163
+ "learning_rate": 5.678899082568807e-05,
1164
+ "loss": 3.2677,
1165
  "step": 8250
1166
  },
1167
  {
1168
  "epoch": 0.7742537313432836,
1169
+ "grad_norm": 0.3480379581451416,
1170
+ "learning_rate": 5.564220183486238e-05,
1171
+ "loss": 3.2749,
1172
  "step": 8300
1173
  },
1174
  {
1175
  "epoch": 0.7789179104477612,
1176
+ "grad_norm": 0.3577636480331421,
1177
+ "learning_rate": 5.44954128440367e-05,
1178
+ "loss": 3.2725,
1179
  "step": 8350
1180
  },
1181
  {
1182
  "epoch": 0.7835820895522388,
1183
+ "grad_norm": 0.36848926544189453,
1184
+ "learning_rate": 5.334862385321101e-05,
1185
+ "loss": 3.2744,
1186
  "step": 8400
1187
  },
1188
  {
1189
  "epoch": 0.7882462686567164,
1190
+ "grad_norm": 0.35368892550468445,
1191
+ "learning_rate": 5.220183486238532e-05,
1192
+ "loss": 3.2616,
1193
  "step": 8450
1194
  },
1195
  {
1196
  "epoch": 0.792910447761194,
1197
+ "grad_norm": 0.41831615567207336,
1198
+ "learning_rate": 5.1055045871559635e-05,
1199
+ "loss": 3.277,
1200
  "step": 8500
1201
  },
1202
  {
1203
  "epoch": 0.7975746268656716,
1204
+ "grad_norm": 0.41986221075057983,
1205
+ "learning_rate": 4.990825688073395e-05,
1206
+ "loss": 3.2522,
1207
  "step": 8550
1208
  },
1209
  {
1210
  "epoch": 0.8022388059701493,
1211
+ "grad_norm": 0.3512188494205475,
1212
+ "learning_rate": 4.876146788990826e-05,
1213
+ "loss": 3.2709,
1214
  "step": 8600
1215
  },
1216
  {
1217
  "epoch": 0.8069029850746269,
1218
+ "grad_norm": 0.3639850318431854,
1219
+ "learning_rate": 4.761467889908257e-05,
1220
+ "loss": 3.2571,
1221
  "step": 8650
1222
  },
1223
  {
1224
  "epoch": 0.8115671641791045,
1225
+ "grad_norm": 0.4546779692173004,
1226
+ "learning_rate": 4.646788990825688e-05,
1227
+ "loss": 3.2599,
1228
  "step": 8700
1229
  },
1230
  {
1231
  "epoch": 0.816231343283582,
1232
+ "grad_norm": 0.35850730538368225,
1233
+ "learning_rate": 4.53211009174312e-05,
1234
+ "loss": 3.2531,
1235
  "step": 8750
1236
  },
1237
  {
1238
  "epoch": 0.8208955223880597,
1239
+ "grad_norm": 0.3843632936477661,
1240
+ "learning_rate": 4.4174311926605504e-05,
1241
+ "loss": 3.2598,
1242
  "step": 8800
1243
  },
1244
  {
1245
  "epoch": 0.8255597014925373,
1246
+ "grad_norm": 0.3926890790462494,
1247
+ "learning_rate": 4.3027522935779816e-05,
1248
+ "loss": 3.2579,
1249
  "step": 8850
1250
  },
1251
  {
1252
  "epoch": 0.8302238805970149,
1253
+ "grad_norm": 0.36206337809562683,
1254
+ "learning_rate": 4.1880733944954134e-05,
1255
+ "loss": 3.2503,
1256
  "step": 8900
1257
  },
1258
  {
1259
  "epoch": 0.8348880597014925,
1260
+ "grad_norm": 0.3471512198448181,
1261
+ "learning_rate": 4.0733944954128446e-05,
1262
+ "loss": 3.2478,
1263
  "step": 8950
1264
  },
1265
  {
1266
  "epoch": 0.8395522388059702,
1267
+ "grad_norm": 0.3711581230163574,
1268
+ "learning_rate": 3.958715596330275e-05,
1269
+ "loss": 3.2461,
1270
  "step": 9000
1271
  }
1272
  ],
 
1287
  "attributes": {}
1288
  }
1289
  },
1290
+ "total_flos": 1912226053095424.0,
1291
  "train_batch_size": 64,
1292
  "trial_name": null,
1293
  "trial_params": null
full/checkpoint-10000/chat_template.jinja ADDED
@@ -0,0 +1 @@
 
 
1
+ {% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% set system_message = false %}{% endif %}{% for message in loop_messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if loop.index0 == 0 and system_message != false %}{% set content = '<<SYS>>\n' + system_message + '\n<</SYS>>\n\n' + message['content'] %}{% else %}{% set content = message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ bos_token + '[INST] ' + content.strip() + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content.strip() + ' ' + eos_token }}{% endif %}{% endfor %}
full/checkpoint-10000/config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LlamaForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 1,
8
+ "eos_token_id": 2,
9
+ "head_dim": 128,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1024,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 2730,
14
+ "max_position_embeddings": 2048,
15
+ "mlp_bias": false,
16
+ "model_type": "llama",
17
+ "num_attention_heads": 8,
18
+ "num_hidden_layers": 16,
19
+ "num_key_value_heads": 8,
20
+ "pretraining_tp": 1,
21
+ "rms_norm_eps": 1e-05,
22
+ "rope_scaling": null,
23
+ "rope_theta": 10000.0,
24
+ "tie_word_embeddings": false,
25
+ "torch_dtype": "float16",
26
+ "transformers_version": "4.55.2",
27
+ "use_cache": false,
28
+ "vocab_size": 32000
29
+ }
full/checkpoint-10000/generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "transformers_version": "4.55.2",
6
+ "use_cache": false
7
+ }
full/checkpoint-10000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30d10fc2f65b25f104d2151e5fb1152390037c7f98e820e1a7917ec2e08ebe0e
3
+ size 1067471056
full/checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ace32f2a668fac0e33c3139e3324a1b087ed7c7d914c1e38edef3c3ce92aa339
3
+ size 1465
full/checkpoint-10000/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": true,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": true,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "</s>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": true,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
full/checkpoint-10000/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
full/checkpoint-10000/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
full/checkpoint-10000/tokenizer_config.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": true,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": true,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": true,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ }
30
+ },
31
+ "bos_token": "<s>",
32
+ "clean_up_tokenization_spaces": false,
33
+ "eos_token": "</s>",
34
+ "extra_special_tokens": {},
35
+ "legacy": true,
36
+ "model_max_length": 2048,
37
+ "pad_token": "</s>",
38
+ "sp_model_kwargs": {},
39
+ "tokenizer_class": "LlamaTokenizer",
40
+ "unk_token": "<unk>",
41
+ "use_default_system_prompt": false
42
+ }
full/checkpoint-10000/trainer_state.json ADDED
@@ -0,0 +1,1434 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.9328358208955224,
6
+ "eval_steps": 500,
7
+ "global_step": 10000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0046641791044776115,
14
+ "grad_norm": 1.907758116722107,
15
+ "learning_rate": 4.9000000000000005e-06,
16
+ "loss": 10.0793,
17
+ "step": 50
18
+ },
19
+ {
20
+ "epoch": 0.009328358208955223,
21
+ "grad_norm": 1.399185061454773,
22
+ "learning_rate": 9.900000000000002e-06,
23
+ "loss": 8.9287,
24
+ "step": 100
25
+ },
26
+ {
27
+ "epoch": 0.013992537313432836,
28
+ "grad_norm": 1.0101089477539062,
29
+ "learning_rate": 1.49e-05,
30
+ "loss": 8.3288,
31
+ "step": 150
32
+ },
33
+ {
34
+ "epoch": 0.018656716417910446,
35
+ "grad_norm": 0.7379341125488281,
36
+ "learning_rate": 1.9900000000000003e-05,
37
+ "loss": 7.6969,
38
+ "step": 200
39
+ },
40
+ {
41
+ "epoch": 0.02332089552238806,
42
+ "grad_norm": 0.9656306505203247,
43
+ "learning_rate": 2.4900000000000002e-05,
44
+ "loss": 7.2698,
45
+ "step": 250
46
+ },
47
+ {
48
+ "epoch": 0.027985074626865673,
49
+ "grad_norm": 1.3790279626846313,
50
+ "learning_rate": 2.9900000000000002e-05,
51
+ "loss": 6.9685,
52
+ "step": 300
53
+ },
54
+ {
55
+ "epoch": 0.03264925373134328,
56
+ "grad_norm": 0.6374709606170654,
57
+ "learning_rate": 3.49e-05,
58
+ "loss": 6.6752,
59
+ "step": 350
60
+ },
61
+ {
62
+ "epoch": 0.03731343283582089,
63
+ "grad_norm": 1.5095465183258057,
64
+ "learning_rate": 3.99e-05,
65
+ "loss": 6.3805,
66
+ "step": 400
67
+ },
68
+ {
69
+ "epoch": 0.04197761194029851,
70
+ "grad_norm": 1.394410490989685,
71
+ "learning_rate": 4.49e-05,
72
+ "loss": 6.1667,
73
+ "step": 450
74
+ },
75
+ {
76
+ "epoch": 0.04664179104477612,
77
+ "grad_norm": 1.7006834745407104,
78
+ "learning_rate": 4.99e-05,
79
+ "loss": 5.9861,
80
+ "step": 500
81
+ },
82
+ {
83
+ "epoch": 0.051305970149253734,
84
+ "grad_norm": 0.9934776425361633,
85
+ "learning_rate": 5.4900000000000006e-05,
86
+ "loss": 5.8166,
87
+ "step": 550
88
+ },
89
+ {
90
+ "epoch": 0.055970149253731345,
91
+ "grad_norm": 2.0550477504730225,
92
+ "learning_rate": 5.99e-05,
93
+ "loss": 5.6646,
94
+ "step": 600
95
+ },
96
+ {
97
+ "epoch": 0.06063432835820896,
98
+ "grad_norm": 2.2504913806915283,
99
+ "learning_rate": 6.49e-05,
100
+ "loss": 5.5353,
101
+ "step": 650
102
+ },
103
+ {
104
+ "epoch": 0.06529850746268656,
105
+ "grad_norm": 2.468212604522705,
106
+ "learning_rate": 6.99e-05,
107
+ "loss": 5.3982,
108
+ "step": 700
109
+ },
110
+ {
111
+ "epoch": 0.06996268656716417,
112
+ "grad_norm": 2.4010939598083496,
113
+ "learning_rate": 7.49e-05,
114
+ "loss": 5.2855,
115
+ "step": 750
116
+ },
117
+ {
118
+ "epoch": 0.07462686567164178,
119
+ "grad_norm": 1.4586199522018433,
120
+ "learning_rate": 7.99e-05,
121
+ "loss": 5.1901,
122
+ "step": 800
123
+ },
124
+ {
125
+ "epoch": 0.07929104477611941,
126
+ "grad_norm": 1.2821402549743652,
127
+ "learning_rate": 8.49e-05,
128
+ "loss": 5.0973,
129
+ "step": 850
130
+ },
131
+ {
132
+ "epoch": 0.08395522388059702,
133
+ "grad_norm": 2.1037180423736572,
134
+ "learning_rate": 8.99e-05,
135
+ "loss": 5.0062,
136
+ "step": 900
137
+ },
138
+ {
139
+ "epoch": 0.08861940298507463,
140
+ "grad_norm": 1.370751976966858,
141
+ "learning_rate": 9.49e-05,
142
+ "loss": 4.9392,
143
+ "step": 950
144
+ },
145
+ {
146
+ "epoch": 0.09328358208955224,
147
+ "grad_norm": 1.276108741760254,
148
+ "learning_rate": 9.99e-05,
149
+ "loss": 4.848,
150
+ "step": 1000
151
+ },
152
+ {
153
+ "epoch": 0.09794776119402986,
154
+ "grad_norm": 1.1077498197555542,
155
+ "learning_rate": 0.0001049,
156
+ "loss": 4.7863,
157
+ "step": 1050
158
+ },
159
+ {
160
+ "epoch": 0.10261194029850747,
161
+ "grad_norm": 1.0212163925170898,
162
+ "learning_rate": 0.0001099,
163
+ "loss": 4.7102,
164
+ "step": 1100
165
+ },
166
+ {
167
+ "epoch": 0.10727611940298508,
168
+ "grad_norm": 1.1480289697647095,
169
+ "learning_rate": 0.0001149,
170
+ "loss": 4.6551,
171
+ "step": 1150
172
+ },
173
+ {
174
+ "epoch": 0.11194029850746269,
175
+ "grad_norm": 1.3106518983840942,
176
+ "learning_rate": 0.00011990000000000001,
177
+ "loss": 4.5627,
178
+ "step": 1200
179
+ },
180
+ {
181
+ "epoch": 0.1166044776119403,
182
+ "grad_norm": 1.4304360151290894,
183
+ "learning_rate": 0.00012490000000000003,
184
+ "loss": 4.4823,
185
+ "step": 1250
186
+ },
187
+ {
188
+ "epoch": 0.12126865671641791,
189
+ "grad_norm": 0.883263885974884,
190
+ "learning_rate": 0.00012989999999999999,
191
+ "loss": 4.4091,
192
+ "step": 1300
193
+ },
194
+ {
195
+ "epoch": 0.1259328358208955,
196
+ "grad_norm": 1.2335679531097412,
197
+ "learning_rate": 0.0001349,
198
+ "loss": 4.3284,
199
+ "step": 1350
200
+ },
201
+ {
202
+ "epoch": 0.13059701492537312,
203
+ "grad_norm": 1.3587566614151,
204
+ "learning_rate": 0.0001399,
205
+ "loss": 4.2523,
206
+ "step": 1400
207
+ },
208
+ {
209
+ "epoch": 0.13526119402985073,
210
+ "grad_norm": 1.4397273063659668,
211
+ "learning_rate": 0.00014490000000000003,
212
+ "loss": 4.1874,
213
+ "step": 1450
214
+ },
215
+ {
216
+ "epoch": 0.13992537313432835,
217
+ "grad_norm": 0.9128643274307251,
218
+ "learning_rate": 0.0001499,
219
+ "loss": 4.1382,
220
+ "step": 1500
221
+ },
222
+ {
223
+ "epoch": 0.14458955223880596,
224
+ "grad_norm": 1.1429646015167236,
225
+ "learning_rate": 0.0001549,
226
+ "loss": 4.0713,
227
+ "step": 1550
228
+ },
229
+ {
230
+ "epoch": 0.14925373134328357,
231
+ "grad_norm": 0.7839804291725159,
232
+ "learning_rate": 0.0001599,
233
+ "loss": 4.0327,
234
+ "step": 1600
235
+ },
236
+ {
237
+ "epoch": 0.15391791044776118,
238
+ "grad_norm": 0.5934556126594543,
239
+ "learning_rate": 0.0001649,
240
+ "loss": 4.0535,
241
+ "step": 1650
242
+ },
243
+ {
244
+ "epoch": 0.15858208955223882,
245
+ "grad_norm": 0.6320812702178955,
246
+ "learning_rate": 0.0001699,
247
+ "loss": 3.9615,
248
+ "step": 1700
249
+ },
250
+ {
251
+ "epoch": 0.16324626865671643,
252
+ "grad_norm": 0.6207342743873596,
253
+ "learning_rate": 0.00017490000000000002,
254
+ "loss": 3.9086,
255
+ "step": 1750
256
+ },
257
+ {
258
+ "epoch": 0.16791044776119404,
259
+ "grad_norm": 0.8524695634841919,
260
+ "learning_rate": 0.0001799,
261
+ "loss": 3.8703,
262
+ "step": 1800
263
+ },
264
+ {
265
+ "epoch": 0.17257462686567165,
266
+ "grad_norm": 0.9148792028427124,
267
+ "learning_rate": 0.0001849,
268
+ "loss": 3.8388,
269
+ "step": 1850
270
+ },
271
+ {
272
+ "epoch": 0.17723880597014927,
273
+ "grad_norm": 0.7455419301986694,
274
+ "learning_rate": 0.0001899,
275
+ "loss": 3.8031,
276
+ "step": 1900
277
+ },
278
+ {
279
+ "epoch": 0.18190298507462688,
280
+ "grad_norm": 0.9777323603630066,
281
+ "learning_rate": 0.00019490000000000002,
282
+ "loss": 3.781,
283
+ "step": 1950
284
+ },
285
+ {
286
+ "epoch": 0.1865671641791045,
287
+ "grad_norm": 0.6018832921981812,
288
+ "learning_rate": 0.0001999,
289
+ "loss": 3.7421,
290
+ "step": 2000
291
+ },
292
+ {
293
+ "epoch": 0.1912313432835821,
294
+ "grad_norm": 0.7007433772087097,
295
+ "learning_rate": 0.00019887614678899083,
296
+ "loss": 3.7148,
297
+ "step": 2050
298
+ },
299
+ {
300
+ "epoch": 0.1958955223880597,
301
+ "grad_norm": 0.7864815592765808,
302
+ "learning_rate": 0.00019772935779816515,
303
+ "loss": 3.6907,
304
+ "step": 2100
305
+ },
306
+ {
307
+ "epoch": 0.20055970149253732,
308
+ "grad_norm": 0.6575994491577148,
309
+ "learning_rate": 0.00019658256880733944,
310
+ "loss": 3.6504,
311
+ "step": 2150
312
+ },
313
+ {
314
+ "epoch": 0.20522388059701493,
315
+ "grad_norm": 0.524401843547821,
316
+ "learning_rate": 0.0001954357798165138,
317
+ "loss": 3.6333,
318
+ "step": 2200
319
+ },
320
+ {
321
+ "epoch": 0.20988805970149255,
322
+ "grad_norm": 0.8242985010147095,
323
+ "learning_rate": 0.00019428899082568808,
324
+ "loss": 3.6084,
325
+ "step": 2250
326
+ },
327
+ {
328
+ "epoch": 0.21455223880597016,
329
+ "grad_norm": 0.5965232849121094,
330
+ "learning_rate": 0.0001931422018348624,
331
+ "loss": 3.5914,
332
+ "step": 2300
333
+ },
334
+ {
335
+ "epoch": 0.21921641791044777,
336
+ "grad_norm": 0.5473845601081848,
337
+ "learning_rate": 0.00019199541284403671,
338
+ "loss": 3.572,
339
+ "step": 2350
340
+ },
341
+ {
342
+ "epoch": 0.22388059701492538,
343
+ "grad_norm": 0.7030166983604431,
344
+ "learning_rate": 0.00019084862385321103,
345
+ "loss": 3.5548,
346
+ "step": 2400
347
+ },
348
+ {
349
+ "epoch": 0.228544776119403,
350
+ "grad_norm": 0.5844017267227173,
351
+ "learning_rate": 0.00018972477064220186,
352
+ "loss": 3.5326,
353
+ "step": 2450
354
+ },
355
+ {
356
+ "epoch": 0.2332089552238806,
357
+ "grad_norm": 0.4722403883934021,
358
+ "learning_rate": 0.00018857798165137615,
359
+ "loss": 3.5113,
360
+ "step": 2500
361
+ },
362
+ {
363
+ "epoch": 0.23787313432835822,
364
+ "grad_norm": 0.4691997468471527,
365
+ "learning_rate": 0.00018743119266055047,
366
+ "loss": 3.4985,
367
+ "step": 2550
368
+ },
369
+ {
370
+ "epoch": 0.24253731343283583,
371
+ "grad_norm": 0.5742476582527161,
372
+ "learning_rate": 0.00018628440366972478,
373
+ "loss": 3.4689,
374
+ "step": 2600
375
+ },
376
+ {
377
+ "epoch": 0.24720149253731344,
378
+ "grad_norm": 0.4867892563343048,
379
+ "learning_rate": 0.0001851376146788991,
380
+ "loss": 3.4737,
381
+ "step": 2650
382
+ },
383
+ {
384
+ "epoch": 0.251865671641791,
385
+ "grad_norm": 0.5876694321632385,
386
+ "learning_rate": 0.0001839908256880734,
387
+ "loss": 3.4516,
388
+ "step": 2700
389
+ },
390
+ {
391
+ "epoch": 0.25652985074626866,
392
+ "grad_norm": 0.4691877067089081,
393
+ "learning_rate": 0.0001828440366972477,
394
+ "loss": 3.4411,
395
+ "step": 2750
396
+ },
397
+ {
398
+ "epoch": 0.26119402985074625,
399
+ "grad_norm": 0.497503399848938,
400
+ "learning_rate": 0.00018169724770642203,
401
+ "loss": 3.4311,
402
+ "step": 2800
403
+ },
404
+ {
405
+ "epoch": 0.2658582089552239,
406
+ "grad_norm": 0.4610927700996399,
407
+ "learning_rate": 0.00018055045871559635,
408
+ "loss": 3.4122,
409
+ "step": 2850
410
+ },
411
+ {
412
+ "epoch": 0.27052238805970147,
413
+ "grad_norm": 0.5312972664833069,
414
+ "learning_rate": 0.00017940366972477064,
415
+ "loss": 3.4004,
416
+ "step": 2900
417
+ },
418
+ {
419
+ "epoch": 0.2751865671641791,
420
+ "grad_norm": 0.5256180167198181,
421
+ "learning_rate": 0.00017825688073394496,
422
+ "loss": 3.3834,
423
+ "step": 2950
424
+ },
425
+ {
426
+ "epoch": 0.2798507462686567,
427
+ "grad_norm": 0.5108594298362732,
428
+ "learning_rate": 0.00017711009174311928,
429
+ "loss": 3.379,
430
+ "step": 3000
431
+ },
432
+ {
433
+ "epoch": 0.28451492537313433,
434
+ "grad_norm": 0.4818528890609741,
435
+ "learning_rate": 0.00017596330275229357,
436
+ "loss": 3.3645,
437
+ "step": 3050
438
+ },
439
+ {
440
+ "epoch": 0.2891791044776119,
441
+ "grad_norm": 0.4077577590942383,
442
+ "learning_rate": 0.0001748165137614679,
443
+ "loss": 3.3659,
444
+ "step": 3100
445
+ },
446
+ {
447
+ "epoch": 0.29384328358208955,
448
+ "grad_norm": 0.4791859984397888,
449
+ "learning_rate": 0.0001736697247706422,
450
+ "loss": 3.3427,
451
+ "step": 3150
452
+ },
453
+ {
454
+ "epoch": 0.29850746268656714,
455
+ "grad_norm": 0.6330239176750183,
456
+ "learning_rate": 0.00017252293577981652,
457
+ "loss": 3.3484,
458
+ "step": 3200
459
+ },
460
+ {
461
+ "epoch": 0.3031716417910448,
462
+ "grad_norm": 0.4511379897594452,
463
+ "learning_rate": 0.00017137614678899084,
464
+ "loss": 3.3302,
465
+ "step": 3250
466
+ },
467
+ {
468
+ "epoch": 0.30783582089552236,
469
+ "grad_norm": 0.4571825861930847,
470
+ "learning_rate": 0.00017022935779816516,
471
+ "loss": 3.3144,
472
+ "step": 3300
473
+ },
474
+ {
475
+ "epoch": 0.3125,
476
+ "grad_norm": 0.44557932019233704,
477
+ "learning_rate": 0.00016908256880733945,
478
+ "loss": 3.3071,
479
+ "step": 3350
480
+ },
481
+ {
482
+ "epoch": 0.31716417910447764,
483
+ "grad_norm": 0.42434263229370117,
484
+ "learning_rate": 0.00016793577981651377,
485
+ "loss": 3.2986,
486
+ "step": 3400
487
+ },
488
+ {
489
+ "epoch": 0.3218283582089552,
490
+ "grad_norm": 0.41813352704048157,
491
+ "learning_rate": 0.0001667889908256881,
492
+ "loss": 3.2902,
493
+ "step": 3450
494
+ },
495
+ {
496
+ "epoch": 0.32649253731343286,
497
+ "grad_norm": 0.4084812104701996,
498
+ "learning_rate": 0.00016564220183486238,
499
+ "loss": 3.283,
500
+ "step": 3500
501
+ },
502
+ {
503
+ "epoch": 0.33115671641791045,
504
+ "grad_norm": 0.38210785388946533,
505
+ "learning_rate": 0.00016449541284403672,
506
+ "loss": 3.2769,
507
+ "step": 3550
508
+ },
509
+ {
510
+ "epoch": 0.3358208955223881,
511
+ "grad_norm": 0.5375980734825134,
512
+ "learning_rate": 0.00016334862385321102,
513
+ "loss": 3.2697,
514
+ "step": 3600
515
+ },
516
+ {
517
+ "epoch": 0.34048507462686567,
518
+ "grad_norm": 0.38781264424324036,
519
+ "learning_rate": 0.00016220183486238533,
520
+ "loss": 3.2556,
521
+ "step": 3650
522
+ },
523
+ {
524
+ "epoch": 0.3451492537313433,
525
+ "grad_norm": 0.41560590267181396,
526
+ "learning_rate": 0.00016105504587155965,
527
+ "loss": 3.2565,
528
+ "step": 3700
529
+ },
530
+ {
531
+ "epoch": 0.3498134328358209,
532
+ "grad_norm": 0.4119664132595062,
533
+ "learning_rate": 0.00015990825688073394,
534
+ "loss": 3.2386,
535
+ "step": 3750
536
+ },
537
+ {
538
+ "epoch": 0.35447761194029853,
539
+ "grad_norm": 0.4190433919429779,
540
+ "learning_rate": 0.00015876146788990826,
541
+ "loss": 3.2345,
542
+ "step": 3800
543
+ },
544
+ {
545
+ "epoch": 0.3591417910447761,
546
+ "grad_norm": 0.37028583884239197,
547
+ "learning_rate": 0.00015761467889908258,
548
+ "loss": 3.226,
549
+ "step": 3850
550
+ },
551
+ {
552
+ "epoch": 0.36380597014925375,
553
+ "grad_norm": 0.3971349000930786,
554
+ "learning_rate": 0.0001564678899082569,
555
+ "loss": 3.2213,
556
+ "step": 3900
557
+ },
558
+ {
559
+ "epoch": 0.36847014925373134,
560
+ "grad_norm": 0.34641316533088684,
561
+ "learning_rate": 0.0001553211009174312,
562
+ "loss": 3.2144,
563
+ "step": 3950
564
+ },
565
+ {
566
+ "epoch": 0.373134328358209,
567
+ "grad_norm": 0.40990573167800903,
568
+ "learning_rate": 0.0001541743119266055,
569
+ "loss": 3.2141,
570
+ "step": 4000
571
+ },
572
+ {
573
+ "epoch": 0.37779850746268656,
574
+ "grad_norm": 0.35828983783721924,
575
+ "learning_rate": 0.00015302752293577983,
576
+ "loss": 3.2054,
577
+ "step": 4050
578
+ },
579
+ {
580
+ "epoch": 0.3824626865671642,
581
+ "grad_norm": 0.37572401762008667,
582
+ "learning_rate": 0.00015188073394495414,
583
+ "loss": 3.1943,
584
+ "step": 4100
585
+ },
586
+ {
587
+ "epoch": 0.3871268656716418,
588
+ "grad_norm": 0.3795240521430969,
589
+ "learning_rate": 0.00015073394495412844,
590
+ "loss": 3.1971,
591
+ "step": 4150
592
+ },
593
+ {
594
+ "epoch": 0.3917910447761194,
595
+ "grad_norm": 0.3884343206882477,
596
+ "learning_rate": 0.00014958715596330275,
597
+ "loss": 3.1877,
598
+ "step": 4200
599
+ },
600
+ {
601
+ "epoch": 0.396455223880597,
602
+ "grad_norm": 0.374661922454834,
603
+ "learning_rate": 0.00014844036697247707,
604
+ "loss": 3.174,
605
+ "step": 4250
606
+ },
607
+ {
608
+ "epoch": 0.40111940298507465,
609
+ "grad_norm": 0.4330432415008545,
610
+ "learning_rate": 0.00014729357798165136,
611
+ "loss": 3.1787,
612
+ "step": 4300
613
+ },
614
+ {
615
+ "epoch": 0.40578358208955223,
616
+ "grad_norm": 0.40475285053253174,
617
+ "learning_rate": 0.0001461467889908257,
618
+ "loss": 3.1781,
619
+ "step": 4350
620
+ },
621
+ {
622
+ "epoch": 0.41044776119402987,
623
+ "grad_norm": 0.36028817296028137,
624
+ "learning_rate": 0.000145,
625
+ "loss": 3.1653,
626
+ "step": 4400
627
+ },
628
+ {
629
+ "epoch": 0.41511194029850745,
630
+ "grad_norm": 0.37242797017097473,
631
+ "learning_rate": 0.00014385321100917432,
632
+ "loss": 3.1608,
633
+ "step": 4450
634
+ },
635
+ {
636
+ "epoch": 0.4197761194029851,
637
+ "grad_norm": 0.42880979180336,
638
+ "learning_rate": 0.00014270642201834864,
639
+ "loss": 3.1656,
640
+ "step": 4500
641
+ },
642
+ {
643
+ "epoch": 0.4244402985074627,
644
+ "grad_norm": 0.3562730848789215,
645
+ "learning_rate": 0.00014155963302752296,
646
+ "loss": 3.1409,
647
+ "step": 4550
648
+ },
649
+ {
650
+ "epoch": 0.4291044776119403,
651
+ "grad_norm": 0.3420323133468628,
652
+ "learning_rate": 0.00014041284403669725,
653
+ "loss": 3.1502,
654
+ "step": 4600
655
+ },
656
+ {
657
+ "epoch": 0.4337686567164179,
658
+ "grad_norm": 0.33985841274261475,
659
+ "learning_rate": 0.00013926605504587157,
660
+ "loss": 3.1437,
661
+ "step": 4650
662
+ },
663
+ {
664
+ "epoch": 0.43843283582089554,
665
+ "grad_norm": 0.35252872109413147,
666
+ "learning_rate": 0.00013811926605504588,
667
+ "loss": 3.1375,
668
+ "step": 4700
669
+ },
670
+ {
671
+ "epoch": 0.4430970149253731,
672
+ "grad_norm": 0.33299019932746887,
673
+ "learning_rate": 0.00013697247706422018,
674
+ "loss": 3.1207,
675
+ "step": 4750
676
+ },
677
+ {
678
+ "epoch": 0.44776119402985076,
679
+ "grad_norm": 0.3340663015842438,
680
+ "learning_rate": 0.00013582568807339452,
681
+ "loss": 3.1285,
682
+ "step": 4800
683
+ },
684
+ {
685
+ "epoch": 0.45242537313432835,
686
+ "grad_norm": 0.3459039330482483,
687
+ "learning_rate": 0.00013470183486238532,
688
+ "loss": 3.1187,
689
+ "step": 4850
690
+ },
691
+ {
692
+ "epoch": 0.457089552238806,
693
+ "grad_norm": 0.34190812706947327,
694
+ "learning_rate": 0.00013355504587155963,
695
+ "loss": 3.1128,
696
+ "step": 4900
697
+ },
698
+ {
699
+ "epoch": 0.46175373134328357,
700
+ "grad_norm": 0.38482001423835754,
701
+ "learning_rate": 0.00013240825688073395,
702
+ "loss": 3.1081,
703
+ "step": 4950
704
+ },
705
+ {
706
+ "epoch": 0.4664179104477612,
707
+ "grad_norm": 0.3969718813896179,
708
+ "learning_rate": 0.00013126146788990827,
709
+ "loss": 3.111,
710
+ "step": 5000
711
+ },
712
+ {
713
+ "epoch": 0.4710820895522388,
714
+ "grad_norm": 0.30978360772132874,
715
+ "learning_rate": 0.0001301146788990826,
716
+ "loss": 3.106,
717
+ "step": 5050
718
+ },
719
+ {
720
+ "epoch": 0.47574626865671643,
721
+ "grad_norm": 0.32029977440834045,
722
+ "learning_rate": 0.00012896788990825688,
723
+ "loss": 3.1095,
724
+ "step": 5100
725
+ },
726
+ {
727
+ "epoch": 0.480410447761194,
728
+ "grad_norm": 0.36698678135871887,
729
+ "learning_rate": 0.0001278440366972477,
730
+ "loss": 3.1017,
731
+ "step": 5150
732
+ },
733
+ {
734
+ "epoch": 0.48507462686567165,
735
+ "grad_norm": 0.3710266947746277,
736
+ "learning_rate": 0.00012669724770642202,
737
+ "loss": 3.0863,
738
+ "step": 5200
739
+ },
740
+ {
741
+ "epoch": 0.48973880597014924,
742
+ "grad_norm": 0.33469706773757935,
743
+ "learning_rate": 0.00012555045871559634,
744
+ "loss": 3.0949,
745
+ "step": 5250
746
+ },
747
+ {
748
+ "epoch": 0.4944029850746269,
749
+ "grad_norm": 0.33493635058403015,
750
+ "learning_rate": 0.00012440366972477063,
751
+ "loss": 3.0794,
752
+ "step": 5300
753
+ },
754
+ {
755
+ "epoch": 0.49906716417910446,
756
+ "grad_norm": 0.3489319682121277,
757
+ "learning_rate": 0.00012325688073394498,
758
+ "loss": 3.0752,
759
+ "step": 5350
760
+ },
761
+ {
762
+ "epoch": 0.503731343283582,
763
+ "grad_norm": 0.32940617203712463,
764
+ "learning_rate": 0.00012211009174311927,
765
+ "loss": 3.0911,
766
+ "step": 5400
767
+ },
768
+ {
769
+ "epoch": 0.5083955223880597,
770
+ "grad_norm": 0.3039519190788269,
771
+ "learning_rate": 0.0001209633027522936,
772
+ "loss": 3.0749,
773
+ "step": 5450
774
+ },
775
+ {
776
+ "epoch": 0.5130597014925373,
777
+ "grad_norm": 0.3196834325790405,
778
+ "learning_rate": 0.00011981651376146789,
779
+ "loss": 3.0685,
780
+ "step": 5500
781
+ },
782
+ {
783
+ "epoch": 0.5177238805970149,
784
+ "grad_norm": 0.35171428322792053,
785
+ "learning_rate": 0.0001186697247706422,
786
+ "loss": 3.0684,
787
+ "step": 5550
788
+ },
789
+ {
790
+ "epoch": 0.5223880597014925,
791
+ "grad_norm": 0.3927328884601593,
792
+ "learning_rate": 0.00011752293577981653,
793
+ "loss": 3.0676,
794
+ "step": 5600
795
+ },
796
+ {
797
+ "epoch": 0.5270522388059702,
798
+ "grad_norm": 0.32428842782974243,
799
+ "learning_rate": 0.00011637614678899082,
800
+ "loss": 3.0635,
801
+ "step": 5650
802
+ },
803
+ {
804
+ "epoch": 0.5317164179104478,
805
+ "grad_norm": 0.3297617435455322,
806
+ "learning_rate": 0.00011522935779816515,
807
+ "loss": 3.0525,
808
+ "step": 5700
809
+ },
810
+ {
811
+ "epoch": 0.5363805970149254,
812
+ "grad_norm": 0.3118542432785034,
813
+ "learning_rate": 0.00011408256880733946,
814
+ "loss": 3.0566,
815
+ "step": 5750
816
+ },
817
+ {
818
+ "epoch": 0.5410447761194029,
819
+ "grad_norm": 0.29949086904525757,
820
+ "learning_rate": 0.00011293577981651377,
821
+ "loss": 3.0454,
822
+ "step": 5800
823
+ },
824
+ {
825
+ "epoch": 0.5457089552238806,
826
+ "grad_norm": 0.3567812740802765,
827
+ "learning_rate": 0.00011178899082568808,
828
+ "loss": 3.0462,
829
+ "step": 5850
830
+ },
831
+ {
832
+ "epoch": 0.5503731343283582,
833
+ "grad_norm": 0.3179719150066376,
834
+ "learning_rate": 0.00011064220183486238,
835
+ "loss": 3.0409,
836
+ "step": 5900
837
+ },
838
+ {
839
+ "epoch": 0.5550373134328358,
840
+ "grad_norm": 0.32883220911026,
841
+ "learning_rate": 0.0001094954128440367,
842
+ "loss": 3.0527,
843
+ "step": 5950
844
+ },
845
+ {
846
+ "epoch": 0.5597014925373134,
847
+ "grad_norm": 0.31516268849372864,
848
+ "learning_rate": 0.000108348623853211,
849
+ "loss": 3.0359,
850
+ "step": 6000
851
+ },
852
+ {
853
+ "epoch": 0.5643656716417911,
854
+ "grad_norm": 0.298753559589386,
855
+ "learning_rate": 0.00010720183486238533,
856
+ "loss": 3.0343,
857
+ "step": 6050
858
+ },
859
+ {
860
+ "epoch": 0.5690298507462687,
861
+ "grad_norm": 0.32886001467704773,
862
+ "learning_rate": 0.00010605504587155963,
863
+ "loss": 3.0229,
864
+ "step": 6100
865
+ },
866
+ {
867
+ "epoch": 0.5736940298507462,
868
+ "grad_norm": 0.3526987135410309,
869
+ "learning_rate": 0.00010490825688073396,
870
+ "loss": 3.0227,
871
+ "step": 6150
872
+ },
873
+ {
874
+ "epoch": 0.5783582089552238,
875
+ "grad_norm": 0.31915777921676636,
876
+ "learning_rate": 0.00010376146788990825,
877
+ "loss": 3.0311,
878
+ "step": 6200
879
+ },
880
+ {
881
+ "epoch": 0.5830223880597015,
882
+ "grad_norm": 0.31342729926109314,
883
+ "learning_rate": 0.00010261467889908258,
884
+ "loss": 3.0257,
885
+ "step": 6250
886
+ },
887
+ {
888
+ "epoch": 0.5876865671641791,
889
+ "grad_norm": 0.3711305856704712,
890
+ "learning_rate": 0.00010146788990825689,
891
+ "loss": 3.0125,
892
+ "step": 6300
893
+ },
894
+ {
895
+ "epoch": 0.5923507462686567,
896
+ "grad_norm": 0.32861173152923584,
897
+ "learning_rate": 0.00010032110091743118,
898
+ "loss": 3.0235,
899
+ "step": 6350
900
+ },
901
+ {
902
+ "epoch": 0.5970149253731343,
903
+ "grad_norm": 0.30850905179977417,
904
+ "learning_rate": 9.917431192660551e-05,
905
+ "loss": 3.0171,
906
+ "step": 6400
907
+ },
908
+ {
909
+ "epoch": 0.601679104477612,
910
+ "grad_norm": 0.31126561760902405,
911
+ "learning_rate": 9.802752293577983e-05,
912
+ "loss": 3.0177,
913
+ "step": 6450
914
+ },
915
+ {
916
+ "epoch": 0.6063432835820896,
917
+ "grad_norm": 0.3373335897922516,
918
+ "learning_rate": 9.688073394495414e-05,
919
+ "loss": 3.0138,
920
+ "step": 6500
921
+ },
922
+ {
923
+ "epoch": 0.6110074626865671,
924
+ "grad_norm": 0.32974258065223694,
925
+ "learning_rate": 9.575688073394496e-05,
926
+ "loss": 3.0092,
927
+ "step": 6550
928
+ },
929
+ {
930
+ "epoch": 0.6156716417910447,
931
+ "grad_norm": 0.29914727807044983,
932
+ "learning_rate": 9.463302752293578e-05,
933
+ "loss": 3.0075,
934
+ "step": 6600
935
+ },
936
+ {
937
+ "epoch": 0.6203358208955224,
938
+ "grad_norm": 0.3006197214126587,
939
+ "learning_rate": 9.34862385321101e-05,
940
+ "loss": 3.015,
941
+ "step": 6650
942
+ },
943
+ {
944
+ "epoch": 0.625,
945
+ "grad_norm": 0.2997278869152069,
946
+ "learning_rate": 9.23394495412844e-05,
947
+ "loss": 3.0037,
948
+ "step": 6700
949
+ },
950
+ {
951
+ "epoch": 0.6296641791044776,
952
+ "grad_norm": 0.31719309091567993,
953
+ "learning_rate": 9.119266055045872e-05,
954
+ "loss": 2.9969,
955
+ "step": 6750
956
+ },
957
+ {
958
+ "epoch": 0.6343283582089553,
959
+ "grad_norm": 0.3004010319709778,
960
+ "learning_rate": 9.004587155963303e-05,
961
+ "loss": 2.9909,
962
+ "step": 6800
963
+ },
964
+ {
965
+ "epoch": 0.6389925373134329,
966
+ "grad_norm": 0.3123314082622528,
967
+ "learning_rate": 8.889908256880735e-05,
968
+ "loss": 2.9853,
969
+ "step": 6850
970
+ },
971
+ {
972
+ "epoch": 0.6436567164179104,
973
+ "grad_norm": 0.334837406873703,
974
+ "learning_rate": 8.775229357798166e-05,
975
+ "loss": 2.9923,
976
+ "step": 6900
977
+ },
978
+ {
979
+ "epoch": 0.648320895522388,
980
+ "grad_norm": 0.3055480420589447,
981
+ "learning_rate": 8.660550458715597e-05,
982
+ "loss": 2.9897,
983
+ "step": 6950
984
+ },
985
+ {
986
+ "epoch": 0.6529850746268657,
987
+ "grad_norm": 0.33151915669441223,
988
+ "learning_rate": 8.545871559633027e-05,
989
+ "loss": 2.9843,
990
+ "step": 7000
991
+ },
992
+ {
993
+ "epoch": 0.6576492537313433,
994
+ "grad_norm": 0.32511621713638306,
995
+ "learning_rate": 8.431192660550459e-05,
996
+ "loss": 2.9797,
997
+ "step": 7050
998
+ },
999
+ {
1000
+ "epoch": 0.6623134328358209,
1001
+ "grad_norm": 0.29789265990257263,
1002
+ "learning_rate": 8.31651376146789e-05,
1003
+ "loss": 2.9789,
1004
+ "step": 7100
1005
+ },
1006
+ {
1007
+ "epoch": 0.6669776119402985,
1008
+ "grad_norm": 0.3272823393344879,
1009
+ "learning_rate": 8.201834862385321e-05,
1010
+ "loss": 2.971,
1011
+ "step": 7150
1012
+ },
1013
+ {
1014
+ "epoch": 0.6716417910447762,
1015
+ "grad_norm": 0.2983226478099823,
1016
+ "learning_rate": 8.087155963302753e-05,
1017
+ "loss": 2.9818,
1018
+ "step": 7200
1019
+ },
1020
+ {
1021
+ "epoch": 0.6763059701492538,
1022
+ "grad_norm": 0.30532923340797424,
1023
+ "learning_rate": 7.972477064220184e-05,
1024
+ "loss": 2.9739,
1025
+ "step": 7250
1026
+ },
1027
+ {
1028
+ "epoch": 0.6809701492537313,
1029
+ "grad_norm": 0.3046092092990875,
1030
+ "learning_rate": 7.857798165137616e-05,
1031
+ "loss": 2.9575,
1032
+ "step": 7300
1033
+ },
1034
+ {
1035
+ "epoch": 0.6856343283582089,
1036
+ "grad_norm": 0.3133400082588196,
1037
+ "learning_rate": 7.743119266055046e-05,
1038
+ "loss": 2.9794,
1039
+ "step": 7350
1040
+ },
1041
+ {
1042
+ "epoch": 0.6902985074626866,
1043
+ "grad_norm": 0.3090495765209198,
1044
+ "learning_rate": 7.628440366972477e-05,
1045
+ "loss": 2.9632,
1046
+ "step": 7400
1047
+ },
1048
+ {
1049
+ "epoch": 0.6949626865671642,
1050
+ "grad_norm": 0.2925854027271271,
1051
+ "learning_rate": 7.513761467889908e-05,
1052
+ "loss": 2.9649,
1053
+ "step": 7450
1054
+ },
1055
+ {
1056
+ "epoch": 0.6996268656716418,
1057
+ "grad_norm": 0.30464622378349304,
1058
+ "learning_rate": 7.399082568807339e-05,
1059
+ "loss": 2.9606,
1060
+ "step": 7500
1061
+ },
1062
+ {
1063
+ "epoch": 0.7042910447761194,
1064
+ "grad_norm": 0.29478827118873596,
1065
+ "learning_rate": 7.284403669724771e-05,
1066
+ "loss": 2.9574,
1067
+ "step": 7550
1068
+ },
1069
+ {
1070
+ "epoch": 0.7089552238805971,
1071
+ "grad_norm": 0.30368906259536743,
1072
+ "learning_rate": 7.169724770642203e-05,
1073
+ "loss": 2.9583,
1074
+ "step": 7600
1075
+ },
1076
+ {
1077
+ "epoch": 0.7136194029850746,
1078
+ "grad_norm": 0.2974824905395508,
1079
+ "learning_rate": 7.055045871559633e-05,
1080
+ "loss": 2.9566,
1081
+ "step": 7650
1082
+ },
1083
+ {
1084
+ "epoch": 0.7182835820895522,
1085
+ "grad_norm": 0.3006385564804077,
1086
+ "learning_rate": 6.940366972477065e-05,
1087
+ "loss": 2.953,
1088
+ "step": 7700
1089
+ },
1090
+ {
1091
+ "epoch": 0.7229477611940298,
1092
+ "grad_norm": 0.2787078022956848,
1093
+ "learning_rate": 6.825688073394497e-05,
1094
+ "loss": 2.9466,
1095
+ "step": 7750
1096
+ },
1097
+ {
1098
+ "epoch": 0.7276119402985075,
1099
+ "grad_norm": 0.2966342568397522,
1100
+ "learning_rate": 6.711009174311927e-05,
1101
+ "loss": 2.9446,
1102
+ "step": 7800
1103
+ },
1104
+ {
1105
+ "epoch": 0.7322761194029851,
1106
+ "grad_norm": 0.32728853821754456,
1107
+ "learning_rate": 6.600917431192661e-05,
1108
+ "loss": 2.9486,
1109
+ "step": 7850
1110
+ },
1111
+ {
1112
+ "epoch": 0.7369402985074627,
1113
+ "grad_norm": 0.325846403837204,
1114
+ "learning_rate": 6.486238532110092e-05,
1115
+ "loss": 2.9456,
1116
+ "step": 7900
1117
+ },
1118
+ {
1119
+ "epoch": 0.7416044776119403,
1120
+ "grad_norm": 0.2933451235294342,
1121
+ "learning_rate": 6.371559633027524e-05,
1122
+ "loss": 2.9404,
1123
+ "step": 7950
1124
+ },
1125
+ {
1126
+ "epoch": 0.746268656716418,
1127
+ "grad_norm": 0.3031325042247772,
1128
+ "learning_rate": 6.256880733944954e-05,
1129
+ "loss": 2.9486,
1130
+ "step": 8000
1131
+ },
1132
+ {
1133
+ "epoch": 0.7509328358208955,
1134
+ "grad_norm": 0.30206266045570374,
1135
+ "learning_rate": 6.142201834862386e-05,
1136
+ "loss": 2.9364,
1137
+ "step": 8050
1138
+ },
1139
+ {
1140
+ "epoch": 0.7555970149253731,
1141
+ "grad_norm": 0.29706859588623047,
1142
+ "learning_rate": 6.027522935779817e-05,
1143
+ "loss": 2.9483,
1144
+ "step": 8100
1145
+ },
1146
+ {
1147
+ "epoch": 0.7602611940298507,
1148
+ "grad_norm": 0.2909660339355469,
1149
+ "learning_rate": 5.912844036697248e-05,
1150
+ "loss": 2.9405,
1151
+ "step": 8150
1152
+ },
1153
+ {
1154
+ "epoch": 0.7649253731343284,
1155
+ "grad_norm": 0.3136695623397827,
1156
+ "learning_rate": 5.798165137614679e-05,
1157
+ "loss": 2.9366,
1158
+ "step": 8200
1159
+ },
1160
+ {
1161
+ "epoch": 0.769589552238806,
1162
+ "grad_norm": 0.29548221826553345,
1163
+ "learning_rate": 5.683486238532111e-05,
1164
+ "loss": 2.9305,
1165
+ "step": 8250
1166
+ },
1167
+ {
1168
+ "epoch": 0.7742537313432836,
1169
+ "grad_norm": 0.2777513265609741,
1170
+ "learning_rate": 5.568807339449541e-05,
1171
+ "loss": 2.9377,
1172
+ "step": 8300
1173
+ },
1174
+ {
1175
+ "epoch": 0.7789179104477612,
1176
+ "grad_norm": 0.28673887252807617,
1177
+ "learning_rate": 5.454128440366972e-05,
1178
+ "loss": 2.9342,
1179
+ "step": 8350
1180
+ },
1181
+ {
1182
+ "epoch": 0.7835820895522388,
1183
+ "grad_norm": 0.2911510169506073,
1184
+ "learning_rate": 5.339449541284404e-05,
1185
+ "loss": 2.938,
1186
+ "step": 8400
1187
+ },
1188
+ {
1189
+ "epoch": 0.7882462686567164,
1190
+ "grad_norm": 0.3052498698234558,
1191
+ "learning_rate": 5.224770642201835e-05,
1192
+ "loss": 2.9235,
1193
+ "step": 8450
1194
+ },
1195
+ {
1196
+ "epoch": 0.792910447761194,
1197
+ "grad_norm": 0.30528631806373596,
1198
+ "learning_rate": 5.110091743119266e-05,
1199
+ "loss": 2.9393,
1200
+ "step": 8500
1201
+ },
1202
+ {
1203
+ "epoch": 0.7975746268656716,
1204
+ "grad_norm": 0.27922648191452026,
1205
+ "learning_rate": 4.9954128440366974e-05,
1206
+ "loss": 2.9153,
1207
+ "step": 8550
1208
+ },
1209
+ {
1210
+ "epoch": 0.8022388059701493,
1211
+ "grad_norm": 0.2799355685710907,
1212
+ "learning_rate": 4.880733944954129e-05,
1213
+ "loss": 2.9329,
1214
+ "step": 8600
1215
+ },
1216
+ {
1217
+ "epoch": 0.8069029850746269,
1218
+ "grad_norm": 0.3089926242828369,
1219
+ "learning_rate": 4.76605504587156e-05,
1220
+ "loss": 2.9203,
1221
+ "step": 8650
1222
+ },
1223
+ {
1224
+ "epoch": 0.8115671641791045,
1225
+ "grad_norm": 0.3086063265800476,
1226
+ "learning_rate": 4.651376146788991e-05,
1227
+ "loss": 2.9223,
1228
+ "step": 8700
1229
+ },
1230
+ {
1231
+ "epoch": 0.816231343283582,
1232
+ "grad_norm": 0.3019801676273346,
1233
+ "learning_rate": 4.536697247706422e-05,
1234
+ "loss": 2.9159,
1235
+ "step": 8750
1236
+ },
1237
+ {
1238
+ "epoch": 0.8208955223880597,
1239
+ "grad_norm": 0.2975527048110962,
1240
+ "learning_rate": 4.422018348623854e-05,
1241
+ "loss": 2.9215,
1242
+ "step": 8800
1243
+ },
1244
+ {
1245
+ "epoch": 0.8255597014925373,
1246
+ "grad_norm": 0.30920466780662537,
1247
+ "learning_rate": 4.3073394495412844e-05,
1248
+ "loss": 2.9213,
1249
+ "step": 8850
1250
+ },
1251
+ {
1252
+ "epoch": 0.8302238805970149,
1253
+ "grad_norm": 0.2723771929740906,
1254
+ "learning_rate": 4.1926605504587155e-05,
1255
+ "loss": 2.9119,
1256
+ "step": 8900
1257
+ },
1258
+ {
1259
+ "epoch": 0.8348880597014925,
1260
+ "grad_norm": 0.28407421708106995,
1261
+ "learning_rate": 4.0779816513761474e-05,
1262
+ "loss": 2.9101,
1263
+ "step": 8950
1264
+ },
1265
+ {
1266
+ "epoch": 0.8395522388059702,
1267
+ "grad_norm": 0.2752555012702942,
1268
+ "learning_rate": 3.9633027522935785e-05,
1269
+ "loss": 2.9087,
1270
+ "step": 9000
1271
+ },
1272
+ {
1273
+ "epoch": 0.8442164179104478,
1274
+ "grad_norm": 0.30377471446990967,
1275
+ "learning_rate": 3.848623853211009e-05,
1276
+ "loss": 2.9003,
1277
+ "step": 9050
1278
+ },
1279
+ {
1280
+ "epoch": 0.8488805970149254,
1281
+ "grad_norm": 0.29200997948646545,
1282
+ "learning_rate": 3.73394495412844e-05,
1283
+ "loss": 2.9064,
1284
+ "step": 9100
1285
+ },
1286
+ {
1287
+ "epoch": 0.8535447761194029,
1288
+ "grad_norm": 0.2830120623111725,
1289
+ "learning_rate": 3.619266055045872e-05,
1290
+ "loss": 2.9085,
1291
+ "step": 9150
1292
+ },
1293
+ {
1294
+ "epoch": 0.8582089552238806,
1295
+ "grad_norm": 0.32428160309791565,
1296
+ "learning_rate": 3.504587155963303e-05,
1297
+ "loss": 2.9116,
1298
+ "step": 9200
1299
+ },
1300
+ {
1301
+ "epoch": 0.8628731343283582,
1302
+ "grad_norm": 0.2852170169353485,
1303
+ "learning_rate": 3.3899082568807337e-05,
1304
+ "loss": 2.9119,
1305
+ "step": 9250
1306
+ },
1307
+ {
1308
+ "epoch": 0.8675373134328358,
1309
+ "grad_norm": 0.2776082754135132,
1310
+ "learning_rate": 3.2752293577981655e-05,
1311
+ "loss": 2.9004,
1312
+ "step": 9300
1313
+ },
1314
+ {
1315
+ "epoch": 0.8722014925373134,
1316
+ "grad_norm": 0.2881547212600708,
1317
+ "learning_rate": 3.1605504587155967e-05,
1318
+ "loss": 2.9024,
1319
+ "step": 9350
1320
+ },
1321
+ {
1322
+ "epoch": 0.8768656716417911,
1323
+ "grad_norm": 0.2810397148132324,
1324
+ "learning_rate": 3.0458715596330278e-05,
1325
+ "loss": 2.9094,
1326
+ "step": 9400
1327
+ },
1328
+ {
1329
+ "epoch": 0.8815298507462687,
1330
+ "grad_norm": 0.2975148558616638,
1331
+ "learning_rate": 2.9311926605504586e-05,
1332
+ "loss": 2.8971,
1333
+ "step": 9450
1334
+ },
1335
+ {
1336
+ "epoch": 0.8861940298507462,
1337
+ "grad_norm": 0.29773736000061035,
1338
+ "learning_rate": 2.8165137614678898e-05,
1339
+ "loss": 2.8943,
1340
+ "step": 9500
1341
+ },
1342
+ {
1343
+ "epoch": 0.8908582089552238,
1344
+ "grad_norm": 0.2762524485588074,
1345
+ "learning_rate": 2.7018348623853213e-05,
1346
+ "loss": 2.8899,
1347
+ "step": 9550
1348
+ },
1349
+ {
1350
+ "epoch": 0.8955223880597015,
1351
+ "grad_norm": 0.2947123348712921,
1352
+ "learning_rate": 2.5871559633027525e-05,
1353
+ "loss": 2.8965,
1354
+ "step": 9600
1355
+ },
1356
+ {
1357
+ "epoch": 0.9001865671641791,
1358
+ "grad_norm": 0.27293965220451355,
1359
+ "learning_rate": 2.4724770642201836e-05,
1360
+ "loss": 2.8869,
1361
+ "step": 9650
1362
+ },
1363
+ {
1364
+ "epoch": 0.9048507462686567,
1365
+ "grad_norm": 0.2870621681213379,
1366
+ "learning_rate": 2.3577981651376148e-05,
1367
+ "loss": 2.8885,
1368
+ "step": 9700
1369
+ },
1370
+ {
1371
+ "epoch": 0.9095149253731343,
1372
+ "grad_norm": 0.2940194606781006,
1373
+ "learning_rate": 2.243119266055046e-05,
1374
+ "loss": 2.8933,
1375
+ "step": 9750
1376
+ },
1377
+ {
1378
+ "epoch": 0.914179104477612,
1379
+ "grad_norm": 0.270706444978714,
1380
+ "learning_rate": 2.1284403669724774e-05,
1381
+ "loss": 2.8889,
1382
+ "step": 9800
1383
+ },
1384
+ {
1385
+ "epoch": 0.9188432835820896,
1386
+ "grad_norm": 0.27721747756004333,
1387
+ "learning_rate": 2.0137614678899082e-05,
1388
+ "loss": 2.8893,
1389
+ "step": 9850
1390
+ },
1391
+ {
1392
+ "epoch": 0.9235074626865671,
1393
+ "grad_norm": 0.2803277373313904,
1394
+ "learning_rate": 1.8990825688073397e-05,
1395
+ "loss": 2.8868,
1396
+ "step": 9900
1397
+ },
1398
+ {
1399
+ "epoch": 0.9281716417910447,
1400
+ "grad_norm": 0.2814282476902008,
1401
+ "learning_rate": 1.7844036697247706e-05,
1402
+ "loss": 2.8906,
1403
+ "step": 9950
1404
+ },
1405
+ {
1406
+ "epoch": 0.9328358208955224,
1407
+ "grad_norm": 0.2730121612548828,
1408
+ "learning_rate": 1.669724770642202e-05,
1409
+ "loss": 2.8886,
1410
+ "step": 10000
1411
+ }
1412
+ ],
1413
+ "logging_steps": 50,
1414
+ "max_steps": 10720,
1415
+ "num_input_tokens_seen": 0,
1416
+ "num_train_epochs": 1,
1417
+ "save_steps": 1000,
1418
+ "stateful_callbacks": {
1419
+ "TrainerControl": {
1420
+ "args": {
1421
+ "should_epoch_stop": false,
1422
+ "should_evaluate": false,
1423
+ "should_log": false,
1424
+ "should_save": true,
1425
+ "should_training_stop": false
1426
+ },
1427
+ "attributes": {}
1428
+ }
1429
+ },
1430
+ "total_flos": 2124398876164096.0,
1431
+ "train_batch_size": 64,
1432
+ "trial_name": null,
1433
+ "trial_params": null
1434
+ }
full/checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0a1e4e68442fcaca00226db2169575a68dcecd4ab600ec17acb8a650bc00a2d
3
+ size 7121
full/checkpoint-10000/zero_to_fp32.py ADDED
@@ -0,0 +1,760 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example:
14
+ # python zero_to_fp32.py . output_dir/
15
+ # or
16
+ # python zero_to_fp32.py . output_dir/ --safe_serialization
17
+
18
+ import argparse
19
+ import torch
20
+ import glob
21
+ import math
22
+ import os
23
+ import re
24
+ import gc
25
+ import json
26
+ import numpy as np
27
+ from tqdm import tqdm
28
+ from collections import OrderedDict
29
+ from dataclasses import dataclass
30
+
31
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
32
+ # DeepSpeed data structures it has to be available in the current python environment.
33
+ from deepspeed.utils import logger
34
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
35
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
36
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
37
+
38
+
39
+ @dataclass
40
+ class zero_model_state:
41
+ buffers: dict()
42
+ param_shapes: dict()
43
+ shared_params: list
44
+ ds_version: int
45
+ frozen_param_shapes: dict()
46
+ frozen_param_fragments: dict()
47
+
48
+
49
+ debug = 0
50
+
51
+ # load to cpu
52
+ device = torch.device('cpu')
53
+
54
+
55
+ def atoi(text):
56
+ return int(text) if text.isdigit() else text
57
+
58
+
59
+ def natural_keys(text):
60
+ '''
61
+ alist.sort(key=natural_keys) sorts in human order
62
+ http://nedbatchelder.com/blog/200712/human_sorting.html
63
+ (See Toothy's implementation in the comments)
64
+ '''
65
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
66
+
67
+
68
+ def get_model_state_file(checkpoint_dir, zero_stage):
69
+ if not os.path.isdir(checkpoint_dir):
70
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
71
+
72
+ # there should be only one file
73
+ if zero_stage <= 2:
74
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
75
+ elif zero_stage == 3:
76
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
77
+
78
+ if not os.path.exists(file):
79
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
80
+
81
+ return file
82
+
83
+
84
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
85
+ # XXX: need to test that this simple glob rule works for multi-node setup too
86
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
87
+
88
+ if len(ckpt_files) == 0:
89
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
90
+
91
+ return ckpt_files
92
+
93
+
94
+ def get_optim_files(checkpoint_dir):
95
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
96
+
97
+
98
+ def get_model_state_files(checkpoint_dir):
99
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
100
+
101
+
102
+ def parse_model_states(files):
103
+ zero_model_states = []
104
+ for file in files:
105
+ state_dict = torch.load(file, map_location=device, weights_only=False)
106
+
107
+ if BUFFER_NAMES not in state_dict:
108
+ raise ValueError(f"{file} is not a model state checkpoint")
109
+ buffer_names = state_dict[BUFFER_NAMES]
110
+ if debug:
111
+ print("Found buffers:", buffer_names)
112
+
113
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
114
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
115
+ param_shapes = state_dict[PARAM_SHAPES]
116
+
117
+ # collect parameters that are included in param_shapes
118
+ param_names = []
119
+ for s in param_shapes:
120
+ for name in s.keys():
121
+ param_names.append(name)
122
+
123
+ # update with frozen parameters
124
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
125
+ if frozen_param_shapes is not None:
126
+ if debug:
127
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
128
+ param_names += list(frozen_param_shapes.keys())
129
+
130
+ # handle shared params
131
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
132
+
133
+ ds_version = state_dict.get(DS_VERSION, None)
134
+
135
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
136
+
137
+ z_model_state = zero_model_state(buffers=buffers,
138
+ param_shapes=param_shapes,
139
+ shared_params=shared_params,
140
+ ds_version=ds_version,
141
+ frozen_param_shapes=frozen_param_shapes,
142
+ frozen_param_fragments=frozen_param_fragments)
143
+ zero_model_states.append(z_model_state)
144
+
145
+ return zero_model_states
146
+
147
+
148
+ def parse_optim_states(files, ds_checkpoint_dir):
149
+ total_files = len(files)
150
+ state_dicts = []
151
+ for f in tqdm(files, desc='Loading checkpoint shards'):
152
+ state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
153
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
154
+ # and also handle the case where it was already removed by another helper script
155
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
156
+ state_dicts.append(state_dict)
157
+
158
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
159
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
160
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
161
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
162
+
163
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
164
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
165
+ # use the max of the partition_count to get the dp world_size.
166
+
167
+ if type(world_size) is list:
168
+ world_size = max(world_size)
169
+
170
+ if world_size != total_files:
171
+ raise ValueError(
172
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
173
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
174
+ )
175
+
176
+ # the groups are named differently in each stage
177
+ if zero_stage <= 2:
178
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
179
+ elif zero_stage == 3:
180
+ fp32_groups_key = FP32_FLAT_GROUPS
181
+ else:
182
+ raise ValueError(f"unknown zero stage {zero_stage}")
183
+
184
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
185
+ return zero_stage, world_size, fp32_flat_groups
186
+
187
+
188
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
189
+ """
190
+ Returns fp32 state_dict reconstructed from ds checkpoint
191
+
192
+ Args:
193
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
194
+
195
+ """
196
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
197
+
198
+ optim_files = get_optim_files(ds_checkpoint_dir)
199
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
200
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
201
+
202
+ model_files = get_model_state_files(ds_checkpoint_dir)
203
+
204
+ zero_model_states = parse_model_states(model_files)
205
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
206
+
207
+ if zero_stage <= 2:
208
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
209
+ exclude_frozen_parameters)
210
+ elif zero_stage == 3:
211
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
212
+ exclude_frozen_parameters)
213
+
214
+
215
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
216
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
217
+ return
218
+
219
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
220
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
221
+
222
+ if debug:
223
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
224
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
225
+
226
+ wanted_params = len(frozen_param_shapes)
227
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
228
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
229
+ print(f'Frozen params: Have {avail_numel} numels to process.')
230
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
231
+
232
+ total_params = 0
233
+ total_numel = 0
234
+ for name, shape in frozen_param_shapes.items():
235
+ total_params += 1
236
+ unpartitioned_numel = shape.numel()
237
+ total_numel += unpartitioned_numel
238
+
239
+ state_dict[name] = frozen_param_fragments[name]
240
+
241
+ if debug:
242
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
243
+
244
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
245
+
246
+
247
+ def _has_callable(obj, fn):
248
+ attr = getattr(obj, fn, None)
249
+ return callable(attr)
250
+
251
+
252
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
253
+ param_shapes = zero_model_states[0].param_shapes
254
+
255
+ # Reconstruction protocol:
256
+ #
257
+ # XXX: document this
258
+
259
+ if debug:
260
+ for i in range(world_size):
261
+ for j in range(len(fp32_flat_groups[0])):
262
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
263
+
264
+ # XXX: memory usage doubles here (zero2)
265
+ num_param_groups = len(fp32_flat_groups[0])
266
+ merged_single_partition_of_fp32_groups = []
267
+ for i in range(num_param_groups):
268
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
269
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
270
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
271
+ avail_numel = sum(
272
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
273
+
274
+ if debug:
275
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
276
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
277
+ # not asserting if there is a mismatch due to possible padding
278
+ print(f"Have {avail_numel} numels to process.")
279
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
280
+
281
+ # params
282
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
283
+ # out-of-core computing solution
284
+ total_numel = 0
285
+ total_params = 0
286
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
287
+ offset = 0
288
+ avail_numel = full_single_fp32_vector.numel()
289
+ for name, shape in shapes.items():
290
+
291
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
292
+ total_numel += unpartitioned_numel
293
+ total_params += 1
294
+
295
+ if debug:
296
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
297
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
298
+ offset += unpartitioned_numel
299
+
300
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
301
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
302
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
303
+ # live optimizer object, so we are checking that the numbers are within the right range
304
+ align_to = 2 * world_size
305
+
306
+ def zero2_align(x):
307
+ return align_to * math.ceil(x / align_to)
308
+
309
+ if debug:
310
+ print(f"original offset={offset}, avail_numel={avail_numel}")
311
+
312
+ offset = zero2_align(offset)
313
+ avail_numel = zero2_align(avail_numel)
314
+
315
+ if debug:
316
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
317
+
318
+ # Sanity check
319
+ if offset != avail_numel:
320
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
321
+
322
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
323
+
324
+
325
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
326
+ exclude_frozen_parameters):
327
+ state_dict = OrderedDict()
328
+
329
+ # buffers
330
+ buffers = zero_model_states[0].buffers
331
+ state_dict.update(buffers)
332
+ if debug:
333
+ print(f"added {len(buffers)} buffers")
334
+
335
+ if not exclude_frozen_parameters:
336
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
337
+
338
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
339
+
340
+ # recover shared parameters
341
+ for pair in zero_model_states[0].shared_params:
342
+ if pair[1] in state_dict:
343
+ state_dict[pair[0]] = state_dict[pair[1]]
344
+
345
+ return state_dict
346
+
347
+
348
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
349
+ remainder = unpartitioned_numel % world_size
350
+ padding_numel = (world_size - remainder) if remainder else 0
351
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
352
+ return partitioned_numel, padding_numel
353
+
354
+
355
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
356
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
357
+ return
358
+
359
+ if debug:
360
+ for i in range(world_size):
361
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
362
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
363
+
364
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
365
+ wanted_params = len(frozen_param_shapes)
366
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
367
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
368
+ print(f'Frozen params: Have {avail_numel} numels to process.')
369
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
370
+
371
+ total_params = 0
372
+ total_numel = 0
373
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
374
+ total_params += 1
375
+ unpartitioned_numel = shape.numel()
376
+ total_numel += unpartitioned_numel
377
+
378
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
379
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
380
+
381
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
382
+
383
+ if debug:
384
+ print(
385
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
386
+ )
387
+
388
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
389
+
390
+
391
+ class GatheredTensor:
392
+ """
393
+ A pseudo tensor that collects partitioned weights.
394
+ It is more memory efficient when there are multiple groups.
395
+ """
396
+
397
+ def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
398
+ self.flat_groups = flat_groups
399
+ self.flat_groups_offset = flat_groups_offset
400
+ self.offset = offset
401
+ self.partitioned_numel = partitioned_numel
402
+ self.shape = shape
403
+ self.dtype = self.flat_groups[0][0].dtype
404
+
405
+ def contiguous(self):
406
+ """
407
+ Merge partitioned weights from flat_groups into a single tensor.
408
+ """
409
+ end_idx = self.offset + self.partitioned_numel
410
+ world_size = len(self.flat_groups)
411
+ pad_flat_param_chunks = []
412
+
413
+ for rank_i in range(world_size):
414
+ # for each rank, we need to collect weights from related group/groups
415
+ flat_groups_at_rank_i = self.flat_groups[rank_i]
416
+ start_group_id = None
417
+ end_group_id = None
418
+ for group_id in range(len(self.flat_groups_offset)):
419
+ if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
420
+ start_group_id = group_id
421
+ if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
422
+ end_group_id = group_id
423
+ break
424
+ # collect weights from related group/groups
425
+ for group_id in range(start_group_id, end_group_id + 1):
426
+ flat_tensor = flat_groups_at_rank_i[group_id]
427
+ start_offset = self.offset - self.flat_groups_offset[group_id]
428
+ end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
429
+ pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
430
+
431
+ # collect weights from all ranks
432
+ pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
433
+ param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
434
+ return param
435
+
436
+
437
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
438
+ param_shapes = zero_model_states[0].param_shapes
439
+ avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
440
+
441
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
442
+ # param, re-consolidating each param, while dealing with padding if any
443
+
444
+ # merge list of dicts, preserving order
445
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
446
+
447
+ if debug:
448
+ for i in range(world_size):
449
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
450
+
451
+ wanted_params = len(param_shapes)
452
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
453
+ # not asserting if there is a mismatch due to possible padding
454
+ avail_numel = fp32_flat_groups[0].numel() * world_size
455
+ print(f"Trainable params: Have {avail_numel} numels to process.")
456
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
457
+
458
+ # params
459
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
460
+ # out-of-core computing solution
461
+ offset = 0
462
+ total_numel = 0
463
+ total_params = 0
464
+ flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
465
+ for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
466
+ unpartitioned_numel = shape.numel()
467
+ total_numel += unpartitioned_numel
468
+ total_params += 1
469
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
470
+
471
+ if debug:
472
+ print(
473
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
474
+ )
475
+
476
+ # memory efficient tensor
477
+ tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
478
+ state_dict[name] = tensor
479
+ offset += partitioned_numel
480
+
481
+ offset *= world_size
482
+
483
+ # Sanity check
484
+ if offset != avail_numel:
485
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
486
+
487
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
488
+
489
+
490
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
491
+ exclude_frozen_parameters):
492
+ state_dict = OrderedDict()
493
+
494
+ # buffers
495
+ buffers = zero_model_states[0].buffers
496
+ state_dict.update(buffers)
497
+ if debug:
498
+ print(f"added {len(buffers)} buffers")
499
+
500
+ if not exclude_frozen_parameters:
501
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
502
+
503
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
504
+
505
+ # recover shared parameters
506
+ for pair in zero_model_states[0].shared_params:
507
+ if pair[1] in state_dict:
508
+ state_dict[pair[0]] = state_dict[pair[1]]
509
+
510
+ return state_dict
511
+
512
+
513
+ def to_torch_tensor(state_dict, return_empty_tensor=False):
514
+ """
515
+ Convert state_dict of GatheredTensor to torch tensor
516
+ """
517
+ torch_state_dict = {}
518
+ converted_tensors = {}
519
+ for name, tensor in state_dict.items():
520
+ tensor_id = id(tensor)
521
+ if tensor_id in converted_tensors: # shared tensors
522
+ shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
523
+ torch_state_dict[name] = shared_tensor
524
+ else:
525
+ converted_tensors[tensor_id] = name
526
+ if return_empty_tensor:
527
+ torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
528
+ else:
529
+ torch_state_dict[name] = tensor.contiguous()
530
+ return torch_state_dict
531
+
532
+
533
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
534
+ tag=None,
535
+ exclude_frozen_parameters=False,
536
+ lazy_mode=False):
537
+ """
538
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
539
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
540
+ via a model hub.
541
+
542
+ Args:
543
+ - ``checkpoint_dir``: path to the desired checkpoint folder
544
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
545
+ - ``exclude_frozen_parameters``: exclude frozen parameters
546
+ - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
547
+ Convert the pesduo tensor to torch tensor by ``.contiguous()``
548
+
549
+ Returns:
550
+ - pytorch ``state_dict``
551
+
552
+ A typical usage might be ::
553
+
554
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
555
+ # do the training and checkpoint saving
556
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
557
+ model = model.cpu() # move to cpu
558
+ model.load_state_dict(state_dict)
559
+ # submit to model hub or save the model to share with others
560
+
561
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
562
+ application. i.e. you will need to re-initialize the deepspeed engine, since
563
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
564
+
565
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
566
+
567
+ Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
568
+ You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
569
+ the checkpoint. Or you can load state_dict in lazy mode ::
570
+
571
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
572
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
573
+ for name, lazy_tensor in state_dict.item():
574
+ tensor = lazy_tensor.contiguous() # to cpu
575
+ print(name, tensor)
576
+ # del tensor to release memory if it no longer in use
577
+ """
578
+ if tag is None:
579
+ latest_path = os.path.join(checkpoint_dir, 'latest')
580
+ if os.path.isfile(latest_path):
581
+ with open(latest_path, 'r') as fd:
582
+ tag = fd.read().strip()
583
+ else:
584
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
585
+
586
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
587
+
588
+ if not os.path.isdir(ds_checkpoint_dir):
589
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
590
+
591
+ state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
592
+ if lazy_mode:
593
+ return state_dict
594
+ else:
595
+ return to_torch_tensor(state_dict)
596
+
597
+
598
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
599
+ output_dir,
600
+ max_shard_size="5GB",
601
+ safe_serialization=False,
602
+ tag=None,
603
+ exclude_frozen_parameters=False):
604
+ """
605
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
606
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
607
+
608
+ Args:
609
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
610
+ - ``output_dir``: directory to the pytorch fp32 state_dict output files
611
+ - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
612
+ - ``safe_serialization``: whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
613
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
614
+ - ``exclude_frozen_parameters``: exclude frozen parameters
615
+ """
616
+
617
+ # Dependency pre-check
618
+ if safe_serialization:
619
+ try:
620
+ from safetensors.torch import save_file
621
+ except ImportError:
622
+ print('If you want to use `safe_serialization`, please `pip install safetensors`')
623
+ raise
624
+ if max_shard_size is not None:
625
+ try:
626
+ from huggingface_hub import split_torch_state_dict_into_shards
627
+ except ImportError:
628
+ print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
629
+ raise
630
+
631
+ # Convert zero checkpoint to state_dict
632
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
633
+ tag,
634
+ exclude_frozen_parameters,
635
+ lazy_mode=True)
636
+
637
+ # Shard the model if it is too big.
638
+ weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
639
+ if max_shard_size is not None:
640
+ filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
641
+ # an memory-efficient approach for sharding
642
+ empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
643
+ state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
644
+ filename_pattern=filename_pattern,
645
+ max_shard_size=max_shard_size)
646
+ else:
647
+ from collections import namedtuple
648
+ StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
649
+ state_dict_split = StateDictSplit(is_sharded=False,
650
+ filename_to_tensors={weights_name: list(state_dict.keys())})
651
+
652
+ # Save the model by shard
653
+ os.makedirs(output_dir, exist_ok=True)
654
+ filename_to_tensors = state_dict_split.filename_to_tensors.items()
655
+ for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
656
+ shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
657
+ shard_state_dict = to_torch_tensor(shard_state_dict)
658
+ output_path = os.path.join(output_dir, shard_file)
659
+ if safe_serialization:
660
+ save_file(shard_state_dict, output_path, metadata={"format": "pt"})
661
+ else:
662
+ torch.save(shard_state_dict, output_path)
663
+ # release the memory of current shard
664
+ for tensor_name in list(shard_state_dict.keys()):
665
+ del state_dict[tensor_name]
666
+ del shard_state_dict[tensor_name]
667
+ del shard_state_dict
668
+ gc.collect()
669
+
670
+ # Save index if sharded
671
+ if state_dict_split.is_sharded:
672
+ index = {
673
+ "metadata": state_dict_split.metadata,
674
+ "weight_map": state_dict_split.tensor_to_filename,
675
+ }
676
+ save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
677
+ save_index_file = os.path.join(output_dir, save_index_file)
678
+ with open(save_index_file, "w", encoding="utf-8") as f:
679
+ content = json.dumps(index, indent=2, sort_keys=True) + "\n"
680
+ f.write(content)
681
+
682
+
683
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
684
+ """
685
+ 1. Put the provided model to cpu
686
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
687
+ 3. Load it into the provided model
688
+
689
+ Args:
690
+ - ``model``: the model object to update
691
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
692
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
693
+
694
+ Returns:
695
+ - ``model`: modified model
696
+
697
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
698
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
699
+ conveniently placed for you in the checkpoint folder.
700
+
701
+ A typical usage might be ::
702
+
703
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
704
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
705
+ # submit to model hub or save the model to share with others
706
+
707
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
708
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
709
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
710
+
711
+ """
712
+ logger.info(f"Extracting fp32 weights")
713
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
714
+
715
+ logger.info(f"Overwriting model with fp32 weights")
716
+ model = model.cpu()
717
+ model.load_state_dict(state_dict, strict=False)
718
+
719
+ return model
720
+
721
+
722
+ if __name__ == "__main__":
723
+ parser = argparse.ArgumentParser()
724
+ parser.add_argument("checkpoint_dir",
725
+ type=str,
726
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
727
+ parser.add_argument("output_dir",
728
+ type=str,
729
+ help="directory to the pytorch fp32 state_dict output files"
730
+ "(e.g. path/checkpoint-12-output/)")
731
+ parser.add_argument(
732
+ "--max_shard_size",
733
+ type=str,
734
+ default="5GB",
735
+ help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
736
+ "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
737
+ "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
738
+ "without CPU OOM issues.")
739
+ parser.add_argument(
740
+ "--safe_serialization",
741
+ default=False,
742
+ action='store_true',
743
+ help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
744
+ parser.add_argument("-t",
745
+ "--tag",
746
+ type=str,
747
+ default=None,
748
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
749
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
750
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
751
+ args = parser.parse_args()
752
+
753
+ debug = args.debug
754
+
755
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
756
+ args.output_dir,
757
+ max_shard_size=args.max_shard_size,
758
+ safe_serialization=args.safe_serialization,
759
+ tag=args.tag,
760
+ exclude_frozen_parameters=args.exclude_frozen_parameters)
full/checkpoint-10720/chat_template.jinja ADDED
@@ -0,0 +1 @@
 
 
1
+ {% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% set system_message = false %}{% endif %}{% for message in loop_messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if loop.index0 == 0 and system_message != false %}{% set content = '<<SYS>>\n' + system_message + '\n<</SYS>>\n\n' + message['content'] %}{% else %}{% set content = message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ bos_token + '[INST] ' + content.strip() + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content.strip() + ' ' + eos_token }}{% endif %}{% endfor %}
full/checkpoint-10720/config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LlamaForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 1,
8
+ "eos_token_id": 2,
9
+ "head_dim": 128,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1024,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 2730,
14
+ "max_position_embeddings": 2048,
15
+ "mlp_bias": false,
16
+ "model_type": "llama",
17
+ "num_attention_heads": 8,
18
+ "num_hidden_layers": 16,
19
+ "num_key_value_heads": 8,
20
+ "pretraining_tp": 1,
21
+ "rms_norm_eps": 1e-05,
22
+ "rope_scaling": null,
23
+ "rope_theta": 10000.0,
24
+ "tie_word_embeddings": false,
25
+ "torch_dtype": "float16",
26
+ "transformers_version": "4.55.2",
27
+ "use_cache": false,
28
+ "vocab_size": 32000
29
+ }
full/checkpoint-10720/generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "transformers_version": "4.55.2",
6
+ "use_cache": false
7
+ }
full/checkpoint-10720/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:801f27268aa61bcbd737f0ac81310a163e93f798aa2731904a74ce142d761898
3
+ size 1067471056
full/checkpoint-10720/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27636c48b899230a9cd8b7e63ecb4b9bd567dd54336eed166836cff4fa52e881
3
+ size 1465
full/checkpoint-10720/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": true,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": true,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "</s>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": true,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
full/checkpoint-10720/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
full/checkpoint-10720/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
full/checkpoint-10720/tokenizer_config.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": true,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": true,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": true,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ }
30
+ },
31
+ "bos_token": "<s>",
32
+ "clean_up_tokenization_spaces": false,
33
+ "eos_token": "</s>",
34
+ "extra_special_tokens": {},
35
+ "legacy": true,
36
+ "model_max_length": 2048,
37
+ "pad_token": "</s>",
38
+ "sp_model_kwargs": {},
39
+ "tokenizer_class": "LlamaTokenizer",
40
+ "unk_token": "<unk>",
41
+ "use_default_system_prompt": false
42
+ }
full/checkpoint-10720/trainer_state.json ADDED
@@ -0,0 +1,1532 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.0,
6
+ "eval_steps": 500,
7
+ "global_step": 10720,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0046641791044776115,
14
+ "grad_norm": 1.907758116722107,
15
+ "learning_rate": 4.9000000000000005e-06,
16
+ "loss": 10.0793,
17
+ "step": 50
18
+ },
19
+ {
20
+ "epoch": 0.009328358208955223,
21
+ "grad_norm": 1.399185061454773,
22
+ "learning_rate": 9.900000000000002e-06,
23
+ "loss": 8.9287,
24
+ "step": 100
25
+ },
26
+ {
27
+ "epoch": 0.013992537313432836,
28
+ "grad_norm": 1.0101089477539062,
29
+ "learning_rate": 1.49e-05,
30
+ "loss": 8.3288,
31
+ "step": 150
32
+ },
33
+ {
34
+ "epoch": 0.018656716417910446,
35
+ "grad_norm": 0.7379341125488281,
36
+ "learning_rate": 1.9900000000000003e-05,
37
+ "loss": 7.6969,
38
+ "step": 200
39
+ },
40
+ {
41
+ "epoch": 0.02332089552238806,
42
+ "grad_norm": 0.9656306505203247,
43
+ "learning_rate": 2.4900000000000002e-05,
44
+ "loss": 7.2698,
45
+ "step": 250
46
+ },
47
+ {
48
+ "epoch": 0.027985074626865673,
49
+ "grad_norm": 1.3790279626846313,
50
+ "learning_rate": 2.9900000000000002e-05,
51
+ "loss": 6.9685,
52
+ "step": 300
53
+ },
54
+ {
55
+ "epoch": 0.03264925373134328,
56
+ "grad_norm": 0.6374709606170654,
57
+ "learning_rate": 3.49e-05,
58
+ "loss": 6.6752,
59
+ "step": 350
60
+ },
61
+ {
62
+ "epoch": 0.03731343283582089,
63
+ "grad_norm": 1.5095465183258057,
64
+ "learning_rate": 3.99e-05,
65
+ "loss": 6.3805,
66
+ "step": 400
67
+ },
68
+ {
69
+ "epoch": 0.04197761194029851,
70
+ "grad_norm": 1.394410490989685,
71
+ "learning_rate": 4.49e-05,
72
+ "loss": 6.1667,
73
+ "step": 450
74
+ },
75
+ {
76
+ "epoch": 0.04664179104477612,
77
+ "grad_norm": 1.7006834745407104,
78
+ "learning_rate": 4.99e-05,
79
+ "loss": 5.9861,
80
+ "step": 500
81
+ },
82
+ {
83
+ "epoch": 0.051305970149253734,
84
+ "grad_norm": 0.9934776425361633,
85
+ "learning_rate": 5.4900000000000006e-05,
86
+ "loss": 5.8166,
87
+ "step": 550
88
+ },
89
+ {
90
+ "epoch": 0.055970149253731345,
91
+ "grad_norm": 2.0550477504730225,
92
+ "learning_rate": 5.99e-05,
93
+ "loss": 5.6646,
94
+ "step": 600
95
+ },
96
+ {
97
+ "epoch": 0.06063432835820896,
98
+ "grad_norm": 2.2504913806915283,
99
+ "learning_rate": 6.49e-05,
100
+ "loss": 5.5353,
101
+ "step": 650
102
+ },
103
+ {
104
+ "epoch": 0.06529850746268656,
105
+ "grad_norm": 2.468212604522705,
106
+ "learning_rate": 6.99e-05,
107
+ "loss": 5.3982,
108
+ "step": 700
109
+ },
110
+ {
111
+ "epoch": 0.06996268656716417,
112
+ "grad_norm": 2.4010939598083496,
113
+ "learning_rate": 7.49e-05,
114
+ "loss": 5.2855,
115
+ "step": 750
116
+ },
117
+ {
118
+ "epoch": 0.07462686567164178,
119
+ "grad_norm": 1.4586199522018433,
120
+ "learning_rate": 7.99e-05,
121
+ "loss": 5.1901,
122
+ "step": 800
123
+ },
124
+ {
125
+ "epoch": 0.07929104477611941,
126
+ "grad_norm": 1.2821402549743652,
127
+ "learning_rate": 8.49e-05,
128
+ "loss": 5.0973,
129
+ "step": 850
130
+ },
131
+ {
132
+ "epoch": 0.08395522388059702,
133
+ "grad_norm": 2.1037180423736572,
134
+ "learning_rate": 8.99e-05,
135
+ "loss": 5.0062,
136
+ "step": 900
137
+ },
138
+ {
139
+ "epoch": 0.08861940298507463,
140
+ "grad_norm": 1.370751976966858,
141
+ "learning_rate": 9.49e-05,
142
+ "loss": 4.9392,
143
+ "step": 950
144
+ },
145
+ {
146
+ "epoch": 0.09328358208955224,
147
+ "grad_norm": 1.276108741760254,
148
+ "learning_rate": 9.99e-05,
149
+ "loss": 4.848,
150
+ "step": 1000
151
+ },
152
+ {
153
+ "epoch": 0.09794776119402986,
154
+ "grad_norm": 1.1077498197555542,
155
+ "learning_rate": 0.0001049,
156
+ "loss": 4.7863,
157
+ "step": 1050
158
+ },
159
+ {
160
+ "epoch": 0.10261194029850747,
161
+ "grad_norm": 1.0212163925170898,
162
+ "learning_rate": 0.0001099,
163
+ "loss": 4.7102,
164
+ "step": 1100
165
+ },
166
+ {
167
+ "epoch": 0.10727611940298508,
168
+ "grad_norm": 1.1480289697647095,
169
+ "learning_rate": 0.0001149,
170
+ "loss": 4.6551,
171
+ "step": 1150
172
+ },
173
+ {
174
+ "epoch": 0.11194029850746269,
175
+ "grad_norm": 1.3106518983840942,
176
+ "learning_rate": 0.00011990000000000001,
177
+ "loss": 4.5627,
178
+ "step": 1200
179
+ },
180
+ {
181
+ "epoch": 0.1166044776119403,
182
+ "grad_norm": 1.4304360151290894,
183
+ "learning_rate": 0.00012490000000000003,
184
+ "loss": 4.4823,
185
+ "step": 1250
186
+ },
187
+ {
188
+ "epoch": 0.12126865671641791,
189
+ "grad_norm": 0.883263885974884,
190
+ "learning_rate": 0.00012989999999999999,
191
+ "loss": 4.4091,
192
+ "step": 1300
193
+ },
194
+ {
195
+ "epoch": 0.1259328358208955,
196
+ "grad_norm": 1.2335679531097412,
197
+ "learning_rate": 0.0001349,
198
+ "loss": 4.3284,
199
+ "step": 1350
200
+ },
201
+ {
202
+ "epoch": 0.13059701492537312,
203
+ "grad_norm": 1.3587566614151,
204
+ "learning_rate": 0.0001399,
205
+ "loss": 4.2523,
206
+ "step": 1400
207
+ },
208
+ {
209
+ "epoch": 0.13526119402985073,
210
+ "grad_norm": 1.4397273063659668,
211
+ "learning_rate": 0.00014490000000000003,
212
+ "loss": 4.1874,
213
+ "step": 1450
214
+ },
215
+ {
216
+ "epoch": 0.13992537313432835,
217
+ "grad_norm": 0.9128643274307251,
218
+ "learning_rate": 0.0001499,
219
+ "loss": 4.1382,
220
+ "step": 1500
221
+ },
222
+ {
223
+ "epoch": 0.14458955223880596,
224
+ "grad_norm": 1.1429646015167236,
225
+ "learning_rate": 0.0001549,
226
+ "loss": 4.0713,
227
+ "step": 1550
228
+ },
229
+ {
230
+ "epoch": 0.14925373134328357,
231
+ "grad_norm": 0.7839804291725159,
232
+ "learning_rate": 0.0001599,
233
+ "loss": 4.0327,
234
+ "step": 1600
235
+ },
236
+ {
237
+ "epoch": 0.15391791044776118,
238
+ "grad_norm": 0.5934556126594543,
239
+ "learning_rate": 0.0001649,
240
+ "loss": 4.0535,
241
+ "step": 1650
242
+ },
243
+ {
244
+ "epoch": 0.15858208955223882,
245
+ "grad_norm": 0.6320812702178955,
246
+ "learning_rate": 0.0001699,
247
+ "loss": 3.9615,
248
+ "step": 1700
249
+ },
250
+ {
251
+ "epoch": 0.16324626865671643,
252
+ "grad_norm": 0.6207342743873596,
253
+ "learning_rate": 0.00017490000000000002,
254
+ "loss": 3.9086,
255
+ "step": 1750
256
+ },
257
+ {
258
+ "epoch": 0.16791044776119404,
259
+ "grad_norm": 0.8524695634841919,
260
+ "learning_rate": 0.0001799,
261
+ "loss": 3.8703,
262
+ "step": 1800
263
+ },
264
+ {
265
+ "epoch": 0.17257462686567165,
266
+ "grad_norm": 0.9148792028427124,
267
+ "learning_rate": 0.0001849,
268
+ "loss": 3.8388,
269
+ "step": 1850
270
+ },
271
+ {
272
+ "epoch": 0.17723880597014927,
273
+ "grad_norm": 0.7455419301986694,
274
+ "learning_rate": 0.0001899,
275
+ "loss": 3.8031,
276
+ "step": 1900
277
+ },
278
+ {
279
+ "epoch": 0.18190298507462688,
280
+ "grad_norm": 0.9777323603630066,
281
+ "learning_rate": 0.00019490000000000002,
282
+ "loss": 3.781,
283
+ "step": 1950
284
+ },
285
+ {
286
+ "epoch": 0.1865671641791045,
287
+ "grad_norm": 0.6018832921981812,
288
+ "learning_rate": 0.0001999,
289
+ "loss": 3.7421,
290
+ "step": 2000
291
+ },
292
+ {
293
+ "epoch": 0.1912313432835821,
294
+ "grad_norm": 0.7007433772087097,
295
+ "learning_rate": 0.00019887614678899083,
296
+ "loss": 3.7148,
297
+ "step": 2050
298
+ },
299
+ {
300
+ "epoch": 0.1958955223880597,
301
+ "grad_norm": 0.7864815592765808,
302
+ "learning_rate": 0.00019772935779816515,
303
+ "loss": 3.6907,
304
+ "step": 2100
305
+ },
306
+ {
307
+ "epoch": 0.20055970149253732,
308
+ "grad_norm": 0.6575994491577148,
309
+ "learning_rate": 0.00019658256880733944,
310
+ "loss": 3.6504,
311
+ "step": 2150
312
+ },
313
+ {
314
+ "epoch": 0.20522388059701493,
315
+ "grad_norm": 0.524401843547821,
316
+ "learning_rate": 0.0001954357798165138,
317
+ "loss": 3.6333,
318
+ "step": 2200
319
+ },
320
+ {
321
+ "epoch": 0.20988805970149255,
322
+ "grad_norm": 0.8242985010147095,
323
+ "learning_rate": 0.00019428899082568808,
324
+ "loss": 3.6084,
325
+ "step": 2250
326
+ },
327
+ {
328
+ "epoch": 0.21455223880597016,
329
+ "grad_norm": 0.5965232849121094,
330
+ "learning_rate": 0.0001931422018348624,
331
+ "loss": 3.5914,
332
+ "step": 2300
333
+ },
334
+ {
335
+ "epoch": 0.21921641791044777,
336
+ "grad_norm": 0.5473845601081848,
337
+ "learning_rate": 0.00019199541284403671,
338
+ "loss": 3.572,
339
+ "step": 2350
340
+ },
341
+ {
342
+ "epoch": 0.22388059701492538,
343
+ "grad_norm": 0.7030166983604431,
344
+ "learning_rate": 0.00019084862385321103,
345
+ "loss": 3.5548,
346
+ "step": 2400
347
+ },
348
+ {
349
+ "epoch": 0.228544776119403,
350
+ "grad_norm": 0.5844017267227173,
351
+ "learning_rate": 0.00018972477064220186,
352
+ "loss": 3.5326,
353
+ "step": 2450
354
+ },
355
+ {
356
+ "epoch": 0.2332089552238806,
357
+ "grad_norm": 0.4722403883934021,
358
+ "learning_rate": 0.00018857798165137615,
359
+ "loss": 3.5113,
360
+ "step": 2500
361
+ },
362
+ {
363
+ "epoch": 0.23787313432835822,
364
+ "grad_norm": 0.4691997468471527,
365
+ "learning_rate": 0.00018743119266055047,
366
+ "loss": 3.4985,
367
+ "step": 2550
368
+ },
369
+ {
370
+ "epoch": 0.24253731343283583,
371
+ "grad_norm": 0.5742476582527161,
372
+ "learning_rate": 0.00018628440366972478,
373
+ "loss": 3.4689,
374
+ "step": 2600
375
+ },
376
+ {
377
+ "epoch": 0.24720149253731344,
378
+ "grad_norm": 0.4867892563343048,
379
+ "learning_rate": 0.0001851376146788991,
380
+ "loss": 3.4737,
381
+ "step": 2650
382
+ },
383
+ {
384
+ "epoch": 0.251865671641791,
385
+ "grad_norm": 0.5876694321632385,
386
+ "learning_rate": 0.0001839908256880734,
387
+ "loss": 3.4516,
388
+ "step": 2700
389
+ },
390
+ {
391
+ "epoch": 0.25652985074626866,
392
+ "grad_norm": 0.4691877067089081,
393
+ "learning_rate": 0.0001828440366972477,
394
+ "loss": 3.4411,
395
+ "step": 2750
396
+ },
397
+ {
398
+ "epoch": 0.26119402985074625,
399
+ "grad_norm": 0.497503399848938,
400
+ "learning_rate": 0.00018169724770642203,
401
+ "loss": 3.4311,
402
+ "step": 2800
403
+ },
404
+ {
405
+ "epoch": 0.2658582089552239,
406
+ "grad_norm": 0.4610927700996399,
407
+ "learning_rate": 0.00018055045871559635,
408
+ "loss": 3.4122,
409
+ "step": 2850
410
+ },
411
+ {
412
+ "epoch": 0.27052238805970147,
413
+ "grad_norm": 0.5312972664833069,
414
+ "learning_rate": 0.00017940366972477064,
415
+ "loss": 3.4004,
416
+ "step": 2900
417
+ },
418
+ {
419
+ "epoch": 0.2751865671641791,
420
+ "grad_norm": 0.5256180167198181,
421
+ "learning_rate": 0.00017825688073394496,
422
+ "loss": 3.3834,
423
+ "step": 2950
424
+ },
425
+ {
426
+ "epoch": 0.2798507462686567,
427
+ "grad_norm": 0.5108594298362732,
428
+ "learning_rate": 0.00017711009174311928,
429
+ "loss": 3.379,
430
+ "step": 3000
431
+ },
432
+ {
433
+ "epoch": 0.28451492537313433,
434
+ "grad_norm": 0.4818528890609741,
435
+ "learning_rate": 0.00017596330275229357,
436
+ "loss": 3.3645,
437
+ "step": 3050
438
+ },
439
+ {
440
+ "epoch": 0.2891791044776119,
441
+ "grad_norm": 0.4077577590942383,
442
+ "learning_rate": 0.0001748165137614679,
443
+ "loss": 3.3659,
444
+ "step": 3100
445
+ },
446
+ {
447
+ "epoch": 0.29384328358208955,
448
+ "grad_norm": 0.4791859984397888,
449
+ "learning_rate": 0.0001736697247706422,
450
+ "loss": 3.3427,
451
+ "step": 3150
452
+ },
453
+ {
454
+ "epoch": 0.29850746268656714,
455
+ "grad_norm": 0.6330239176750183,
456
+ "learning_rate": 0.00017252293577981652,
457
+ "loss": 3.3484,
458
+ "step": 3200
459
+ },
460
+ {
461
+ "epoch": 0.3031716417910448,
462
+ "grad_norm": 0.4511379897594452,
463
+ "learning_rate": 0.00017137614678899084,
464
+ "loss": 3.3302,
465
+ "step": 3250
466
+ },
467
+ {
468
+ "epoch": 0.30783582089552236,
469
+ "grad_norm": 0.4571825861930847,
470
+ "learning_rate": 0.00017022935779816516,
471
+ "loss": 3.3144,
472
+ "step": 3300
473
+ },
474
+ {
475
+ "epoch": 0.3125,
476
+ "grad_norm": 0.44557932019233704,
477
+ "learning_rate": 0.00016908256880733945,
478
+ "loss": 3.3071,
479
+ "step": 3350
480
+ },
481
+ {
482
+ "epoch": 0.31716417910447764,
483
+ "grad_norm": 0.42434263229370117,
484
+ "learning_rate": 0.00016793577981651377,
485
+ "loss": 3.2986,
486
+ "step": 3400
487
+ },
488
+ {
489
+ "epoch": 0.3218283582089552,
490
+ "grad_norm": 0.41813352704048157,
491
+ "learning_rate": 0.0001667889908256881,
492
+ "loss": 3.2902,
493
+ "step": 3450
494
+ },
495
+ {
496
+ "epoch": 0.32649253731343286,
497
+ "grad_norm": 0.4084812104701996,
498
+ "learning_rate": 0.00016564220183486238,
499
+ "loss": 3.283,
500
+ "step": 3500
501
+ },
502
+ {
503
+ "epoch": 0.33115671641791045,
504
+ "grad_norm": 0.38210785388946533,
505
+ "learning_rate": 0.00016449541284403672,
506
+ "loss": 3.2769,
507
+ "step": 3550
508
+ },
509
+ {
510
+ "epoch": 0.3358208955223881,
511
+ "grad_norm": 0.5375980734825134,
512
+ "learning_rate": 0.00016334862385321102,
513
+ "loss": 3.2697,
514
+ "step": 3600
515
+ },
516
+ {
517
+ "epoch": 0.34048507462686567,
518
+ "grad_norm": 0.38781264424324036,
519
+ "learning_rate": 0.00016220183486238533,
520
+ "loss": 3.2556,
521
+ "step": 3650
522
+ },
523
+ {
524
+ "epoch": 0.3451492537313433,
525
+ "grad_norm": 0.41560590267181396,
526
+ "learning_rate": 0.00016105504587155965,
527
+ "loss": 3.2565,
528
+ "step": 3700
529
+ },
530
+ {
531
+ "epoch": 0.3498134328358209,
532
+ "grad_norm": 0.4119664132595062,
533
+ "learning_rate": 0.00015990825688073394,
534
+ "loss": 3.2386,
535
+ "step": 3750
536
+ },
537
+ {
538
+ "epoch": 0.35447761194029853,
539
+ "grad_norm": 0.4190433919429779,
540
+ "learning_rate": 0.00015876146788990826,
541
+ "loss": 3.2345,
542
+ "step": 3800
543
+ },
544
+ {
545
+ "epoch": 0.3591417910447761,
546
+ "grad_norm": 0.37028583884239197,
547
+ "learning_rate": 0.00015761467889908258,
548
+ "loss": 3.226,
549
+ "step": 3850
550
+ },
551
+ {
552
+ "epoch": 0.36380597014925375,
553
+ "grad_norm": 0.3971349000930786,
554
+ "learning_rate": 0.0001564678899082569,
555
+ "loss": 3.2213,
556
+ "step": 3900
557
+ },
558
+ {
559
+ "epoch": 0.36847014925373134,
560
+ "grad_norm": 0.34641316533088684,
561
+ "learning_rate": 0.0001553211009174312,
562
+ "loss": 3.2144,
563
+ "step": 3950
564
+ },
565
+ {
566
+ "epoch": 0.373134328358209,
567
+ "grad_norm": 0.40990573167800903,
568
+ "learning_rate": 0.0001541743119266055,
569
+ "loss": 3.2141,
570
+ "step": 4000
571
+ },
572
+ {
573
+ "epoch": 0.37779850746268656,
574
+ "grad_norm": 0.35828983783721924,
575
+ "learning_rate": 0.00015302752293577983,
576
+ "loss": 3.2054,
577
+ "step": 4050
578
+ },
579
+ {
580
+ "epoch": 0.3824626865671642,
581
+ "grad_norm": 0.37572401762008667,
582
+ "learning_rate": 0.00015188073394495414,
583
+ "loss": 3.1943,
584
+ "step": 4100
585
+ },
586
+ {
587
+ "epoch": 0.3871268656716418,
588
+ "grad_norm": 0.3795240521430969,
589
+ "learning_rate": 0.00015073394495412844,
590
+ "loss": 3.1971,
591
+ "step": 4150
592
+ },
593
+ {
594
+ "epoch": 0.3917910447761194,
595
+ "grad_norm": 0.3884343206882477,
596
+ "learning_rate": 0.00014958715596330275,
597
+ "loss": 3.1877,
598
+ "step": 4200
599
+ },
600
+ {
601
+ "epoch": 0.396455223880597,
602
+ "grad_norm": 0.374661922454834,
603
+ "learning_rate": 0.00014844036697247707,
604
+ "loss": 3.174,
605
+ "step": 4250
606
+ },
607
+ {
608
+ "epoch": 0.40111940298507465,
609
+ "grad_norm": 0.4330432415008545,
610
+ "learning_rate": 0.00014729357798165136,
611
+ "loss": 3.1787,
612
+ "step": 4300
613
+ },
614
+ {
615
+ "epoch": 0.40578358208955223,
616
+ "grad_norm": 0.40475285053253174,
617
+ "learning_rate": 0.0001461467889908257,
618
+ "loss": 3.1781,
619
+ "step": 4350
620
+ },
621
+ {
622
+ "epoch": 0.41044776119402987,
623
+ "grad_norm": 0.36028817296028137,
624
+ "learning_rate": 0.000145,
625
+ "loss": 3.1653,
626
+ "step": 4400
627
+ },
628
+ {
629
+ "epoch": 0.41511194029850745,
630
+ "grad_norm": 0.37242797017097473,
631
+ "learning_rate": 0.00014385321100917432,
632
+ "loss": 3.1608,
633
+ "step": 4450
634
+ },
635
+ {
636
+ "epoch": 0.4197761194029851,
637
+ "grad_norm": 0.42880979180336,
638
+ "learning_rate": 0.00014270642201834864,
639
+ "loss": 3.1656,
640
+ "step": 4500
641
+ },
642
+ {
643
+ "epoch": 0.4244402985074627,
644
+ "grad_norm": 0.3562730848789215,
645
+ "learning_rate": 0.00014155963302752296,
646
+ "loss": 3.1409,
647
+ "step": 4550
648
+ },
649
+ {
650
+ "epoch": 0.4291044776119403,
651
+ "grad_norm": 0.3420323133468628,
652
+ "learning_rate": 0.00014041284403669725,
653
+ "loss": 3.1502,
654
+ "step": 4600
655
+ },
656
+ {
657
+ "epoch": 0.4337686567164179,
658
+ "grad_norm": 0.33985841274261475,
659
+ "learning_rate": 0.00013926605504587157,
660
+ "loss": 3.1437,
661
+ "step": 4650
662
+ },
663
+ {
664
+ "epoch": 0.43843283582089554,
665
+ "grad_norm": 0.35252872109413147,
666
+ "learning_rate": 0.00013811926605504588,
667
+ "loss": 3.1375,
668
+ "step": 4700
669
+ },
670
+ {
671
+ "epoch": 0.4430970149253731,
672
+ "grad_norm": 0.33299019932746887,
673
+ "learning_rate": 0.00013697247706422018,
674
+ "loss": 3.1207,
675
+ "step": 4750
676
+ },
677
+ {
678
+ "epoch": 0.44776119402985076,
679
+ "grad_norm": 0.3340663015842438,
680
+ "learning_rate": 0.00013582568807339452,
681
+ "loss": 3.1285,
682
+ "step": 4800
683
+ },
684
+ {
685
+ "epoch": 0.45242537313432835,
686
+ "grad_norm": 0.3459039330482483,
687
+ "learning_rate": 0.00013470183486238532,
688
+ "loss": 3.1187,
689
+ "step": 4850
690
+ },
691
+ {
692
+ "epoch": 0.457089552238806,
693
+ "grad_norm": 0.34190812706947327,
694
+ "learning_rate": 0.00013355504587155963,
695
+ "loss": 3.1128,
696
+ "step": 4900
697
+ },
698
+ {
699
+ "epoch": 0.46175373134328357,
700
+ "grad_norm": 0.38482001423835754,
701
+ "learning_rate": 0.00013240825688073395,
702
+ "loss": 3.1081,
703
+ "step": 4950
704
+ },
705
+ {
706
+ "epoch": 0.4664179104477612,
707
+ "grad_norm": 0.3969718813896179,
708
+ "learning_rate": 0.00013126146788990827,
709
+ "loss": 3.111,
710
+ "step": 5000
711
+ },
712
+ {
713
+ "epoch": 0.4710820895522388,
714
+ "grad_norm": 0.30978360772132874,
715
+ "learning_rate": 0.0001301146788990826,
716
+ "loss": 3.106,
717
+ "step": 5050
718
+ },
719
+ {
720
+ "epoch": 0.47574626865671643,
721
+ "grad_norm": 0.32029977440834045,
722
+ "learning_rate": 0.00012896788990825688,
723
+ "loss": 3.1095,
724
+ "step": 5100
725
+ },
726
+ {
727
+ "epoch": 0.480410447761194,
728
+ "grad_norm": 0.36698678135871887,
729
+ "learning_rate": 0.0001278440366972477,
730
+ "loss": 3.1017,
731
+ "step": 5150
732
+ },
733
+ {
734
+ "epoch": 0.48507462686567165,
735
+ "grad_norm": 0.3710266947746277,
736
+ "learning_rate": 0.00012669724770642202,
737
+ "loss": 3.0863,
738
+ "step": 5200
739
+ },
740
+ {
741
+ "epoch": 0.48973880597014924,
742
+ "grad_norm": 0.33469706773757935,
743
+ "learning_rate": 0.00012555045871559634,
744
+ "loss": 3.0949,
745
+ "step": 5250
746
+ },
747
+ {
748
+ "epoch": 0.4944029850746269,
749
+ "grad_norm": 0.33493635058403015,
750
+ "learning_rate": 0.00012440366972477063,
751
+ "loss": 3.0794,
752
+ "step": 5300
753
+ },
754
+ {
755
+ "epoch": 0.49906716417910446,
756
+ "grad_norm": 0.3489319682121277,
757
+ "learning_rate": 0.00012325688073394498,
758
+ "loss": 3.0752,
759
+ "step": 5350
760
+ },
761
+ {
762
+ "epoch": 0.503731343283582,
763
+ "grad_norm": 0.32940617203712463,
764
+ "learning_rate": 0.00012211009174311927,
765
+ "loss": 3.0911,
766
+ "step": 5400
767
+ },
768
+ {
769
+ "epoch": 0.5083955223880597,
770
+ "grad_norm": 0.3039519190788269,
771
+ "learning_rate": 0.0001209633027522936,
772
+ "loss": 3.0749,
773
+ "step": 5450
774
+ },
775
+ {
776
+ "epoch": 0.5130597014925373,
777
+ "grad_norm": 0.3196834325790405,
778
+ "learning_rate": 0.00011981651376146789,
779
+ "loss": 3.0685,
780
+ "step": 5500
781
+ },
782
+ {
783
+ "epoch": 0.5177238805970149,
784
+ "grad_norm": 0.35171428322792053,
785
+ "learning_rate": 0.0001186697247706422,
786
+ "loss": 3.0684,
787
+ "step": 5550
788
+ },
789
+ {
790
+ "epoch": 0.5223880597014925,
791
+ "grad_norm": 0.3927328884601593,
792
+ "learning_rate": 0.00011752293577981653,
793
+ "loss": 3.0676,
794
+ "step": 5600
795
+ },
796
+ {
797
+ "epoch": 0.5270522388059702,
798
+ "grad_norm": 0.32428842782974243,
799
+ "learning_rate": 0.00011637614678899082,
800
+ "loss": 3.0635,
801
+ "step": 5650
802
+ },
803
+ {
804
+ "epoch": 0.5317164179104478,
805
+ "grad_norm": 0.3297617435455322,
806
+ "learning_rate": 0.00011522935779816515,
807
+ "loss": 3.0525,
808
+ "step": 5700
809
+ },
810
+ {
811
+ "epoch": 0.5363805970149254,
812
+ "grad_norm": 0.3118542432785034,
813
+ "learning_rate": 0.00011408256880733946,
814
+ "loss": 3.0566,
815
+ "step": 5750
816
+ },
817
+ {
818
+ "epoch": 0.5410447761194029,
819
+ "grad_norm": 0.29949086904525757,
820
+ "learning_rate": 0.00011293577981651377,
821
+ "loss": 3.0454,
822
+ "step": 5800
823
+ },
824
+ {
825
+ "epoch": 0.5457089552238806,
826
+ "grad_norm": 0.3567812740802765,
827
+ "learning_rate": 0.00011178899082568808,
828
+ "loss": 3.0462,
829
+ "step": 5850
830
+ },
831
+ {
832
+ "epoch": 0.5503731343283582,
833
+ "grad_norm": 0.3179719150066376,
834
+ "learning_rate": 0.00011064220183486238,
835
+ "loss": 3.0409,
836
+ "step": 5900
837
+ },
838
+ {
839
+ "epoch": 0.5550373134328358,
840
+ "grad_norm": 0.32883220911026,
841
+ "learning_rate": 0.0001094954128440367,
842
+ "loss": 3.0527,
843
+ "step": 5950
844
+ },
845
+ {
846
+ "epoch": 0.5597014925373134,
847
+ "grad_norm": 0.31516268849372864,
848
+ "learning_rate": 0.000108348623853211,
849
+ "loss": 3.0359,
850
+ "step": 6000
851
+ },
852
+ {
853
+ "epoch": 0.5643656716417911,
854
+ "grad_norm": 0.298753559589386,
855
+ "learning_rate": 0.00010720183486238533,
856
+ "loss": 3.0343,
857
+ "step": 6050
858
+ },
859
+ {
860
+ "epoch": 0.5690298507462687,
861
+ "grad_norm": 0.32886001467704773,
862
+ "learning_rate": 0.00010605504587155963,
863
+ "loss": 3.0229,
864
+ "step": 6100
865
+ },
866
+ {
867
+ "epoch": 0.5736940298507462,
868
+ "grad_norm": 0.3526987135410309,
869
+ "learning_rate": 0.00010490825688073396,
870
+ "loss": 3.0227,
871
+ "step": 6150
872
+ },
873
+ {
874
+ "epoch": 0.5783582089552238,
875
+ "grad_norm": 0.31915777921676636,
876
+ "learning_rate": 0.00010376146788990825,
877
+ "loss": 3.0311,
878
+ "step": 6200
879
+ },
880
+ {
881
+ "epoch": 0.5830223880597015,
882
+ "grad_norm": 0.31342729926109314,
883
+ "learning_rate": 0.00010261467889908258,
884
+ "loss": 3.0257,
885
+ "step": 6250
886
+ },
887
+ {
888
+ "epoch": 0.5876865671641791,
889
+ "grad_norm": 0.3711305856704712,
890
+ "learning_rate": 0.00010146788990825689,
891
+ "loss": 3.0125,
892
+ "step": 6300
893
+ },
894
+ {
895
+ "epoch": 0.5923507462686567,
896
+ "grad_norm": 0.32861173152923584,
897
+ "learning_rate": 0.00010032110091743118,
898
+ "loss": 3.0235,
899
+ "step": 6350
900
+ },
901
+ {
902
+ "epoch": 0.5970149253731343,
903
+ "grad_norm": 0.30850905179977417,
904
+ "learning_rate": 9.917431192660551e-05,
905
+ "loss": 3.0171,
906
+ "step": 6400
907
+ },
908
+ {
909
+ "epoch": 0.601679104477612,
910
+ "grad_norm": 0.31126561760902405,
911
+ "learning_rate": 9.802752293577983e-05,
912
+ "loss": 3.0177,
913
+ "step": 6450
914
+ },
915
+ {
916
+ "epoch": 0.6063432835820896,
917
+ "grad_norm": 0.3373335897922516,
918
+ "learning_rate": 9.688073394495414e-05,
919
+ "loss": 3.0138,
920
+ "step": 6500
921
+ },
922
+ {
923
+ "epoch": 0.6110074626865671,
924
+ "grad_norm": 0.32974258065223694,
925
+ "learning_rate": 9.575688073394496e-05,
926
+ "loss": 3.0092,
927
+ "step": 6550
928
+ },
929
+ {
930
+ "epoch": 0.6156716417910447,
931
+ "grad_norm": 0.29914727807044983,
932
+ "learning_rate": 9.463302752293578e-05,
933
+ "loss": 3.0075,
934
+ "step": 6600
935
+ },
936
+ {
937
+ "epoch": 0.6203358208955224,
938
+ "grad_norm": 0.3006197214126587,
939
+ "learning_rate": 9.34862385321101e-05,
940
+ "loss": 3.015,
941
+ "step": 6650
942
+ },
943
+ {
944
+ "epoch": 0.625,
945
+ "grad_norm": 0.2997278869152069,
946
+ "learning_rate": 9.23394495412844e-05,
947
+ "loss": 3.0037,
948
+ "step": 6700
949
+ },
950
+ {
951
+ "epoch": 0.6296641791044776,
952
+ "grad_norm": 0.31719309091567993,
953
+ "learning_rate": 9.119266055045872e-05,
954
+ "loss": 2.9969,
955
+ "step": 6750
956
+ },
957
+ {
958
+ "epoch": 0.6343283582089553,
959
+ "grad_norm": 0.3004010319709778,
960
+ "learning_rate": 9.004587155963303e-05,
961
+ "loss": 2.9909,
962
+ "step": 6800
963
+ },
964
+ {
965
+ "epoch": 0.6389925373134329,
966
+ "grad_norm": 0.3123314082622528,
967
+ "learning_rate": 8.889908256880735e-05,
968
+ "loss": 2.9853,
969
+ "step": 6850
970
+ },
971
+ {
972
+ "epoch": 0.6436567164179104,
973
+ "grad_norm": 0.334837406873703,
974
+ "learning_rate": 8.775229357798166e-05,
975
+ "loss": 2.9923,
976
+ "step": 6900
977
+ },
978
+ {
979
+ "epoch": 0.648320895522388,
980
+ "grad_norm": 0.3055480420589447,
981
+ "learning_rate": 8.660550458715597e-05,
982
+ "loss": 2.9897,
983
+ "step": 6950
984
+ },
985
+ {
986
+ "epoch": 0.6529850746268657,
987
+ "grad_norm": 0.33151915669441223,
988
+ "learning_rate": 8.545871559633027e-05,
989
+ "loss": 2.9843,
990
+ "step": 7000
991
+ },
992
+ {
993
+ "epoch": 0.6576492537313433,
994
+ "grad_norm": 0.32511621713638306,
995
+ "learning_rate": 8.431192660550459e-05,
996
+ "loss": 2.9797,
997
+ "step": 7050
998
+ },
999
+ {
1000
+ "epoch": 0.6623134328358209,
1001
+ "grad_norm": 0.29789265990257263,
1002
+ "learning_rate": 8.31651376146789e-05,
1003
+ "loss": 2.9789,
1004
+ "step": 7100
1005
+ },
1006
+ {
1007
+ "epoch": 0.6669776119402985,
1008
+ "grad_norm": 0.3272823393344879,
1009
+ "learning_rate": 8.201834862385321e-05,
1010
+ "loss": 2.971,
1011
+ "step": 7150
1012
+ },
1013
+ {
1014
+ "epoch": 0.6716417910447762,
1015
+ "grad_norm": 0.2983226478099823,
1016
+ "learning_rate": 8.087155963302753e-05,
1017
+ "loss": 2.9818,
1018
+ "step": 7200
1019
+ },
1020
+ {
1021
+ "epoch": 0.6763059701492538,
1022
+ "grad_norm": 0.30532923340797424,
1023
+ "learning_rate": 7.972477064220184e-05,
1024
+ "loss": 2.9739,
1025
+ "step": 7250
1026
+ },
1027
+ {
1028
+ "epoch": 0.6809701492537313,
1029
+ "grad_norm": 0.3046092092990875,
1030
+ "learning_rate": 7.857798165137616e-05,
1031
+ "loss": 2.9575,
1032
+ "step": 7300
1033
+ },
1034
+ {
1035
+ "epoch": 0.6856343283582089,
1036
+ "grad_norm": 0.3133400082588196,
1037
+ "learning_rate": 7.743119266055046e-05,
1038
+ "loss": 2.9794,
1039
+ "step": 7350
1040
+ },
1041
+ {
1042
+ "epoch": 0.6902985074626866,
1043
+ "grad_norm": 0.3090495765209198,
1044
+ "learning_rate": 7.628440366972477e-05,
1045
+ "loss": 2.9632,
1046
+ "step": 7400
1047
+ },
1048
+ {
1049
+ "epoch": 0.6949626865671642,
1050
+ "grad_norm": 0.2925854027271271,
1051
+ "learning_rate": 7.513761467889908e-05,
1052
+ "loss": 2.9649,
1053
+ "step": 7450
1054
+ },
1055
+ {
1056
+ "epoch": 0.6996268656716418,
1057
+ "grad_norm": 0.30464622378349304,
1058
+ "learning_rate": 7.399082568807339e-05,
1059
+ "loss": 2.9606,
1060
+ "step": 7500
1061
+ },
1062
+ {
1063
+ "epoch": 0.7042910447761194,
1064
+ "grad_norm": 0.29478827118873596,
1065
+ "learning_rate": 7.284403669724771e-05,
1066
+ "loss": 2.9574,
1067
+ "step": 7550
1068
+ },
1069
+ {
1070
+ "epoch": 0.7089552238805971,
1071
+ "grad_norm": 0.30368906259536743,
1072
+ "learning_rate": 7.169724770642203e-05,
1073
+ "loss": 2.9583,
1074
+ "step": 7600
1075
+ },
1076
+ {
1077
+ "epoch": 0.7136194029850746,
1078
+ "grad_norm": 0.2974824905395508,
1079
+ "learning_rate": 7.055045871559633e-05,
1080
+ "loss": 2.9566,
1081
+ "step": 7650
1082
+ },
1083
+ {
1084
+ "epoch": 0.7182835820895522,
1085
+ "grad_norm": 0.3006385564804077,
1086
+ "learning_rate": 6.940366972477065e-05,
1087
+ "loss": 2.953,
1088
+ "step": 7700
1089
+ },
1090
+ {
1091
+ "epoch": 0.7229477611940298,
1092
+ "grad_norm": 0.2787078022956848,
1093
+ "learning_rate": 6.825688073394497e-05,
1094
+ "loss": 2.9466,
1095
+ "step": 7750
1096
+ },
1097
+ {
1098
+ "epoch": 0.7276119402985075,
1099
+ "grad_norm": 0.2966342568397522,
1100
+ "learning_rate": 6.711009174311927e-05,
1101
+ "loss": 2.9446,
1102
+ "step": 7800
1103
+ },
1104
+ {
1105
+ "epoch": 0.7322761194029851,
1106
+ "grad_norm": 0.32728853821754456,
1107
+ "learning_rate": 6.600917431192661e-05,
1108
+ "loss": 2.9486,
1109
+ "step": 7850
1110
+ },
1111
+ {
1112
+ "epoch": 0.7369402985074627,
1113
+ "grad_norm": 0.325846403837204,
1114
+ "learning_rate": 6.486238532110092e-05,
1115
+ "loss": 2.9456,
1116
+ "step": 7900
1117
+ },
1118
+ {
1119
+ "epoch": 0.7416044776119403,
1120
+ "grad_norm": 0.2933451235294342,
1121
+ "learning_rate": 6.371559633027524e-05,
1122
+ "loss": 2.9404,
1123
+ "step": 7950
1124
+ },
1125
+ {
1126
+ "epoch": 0.746268656716418,
1127
+ "grad_norm": 0.3031325042247772,
1128
+ "learning_rate": 6.256880733944954e-05,
1129
+ "loss": 2.9486,
1130
+ "step": 8000
1131
+ },
1132
+ {
1133
+ "epoch": 0.7509328358208955,
1134
+ "grad_norm": 0.30206266045570374,
1135
+ "learning_rate": 6.142201834862386e-05,
1136
+ "loss": 2.9364,
1137
+ "step": 8050
1138
+ },
1139
+ {
1140
+ "epoch": 0.7555970149253731,
1141
+ "grad_norm": 0.29706859588623047,
1142
+ "learning_rate": 6.027522935779817e-05,
1143
+ "loss": 2.9483,
1144
+ "step": 8100
1145
+ },
1146
+ {
1147
+ "epoch": 0.7602611940298507,
1148
+ "grad_norm": 0.2909660339355469,
1149
+ "learning_rate": 5.912844036697248e-05,
1150
+ "loss": 2.9405,
1151
+ "step": 8150
1152
+ },
1153
+ {
1154
+ "epoch": 0.7649253731343284,
1155
+ "grad_norm": 0.3136695623397827,
1156
+ "learning_rate": 5.798165137614679e-05,
1157
+ "loss": 2.9366,
1158
+ "step": 8200
1159
+ },
1160
+ {
1161
+ "epoch": 0.769589552238806,
1162
+ "grad_norm": 0.29548221826553345,
1163
+ "learning_rate": 5.683486238532111e-05,
1164
+ "loss": 2.9305,
1165
+ "step": 8250
1166
+ },
1167
+ {
1168
+ "epoch": 0.7742537313432836,
1169
+ "grad_norm": 0.2777513265609741,
1170
+ "learning_rate": 5.568807339449541e-05,
1171
+ "loss": 2.9377,
1172
+ "step": 8300
1173
+ },
1174
+ {
1175
+ "epoch": 0.7789179104477612,
1176
+ "grad_norm": 0.28673887252807617,
1177
+ "learning_rate": 5.454128440366972e-05,
1178
+ "loss": 2.9342,
1179
+ "step": 8350
1180
+ },
1181
+ {
1182
+ "epoch": 0.7835820895522388,
1183
+ "grad_norm": 0.2911510169506073,
1184
+ "learning_rate": 5.339449541284404e-05,
1185
+ "loss": 2.938,
1186
+ "step": 8400
1187
+ },
1188
+ {
1189
+ "epoch": 0.7882462686567164,
1190
+ "grad_norm": 0.3052498698234558,
1191
+ "learning_rate": 5.224770642201835e-05,
1192
+ "loss": 2.9235,
1193
+ "step": 8450
1194
+ },
1195
+ {
1196
+ "epoch": 0.792910447761194,
1197
+ "grad_norm": 0.30528631806373596,
1198
+ "learning_rate": 5.110091743119266e-05,
1199
+ "loss": 2.9393,
1200
+ "step": 8500
1201
+ },
1202
+ {
1203
+ "epoch": 0.7975746268656716,
1204
+ "grad_norm": 0.27922648191452026,
1205
+ "learning_rate": 4.9954128440366974e-05,
1206
+ "loss": 2.9153,
1207
+ "step": 8550
1208
+ },
1209
+ {
1210
+ "epoch": 0.8022388059701493,
1211
+ "grad_norm": 0.2799355685710907,
1212
+ "learning_rate": 4.880733944954129e-05,
1213
+ "loss": 2.9329,
1214
+ "step": 8600
1215
+ },
1216
+ {
1217
+ "epoch": 0.8069029850746269,
1218
+ "grad_norm": 0.3089926242828369,
1219
+ "learning_rate": 4.76605504587156e-05,
1220
+ "loss": 2.9203,
1221
+ "step": 8650
1222
+ },
1223
+ {
1224
+ "epoch": 0.8115671641791045,
1225
+ "grad_norm": 0.3086063265800476,
1226
+ "learning_rate": 4.651376146788991e-05,
1227
+ "loss": 2.9223,
1228
+ "step": 8700
1229
+ },
1230
+ {
1231
+ "epoch": 0.816231343283582,
1232
+ "grad_norm": 0.3019801676273346,
1233
+ "learning_rate": 4.536697247706422e-05,
1234
+ "loss": 2.9159,
1235
+ "step": 8750
1236
+ },
1237
+ {
1238
+ "epoch": 0.8208955223880597,
1239
+ "grad_norm": 0.2975527048110962,
1240
+ "learning_rate": 4.422018348623854e-05,
1241
+ "loss": 2.9215,
1242
+ "step": 8800
1243
+ },
1244
+ {
1245
+ "epoch": 0.8255597014925373,
1246
+ "grad_norm": 0.30920466780662537,
1247
+ "learning_rate": 4.3073394495412844e-05,
1248
+ "loss": 2.9213,
1249
+ "step": 8850
1250
+ },
1251
+ {
1252
+ "epoch": 0.8302238805970149,
1253
+ "grad_norm": 0.2723771929740906,
1254
+ "learning_rate": 4.1926605504587155e-05,
1255
+ "loss": 2.9119,
1256
+ "step": 8900
1257
+ },
1258
+ {
1259
+ "epoch": 0.8348880597014925,
1260
+ "grad_norm": 0.28407421708106995,
1261
+ "learning_rate": 4.0779816513761474e-05,
1262
+ "loss": 2.9101,
1263
+ "step": 8950
1264
+ },
1265
+ {
1266
+ "epoch": 0.8395522388059702,
1267
+ "grad_norm": 0.2752555012702942,
1268
+ "learning_rate": 3.9633027522935785e-05,
1269
+ "loss": 2.9087,
1270
+ "step": 9000
1271
+ },
1272
+ {
1273
+ "epoch": 0.8442164179104478,
1274
+ "grad_norm": 0.30377471446990967,
1275
+ "learning_rate": 3.848623853211009e-05,
1276
+ "loss": 2.9003,
1277
+ "step": 9050
1278
+ },
1279
+ {
1280
+ "epoch": 0.8488805970149254,
1281
+ "grad_norm": 0.29200997948646545,
1282
+ "learning_rate": 3.73394495412844e-05,
1283
+ "loss": 2.9064,
1284
+ "step": 9100
1285
+ },
1286
+ {
1287
+ "epoch": 0.8535447761194029,
1288
+ "grad_norm": 0.2830120623111725,
1289
+ "learning_rate": 3.619266055045872e-05,
1290
+ "loss": 2.9085,
1291
+ "step": 9150
1292
+ },
1293
+ {
1294
+ "epoch": 0.8582089552238806,
1295
+ "grad_norm": 0.32428160309791565,
1296
+ "learning_rate": 3.504587155963303e-05,
1297
+ "loss": 2.9116,
1298
+ "step": 9200
1299
+ },
1300
+ {
1301
+ "epoch": 0.8628731343283582,
1302
+ "grad_norm": 0.2852170169353485,
1303
+ "learning_rate": 3.3899082568807337e-05,
1304
+ "loss": 2.9119,
1305
+ "step": 9250
1306
+ },
1307
+ {
1308
+ "epoch": 0.8675373134328358,
1309
+ "grad_norm": 0.2776082754135132,
1310
+ "learning_rate": 3.2752293577981655e-05,
1311
+ "loss": 2.9004,
1312
+ "step": 9300
1313
+ },
1314
+ {
1315
+ "epoch": 0.8722014925373134,
1316
+ "grad_norm": 0.2881547212600708,
1317
+ "learning_rate": 3.1605504587155967e-05,
1318
+ "loss": 2.9024,
1319
+ "step": 9350
1320
+ },
1321
+ {
1322
+ "epoch": 0.8768656716417911,
1323
+ "grad_norm": 0.2810397148132324,
1324
+ "learning_rate": 3.0458715596330278e-05,
1325
+ "loss": 2.9094,
1326
+ "step": 9400
1327
+ },
1328
+ {
1329
+ "epoch": 0.8815298507462687,
1330
+ "grad_norm": 0.2975148558616638,
1331
+ "learning_rate": 2.9311926605504586e-05,
1332
+ "loss": 2.8971,
1333
+ "step": 9450
1334
+ },
1335
+ {
1336
+ "epoch": 0.8861940298507462,
1337
+ "grad_norm": 0.29773736000061035,
1338
+ "learning_rate": 2.8165137614678898e-05,
1339
+ "loss": 2.8943,
1340
+ "step": 9500
1341
+ },
1342
+ {
1343
+ "epoch": 0.8908582089552238,
1344
+ "grad_norm": 0.2762524485588074,
1345
+ "learning_rate": 2.7018348623853213e-05,
1346
+ "loss": 2.8899,
1347
+ "step": 9550
1348
+ },
1349
+ {
1350
+ "epoch": 0.8955223880597015,
1351
+ "grad_norm": 0.2947123348712921,
1352
+ "learning_rate": 2.5871559633027525e-05,
1353
+ "loss": 2.8965,
1354
+ "step": 9600
1355
+ },
1356
+ {
1357
+ "epoch": 0.9001865671641791,
1358
+ "grad_norm": 0.27293965220451355,
1359
+ "learning_rate": 2.4724770642201836e-05,
1360
+ "loss": 2.8869,
1361
+ "step": 9650
1362
+ },
1363
+ {
1364
+ "epoch": 0.9048507462686567,
1365
+ "grad_norm": 0.2870621681213379,
1366
+ "learning_rate": 2.3577981651376148e-05,
1367
+ "loss": 2.8885,
1368
+ "step": 9700
1369
+ },
1370
+ {
1371
+ "epoch": 0.9095149253731343,
1372
+ "grad_norm": 0.2940194606781006,
1373
+ "learning_rate": 2.243119266055046e-05,
1374
+ "loss": 2.8933,
1375
+ "step": 9750
1376
+ },
1377
+ {
1378
+ "epoch": 0.914179104477612,
1379
+ "grad_norm": 0.270706444978714,
1380
+ "learning_rate": 2.1284403669724774e-05,
1381
+ "loss": 2.8889,
1382
+ "step": 9800
1383
+ },
1384
+ {
1385
+ "epoch": 0.9188432835820896,
1386
+ "grad_norm": 0.27721747756004333,
1387
+ "learning_rate": 2.0137614678899082e-05,
1388
+ "loss": 2.8893,
1389
+ "step": 9850
1390
+ },
1391
+ {
1392
+ "epoch": 0.9235074626865671,
1393
+ "grad_norm": 0.2803277373313904,
1394
+ "learning_rate": 1.8990825688073397e-05,
1395
+ "loss": 2.8868,
1396
+ "step": 9900
1397
+ },
1398
+ {
1399
+ "epoch": 0.9281716417910447,
1400
+ "grad_norm": 0.2814282476902008,
1401
+ "learning_rate": 1.7844036697247706e-05,
1402
+ "loss": 2.8906,
1403
+ "step": 9950
1404
+ },
1405
+ {
1406
+ "epoch": 0.9328358208955224,
1407
+ "grad_norm": 0.2730121612548828,
1408
+ "learning_rate": 1.669724770642202e-05,
1409
+ "loss": 2.8886,
1410
+ "step": 10000
1411
+ },
1412
+ {
1413
+ "epoch": 0.9375,
1414
+ "grad_norm": 0.2759721577167511,
1415
+ "learning_rate": 1.555045871559633e-05,
1416
+ "loss": 2.8906,
1417
+ "step": 10050
1418
+ },
1419
+ {
1420
+ "epoch": 0.9421641791044776,
1421
+ "grad_norm": 0.28318414092063904,
1422
+ "learning_rate": 1.4403669724770644e-05,
1423
+ "loss": 2.8834,
1424
+ "step": 10100
1425
+ },
1426
+ {
1427
+ "epoch": 0.9468283582089553,
1428
+ "grad_norm": 0.27038970589637756,
1429
+ "learning_rate": 1.3256880733944954e-05,
1430
+ "loss": 2.8866,
1431
+ "step": 10150
1432
+ },
1433
+ {
1434
+ "epoch": 0.9514925373134329,
1435
+ "grad_norm": 0.29715093970298767,
1436
+ "learning_rate": 1.2110091743119267e-05,
1437
+ "loss": 2.8749,
1438
+ "step": 10200
1439
+ },
1440
+ {
1441
+ "epoch": 0.9561567164179104,
1442
+ "grad_norm": 0.27278682589530945,
1443
+ "learning_rate": 1.0986238532110093e-05,
1444
+ "loss": 2.8771,
1445
+ "step": 10250
1446
+ },
1447
+ {
1448
+ "epoch": 0.960820895522388,
1449
+ "grad_norm": 0.26732149720191956,
1450
+ "learning_rate": 9.839449541284404e-06,
1451
+ "loss": 2.8857,
1452
+ "step": 10300
1453
+ },
1454
+ {
1455
+ "epoch": 0.9654850746268657,
1456
+ "grad_norm": 0.2564278244972229,
1457
+ "learning_rate": 8.692660550458716e-06,
1458
+ "loss": 2.8801,
1459
+ "step": 10350
1460
+ },
1461
+ {
1462
+ "epoch": 0.9701492537313433,
1463
+ "grad_norm": 0.2738514244556427,
1464
+ "learning_rate": 7.545871559633027e-06,
1465
+ "loss": 2.8833,
1466
+ "step": 10400
1467
+ },
1468
+ {
1469
+ "epoch": 0.9748134328358209,
1470
+ "grad_norm": 0.26929736137390137,
1471
+ "learning_rate": 6.399082568807339e-06,
1472
+ "loss": 2.8706,
1473
+ "step": 10450
1474
+ },
1475
+ {
1476
+ "epoch": 0.9794776119402985,
1477
+ "grad_norm": 0.2783481180667877,
1478
+ "learning_rate": 5.252293577981652e-06,
1479
+ "loss": 2.8758,
1480
+ "step": 10500
1481
+ },
1482
+ {
1483
+ "epoch": 0.9841417910447762,
1484
+ "grad_norm": 0.2605401575565338,
1485
+ "learning_rate": 4.105504587155964e-06,
1486
+ "loss": 2.8718,
1487
+ "step": 10550
1488
+ },
1489
+ {
1490
+ "epoch": 0.9888059701492538,
1491
+ "grad_norm": 0.2604130506515503,
1492
+ "learning_rate": 2.9587155963302755e-06,
1493
+ "loss": 2.87,
1494
+ "step": 10600
1495
+ },
1496
+ {
1497
+ "epoch": 0.9934701492537313,
1498
+ "grad_norm": 0.259562224149704,
1499
+ "learning_rate": 1.8119266055045873e-06,
1500
+ "loss": 2.8781,
1501
+ "step": 10650
1502
+ },
1503
+ {
1504
+ "epoch": 0.9981343283582089,
1505
+ "grad_norm": 0.2627047300338745,
1506
+ "learning_rate": 6.651376146788992e-07,
1507
+ "loss": 2.8782,
1508
+ "step": 10700
1509
+ }
1510
+ ],
1511
+ "logging_steps": 50,
1512
+ "max_steps": 10720,
1513
+ "num_input_tokens_seen": 0,
1514
+ "num_train_epochs": 1,
1515
+ "save_steps": 1000,
1516
+ "stateful_callbacks": {
1517
+ "TrainerControl": {
1518
+ "args": {
1519
+ "should_epoch_stop": false,
1520
+ "should_evaluate": false,
1521
+ "should_log": false,
1522
+ "should_save": true,
1523
+ "should_training_stop": true
1524
+ },
1525
+ "attributes": {}
1526
+ }
1527
+ },
1528
+ "total_flos": 2277239225843712.0,
1529
+ "train_batch_size": 64,
1530
+ "trial_name": null,
1531
+ "trial_params": null
1532
+ }
full/checkpoint-10720/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0a1e4e68442fcaca00226db2169575a68dcecd4ab600ec17acb8a650bc00a2d
3
+ size 7121
full/checkpoint-10720/zero_to_fp32.py ADDED
@@ -0,0 +1,760 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example:
14
+ # python zero_to_fp32.py . output_dir/
15
+ # or
16
+ # python zero_to_fp32.py . output_dir/ --safe_serialization
17
+
18
+ import argparse
19
+ import torch
20
+ import glob
21
+ import math
22
+ import os
23
+ import re
24
+ import gc
25
+ import json
26
+ import numpy as np
27
+ from tqdm import tqdm
28
+ from collections import OrderedDict
29
+ from dataclasses import dataclass
30
+
31
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
32
+ # DeepSpeed data structures it has to be available in the current python environment.
33
+ from deepspeed.utils import logger
34
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
35
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
36
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
37
+
38
+
39
+ @dataclass
40
+ class zero_model_state:
41
+ buffers: dict()
42
+ param_shapes: dict()
43
+ shared_params: list
44
+ ds_version: int
45
+ frozen_param_shapes: dict()
46
+ frozen_param_fragments: dict()
47
+
48
+
49
+ debug = 0
50
+
51
+ # load to cpu
52
+ device = torch.device('cpu')
53
+
54
+
55
+ def atoi(text):
56
+ return int(text) if text.isdigit() else text
57
+
58
+
59
+ def natural_keys(text):
60
+ '''
61
+ alist.sort(key=natural_keys) sorts in human order
62
+ http://nedbatchelder.com/blog/200712/human_sorting.html
63
+ (See Toothy's implementation in the comments)
64
+ '''
65
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
66
+
67
+
68
+ def get_model_state_file(checkpoint_dir, zero_stage):
69
+ if not os.path.isdir(checkpoint_dir):
70
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
71
+
72
+ # there should be only one file
73
+ if zero_stage <= 2:
74
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
75
+ elif zero_stage == 3:
76
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
77
+
78
+ if not os.path.exists(file):
79
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
80
+
81
+ return file
82
+
83
+
84
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
85
+ # XXX: need to test that this simple glob rule works for multi-node setup too
86
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
87
+
88
+ if len(ckpt_files) == 0:
89
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
90
+
91
+ return ckpt_files
92
+
93
+
94
+ def get_optim_files(checkpoint_dir):
95
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
96
+
97
+
98
+ def get_model_state_files(checkpoint_dir):
99
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
100
+
101
+
102
+ def parse_model_states(files):
103
+ zero_model_states = []
104
+ for file in files:
105
+ state_dict = torch.load(file, map_location=device, weights_only=False)
106
+
107
+ if BUFFER_NAMES not in state_dict:
108
+ raise ValueError(f"{file} is not a model state checkpoint")
109
+ buffer_names = state_dict[BUFFER_NAMES]
110
+ if debug:
111
+ print("Found buffers:", buffer_names)
112
+
113
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
114
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
115
+ param_shapes = state_dict[PARAM_SHAPES]
116
+
117
+ # collect parameters that are included in param_shapes
118
+ param_names = []
119
+ for s in param_shapes:
120
+ for name in s.keys():
121
+ param_names.append(name)
122
+
123
+ # update with frozen parameters
124
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
125
+ if frozen_param_shapes is not None:
126
+ if debug:
127
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
128
+ param_names += list(frozen_param_shapes.keys())
129
+
130
+ # handle shared params
131
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
132
+
133
+ ds_version = state_dict.get(DS_VERSION, None)
134
+
135
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
136
+
137
+ z_model_state = zero_model_state(buffers=buffers,
138
+ param_shapes=param_shapes,
139
+ shared_params=shared_params,
140
+ ds_version=ds_version,
141
+ frozen_param_shapes=frozen_param_shapes,
142
+ frozen_param_fragments=frozen_param_fragments)
143
+ zero_model_states.append(z_model_state)
144
+
145
+ return zero_model_states
146
+
147
+
148
+ def parse_optim_states(files, ds_checkpoint_dir):
149
+ total_files = len(files)
150
+ state_dicts = []
151
+ for f in tqdm(files, desc='Loading checkpoint shards'):
152
+ state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
153
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
154
+ # and also handle the case where it was already removed by another helper script
155
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
156
+ state_dicts.append(state_dict)
157
+
158
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
159
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
160
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
161
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
162
+
163
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
164
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
165
+ # use the max of the partition_count to get the dp world_size.
166
+
167
+ if type(world_size) is list:
168
+ world_size = max(world_size)
169
+
170
+ if world_size != total_files:
171
+ raise ValueError(
172
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
173
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
174
+ )
175
+
176
+ # the groups are named differently in each stage
177
+ if zero_stage <= 2:
178
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
179
+ elif zero_stage == 3:
180
+ fp32_groups_key = FP32_FLAT_GROUPS
181
+ else:
182
+ raise ValueError(f"unknown zero stage {zero_stage}")
183
+
184
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
185
+ return zero_stage, world_size, fp32_flat_groups
186
+
187
+
188
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
189
+ """
190
+ Returns fp32 state_dict reconstructed from ds checkpoint
191
+
192
+ Args:
193
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
194
+
195
+ """
196
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
197
+
198
+ optim_files = get_optim_files(ds_checkpoint_dir)
199
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
200
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
201
+
202
+ model_files = get_model_state_files(ds_checkpoint_dir)
203
+
204
+ zero_model_states = parse_model_states(model_files)
205
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
206
+
207
+ if zero_stage <= 2:
208
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
209
+ exclude_frozen_parameters)
210
+ elif zero_stage == 3:
211
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
212
+ exclude_frozen_parameters)
213
+
214
+
215
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
216
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
217
+ return
218
+
219
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
220
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
221
+
222
+ if debug:
223
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
224
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
225
+
226
+ wanted_params = len(frozen_param_shapes)
227
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
228
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
229
+ print(f'Frozen params: Have {avail_numel} numels to process.')
230
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
231
+
232
+ total_params = 0
233
+ total_numel = 0
234
+ for name, shape in frozen_param_shapes.items():
235
+ total_params += 1
236
+ unpartitioned_numel = shape.numel()
237
+ total_numel += unpartitioned_numel
238
+
239
+ state_dict[name] = frozen_param_fragments[name]
240
+
241
+ if debug:
242
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
243
+
244
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
245
+
246
+
247
+ def _has_callable(obj, fn):
248
+ attr = getattr(obj, fn, None)
249
+ return callable(attr)
250
+
251
+
252
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
253
+ param_shapes = zero_model_states[0].param_shapes
254
+
255
+ # Reconstruction protocol:
256
+ #
257
+ # XXX: document this
258
+
259
+ if debug:
260
+ for i in range(world_size):
261
+ for j in range(len(fp32_flat_groups[0])):
262
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
263
+
264
+ # XXX: memory usage doubles here (zero2)
265
+ num_param_groups = len(fp32_flat_groups[0])
266
+ merged_single_partition_of_fp32_groups = []
267
+ for i in range(num_param_groups):
268
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
269
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
270
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
271
+ avail_numel = sum(
272
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
273
+
274
+ if debug:
275
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
276
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
277
+ # not asserting if there is a mismatch due to possible padding
278
+ print(f"Have {avail_numel} numels to process.")
279
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
280
+
281
+ # params
282
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
283
+ # out-of-core computing solution
284
+ total_numel = 0
285
+ total_params = 0
286
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
287
+ offset = 0
288
+ avail_numel = full_single_fp32_vector.numel()
289
+ for name, shape in shapes.items():
290
+
291
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
292
+ total_numel += unpartitioned_numel
293
+ total_params += 1
294
+
295
+ if debug:
296
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
297
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
298
+ offset += unpartitioned_numel
299
+
300
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
301
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
302
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
303
+ # live optimizer object, so we are checking that the numbers are within the right range
304
+ align_to = 2 * world_size
305
+
306
+ def zero2_align(x):
307
+ return align_to * math.ceil(x / align_to)
308
+
309
+ if debug:
310
+ print(f"original offset={offset}, avail_numel={avail_numel}")
311
+
312
+ offset = zero2_align(offset)
313
+ avail_numel = zero2_align(avail_numel)
314
+
315
+ if debug:
316
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
317
+
318
+ # Sanity check
319
+ if offset != avail_numel:
320
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
321
+
322
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
323
+
324
+
325
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
326
+ exclude_frozen_parameters):
327
+ state_dict = OrderedDict()
328
+
329
+ # buffers
330
+ buffers = zero_model_states[0].buffers
331
+ state_dict.update(buffers)
332
+ if debug:
333
+ print(f"added {len(buffers)} buffers")
334
+
335
+ if not exclude_frozen_parameters:
336
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
337
+
338
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
339
+
340
+ # recover shared parameters
341
+ for pair in zero_model_states[0].shared_params:
342
+ if pair[1] in state_dict:
343
+ state_dict[pair[0]] = state_dict[pair[1]]
344
+
345
+ return state_dict
346
+
347
+
348
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
349
+ remainder = unpartitioned_numel % world_size
350
+ padding_numel = (world_size - remainder) if remainder else 0
351
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
352
+ return partitioned_numel, padding_numel
353
+
354
+
355
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
356
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
357
+ return
358
+
359
+ if debug:
360
+ for i in range(world_size):
361
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
362
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
363
+
364
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
365
+ wanted_params = len(frozen_param_shapes)
366
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
367
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
368
+ print(f'Frozen params: Have {avail_numel} numels to process.')
369
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
370
+
371
+ total_params = 0
372
+ total_numel = 0
373
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
374
+ total_params += 1
375
+ unpartitioned_numel = shape.numel()
376
+ total_numel += unpartitioned_numel
377
+
378
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
379
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
380
+
381
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
382
+
383
+ if debug:
384
+ print(
385
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
386
+ )
387
+
388
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
389
+
390
+
391
+ class GatheredTensor:
392
+ """
393
+ A pseudo tensor that collects partitioned weights.
394
+ It is more memory efficient when there are multiple groups.
395
+ """
396
+
397
+ def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
398
+ self.flat_groups = flat_groups
399
+ self.flat_groups_offset = flat_groups_offset
400
+ self.offset = offset
401
+ self.partitioned_numel = partitioned_numel
402
+ self.shape = shape
403
+ self.dtype = self.flat_groups[0][0].dtype
404
+
405
+ def contiguous(self):
406
+ """
407
+ Merge partitioned weights from flat_groups into a single tensor.
408
+ """
409
+ end_idx = self.offset + self.partitioned_numel
410
+ world_size = len(self.flat_groups)
411
+ pad_flat_param_chunks = []
412
+
413
+ for rank_i in range(world_size):
414
+ # for each rank, we need to collect weights from related group/groups
415
+ flat_groups_at_rank_i = self.flat_groups[rank_i]
416
+ start_group_id = None
417
+ end_group_id = None
418
+ for group_id in range(len(self.flat_groups_offset)):
419
+ if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
420
+ start_group_id = group_id
421
+ if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
422
+ end_group_id = group_id
423
+ break
424
+ # collect weights from related group/groups
425
+ for group_id in range(start_group_id, end_group_id + 1):
426
+ flat_tensor = flat_groups_at_rank_i[group_id]
427
+ start_offset = self.offset - self.flat_groups_offset[group_id]
428
+ end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
429
+ pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
430
+
431
+ # collect weights from all ranks
432
+ pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
433
+ param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
434
+ return param
435
+
436
+
437
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
438
+ param_shapes = zero_model_states[0].param_shapes
439
+ avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
440
+
441
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
442
+ # param, re-consolidating each param, while dealing with padding if any
443
+
444
+ # merge list of dicts, preserving order
445
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
446
+
447
+ if debug:
448
+ for i in range(world_size):
449
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
450
+
451
+ wanted_params = len(param_shapes)
452
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
453
+ # not asserting if there is a mismatch due to possible padding
454
+ avail_numel = fp32_flat_groups[0].numel() * world_size
455
+ print(f"Trainable params: Have {avail_numel} numels to process.")
456
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
457
+
458
+ # params
459
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
460
+ # out-of-core computing solution
461
+ offset = 0
462
+ total_numel = 0
463
+ total_params = 0
464
+ flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
465
+ for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
466
+ unpartitioned_numel = shape.numel()
467
+ total_numel += unpartitioned_numel
468
+ total_params += 1
469
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
470
+
471
+ if debug:
472
+ print(
473
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
474
+ )
475
+
476
+ # memory efficient tensor
477
+ tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
478
+ state_dict[name] = tensor
479
+ offset += partitioned_numel
480
+
481
+ offset *= world_size
482
+
483
+ # Sanity check
484
+ if offset != avail_numel:
485
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
486
+
487
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
488
+
489
+
490
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
491
+ exclude_frozen_parameters):
492
+ state_dict = OrderedDict()
493
+
494
+ # buffers
495
+ buffers = zero_model_states[0].buffers
496
+ state_dict.update(buffers)
497
+ if debug:
498
+ print(f"added {len(buffers)} buffers")
499
+
500
+ if not exclude_frozen_parameters:
501
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
502
+
503
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
504
+
505
+ # recover shared parameters
506
+ for pair in zero_model_states[0].shared_params:
507
+ if pair[1] in state_dict:
508
+ state_dict[pair[0]] = state_dict[pair[1]]
509
+
510
+ return state_dict
511
+
512
+
513
+ def to_torch_tensor(state_dict, return_empty_tensor=False):
514
+ """
515
+ Convert state_dict of GatheredTensor to torch tensor
516
+ """
517
+ torch_state_dict = {}
518
+ converted_tensors = {}
519
+ for name, tensor in state_dict.items():
520
+ tensor_id = id(tensor)
521
+ if tensor_id in converted_tensors: # shared tensors
522
+ shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
523
+ torch_state_dict[name] = shared_tensor
524
+ else:
525
+ converted_tensors[tensor_id] = name
526
+ if return_empty_tensor:
527
+ torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
528
+ else:
529
+ torch_state_dict[name] = tensor.contiguous()
530
+ return torch_state_dict
531
+
532
+
533
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
534
+ tag=None,
535
+ exclude_frozen_parameters=False,
536
+ lazy_mode=False):
537
+ """
538
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
539
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
540
+ via a model hub.
541
+
542
+ Args:
543
+ - ``checkpoint_dir``: path to the desired checkpoint folder
544
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
545
+ - ``exclude_frozen_parameters``: exclude frozen parameters
546
+ - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
547
+ Convert the pesduo tensor to torch tensor by ``.contiguous()``
548
+
549
+ Returns:
550
+ - pytorch ``state_dict``
551
+
552
+ A typical usage might be ::
553
+
554
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
555
+ # do the training and checkpoint saving
556
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
557
+ model = model.cpu() # move to cpu
558
+ model.load_state_dict(state_dict)
559
+ # submit to model hub or save the model to share with others
560
+
561
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
562
+ application. i.e. you will need to re-initialize the deepspeed engine, since
563
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
564
+
565
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
566
+
567
+ Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
568
+ You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
569
+ the checkpoint. Or you can load state_dict in lazy mode ::
570
+
571
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
572
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
573
+ for name, lazy_tensor in state_dict.item():
574
+ tensor = lazy_tensor.contiguous() # to cpu
575
+ print(name, tensor)
576
+ # del tensor to release memory if it no longer in use
577
+ """
578
+ if tag is None:
579
+ latest_path = os.path.join(checkpoint_dir, 'latest')
580
+ if os.path.isfile(latest_path):
581
+ with open(latest_path, 'r') as fd:
582
+ tag = fd.read().strip()
583
+ else:
584
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
585
+
586
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
587
+
588
+ if not os.path.isdir(ds_checkpoint_dir):
589
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
590
+
591
+ state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
592
+ if lazy_mode:
593
+ return state_dict
594
+ else:
595
+ return to_torch_tensor(state_dict)
596
+
597
+
598
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
599
+ output_dir,
600
+ max_shard_size="5GB",
601
+ safe_serialization=False,
602
+ tag=None,
603
+ exclude_frozen_parameters=False):
604
+ """
605
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
606
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
607
+
608
+ Args:
609
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
610
+ - ``output_dir``: directory to the pytorch fp32 state_dict output files
611
+ - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
612
+ - ``safe_serialization``: whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
613
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
614
+ - ``exclude_frozen_parameters``: exclude frozen parameters
615
+ """
616
+
617
+ # Dependency pre-check
618
+ if safe_serialization:
619
+ try:
620
+ from safetensors.torch import save_file
621
+ except ImportError:
622
+ print('If you want to use `safe_serialization`, please `pip install safetensors`')
623
+ raise
624
+ if max_shard_size is not None:
625
+ try:
626
+ from huggingface_hub import split_torch_state_dict_into_shards
627
+ except ImportError:
628
+ print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
629
+ raise
630
+
631
+ # Convert zero checkpoint to state_dict
632
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
633
+ tag,
634
+ exclude_frozen_parameters,
635
+ lazy_mode=True)
636
+
637
+ # Shard the model if it is too big.
638
+ weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
639
+ if max_shard_size is not None:
640
+ filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
641
+ # an memory-efficient approach for sharding
642
+ empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
643
+ state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
644
+ filename_pattern=filename_pattern,
645
+ max_shard_size=max_shard_size)
646
+ else:
647
+ from collections import namedtuple
648
+ StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
649
+ state_dict_split = StateDictSplit(is_sharded=False,
650
+ filename_to_tensors={weights_name: list(state_dict.keys())})
651
+
652
+ # Save the model by shard
653
+ os.makedirs(output_dir, exist_ok=True)
654
+ filename_to_tensors = state_dict_split.filename_to_tensors.items()
655
+ for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
656
+ shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
657
+ shard_state_dict = to_torch_tensor(shard_state_dict)
658
+ output_path = os.path.join(output_dir, shard_file)
659
+ if safe_serialization:
660
+ save_file(shard_state_dict, output_path, metadata={"format": "pt"})
661
+ else:
662
+ torch.save(shard_state_dict, output_path)
663
+ # release the memory of current shard
664
+ for tensor_name in list(shard_state_dict.keys()):
665
+ del state_dict[tensor_name]
666
+ del shard_state_dict[tensor_name]
667
+ del shard_state_dict
668
+ gc.collect()
669
+
670
+ # Save index if sharded
671
+ if state_dict_split.is_sharded:
672
+ index = {
673
+ "metadata": state_dict_split.metadata,
674
+ "weight_map": state_dict_split.tensor_to_filename,
675
+ }
676
+ save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
677
+ save_index_file = os.path.join(output_dir, save_index_file)
678
+ with open(save_index_file, "w", encoding="utf-8") as f:
679
+ content = json.dumps(index, indent=2, sort_keys=True) + "\n"
680
+ f.write(content)
681
+
682
+
683
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
684
+ """
685
+ 1. Put the provided model to cpu
686
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
687
+ 3. Load it into the provided model
688
+
689
+ Args:
690
+ - ``model``: the model object to update
691
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
692
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
693
+
694
+ Returns:
695
+ - ``model`: modified model
696
+
697
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
698
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
699
+ conveniently placed for you in the checkpoint folder.
700
+
701
+ A typical usage might be ::
702
+
703
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
704
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
705
+ # submit to model hub or save the model to share with others
706
+
707
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
708
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
709
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
710
+
711
+ """
712
+ logger.info(f"Extracting fp32 weights")
713
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
714
+
715
+ logger.info(f"Overwriting model with fp32 weights")
716
+ model = model.cpu()
717
+ model.load_state_dict(state_dict, strict=False)
718
+
719
+ return model
720
+
721
+
722
+ if __name__ == "__main__":
723
+ parser = argparse.ArgumentParser()
724
+ parser.add_argument("checkpoint_dir",
725
+ type=str,
726
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
727
+ parser.add_argument("output_dir",
728
+ type=str,
729
+ help="directory to the pytorch fp32 state_dict output files"
730
+ "(e.g. path/checkpoint-12-output/)")
731
+ parser.add_argument(
732
+ "--max_shard_size",
733
+ type=str,
734
+ default="5GB",
735
+ help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
736
+ "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
737
+ "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
738
+ "without CPU OOM issues.")
739
+ parser.add_argument(
740
+ "--safe_serialization",
741
+ default=False,
742
+ action='store_true',
743
+ help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
744
+ parser.add_argument("-t",
745
+ "--tag",
746
+ type=str,
747
+ default=None,
748
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
749
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
750
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
751
+ args = parser.parse_args()
752
+
753
+ debug = args.debug
754
+
755
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
756
+ args.output_dir,
757
+ max_shard_size=args.max_shard_size,
758
+ safe_serialization=args.safe_serialization,
759
+ tag=args.tag,
760
+ exclude_frozen_parameters=args.exclude_frozen_parameters)
full/checkpoint-9000/chat_template.jinja ADDED
@@ -0,0 +1 @@
 
 
1
+ {% if messages[0]['role'] == 'system' %}{% set loop_messages = messages[1:] %}{% set system_message = messages[0]['content'] %}{% else %}{% set loop_messages = messages %}{% set system_message = false %}{% endif %}{% for message in loop_messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if loop.index0 == 0 and system_message != false %}{% set content = '<<SYS>>\n' + system_message + '\n<</SYS>>\n\n' + message['content'] %}{% else %}{% set content = message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ bos_token + '[INST] ' + content.strip() + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ ' ' + content.strip() + ' ' + eos_token }}{% endif %}{% endfor %}
full/checkpoint-9000/config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LlamaForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 1,
8
+ "eos_token_id": 2,
9
+ "head_dim": 128,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1024,
12
+ "initializer_range": 0.02,
13
+ "intermediate_size": 2730,
14
+ "max_position_embeddings": 2048,
15
+ "mlp_bias": false,
16
+ "model_type": "llama",
17
+ "num_attention_heads": 8,
18
+ "num_hidden_layers": 16,
19
+ "num_key_value_heads": 8,
20
+ "pretraining_tp": 1,
21
+ "rms_norm_eps": 1e-05,
22
+ "rope_scaling": null,
23
+ "rope_theta": 10000.0,
24
+ "tie_word_embeddings": false,
25
+ "torch_dtype": "float16",
26
+ "transformers_version": "4.55.2",
27
+ "use_cache": false,
28
+ "vocab_size": 32000
29
+ }
full/checkpoint-9000/generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "transformers_version": "4.55.2",
6
+ "use_cache": false
7
+ }
full/checkpoint-9000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb3b84d468444da94f4aca832d8968c5168101a33c276e95372b4c6b1c2203ad
3
+ size 1067471056
full/checkpoint-9000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cae9caf33f74c42ea1c1debccd7a6aeeb6572a94f1f322fe88d1c2a83ac3e3e7
3
+ size 1465
full/checkpoint-9000/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": true,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": true,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "</s>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": true,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
full/checkpoint-9000/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
full/checkpoint-9000/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
full/checkpoint-9000/tokenizer_config.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": true,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": true,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": true,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ }
30
+ },
31
+ "bos_token": "<s>",
32
+ "clean_up_tokenization_spaces": false,
33
+ "eos_token": "</s>",
34
+ "extra_special_tokens": {},
35
+ "legacy": true,
36
+ "model_max_length": 2048,
37
+ "pad_token": "</s>",
38
+ "sp_model_kwargs": {},
39
+ "tokenizer_class": "LlamaTokenizer",
40
+ "unk_token": "<unk>",
41
+ "use_default_system_prompt": false
42
+ }
full/checkpoint-9000/trainer_state.json ADDED
@@ -0,0 +1,1294 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.8395522388059702,
6
+ "eval_steps": 500,
7
+ "global_step": 9000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0046641791044776115,
14
+ "grad_norm": 1.907758116722107,
15
+ "learning_rate": 4.9000000000000005e-06,
16
+ "loss": 10.0793,
17
+ "step": 50
18
+ },
19
+ {
20
+ "epoch": 0.009328358208955223,
21
+ "grad_norm": 1.399185061454773,
22
+ "learning_rate": 9.900000000000002e-06,
23
+ "loss": 8.9287,
24
+ "step": 100
25
+ },
26
+ {
27
+ "epoch": 0.013992537313432836,
28
+ "grad_norm": 1.0101089477539062,
29
+ "learning_rate": 1.49e-05,
30
+ "loss": 8.3288,
31
+ "step": 150
32
+ },
33
+ {
34
+ "epoch": 0.018656716417910446,
35
+ "grad_norm": 0.7379341125488281,
36
+ "learning_rate": 1.9900000000000003e-05,
37
+ "loss": 7.6969,
38
+ "step": 200
39
+ },
40
+ {
41
+ "epoch": 0.02332089552238806,
42
+ "grad_norm": 0.9656306505203247,
43
+ "learning_rate": 2.4900000000000002e-05,
44
+ "loss": 7.2698,
45
+ "step": 250
46
+ },
47
+ {
48
+ "epoch": 0.027985074626865673,
49
+ "grad_norm": 1.3790279626846313,
50
+ "learning_rate": 2.9900000000000002e-05,
51
+ "loss": 6.9685,
52
+ "step": 300
53
+ },
54
+ {
55
+ "epoch": 0.03264925373134328,
56
+ "grad_norm": 0.6374709606170654,
57
+ "learning_rate": 3.49e-05,
58
+ "loss": 6.6752,
59
+ "step": 350
60
+ },
61
+ {
62
+ "epoch": 0.03731343283582089,
63
+ "grad_norm": 1.5095465183258057,
64
+ "learning_rate": 3.99e-05,
65
+ "loss": 6.3805,
66
+ "step": 400
67
+ },
68
+ {
69
+ "epoch": 0.04197761194029851,
70
+ "grad_norm": 1.394410490989685,
71
+ "learning_rate": 4.49e-05,
72
+ "loss": 6.1667,
73
+ "step": 450
74
+ },
75
+ {
76
+ "epoch": 0.04664179104477612,
77
+ "grad_norm": 1.7006834745407104,
78
+ "learning_rate": 4.99e-05,
79
+ "loss": 5.9861,
80
+ "step": 500
81
+ },
82
+ {
83
+ "epoch": 0.051305970149253734,
84
+ "grad_norm": 0.9934776425361633,
85
+ "learning_rate": 5.4900000000000006e-05,
86
+ "loss": 5.8166,
87
+ "step": 550
88
+ },
89
+ {
90
+ "epoch": 0.055970149253731345,
91
+ "grad_norm": 2.0550477504730225,
92
+ "learning_rate": 5.99e-05,
93
+ "loss": 5.6646,
94
+ "step": 600
95
+ },
96
+ {
97
+ "epoch": 0.06063432835820896,
98
+ "grad_norm": 2.2504913806915283,
99
+ "learning_rate": 6.49e-05,
100
+ "loss": 5.5353,
101
+ "step": 650
102
+ },
103
+ {
104
+ "epoch": 0.06529850746268656,
105
+ "grad_norm": 2.468212604522705,
106
+ "learning_rate": 6.99e-05,
107
+ "loss": 5.3982,
108
+ "step": 700
109
+ },
110
+ {
111
+ "epoch": 0.06996268656716417,
112
+ "grad_norm": 2.4010939598083496,
113
+ "learning_rate": 7.49e-05,
114
+ "loss": 5.2855,
115
+ "step": 750
116
+ },
117
+ {
118
+ "epoch": 0.07462686567164178,
119
+ "grad_norm": 1.4586199522018433,
120
+ "learning_rate": 7.99e-05,
121
+ "loss": 5.1901,
122
+ "step": 800
123
+ },
124
+ {
125
+ "epoch": 0.07929104477611941,
126
+ "grad_norm": 1.2821402549743652,
127
+ "learning_rate": 8.49e-05,
128
+ "loss": 5.0973,
129
+ "step": 850
130
+ },
131
+ {
132
+ "epoch": 0.08395522388059702,
133
+ "grad_norm": 2.1037180423736572,
134
+ "learning_rate": 8.99e-05,
135
+ "loss": 5.0062,
136
+ "step": 900
137
+ },
138
+ {
139
+ "epoch": 0.08861940298507463,
140
+ "grad_norm": 1.370751976966858,
141
+ "learning_rate": 9.49e-05,
142
+ "loss": 4.9392,
143
+ "step": 950
144
+ },
145
+ {
146
+ "epoch": 0.09328358208955224,
147
+ "grad_norm": 1.276108741760254,
148
+ "learning_rate": 9.99e-05,
149
+ "loss": 4.848,
150
+ "step": 1000
151
+ },
152
+ {
153
+ "epoch": 0.09794776119402986,
154
+ "grad_norm": 1.1077498197555542,
155
+ "learning_rate": 0.0001049,
156
+ "loss": 4.7863,
157
+ "step": 1050
158
+ },
159
+ {
160
+ "epoch": 0.10261194029850747,
161
+ "grad_norm": 1.0212163925170898,
162
+ "learning_rate": 0.0001099,
163
+ "loss": 4.7102,
164
+ "step": 1100
165
+ },
166
+ {
167
+ "epoch": 0.10727611940298508,
168
+ "grad_norm": 1.1480289697647095,
169
+ "learning_rate": 0.0001149,
170
+ "loss": 4.6551,
171
+ "step": 1150
172
+ },
173
+ {
174
+ "epoch": 0.11194029850746269,
175
+ "grad_norm": 1.3106518983840942,
176
+ "learning_rate": 0.00011990000000000001,
177
+ "loss": 4.5627,
178
+ "step": 1200
179
+ },
180
+ {
181
+ "epoch": 0.1166044776119403,
182
+ "grad_norm": 1.4304360151290894,
183
+ "learning_rate": 0.00012490000000000003,
184
+ "loss": 4.4823,
185
+ "step": 1250
186
+ },
187
+ {
188
+ "epoch": 0.12126865671641791,
189
+ "grad_norm": 0.883263885974884,
190
+ "learning_rate": 0.00012989999999999999,
191
+ "loss": 4.4091,
192
+ "step": 1300
193
+ },
194
+ {
195
+ "epoch": 0.1259328358208955,
196
+ "grad_norm": 1.2335679531097412,
197
+ "learning_rate": 0.0001349,
198
+ "loss": 4.3284,
199
+ "step": 1350
200
+ },
201
+ {
202
+ "epoch": 0.13059701492537312,
203
+ "grad_norm": 1.3587566614151,
204
+ "learning_rate": 0.0001399,
205
+ "loss": 4.2523,
206
+ "step": 1400
207
+ },
208
+ {
209
+ "epoch": 0.13526119402985073,
210
+ "grad_norm": 1.4397273063659668,
211
+ "learning_rate": 0.00014490000000000003,
212
+ "loss": 4.1874,
213
+ "step": 1450
214
+ },
215
+ {
216
+ "epoch": 0.13992537313432835,
217
+ "grad_norm": 0.9128643274307251,
218
+ "learning_rate": 0.0001499,
219
+ "loss": 4.1382,
220
+ "step": 1500
221
+ },
222
+ {
223
+ "epoch": 0.14458955223880596,
224
+ "grad_norm": 1.1429646015167236,
225
+ "learning_rate": 0.0001549,
226
+ "loss": 4.0713,
227
+ "step": 1550
228
+ },
229
+ {
230
+ "epoch": 0.14925373134328357,
231
+ "grad_norm": 0.7839804291725159,
232
+ "learning_rate": 0.0001599,
233
+ "loss": 4.0327,
234
+ "step": 1600
235
+ },
236
+ {
237
+ "epoch": 0.15391791044776118,
238
+ "grad_norm": 0.5934556126594543,
239
+ "learning_rate": 0.0001649,
240
+ "loss": 4.0535,
241
+ "step": 1650
242
+ },
243
+ {
244
+ "epoch": 0.15858208955223882,
245
+ "grad_norm": 0.6320812702178955,
246
+ "learning_rate": 0.0001699,
247
+ "loss": 3.9615,
248
+ "step": 1700
249
+ },
250
+ {
251
+ "epoch": 0.16324626865671643,
252
+ "grad_norm": 0.6207342743873596,
253
+ "learning_rate": 0.00017490000000000002,
254
+ "loss": 3.9086,
255
+ "step": 1750
256
+ },
257
+ {
258
+ "epoch": 0.16791044776119404,
259
+ "grad_norm": 0.8524695634841919,
260
+ "learning_rate": 0.0001799,
261
+ "loss": 3.8703,
262
+ "step": 1800
263
+ },
264
+ {
265
+ "epoch": 0.17257462686567165,
266
+ "grad_norm": 0.9148792028427124,
267
+ "learning_rate": 0.0001849,
268
+ "loss": 3.8388,
269
+ "step": 1850
270
+ },
271
+ {
272
+ "epoch": 0.17723880597014927,
273
+ "grad_norm": 0.7455419301986694,
274
+ "learning_rate": 0.0001899,
275
+ "loss": 3.8031,
276
+ "step": 1900
277
+ },
278
+ {
279
+ "epoch": 0.18190298507462688,
280
+ "grad_norm": 0.9777323603630066,
281
+ "learning_rate": 0.00019490000000000002,
282
+ "loss": 3.781,
283
+ "step": 1950
284
+ },
285
+ {
286
+ "epoch": 0.1865671641791045,
287
+ "grad_norm": 0.6018832921981812,
288
+ "learning_rate": 0.0001999,
289
+ "loss": 3.7421,
290
+ "step": 2000
291
+ },
292
+ {
293
+ "epoch": 0.1912313432835821,
294
+ "grad_norm": 0.7007433772087097,
295
+ "learning_rate": 0.00019887614678899083,
296
+ "loss": 3.7148,
297
+ "step": 2050
298
+ },
299
+ {
300
+ "epoch": 0.1958955223880597,
301
+ "grad_norm": 0.7864815592765808,
302
+ "learning_rate": 0.00019772935779816515,
303
+ "loss": 3.6907,
304
+ "step": 2100
305
+ },
306
+ {
307
+ "epoch": 0.20055970149253732,
308
+ "grad_norm": 0.6575994491577148,
309
+ "learning_rate": 0.00019658256880733944,
310
+ "loss": 3.6504,
311
+ "step": 2150
312
+ },
313
+ {
314
+ "epoch": 0.20522388059701493,
315
+ "grad_norm": 0.524401843547821,
316
+ "learning_rate": 0.0001954357798165138,
317
+ "loss": 3.6333,
318
+ "step": 2200
319
+ },
320
+ {
321
+ "epoch": 0.20988805970149255,
322
+ "grad_norm": 0.8242985010147095,
323
+ "learning_rate": 0.00019428899082568808,
324
+ "loss": 3.6084,
325
+ "step": 2250
326
+ },
327
+ {
328
+ "epoch": 0.21455223880597016,
329
+ "grad_norm": 0.5965232849121094,
330
+ "learning_rate": 0.0001931422018348624,
331
+ "loss": 3.5914,
332
+ "step": 2300
333
+ },
334
+ {
335
+ "epoch": 0.21921641791044777,
336
+ "grad_norm": 0.5473845601081848,
337
+ "learning_rate": 0.00019199541284403671,
338
+ "loss": 3.572,
339
+ "step": 2350
340
+ },
341
+ {
342
+ "epoch": 0.22388059701492538,
343
+ "grad_norm": 0.7030166983604431,
344
+ "learning_rate": 0.00019084862385321103,
345
+ "loss": 3.5548,
346
+ "step": 2400
347
+ },
348
+ {
349
+ "epoch": 0.228544776119403,
350
+ "grad_norm": 0.5844017267227173,
351
+ "learning_rate": 0.00018972477064220186,
352
+ "loss": 3.5326,
353
+ "step": 2450
354
+ },
355
+ {
356
+ "epoch": 0.2332089552238806,
357
+ "grad_norm": 0.4722403883934021,
358
+ "learning_rate": 0.00018857798165137615,
359
+ "loss": 3.5113,
360
+ "step": 2500
361
+ },
362
+ {
363
+ "epoch": 0.23787313432835822,
364
+ "grad_norm": 0.4691997468471527,
365
+ "learning_rate": 0.00018743119266055047,
366
+ "loss": 3.4985,
367
+ "step": 2550
368
+ },
369
+ {
370
+ "epoch": 0.24253731343283583,
371
+ "grad_norm": 0.5742476582527161,
372
+ "learning_rate": 0.00018628440366972478,
373
+ "loss": 3.4689,
374
+ "step": 2600
375
+ },
376
+ {
377
+ "epoch": 0.24720149253731344,
378
+ "grad_norm": 0.4867892563343048,
379
+ "learning_rate": 0.0001851376146788991,
380
+ "loss": 3.4737,
381
+ "step": 2650
382
+ },
383
+ {
384
+ "epoch": 0.251865671641791,
385
+ "grad_norm": 0.5876694321632385,
386
+ "learning_rate": 0.0001839908256880734,
387
+ "loss": 3.4516,
388
+ "step": 2700
389
+ },
390
+ {
391
+ "epoch": 0.25652985074626866,
392
+ "grad_norm": 0.4691877067089081,
393
+ "learning_rate": 0.0001828440366972477,
394
+ "loss": 3.4411,
395
+ "step": 2750
396
+ },
397
+ {
398
+ "epoch": 0.26119402985074625,
399
+ "grad_norm": 0.497503399848938,
400
+ "learning_rate": 0.00018169724770642203,
401
+ "loss": 3.4311,
402
+ "step": 2800
403
+ },
404
+ {
405
+ "epoch": 0.2658582089552239,
406
+ "grad_norm": 0.4610927700996399,
407
+ "learning_rate": 0.00018055045871559635,
408
+ "loss": 3.4122,
409
+ "step": 2850
410
+ },
411
+ {
412
+ "epoch": 0.27052238805970147,
413
+ "grad_norm": 0.5312972664833069,
414
+ "learning_rate": 0.00017940366972477064,
415
+ "loss": 3.4004,
416
+ "step": 2900
417
+ },
418
+ {
419
+ "epoch": 0.2751865671641791,
420
+ "grad_norm": 0.5256180167198181,
421
+ "learning_rate": 0.00017825688073394496,
422
+ "loss": 3.3834,
423
+ "step": 2950
424
+ },
425
+ {
426
+ "epoch": 0.2798507462686567,
427
+ "grad_norm": 0.5108594298362732,
428
+ "learning_rate": 0.00017711009174311928,
429
+ "loss": 3.379,
430
+ "step": 3000
431
+ },
432
+ {
433
+ "epoch": 0.28451492537313433,
434
+ "grad_norm": 0.4818528890609741,
435
+ "learning_rate": 0.00017596330275229357,
436
+ "loss": 3.3645,
437
+ "step": 3050
438
+ },
439
+ {
440
+ "epoch": 0.2891791044776119,
441
+ "grad_norm": 0.4077577590942383,
442
+ "learning_rate": 0.0001748165137614679,
443
+ "loss": 3.3659,
444
+ "step": 3100
445
+ },
446
+ {
447
+ "epoch": 0.29384328358208955,
448
+ "grad_norm": 0.4791859984397888,
449
+ "learning_rate": 0.0001736697247706422,
450
+ "loss": 3.3427,
451
+ "step": 3150
452
+ },
453
+ {
454
+ "epoch": 0.29850746268656714,
455
+ "grad_norm": 0.6330239176750183,
456
+ "learning_rate": 0.00017252293577981652,
457
+ "loss": 3.3484,
458
+ "step": 3200
459
+ },
460
+ {
461
+ "epoch": 0.3031716417910448,
462
+ "grad_norm": 0.4511379897594452,
463
+ "learning_rate": 0.00017137614678899084,
464
+ "loss": 3.3302,
465
+ "step": 3250
466
+ },
467
+ {
468
+ "epoch": 0.30783582089552236,
469
+ "grad_norm": 0.4571825861930847,
470
+ "learning_rate": 0.00017022935779816516,
471
+ "loss": 3.3144,
472
+ "step": 3300
473
+ },
474
+ {
475
+ "epoch": 0.3125,
476
+ "grad_norm": 0.44557932019233704,
477
+ "learning_rate": 0.00016908256880733945,
478
+ "loss": 3.3071,
479
+ "step": 3350
480
+ },
481
+ {
482
+ "epoch": 0.31716417910447764,
483
+ "grad_norm": 0.42434263229370117,
484
+ "learning_rate": 0.00016793577981651377,
485
+ "loss": 3.2986,
486
+ "step": 3400
487
+ },
488
+ {
489
+ "epoch": 0.3218283582089552,
490
+ "grad_norm": 0.41813352704048157,
491
+ "learning_rate": 0.0001667889908256881,
492
+ "loss": 3.2902,
493
+ "step": 3450
494
+ },
495
+ {
496
+ "epoch": 0.32649253731343286,
497
+ "grad_norm": 0.4084812104701996,
498
+ "learning_rate": 0.00016564220183486238,
499
+ "loss": 3.283,
500
+ "step": 3500
501
+ },
502
+ {
503
+ "epoch": 0.33115671641791045,
504
+ "grad_norm": 0.38210785388946533,
505
+ "learning_rate": 0.00016449541284403672,
506
+ "loss": 3.2769,
507
+ "step": 3550
508
+ },
509
+ {
510
+ "epoch": 0.3358208955223881,
511
+ "grad_norm": 0.5375980734825134,
512
+ "learning_rate": 0.00016334862385321102,
513
+ "loss": 3.2697,
514
+ "step": 3600
515
+ },
516
+ {
517
+ "epoch": 0.34048507462686567,
518
+ "grad_norm": 0.38781264424324036,
519
+ "learning_rate": 0.00016220183486238533,
520
+ "loss": 3.2556,
521
+ "step": 3650
522
+ },
523
+ {
524
+ "epoch": 0.3451492537313433,
525
+ "grad_norm": 0.41560590267181396,
526
+ "learning_rate": 0.00016105504587155965,
527
+ "loss": 3.2565,
528
+ "step": 3700
529
+ },
530
+ {
531
+ "epoch": 0.3498134328358209,
532
+ "grad_norm": 0.4119664132595062,
533
+ "learning_rate": 0.00015990825688073394,
534
+ "loss": 3.2386,
535
+ "step": 3750
536
+ },
537
+ {
538
+ "epoch": 0.35447761194029853,
539
+ "grad_norm": 0.4190433919429779,
540
+ "learning_rate": 0.00015876146788990826,
541
+ "loss": 3.2345,
542
+ "step": 3800
543
+ },
544
+ {
545
+ "epoch": 0.3591417910447761,
546
+ "grad_norm": 0.37028583884239197,
547
+ "learning_rate": 0.00015761467889908258,
548
+ "loss": 3.226,
549
+ "step": 3850
550
+ },
551
+ {
552
+ "epoch": 0.36380597014925375,
553
+ "grad_norm": 0.3971349000930786,
554
+ "learning_rate": 0.0001564678899082569,
555
+ "loss": 3.2213,
556
+ "step": 3900
557
+ },
558
+ {
559
+ "epoch": 0.36847014925373134,
560
+ "grad_norm": 0.34641316533088684,
561
+ "learning_rate": 0.0001553211009174312,
562
+ "loss": 3.2144,
563
+ "step": 3950
564
+ },
565
+ {
566
+ "epoch": 0.373134328358209,
567
+ "grad_norm": 0.40990573167800903,
568
+ "learning_rate": 0.0001541743119266055,
569
+ "loss": 3.2141,
570
+ "step": 4000
571
+ },
572
+ {
573
+ "epoch": 0.37779850746268656,
574
+ "grad_norm": 0.35828983783721924,
575
+ "learning_rate": 0.00015302752293577983,
576
+ "loss": 3.2054,
577
+ "step": 4050
578
+ },
579
+ {
580
+ "epoch": 0.3824626865671642,
581
+ "grad_norm": 0.37572401762008667,
582
+ "learning_rate": 0.00015188073394495414,
583
+ "loss": 3.1943,
584
+ "step": 4100
585
+ },
586
+ {
587
+ "epoch": 0.3871268656716418,
588
+ "grad_norm": 0.3795240521430969,
589
+ "learning_rate": 0.00015073394495412844,
590
+ "loss": 3.1971,
591
+ "step": 4150
592
+ },
593
+ {
594
+ "epoch": 0.3917910447761194,
595
+ "grad_norm": 0.3884343206882477,
596
+ "learning_rate": 0.00014958715596330275,
597
+ "loss": 3.1877,
598
+ "step": 4200
599
+ },
600
+ {
601
+ "epoch": 0.396455223880597,
602
+ "grad_norm": 0.374661922454834,
603
+ "learning_rate": 0.00014844036697247707,
604
+ "loss": 3.174,
605
+ "step": 4250
606
+ },
607
+ {
608
+ "epoch": 0.40111940298507465,
609
+ "grad_norm": 0.4330432415008545,
610
+ "learning_rate": 0.00014729357798165136,
611
+ "loss": 3.1787,
612
+ "step": 4300
613
+ },
614
+ {
615
+ "epoch": 0.40578358208955223,
616
+ "grad_norm": 0.40475285053253174,
617
+ "learning_rate": 0.0001461467889908257,
618
+ "loss": 3.1781,
619
+ "step": 4350
620
+ },
621
+ {
622
+ "epoch": 0.41044776119402987,
623
+ "grad_norm": 0.36028817296028137,
624
+ "learning_rate": 0.000145,
625
+ "loss": 3.1653,
626
+ "step": 4400
627
+ },
628
+ {
629
+ "epoch": 0.41511194029850745,
630
+ "grad_norm": 0.37242797017097473,
631
+ "learning_rate": 0.00014385321100917432,
632
+ "loss": 3.1608,
633
+ "step": 4450
634
+ },
635
+ {
636
+ "epoch": 0.4197761194029851,
637
+ "grad_norm": 0.42880979180336,
638
+ "learning_rate": 0.00014270642201834864,
639
+ "loss": 3.1656,
640
+ "step": 4500
641
+ },
642
+ {
643
+ "epoch": 0.4244402985074627,
644
+ "grad_norm": 0.3562730848789215,
645
+ "learning_rate": 0.00014155963302752296,
646
+ "loss": 3.1409,
647
+ "step": 4550
648
+ },
649
+ {
650
+ "epoch": 0.4291044776119403,
651
+ "grad_norm": 0.3420323133468628,
652
+ "learning_rate": 0.00014041284403669725,
653
+ "loss": 3.1502,
654
+ "step": 4600
655
+ },
656
+ {
657
+ "epoch": 0.4337686567164179,
658
+ "grad_norm": 0.33985841274261475,
659
+ "learning_rate": 0.00013926605504587157,
660
+ "loss": 3.1437,
661
+ "step": 4650
662
+ },
663
+ {
664
+ "epoch": 0.43843283582089554,
665
+ "grad_norm": 0.35252872109413147,
666
+ "learning_rate": 0.00013811926605504588,
667
+ "loss": 3.1375,
668
+ "step": 4700
669
+ },
670
+ {
671
+ "epoch": 0.4430970149253731,
672
+ "grad_norm": 0.33299019932746887,
673
+ "learning_rate": 0.00013697247706422018,
674
+ "loss": 3.1207,
675
+ "step": 4750
676
+ },
677
+ {
678
+ "epoch": 0.44776119402985076,
679
+ "grad_norm": 0.3340663015842438,
680
+ "learning_rate": 0.00013582568807339452,
681
+ "loss": 3.1285,
682
+ "step": 4800
683
+ },
684
+ {
685
+ "epoch": 0.45242537313432835,
686
+ "grad_norm": 0.3459039330482483,
687
+ "learning_rate": 0.00013470183486238532,
688
+ "loss": 3.1187,
689
+ "step": 4850
690
+ },
691
+ {
692
+ "epoch": 0.457089552238806,
693
+ "grad_norm": 0.34190812706947327,
694
+ "learning_rate": 0.00013355504587155963,
695
+ "loss": 3.1128,
696
+ "step": 4900
697
+ },
698
+ {
699
+ "epoch": 0.46175373134328357,
700
+ "grad_norm": 0.38482001423835754,
701
+ "learning_rate": 0.00013240825688073395,
702
+ "loss": 3.1081,
703
+ "step": 4950
704
+ },
705
+ {
706
+ "epoch": 0.4664179104477612,
707
+ "grad_norm": 0.3969718813896179,
708
+ "learning_rate": 0.00013126146788990827,
709
+ "loss": 3.111,
710
+ "step": 5000
711
+ },
712
+ {
713
+ "epoch": 0.4710820895522388,
714
+ "grad_norm": 0.30978360772132874,
715
+ "learning_rate": 0.0001301146788990826,
716
+ "loss": 3.106,
717
+ "step": 5050
718
+ },
719
+ {
720
+ "epoch": 0.47574626865671643,
721
+ "grad_norm": 0.32029977440834045,
722
+ "learning_rate": 0.00012896788990825688,
723
+ "loss": 3.1095,
724
+ "step": 5100
725
+ },
726
+ {
727
+ "epoch": 0.480410447761194,
728
+ "grad_norm": 0.36698678135871887,
729
+ "learning_rate": 0.0001278440366972477,
730
+ "loss": 3.1017,
731
+ "step": 5150
732
+ },
733
+ {
734
+ "epoch": 0.48507462686567165,
735
+ "grad_norm": 0.3710266947746277,
736
+ "learning_rate": 0.00012669724770642202,
737
+ "loss": 3.0863,
738
+ "step": 5200
739
+ },
740
+ {
741
+ "epoch": 0.48973880597014924,
742
+ "grad_norm": 0.33469706773757935,
743
+ "learning_rate": 0.00012555045871559634,
744
+ "loss": 3.0949,
745
+ "step": 5250
746
+ },
747
+ {
748
+ "epoch": 0.4944029850746269,
749
+ "grad_norm": 0.33493635058403015,
750
+ "learning_rate": 0.00012440366972477063,
751
+ "loss": 3.0794,
752
+ "step": 5300
753
+ },
754
+ {
755
+ "epoch": 0.49906716417910446,
756
+ "grad_norm": 0.3489319682121277,
757
+ "learning_rate": 0.00012325688073394498,
758
+ "loss": 3.0752,
759
+ "step": 5350
760
+ },
761
+ {
762
+ "epoch": 0.503731343283582,
763
+ "grad_norm": 0.32940617203712463,
764
+ "learning_rate": 0.00012211009174311927,
765
+ "loss": 3.0911,
766
+ "step": 5400
767
+ },
768
+ {
769
+ "epoch": 0.5083955223880597,
770
+ "grad_norm": 0.3039519190788269,
771
+ "learning_rate": 0.0001209633027522936,
772
+ "loss": 3.0749,
773
+ "step": 5450
774
+ },
775
+ {
776
+ "epoch": 0.5130597014925373,
777
+ "grad_norm": 0.3196834325790405,
778
+ "learning_rate": 0.00011981651376146789,
779
+ "loss": 3.0685,
780
+ "step": 5500
781
+ },
782
+ {
783
+ "epoch": 0.5177238805970149,
784
+ "grad_norm": 0.35171428322792053,
785
+ "learning_rate": 0.0001186697247706422,
786
+ "loss": 3.0684,
787
+ "step": 5550
788
+ },
789
+ {
790
+ "epoch": 0.5223880597014925,
791
+ "grad_norm": 0.3927328884601593,
792
+ "learning_rate": 0.00011752293577981653,
793
+ "loss": 3.0676,
794
+ "step": 5600
795
+ },
796
+ {
797
+ "epoch": 0.5270522388059702,
798
+ "grad_norm": 0.32428842782974243,
799
+ "learning_rate": 0.00011637614678899082,
800
+ "loss": 3.0635,
801
+ "step": 5650
802
+ },
803
+ {
804
+ "epoch": 0.5317164179104478,
805
+ "grad_norm": 0.3297617435455322,
806
+ "learning_rate": 0.00011522935779816515,
807
+ "loss": 3.0525,
808
+ "step": 5700
809
+ },
810
+ {
811
+ "epoch": 0.5363805970149254,
812
+ "grad_norm": 0.3118542432785034,
813
+ "learning_rate": 0.00011408256880733946,
814
+ "loss": 3.0566,
815
+ "step": 5750
816
+ },
817
+ {
818
+ "epoch": 0.5410447761194029,
819
+ "grad_norm": 0.29949086904525757,
820
+ "learning_rate": 0.00011293577981651377,
821
+ "loss": 3.0454,
822
+ "step": 5800
823
+ },
824
+ {
825
+ "epoch": 0.5457089552238806,
826
+ "grad_norm": 0.3567812740802765,
827
+ "learning_rate": 0.00011178899082568808,
828
+ "loss": 3.0462,
829
+ "step": 5850
830
+ },
831
+ {
832
+ "epoch": 0.5503731343283582,
833
+ "grad_norm": 0.3179719150066376,
834
+ "learning_rate": 0.00011064220183486238,
835
+ "loss": 3.0409,
836
+ "step": 5900
837
+ },
838
+ {
839
+ "epoch": 0.5550373134328358,
840
+ "grad_norm": 0.32883220911026,
841
+ "learning_rate": 0.0001094954128440367,
842
+ "loss": 3.0527,
843
+ "step": 5950
844
+ },
845
+ {
846
+ "epoch": 0.5597014925373134,
847
+ "grad_norm": 0.31516268849372864,
848
+ "learning_rate": 0.000108348623853211,
849
+ "loss": 3.0359,
850
+ "step": 6000
851
+ },
852
+ {
853
+ "epoch": 0.5643656716417911,
854
+ "grad_norm": 0.298753559589386,
855
+ "learning_rate": 0.00010720183486238533,
856
+ "loss": 3.0343,
857
+ "step": 6050
858
+ },
859
+ {
860
+ "epoch": 0.5690298507462687,
861
+ "grad_norm": 0.32886001467704773,
862
+ "learning_rate": 0.00010605504587155963,
863
+ "loss": 3.0229,
864
+ "step": 6100
865
+ },
866
+ {
867
+ "epoch": 0.5736940298507462,
868
+ "grad_norm": 0.3526987135410309,
869
+ "learning_rate": 0.00010490825688073396,
870
+ "loss": 3.0227,
871
+ "step": 6150
872
+ },
873
+ {
874
+ "epoch": 0.5783582089552238,
875
+ "grad_norm": 0.31915777921676636,
876
+ "learning_rate": 0.00010376146788990825,
877
+ "loss": 3.0311,
878
+ "step": 6200
879
+ },
880
+ {
881
+ "epoch": 0.5830223880597015,
882
+ "grad_norm": 0.31342729926109314,
883
+ "learning_rate": 0.00010261467889908258,
884
+ "loss": 3.0257,
885
+ "step": 6250
886
+ },
887
+ {
888
+ "epoch": 0.5876865671641791,
889
+ "grad_norm": 0.3711305856704712,
890
+ "learning_rate": 0.00010146788990825689,
891
+ "loss": 3.0125,
892
+ "step": 6300
893
+ },
894
+ {
895
+ "epoch": 0.5923507462686567,
896
+ "grad_norm": 0.32861173152923584,
897
+ "learning_rate": 0.00010032110091743118,
898
+ "loss": 3.0235,
899
+ "step": 6350
900
+ },
901
+ {
902
+ "epoch": 0.5970149253731343,
903
+ "grad_norm": 0.30850905179977417,
904
+ "learning_rate": 9.917431192660551e-05,
905
+ "loss": 3.0171,
906
+ "step": 6400
907
+ },
908
+ {
909
+ "epoch": 0.601679104477612,
910
+ "grad_norm": 0.31126561760902405,
911
+ "learning_rate": 9.802752293577983e-05,
912
+ "loss": 3.0177,
913
+ "step": 6450
914
+ },
915
+ {
916
+ "epoch": 0.6063432835820896,
917
+ "grad_norm": 0.3373335897922516,
918
+ "learning_rate": 9.688073394495414e-05,
919
+ "loss": 3.0138,
920
+ "step": 6500
921
+ },
922
+ {
923
+ "epoch": 0.6110074626865671,
924
+ "grad_norm": 0.32974258065223694,
925
+ "learning_rate": 9.575688073394496e-05,
926
+ "loss": 3.0092,
927
+ "step": 6550
928
+ },
929
+ {
930
+ "epoch": 0.6156716417910447,
931
+ "grad_norm": 0.29914727807044983,
932
+ "learning_rate": 9.463302752293578e-05,
933
+ "loss": 3.0075,
934
+ "step": 6600
935
+ },
936
+ {
937
+ "epoch": 0.6203358208955224,
938
+ "grad_norm": 0.3006197214126587,
939
+ "learning_rate": 9.34862385321101e-05,
940
+ "loss": 3.015,
941
+ "step": 6650
942
+ },
943
+ {
944
+ "epoch": 0.625,
945
+ "grad_norm": 0.2997278869152069,
946
+ "learning_rate": 9.23394495412844e-05,
947
+ "loss": 3.0037,
948
+ "step": 6700
949
+ },
950
+ {
951
+ "epoch": 0.6296641791044776,
952
+ "grad_norm": 0.31719309091567993,
953
+ "learning_rate": 9.119266055045872e-05,
954
+ "loss": 2.9969,
955
+ "step": 6750
956
+ },
957
+ {
958
+ "epoch": 0.6343283582089553,
959
+ "grad_norm": 0.3004010319709778,
960
+ "learning_rate": 9.004587155963303e-05,
961
+ "loss": 2.9909,
962
+ "step": 6800
963
+ },
964
+ {
965
+ "epoch": 0.6389925373134329,
966
+ "grad_norm": 0.3123314082622528,
967
+ "learning_rate": 8.889908256880735e-05,
968
+ "loss": 2.9853,
969
+ "step": 6850
970
+ },
971
+ {
972
+ "epoch": 0.6436567164179104,
973
+ "grad_norm": 0.334837406873703,
974
+ "learning_rate": 8.775229357798166e-05,
975
+ "loss": 2.9923,
976
+ "step": 6900
977
+ },
978
+ {
979
+ "epoch": 0.648320895522388,
980
+ "grad_norm": 0.3055480420589447,
981
+ "learning_rate": 8.660550458715597e-05,
982
+ "loss": 2.9897,
983
+ "step": 6950
984
+ },
985
+ {
986
+ "epoch": 0.6529850746268657,
987
+ "grad_norm": 0.33151915669441223,
988
+ "learning_rate": 8.545871559633027e-05,
989
+ "loss": 2.9843,
990
+ "step": 7000
991
+ },
992
+ {
993
+ "epoch": 0.6576492537313433,
994
+ "grad_norm": 0.32511621713638306,
995
+ "learning_rate": 8.431192660550459e-05,
996
+ "loss": 2.9797,
997
+ "step": 7050
998
+ },
999
+ {
1000
+ "epoch": 0.6623134328358209,
1001
+ "grad_norm": 0.29789265990257263,
1002
+ "learning_rate": 8.31651376146789e-05,
1003
+ "loss": 2.9789,
1004
+ "step": 7100
1005
+ },
1006
+ {
1007
+ "epoch": 0.6669776119402985,
1008
+ "grad_norm": 0.3272823393344879,
1009
+ "learning_rate": 8.201834862385321e-05,
1010
+ "loss": 2.971,
1011
+ "step": 7150
1012
+ },
1013
+ {
1014
+ "epoch": 0.6716417910447762,
1015
+ "grad_norm": 0.2983226478099823,
1016
+ "learning_rate": 8.087155963302753e-05,
1017
+ "loss": 2.9818,
1018
+ "step": 7200
1019
+ },
1020
+ {
1021
+ "epoch": 0.6763059701492538,
1022
+ "grad_norm": 0.30532923340797424,
1023
+ "learning_rate": 7.972477064220184e-05,
1024
+ "loss": 2.9739,
1025
+ "step": 7250
1026
+ },
1027
+ {
1028
+ "epoch": 0.6809701492537313,
1029
+ "grad_norm": 0.3046092092990875,
1030
+ "learning_rate": 7.857798165137616e-05,
1031
+ "loss": 2.9575,
1032
+ "step": 7300
1033
+ },
1034
+ {
1035
+ "epoch": 0.6856343283582089,
1036
+ "grad_norm": 0.3133400082588196,
1037
+ "learning_rate": 7.743119266055046e-05,
1038
+ "loss": 2.9794,
1039
+ "step": 7350
1040
+ },
1041
+ {
1042
+ "epoch": 0.6902985074626866,
1043
+ "grad_norm": 0.3090495765209198,
1044
+ "learning_rate": 7.628440366972477e-05,
1045
+ "loss": 2.9632,
1046
+ "step": 7400
1047
+ },
1048
+ {
1049
+ "epoch": 0.6949626865671642,
1050
+ "grad_norm": 0.2925854027271271,
1051
+ "learning_rate": 7.513761467889908e-05,
1052
+ "loss": 2.9649,
1053
+ "step": 7450
1054
+ },
1055
+ {
1056
+ "epoch": 0.6996268656716418,
1057
+ "grad_norm": 0.30464622378349304,
1058
+ "learning_rate": 7.399082568807339e-05,
1059
+ "loss": 2.9606,
1060
+ "step": 7500
1061
+ },
1062
+ {
1063
+ "epoch": 0.7042910447761194,
1064
+ "grad_norm": 0.29478827118873596,
1065
+ "learning_rate": 7.284403669724771e-05,
1066
+ "loss": 2.9574,
1067
+ "step": 7550
1068
+ },
1069
+ {
1070
+ "epoch": 0.7089552238805971,
1071
+ "grad_norm": 0.30368906259536743,
1072
+ "learning_rate": 7.169724770642203e-05,
1073
+ "loss": 2.9583,
1074
+ "step": 7600
1075
+ },
1076
+ {
1077
+ "epoch": 0.7136194029850746,
1078
+ "grad_norm": 0.2974824905395508,
1079
+ "learning_rate": 7.055045871559633e-05,
1080
+ "loss": 2.9566,
1081
+ "step": 7650
1082
+ },
1083
+ {
1084
+ "epoch": 0.7182835820895522,
1085
+ "grad_norm": 0.3006385564804077,
1086
+ "learning_rate": 6.940366972477065e-05,
1087
+ "loss": 2.953,
1088
+ "step": 7700
1089
+ },
1090
+ {
1091
+ "epoch": 0.7229477611940298,
1092
+ "grad_norm": 0.2787078022956848,
1093
+ "learning_rate": 6.825688073394497e-05,
1094
+ "loss": 2.9466,
1095
+ "step": 7750
1096
+ },
1097
+ {
1098
+ "epoch": 0.7276119402985075,
1099
+ "grad_norm": 0.2966342568397522,
1100
+ "learning_rate": 6.711009174311927e-05,
1101
+ "loss": 2.9446,
1102
+ "step": 7800
1103
+ },
1104
+ {
1105
+ "epoch": 0.7322761194029851,
1106
+ "grad_norm": 0.32728853821754456,
1107
+ "learning_rate": 6.600917431192661e-05,
1108
+ "loss": 2.9486,
1109
+ "step": 7850
1110
+ },
1111
+ {
1112
+ "epoch": 0.7369402985074627,
1113
+ "grad_norm": 0.325846403837204,
1114
+ "learning_rate": 6.486238532110092e-05,
1115
+ "loss": 2.9456,
1116
+ "step": 7900
1117
+ },
1118
+ {
1119
+ "epoch": 0.7416044776119403,
1120
+ "grad_norm": 0.2933451235294342,
1121
+ "learning_rate": 6.371559633027524e-05,
1122
+ "loss": 2.9404,
1123
+ "step": 7950
1124
+ },
1125
+ {
1126
+ "epoch": 0.746268656716418,
1127
+ "grad_norm": 0.3031325042247772,
1128
+ "learning_rate": 6.256880733944954e-05,
1129
+ "loss": 2.9486,
1130
+ "step": 8000
1131
+ },
1132
+ {
1133
+ "epoch": 0.7509328358208955,
1134
+ "grad_norm": 0.30206266045570374,
1135
+ "learning_rate": 6.142201834862386e-05,
1136
+ "loss": 2.9364,
1137
+ "step": 8050
1138
+ },
1139
+ {
1140
+ "epoch": 0.7555970149253731,
1141
+ "grad_norm": 0.29706859588623047,
1142
+ "learning_rate": 6.027522935779817e-05,
1143
+ "loss": 2.9483,
1144
+ "step": 8100
1145
+ },
1146
+ {
1147
+ "epoch": 0.7602611940298507,
1148
+ "grad_norm": 0.2909660339355469,
1149
+ "learning_rate": 5.912844036697248e-05,
1150
+ "loss": 2.9405,
1151
+ "step": 8150
1152
+ },
1153
+ {
1154
+ "epoch": 0.7649253731343284,
1155
+ "grad_norm": 0.3136695623397827,
1156
+ "learning_rate": 5.798165137614679e-05,
1157
+ "loss": 2.9366,
1158
+ "step": 8200
1159
+ },
1160
+ {
1161
+ "epoch": 0.769589552238806,
1162
+ "grad_norm": 0.29548221826553345,
1163
+ "learning_rate": 5.683486238532111e-05,
1164
+ "loss": 2.9305,
1165
+ "step": 8250
1166
+ },
1167
+ {
1168
+ "epoch": 0.7742537313432836,
1169
+ "grad_norm": 0.2777513265609741,
1170
+ "learning_rate": 5.568807339449541e-05,
1171
+ "loss": 2.9377,
1172
+ "step": 8300
1173
+ },
1174
+ {
1175
+ "epoch": 0.7789179104477612,
1176
+ "grad_norm": 0.28673887252807617,
1177
+ "learning_rate": 5.454128440366972e-05,
1178
+ "loss": 2.9342,
1179
+ "step": 8350
1180
+ },
1181
+ {
1182
+ "epoch": 0.7835820895522388,
1183
+ "grad_norm": 0.2911510169506073,
1184
+ "learning_rate": 5.339449541284404e-05,
1185
+ "loss": 2.938,
1186
+ "step": 8400
1187
+ },
1188
+ {
1189
+ "epoch": 0.7882462686567164,
1190
+ "grad_norm": 0.3052498698234558,
1191
+ "learning_rate": 5.224770642201835e-05,
1192
+ "loss": 2.9235,
1193
+ "step": 8450
1194
+ },
1195
+ {
1196
+ "epoch": 0.792910447761194,
1197
+ "grad_norm": 0.30528631806373596,
1198
+ "learning_rate": 5.110091743119266e-05,
1199
+ "loss": 2.9393,
1200
+ "step": 8500
1201
+ },
1202
+ {
1203
+ "epoch": 0.7975746268656716,
1204
+ "grad_norm": 0.27922648191452026,
1205
+ "learning_rate": 4.9954128440366974e-05,
1206
+ "loss": 2.9153,
1207
+ "step": 8550
1208
+ },
1209
+ {
1210
+ "epoch": 0.8022388059701493,
1211
+ "grad_norm": 0.2799355685710907,
1212
+ "learning_rate": 4.880733944954129e-05,
1213
+ "loss": 2.9329,
1214
+ "step": 8600
1215
+ },
1216
+ {
1217
+ "epoch": 0.8069029850746269,
1218
+ "grad_norm": 0.3089926242828369,
1219
+ "learning_rate": 4.76605504587156e-05,
1220
+ "loss": 2.9203,
1221
+ "step": 8650
1222
+ },
1223
+ {
1224
+ "epoch": 0.8115671641791045,
1225
+ "grad_norm": 0.3086063265800476,
1226
+ "learning_rate": 4.651376146788991e-05,
1227
+ "loss": 2.9223,
1228
+ "step": 8700
1229
+ },
1230
+ {
1231
+ "epoch": 0.816231343283582,
1232
+ "grad_norm": 0.3019801676273346,
1233
+ "learning_rate": 4.536697247706422e-05,
1234
+ "loss": 2.9159,
1235
+ "step": 8750
1236
+ },
1237
+ {
1238
+ "epoch": 0.8208955223880597,
1239
+ "grad_norm": 0.2975527048110962,
1240
+ "learning_rate": 4.422018348623854e-05,
1241
+ "loss": 2.9215,
1242
+ "step": 8800
1243
+ },
1244
+ {
1245
+ "epoch": 0.8255597014925373,
1246
+ "grad_norm": 0.30920466780662537,
1247
+ "learning_rate": 4.3073394495412844e-05,
1248
+ "loss": 2.9213,
1249
+ "step": 8850
1250
+ },
1251
+ {
1252
+ "epoch": 0.8302238805970149,
1253
+ "grad_norm": 0.2723771929740906,
1254
+ "learning_rate": 4.1926605504587155e-05,
1255
+ "loss": 2.9119,
1256
+ "step": 8900
1257
+ },
1258
+ {
1259
+ "epoch": 0.8348880597014925,
1260
+ "grad_norm": 0.28407421708106995,
1261
+ "learning_rate": 4.0779816513761474e-05,
1262
+ "loss": 2.9101,
1263
+ "step": 8950
1264
+ },
1265
+ {
1266
+ "epoch": 0.8395522388059702,
1267
+ "grad_norm": 0.2752555012702942,
1268
+ "learning_rate": 3.9633027522935785e-05,
1269
+ "loss": 2.9087,
1270
+ "step": 9000
1271
+ }
1272
+ ],
1273
+ "logging_steps": 50,
1274
+ "max_steps": 10720,
1275
+ "num_input_tokens_seen": 0,
1276
+ "num_train_epochs": 1,
1277
+ "save_steps": 1000,
1278
+ "stateful_callbacks": {
1279
+ "TrainerControl": {
1280
+ "args": {
1281
+ "should_epoch_stop": false,
1282
+ "should_evaluate": false,
1283
+ "should_log": false,
1284
+ "should_save": true,
1285
+ "should_training_stop": false
1286
+ },
1287
+ "attributes": {}
1288
+ }
1289
+ },
1290
+ "total_flos": 1911800851333120.0,
1291
+ "train_batch_size": 64,
1292
+ "trial_name": null,
1293
+ "trial_params": null
1294
+ }
full/checkpoint-9000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0a1e4e68442fcaca00226db2169575a68dcecd4ab600ec17acb8a650bc00a2d
3
+ size 7121
full/checkpoint-9000/zero_to_fp32.py ADDED
@@ -0,0 +1,760 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example:
14
+ # python zero_to_fp32.py . output_dir/
15
+ # or
16
+ # python zero_to_fp32.py . output_dir/ --safe_serialization
17
+
18
+ import argparse
19
+ import torch
20
+ import glob
21
+ import math
22
+ import os
23
+ import re
24
+ import gc
25
+ import json
26
+ import numpy as np
27
+ from tqdm import tqdm
28
+ from collections import OrderedDict
29
+ from dataclasses import dataclass
30
+
31
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
32
+ # DeepSpeed data structures it has to be available in the current python environment.
33
+ from deepspeed.utils import logger
34
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
35
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
36
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
37
+
38
+
39
+ @dataclass
40
+ class zero_model_state:
41
+ buffers: dict()
42
+ param_shapes: dict()
43
+ shared_params: list
44
+ ds_version: int
45
+ frozen_param_shapes: dict()
46
+ frozen_param_fragments: dict()
47
+
48
+
49
+ debug = 0
50
+
51
+ # load to cpu
52
+ device = torch.device('cpu')
53
+
54
+
55
+ def atoi(text):
56
+ return int(text) if text.isdigit() else text
57
+
58
+
59
+ def natural_keys(text):
60
+ '''
61
+ alist.sort(key=natural_keys) sorts in human order
62
+ http://nedbatchelder.com/blog/200712/human_sorting.html
63
+ (See Toothy's implementation in the comments)
64
+ '''
65
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
66
+
67
+
68
+ def get_model_state_file(checkpoint_dir, zero_stage):
69
+ if not os.path.isdir(checkpoint_dir):
70
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
71
+
72
+ # there should be only one file
73
+ if zero_stage <= 2:
74
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
75
+ elif zero_stage == 3:
76
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
77
+
78
+ if not os.path.exists(file):
79
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
80
+
81
+ return file
82
+
83
+
84
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
85
+ # XXX: need to test that this simple glob rule works for multi-node setup too
86
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
87
+
88
+ if len(ckpt_files) == 0:
89
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
90
+
91
+ return ckpt_files
92
+
93
+
94
+ def get_optim_files(checkpoint_dir):
95
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
96
+
97
+
98
+ def get_model_state_files(checkpoint_dir):
99
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
100
+
101
+
102
+ def parse_model_states(files):
103
+ zero_model_states = []
104
+ for file in files:
105
+ state_dict = torch.load(file, map_location=device, weights_only=False)
106
+
107
+ if BUFFER_NAMES not in state_dict:
108
+ raise ValueError(f"{file} is not a model state checkpoint")
109
+ buffer_names = state_dict[BUFFER_NAMES]
110
+ if debug:
111
+ print("Found buffers:", buffer_names)
112
+
113
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
114
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
115
+ param_shapes = state_dict[PARAM_SHAPES]
116
+
117
+ # collect parameters that are included in param_shapes
118
+ param_names = []
119
+ for s in param_shapes:
120
+ for name in s.keys():
121
+ param_names.append(name)
122
+
123
+ # update with frozen parameters
124
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
125
+ if frozen_param_shapes is not None:
126
+ if debug:
127
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
128
+ param_names += list(frozen_param_shapes.keys())
129
+
130
+ # handle shared params
131
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
132
+
133
+ ds_version = state_dict.get(DS_VERSION, None)
134
+
135
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
136
+
137
+ z_model_state = zero_model_state(buffers=buffers,
138
+ param_shapes=param_shapes,
139
+ shared_params=shared_params,
140
+ ds_version=ds_version,
141
+ frozen_param_shapes=frozen_param_shapes,
142
+ frozen_param_fragments=frozen_param_fragments)
143
+ zero_model_states.append(z_model_state)
144
+
145
+ return zero_model_states
146
+
147
+
148
+ def parse_optim_states(files, ds_checkpoint_dir):
149
+ total_files = len(files)
150
+ state_dicts = []
151
+ for f in tqdm(files, desc='Loading checkpoint shards'):
152
+ state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
153
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
154
+ # and also handle the case where it was already removed by another helper script
155
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
156
+ state_dicts.append(state_dict)
157
+
158
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
159
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
160
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
161
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
162
+
163
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
164
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
165
+ # use the max of the partition_count to get the dp world_size.
166
+
167
+ if type(world_size) is list:
168
+ world_size = max(world_size)
169
+
170
+ if world_size != total_files:
171
+ raise ValueError(
172
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
173
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
174
+ )
175
+
176
+ # the groups are named differently in each stage
177
+ if zero_stage <= 2:
178
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
179
+ elif zero_stage == 3:
180
+ fp32_groups_key = FP32_FLAT_GROUPS
181
+ else:
182
+ raise ValueError(f"unknown zero stage {zero_stage}")
183
+
184
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
185
+ return zero_stage, world_size, fp32_flat_groups
186
+
187
+
188
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
189
+ """
190
+ Returns fp32 state_dict reconstructed from ds checkpoint
191
+
192
+ Args:
193
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
194
+
195
+ """
196
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
197
+
198
+ optim_files = get_optim_files(ds_checkpoint_dir)
199
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
200
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
201
+
202
+ model_files = get_model_state_files(ds_checkpoint_dir)
203
+
204
+ zero_model_states = parse_model_states(model_files)
205
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
206
+
207
+ if zero_stage <= 2:
208
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
209
+ exclude_frozen_parameters)
210
+ elif zero_stage == 3:
211
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
212
+ exclude_frozen_parameters)
213
+
214
+
215
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
216
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
217
+ return
218
+
219
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
220
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
221
+
222
+ if debug:
223
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
224
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
225
+
226
+ wanted_params = len(frozen_param_shapes)
227
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
228
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
229
+ print(f'Frozen params: Have {avail_numel} numels to process.')
230
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
231
+
232
+ total_params = 0
233
+ total_numel = 0
234
+ for name, shape in frozen_param_shapes.items():
235
+ total_params += 1
236
+ unpartitioned_numel = shape.numel()
237
+ total_numel += unpartitioned_numel
238
+
239
+ state_dict[name] = frozen_param_fragments[name]
240
+
241
+ if debug:
242
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
243
+
244
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
245
+
246
+
247
+ def _has_callable(obj, fn):
248
+ attr = getattr(obj, fn, None)
249
+ return callable(attr)
250
+
251
+
252
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
253
+ param_shapes = zero_model_states[0].param_shapes
254
+
255
+ # Reconstruction protocol:
256
+ #
257
+ # XXX: document this
258
+
259
+ if debug:
260
+ for i in range(world_size):
261
+ for j in range(len(fp32_flat_groups[0])):
262
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
263
+
264
+ # XXX: memory usage doubles here (zero2)
265
+ num_param_groups = len(fp32_flat_groups[0])
266
+ merged_single_partition_of_fp32_groups = []
267
+ for i in range(num_param_groups):
268
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
269
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
270
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
271
+ avail_numel = sum(
272
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
273
+
274
+ if debug:
275
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
276
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
277
+ # not asserting if there is a mismatch due to possible padding
278
+ print(f"Have {avail_numel} numels to process.")
279
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
280
+
281
+ # params
282
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
283
+ # out-of-core computing solution
284
+ total_numel = 0
285
+ total_params = 0
286
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
287
+ offset = 0
288
+ avail_numel = full_single_fp32_vector.numel()
289
+ for name, shape in shapes.items():
290
+
291
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
292
+ total_numel += unpartitioned_numel
293
+ total_params += 1
294
+
295
+ if debug:
296
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
297
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
298
+ offset += unpartitioned_numel
299
+
300
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
301
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
302
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
303
+ # live optimizer object, so we are checking that the numbers are within the right range
304
+ align_to = 2 * world_size
305
+
306
+ def zero2_align(x):
307
+ return align_to * math.ceil(x / align_to)
308
+
309
+ if debug:
310
+ print(f"original offset={offset}, avail_numel={avail_numel}")
311
+
312
+ offset = zero2_align(offset)
313
+ avail_numel = zero2_align(avail_numel)
314
+
315
+ if debug:
316
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
317
+
318
+ # Sanity check
319
+ if offset != avail_numel:
320
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
321
+
322
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
323
+
324
+
325
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
326
+ exclude_frozen_parameters):
327
+ state_dict = OrderedDict()
328
+
329
+ # buffers
330
+ buffers = zero_model_states[0].buffers
331
+ state_dict.update(buffers)
332
+ if debug:
333
+ print(f"added {len(buffers)} buffers")
334
+
335
+ if not exclude_frozen_parameters:
336
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
337
+
338
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
339
+
340
+ # recover shared parameters
341
+ for pair in zero_model_states[0].shared_params:
342
+ if pair[1] in state_dict:
343
+ state_dict[pair[0]] = state_dict[pair[1]]
344
+
345
+ return state_dict
346
+
347
+
348
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
349
+ remainder = unpartitioned_numel % world_size
350
+ padding_numel = (world_size - remainder) if remainder else 0
351
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
352
+ return partitioned_numel, padding_numel
353
+
354
+
355
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
356
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
357
+ return
358
+
359
+ if debug:
360
+ for i in range(world_size):
361
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
362
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
363
+
364
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
365
+ wanted_params = len(frozen_param_shapes)
366
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
367
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
368
+ print(f'Frozen params: Have {avail_numel} numels to process.')
369
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
370
+
371
+ total_params = 0
372
+ total_numel = 0
373
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
374
+ total_params += 1
375
+ unpartitioned_numel = shape.numel()
376
+ total_numel += unpartitioned_numel
377
+
378
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
379
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
380
+
381
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
382
+
383
+ if debug:
384
+ print(
385
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
386
+ )
387
+
388
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
389
+
390
+
391
+ class GatheredTensor:
392
+ """
393
+ A pseudo tensor that collects partitioned weights.
394
+ It is more memory efficient when there are multiple groups.
395
+ """
396
+
397
+ def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
398
+ self.flat_groups = flat_groups
399
+ self.flat_groups_offset = flat_groups_offset
400
+ self.offset = offset
401
+ self.partitioned_numel = partitioned_numel
402
+ self.shape = shape
403
+ self.dtype = self.flat_groups[0][0].dtype
404
+
405
+ def contiguous(self):
406
+ """
407
+ Merge partitioned weights from flat_groups into a single tensor.
408
+ """
409
+ end_idx = self.offset + self.partitioned_numel
410
+ world_size = len(self.flat_groups)
411
+ pad_flat_param_chunks = []
412
+
413
+ for rank_i in range(world_size):
414
+ # for each rank, we need to collect weights from related group/groups
415
+ flat_groups_at_rank_i = self.flat_groups[rank_i]
416
+ start_group_id = None
417
+ end_group_id = None
418
+ for group_id in range(len(self.flat_groups_offset)):
419
+ if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
420
+ start_group_id = group_id
421
+ if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
422
+ end_group_id = group_id
423
+ break
424
+ # collect weights from related group/groups
425
+ for group_id in range(start_group_id, end_group_id + 1):
426
+ flat_tensor = flat_groups_at_rank_i[group_id]
427
+ start_offset = self.offset - self.flat_groups_offset[group_id]
428
+ end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
429
+ pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
430
+
431
+ # collect weights from all ranks
432
+ pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
433
+ param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
434
+ return param
435
+
436
+
437
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
438
+ param_shapes = zero_model_states[0].param_shapes
439
+ avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
440
+
441
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
442
+ # param, re-consolidating each param, while dealing with padding if any
443
+
444
+ # merge list of dicts, preserving order
445
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
446
+
447
+ if debug:
448
+ for i in range(world_size):
449
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
450
+
451
+ wanted_params = len(param_shapes)
452
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
453
+ # not asserting if there is a mismatch due to possible padding
454
+ avail_numel = fp32_flat_groups[0].numel() * world_size
455
+ print(f"Trainable params: Have {avail_numel} numels to process.")
456
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
457
+
458
+ # params
459
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
460
+ # out-of-core computing solution
461
+ offset = 0
462
+ total_numel = 0
463
+ total_params = 0
464
+ flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
465
+ for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
466
+ unpartitioned_numel = shape.numel()
467
+ total_numel += unpartitioned_numel
468
+ total_params += 1
469
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
470
+
471
+ if debug:
472
+ print(
473
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
474
+ )
475
+
476
+ # memory efficient tensor
477
+ tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
478
+ state_dict[name] = tensor
479
+ offset += partitioned_numel
480
+
481
+ offset *= world_size
482
+
483
+ # Sanity check
484
+ if offset != avail_numel:
485
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
486
+
487
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
488
+
489
+
490
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
491
+ exclude_frozen_parameters):
492
+ state_dict = OrderedDict()
493
+
494
+ # buffers
495
+ buffers = zero_model_states[0].buffers
496
+ state_dict.update(buffers)
497
+ if debug:
498
+ print(f"added {len(buffers)} buffers")
499
+
500
+ if not exclude_frozen_parameters:
501
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
502
+
503
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
504
+
505
+ # recover shared parameters
506
+ for pair in zero_model_states[0].shared_params:
507
+ if pair[1] in state_dict:
508
+ state_dict[pair[0]] = state_dict[pair[1]]
509
+
510
+ return state_dict
511
+
512
+
513
+ def to_torch_tensor(state_dict, return_empty_tensor=False):
514
+ """
515
+ Convert state_dict of GatheredTensor to torch tensor
516
+ """
517
+ torch_state_dict = {}
518
+ converted_tensors = {}
519
+ for name, tensor in state_dict.items():
520
+ tensor_id = id(tensor)
521
+ if tensor_id in converted_tensors: # shared tensors
522
+ shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
523
+ torch_state_dict[name] = shared_tensor
524
+ else:
525
+ converted_tensors[tensor_id] = name
526
+ if return_empty_tensor:
527
+ torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
528
+ else:
529
+ torch_state_dict[name] = tensor.contiguous()
530
+ return torch_state_dict
531
+
532
+
533
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
534
+ tag=None,
535
+ exclude_frozen_parameters=False,
536
+ lazy_mode=False):
537
+ """
538
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
539
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
540
+ via a model hub.
541
+
542
+ Args:
543
+ - ``checkpoint_dir``: path to the desired checkpoint folder
544
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
545
+ - ``exclude_frozen_parameters``: exclude frozen parameters
546
+ - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
547
+ Convert the pesduo tensor to torch tensor by ``.contiguous()``
548
+
549
+ Returns:
550
+ - pytorch ``state_dict``
551
+
552
+ A typical usage might be ::
553
+
554
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
555
+ # do the training and checkpoint saving
556
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
557
+ model = model.cpu() # move to cpu
558
+ model.load_state_dict(state_dict)
559
+ # submit to model hub or save the model to share with others
560
+
561
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
562
+ application. i.e. you will need to re-initialize the deepspeed engine, since
563
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
564
+
565
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
566
+
567
+ Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
568
+ You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
569
+ the checkpoint. Or you can load state_dict in lazy mode ::
570
+
571
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
572
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
573
+ for name, lazy_tensor in state_dict.item():
574
+ tensor = lazy_tensor.contiguous() # to cpu
575
+ print(name, tensor)
576
+ # del tensor to release memory if it no longer in use
577
+ """
578
+ if tag is None:
579
+ latest_path = os.path.join(checkpoint_dir, 'latest')
580
+ if os.path.isfile(latest_path):
581
+ with open(latest_path, 'r') as fd:
582
+ tag = fd.read().strip()
583
+ else:
584
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
585
+
586
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
587
+
588
+ if not os.path.isdir(ds_checkpoint_dir):
589
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
590
+
591
+ state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
592
+ if lazy_mode:
593
+ return state_dict
594
+ else:
595
+ return to_torch_tensor(state_dict)
596
+
597
+
598
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
599
+ output_dir,
600
+ max_shard_size="5GB",
601
+ safe_serialization=False,
602
+ tag=None,
603
+ exclude_frozen_parameters=False):
604
+ """
605
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
606
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
607
+
608
+ Args:
609
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
610
+ - ``output_dir``: directory to the pytorch fp32 state_dict output files
611
+ - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
612
+ - ``safe_serialization``: whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
613
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
614
+ - ``exclude_frozen_parameters``: exclude frozen parameters
615
+ """
616
+
617
+ # Dependency pre-check
618
+ if safe_serialization:
619
+ try:
620
+ from safetensors.torch import save_file
621
+ except ImportError:
622
+ print('If you want to use `safe_serialization`, please `pip install safetensors`')
623
+ raise
624
+ if max_shard_size is not None:
625
+ try:
626
+ from huggingface_hub import split_torch_state_dict_into_shards
627
+ except ImportError:
628
+ print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
629
+ raise
630
+
631
+ # Convert zero checkpoint to state_dict
632
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
633
+ tag,
634
+ exclude_frozen_parameters,
635
+ lazy_mode=True)
636
+
637
+ # Shard the model if it is too big.
638
+ weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
639
+ if max_shard_size is not None:
640
+ filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
641
+ # an memory-efficient approach for sharding
642
+ empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
643
+ state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
644
+ filename_pattern=filename_pattern,
645
+ max_shard_size=max_shard_size)
646
+ else:
647
+ from collections import namedtuple
648
+ StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
649
+ state_dict_split = StateDictSplit(is_sharded=False,
650
+ filename_to_tensors={weights_name: list(state_dict.keys())})
651
+
652
+ # Save the model by shard
653
+ os.makedirs(output_dir, exist_ok=True)
654
+ filename_to_tensors = state_dict_split.filename_to_tensors.items()
655
+ for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
656
+ shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
657
+ shard_state_dict = to_torch_tensor(shard_state_dict)
658
+ output_path = os.path.join(output_dir, shard_file)
659
+ if safe_serialization:
660
+ save_file(shard_state_dict, output_path, metadata={"format": "pt"})
661
+ else:
662
+ torch.save(shard_state_dict, output_path)
663
+ # release the memory of current shard
664
+ for tensor_name in list(shard_state_dict.keys()):
665
+ del state_dict[tensor_name]
666
+ del shard_state_dict[tensor_name]
667
+ del shard_state_dict
668
+ gc.collect()
669
+
670
+ # Save index if sharded
671
+ if state_dict_split.is_sharded:
672
+ index = {
673
+ "metadata": state_dict_split.metadata,
674
+ "weight_map": state_dict_split.tensor_to_filename,
675
+ }
676
+ save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
677
+ save_index_file = os.path.join(output_dir, save_index_file)
678
+ with open(save_index_file, "w", encoding="utf-8") as f:
679
+ content = json.dumps(index, indent=2, sort_keys=True) + "\n"
680
+ f.write(content)
681
+
682
+
683
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
684
+ """
685
+ 1. Put the provided model to cpu
686
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
687
+ 3. Load it into the provided model
688
+
689
+ Args:
690
+ - ``model``: the model object to update
691
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
692
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
693
+
694
+ Returns:
695
+ - ``model`: modified model
696
+
697
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
698
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
699
+ conveniently placed for you in the checkpoint folder.
700
+
701
+ A typical usage might be ::
702
+
703
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
704
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
705
+ # submit to model hub or save the model to share with others
706
+
707
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
708
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
709
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
710
+
711
+ """
712
+ logger.info(f"Extracting fp32 weights")
713
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
714
+
715
+ logger.info(f"Overwriting model with fp32 weights")
716
+ model = model.cpu()
717
+ model.load_state_dict(state_dict, strict=False)
718
+
719
+ return model
720
+
721
+
722
+ if __name__ == "__main__":
723
+ parser = argparse.ArgumentParser()
724
+ parser.add_argument("checkpoint_dir",
725
+ type=str,
726
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
727
+ parser.add_argument("output_dir",
728
+ type=str,
729
+ help="directory to the pytorch fp32 state_dict output files"
730
+ "(e.g. path/checkpoint-12-output/)")
731
+ parser.add_argument(
732
+ "--max_shard_size",
733
+ type=str,
734
+ default="5GB",
735
+ help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
736
+ "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
737
+ "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
738
+ "without CPU OOM issues.")
739
+ parser.add_argument(
740
+ "--safe_serialization",
741
+ default=False,
742
+ action='store_true',
743
+ help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
744
+ parser.add_argument("-t",
745
+ "--tag",
746
+ type=str,
747
+ default=None,
748
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
749
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
750
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
751
+ args = parser.parse_args()
752
+
753
+ debug = args.debug
754
+
755
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
756
+ args.output_dir,
757
+ max_shard_size=args.max_shard_size,
758
+ safe_serialization=args.safe_serialization,
759
+ tag=args.tag,
760
+ exclude_frozen_parameters=args.exclude_frozen_parameters)