Hugging Face's logo

Xtra-Computing
/
XtraGPT-14B

Text Generation
Transformers
Safetensors
phi3
chat
conversational
custom_code
text-generation-inference
Model card Files
xet
 Community
2
nuojohnchen commited on
Commit
ac3efac
·
verified ·
1 Parent(s): 291fc66

Clean remaining training artifacts

Browse files
Files changed (4) hide show
  1. scheduler.pt +0 -3
  2. trainer_state.json +0 -3533
  3. training_args.bin +0 -3
  4. zero_to_fp32.py +0 -760
scheduler.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:1590c1c5090942d2d4d2cecd50f9e37516d1d2b965656284b2785a2acdbeeb5b
3
- size 1064
 
 
 
 
trainer_state.json DELETED
@@ -1,3533 +0,0 @@
1
- {
2
- "best_metric": null,
3
- "best_model_checkpoint": null,
4
- "epoch": 0.6020469596628537,
5
- "eval_steps": 500,
6
- "global_step": 5000,
7
- "is_hyper_param_search": false,
8
- "is_local_process_zero": true,
9
- "is_world_process_zero": true,
10
- "log_history": [
11
- {
12
- "epoch": 0.0012040939193257074,
13
- "grad_norm": 2.0694425106048584,
14
- "learning_rate": 1.2033694344163658e-08,
15
- "loss": 0.6897,
16
- "step": 10
17
- },
18
- {
19
- "epoch": 0.002408187838651415,
20
- "grad_norm": 2.151496171951294,
21
- "learning_rate": 2.4067388688327316e-08,
22
- "loss": 0.6787,
23
- "step": 20
24
- },
25
- {
26
- "epoch": 0.003612281757977122,
27
- "grad_norm": 2.640268564224243,
28
- "learning_rate": 3.610108303249097e-08,
29
- "loss": 0.6639,
30
- "step": 30
31
- },
32
- {
33
- "epoch": 0.00481637567730283,
34
- "grad_norm": 2.6572210788726807,
35
- "learning_rate": 4.813477737665463e-08,
36
- "loss": 0.7152,
37
- "step": 40
38
- },
39
- {
40
- "epoch": 0.006020469596628537,
41
- "grad_norm": 1.7933714389801025,
42
- "learning_rate": 6.016847172081829e-08,
43
- "loss": 0.6503,
44
- "step": 50
45
- },
46
- {
47
- "epoch": 0.007224563515954244,
48
- "grad_norm": 2.3688879013061523,
49
- "learning_rate": 7.220216606498194e-08,
50
- "loss": 0.6827,
51
- "step": 60
52
- },
53
- {
54
- "epoch": 0.008428657435279952,
55
- "grad_norm": 2.220139265060425,
56
- "learning_rate": 8.42358604091456e-08,
57
- "loss": 0.6443,
58
- "step": 70
59
- },
60
- {
61
- "epoch": 0.00963275135460566,
62
- "grad_norm": 2.4725093841552734,
63
- "learning_rate": 9.626955475330927e-08,
64
- "loss": 0.6681,
65
- "step": 80
66
- },
67
- {
68
- "epoch": 0.010836845273931367,
69
- "grad_norm": 1.4149224758148193,
70
- "learning_rate": 1.0830324909747292e-07,
71
- "loss": 0.5592,
72
- "step": 90
73
- },
74
- {
75
- "epoch": 0.012040939193257074,
76
- "grad_norm": 0.9355699419975281,
77
- "learning_rate": 1.2033694344163658e-07,
78
- "loss": 0.5802,
79
- "step": 100
80
- },
81
- {
82
- "epoch": 0.013245033112582781,
83
- "grad_norm": 1.0211461782455444,
84
- "learning_rate": 1.3237063778580024e-07,
85
- "loss": 0.5589,
86
- "step": 110
87
- },
88
- {
89
- "epoch": 0.014449127031908489,
90
- "grad_norm": 1.0006492137908936,
91
- "learning_rate": 1.4440433212996388e-07,
92
- "loss": 0.5421,
93
- "step": 120
94
- },
95
- {
96
- "epoch": 0.015653220951234198,
97
- "grad_norm": 0.8444674015045166,
98
- "learning_rate": 1.5643802647412754e-07,
99
- "loss": 0.5079,
100
- "step": 130
101
- },
102
- {
103
- "epoch": 0.016857314870559904,
104
- "grad_norm": 0.7920398712158203,
105
- "learning_rate": 1.684717208182912e-07,
106
- "loss": 0.4898,
107
- "step": 140
108
- },
109
- {
110
- "epoch": 0.018061408789885613,
111
- "grad_norm": 0.6817948818206787,
112
- "learning_rate": 1.8050541516245487e-07,
113
- "loss": 0.4645,
114
- "step": 150
115
- },
116
- {
117
- "epoch": 0.01926550270921132,
118
- "grad_norm": 0.9353106021881104,
119
- "learning_rate": 1.9253910950661853e-07,
120
- "loss": 0.485,
121
- "step": 160
122
- },
123
- {
124
- "epoch": 0.020469596628537028,
125
- "grad_norm": 0.6695616841316223,
126
- "learning_rate": 2.045728038507822e-07,
127
- "loss": 0.4647,
128
- "step": 170
129
- },
130
- {
131
- "epoch": 0.021673690547862733,
132
- "grad_norm": 0.6993837952613831,
133
- "learning_rate": 2.1660649819494583e-07,
134
- "loss": 0.4378,
135
- "step": 180
136
- },
137
- {
138
- "epoch": 0.022877784467188442,
139
- "grad_norm": 0.7333642244338989,
140
- "learning_rate": 2.286401925391095e-07,
141
- "loss": 0.4288,
142
- "step": 190
143
- },
144
- {
145
- "epoch": 0.024081878386514148,
146
- "grad_norm": 0.707914412021637,
147
- "learning_rate": 2.4067388688327316e-07,
148
- "loss": 0.4601,
149
- "step": 200
150
- },
151
- {
152
- "epoch": 0.025285972305839857,
153
- "grad_norm": 0.7626605033874512,
154
- "learning_rate": 2.527075812274368e-07,
155
- "loss": 0.4454,
156
- "step": 210
157
- },
158
- {
159
- "epoch": 0.026490066225165563,
160
- "grad_norm": 1.2267224788665771,
161
- "learning_rate": 2.647412755716005e-07,
162
- "loss": 0.4398,
163
- "step": 220
164
- },
165
- {
166
- "epoch": 0.027694160144491272,
167
- "grad_norm": 0.7376552224159241,
168
- "learning_rate": 2.767749699157641e-07,
169
- "loss": 0.4275,
170
- "step": 230
171
- },
172
- {
173
- "epoch": 0.028898254063816978,
174
- "grad_norm": 0.7109339237213135,
175
- "learning_rate": 2.8880866425992776e-07,
176
- "loss": 0.3996,
177
- "step": 240
178
- },
179
- {
180
- "epoch": 0.030102347983142687,
181
- "grad_norm": 0.6406791806221008,
182
- "learning_rate": 3.008423586040915e-07,
183
- "loss": 0.4337,
184
- "step": 250
185
- },
186
- {
187
- "epoch": 0.031306441902468396,
188
- "grad_norm": 0.6780328154563904,
189
- "learning_rate": 3.128760529482551e-07,
190
- "loss": 0.4296,
191
- "step": 260
192
- },
193
- {
194
- "epoch": 0.0325105358217941,
195
- "grad_norm": 0.5574681162834167,
196
- "learning_rate": 3.2490974729241875e-07,
197
- "loss": 0.4123,
198
- "step": 270
199
- },
200
- {
201
- "epoch": 0.03371462974111981,
202
- "grad_norm": 0.6190093755722046,
203
- "learning_rate": 3.369434416365824e-07,
204
- "loss": 0.3959,
205
- "step": 280
206
- },
207
- {
208
- "epoch": 0.034918723660445516,
209
- "grad_norm": 0.6488677859306335,
210
- "learning_rate": 3.4897713598074607e-07,
211
- "loss": 0.3883,
212
- "step": 290
213
- },
214
- {
215
- "epoch": 0.036122817579771226,
216
- "grad_norm": 0.6014848351478577,
217
- "learning_rate": 3.6101083032490974e-07,
218
- "loss": 0.4222,
219
- "step": 300
220
- },
221
- {
222
- "epoch": 0.03732691149909693,
223
- "grad_norm": 0.5347362160682678,
224
- "learning_rate": 3.730445246690734e-07,
225
- "loss": 0.3929,
226
- "step": 310
227
- },
228
- {
229
- "epoch": 0.03853100541842264,
230
- "grad_norm": 1.4445090293884277,
231
- "learning_rate": 3.8507821901323706e-07,
232
- "loss": 0.3798,
233
- "step": 320
234
- },
235
- {
236
- "epoch": 0.039735099337748346,
237
- "grad_norm": 0.6319730877876282,
238
- "learning_rate": 3.9711191335740067e-07,
239
- "loss": 0.386,
240
- "step": 330
241
- },
242
- {
243
- "epoch": 0.040939193257074055,
244
- "grad_norm": 0.9257851243019104,
245
- "learning_rate": 4.091456077015644e-07,
246
- "loss": 0.393,
247
- "step": 340
248
- },
249
- {
250
- "epoch": 0.04214328717639976,
251
- "grad_norm": 0.5936801433563232,
252
- "learning_rate": 4.2117930204572805e-07,
253
- "loss": 0.3912,
254
- "step": 350
255
- },
256
- {
257
- "epoch": 0.04334738109572547,
258
- "grad_norm": 0.686888575553894,
259
- "learning_rate": 4.3321299638989166e-07,
260
- "loss": 0.4015,
261
- "step": 360
262
- },
263
- {
264
- "epoch": 0.044551475015051176,
265
- "grad_norm": 0.5986278653144836,
266
- "learning_rate": 4.452466907340554e-07,
267
- "loss": 0.3622,
268
- "step": 370
269
- },
270
- {
271
- "epoch": 0.045755568934376885,
272
- "grad_norm": 0.5603286623954773,
273
- "learning_rate": 4.57280385078219e-07,
274
- "loss": 0.3774,
275
- "step": 380
276
- },
277
- {
278
- "epoch": 0.04695966285370259,
279
- "grad_norm": 1.2507776021957397,
280
- "learning_rate": 4.6931407942238265e-07,
281
- "loss": 0.3681,
282
- "step": 390
283
- },
284
- {
285
- "epoch": 0.048163756773028296,
286
- "grad_norm": 0.5886845588684082,
287
- "learning_rate": 4.813477737665463e-07,
288
- "loss": 0.371,
289
- "step": 400
290
- },
291
- {
292
- "epoch": 0.049367850692354005,
293
- "grad_norm": 0.5690301656723022,
294
- "learning_rate": 4.9338146811071e-07,
295
- "loss": 0.3454,
296
- "step": 410
297
- },
298
- {
299
- "epoch": 0.050571944611679714,
300
- "grad_norm": 0.6363804340362549,
301
- "learning_rate": 5.054151624548736e-07,
302
- "loss": 0.3477,
303
- "step": 420
304
- },
305
- {
306
- "epoch": 0.05177603853100542,
307
- "grad_norm": 0.49289166927337646,
308
- "learning_rate": 5.174488567990373e-07,
309
- "loss": 0.352,
310
- "step": 430
311
- },
312
- {
313
- "epoch": 0.052980132450331126,
314
- "grad_norm": 0.5901724696159363,
315
- "learning_rate": 5.29482551143201e-07,
316
- "loss": 0.3514,
317
- "step": 440
318
- },
319
- {
320
- "epoch": 0.054184226369656835,
321
- "grad_norm": 0.6019484996795654,
322
- "learning_rate": 5.415162454873646e-07,
323
- "loss": 0.3713,
324
- "step": 450
325
- },
326
- {
327
- "epoch": 0.055388320288982544,
328
- "grad_norm": 0.5057175755500793,
329
- "learning_rate": 5.535499398315282e-07,
330
- "loss": 0.3346,
331
- "step": 460
332
- },
333
- {
334
- "epoch": 0.056592414208308246,
335
- "grad_norm": 0.4834252893924713,
336
- "learning_rate": 5.655836341756919e-07,
337
- "loss": 0.3638,
338
- "step": 470
339
- },
340
- {
341
- "epoch": 0.057796508127633955,
342
- "grad_norm": 0.6098750233650208,
343
- "learning_rate": 5.776173285198555e-07,
344
- "loss": 0.3622,
345
- "step": 480
346
- },
347
- {
348
- "epoch": 0.059000602046959665,
349
- "grad_norm": 0.6201721429824829,
350
- "learning_rate": 5.896510228640193e-07,
351
- "loss": 0.3329,
352
- "step": 490
353
- },
354
- {
355
- "epoch": 0.060204695966285374,
356
- "grad_norm": 0.7006021738052368,
357
- "learning_rate": 6.01684717208183e-07,
358
- "loss": 0.3487,
359
- "step": 500
360
- },
361
- {
362
- "epoch": 0.061408789885611076,
363
- "grad_norm": 0.708990216255188,
364
- "learning_rate": 6.137184115523465e-07,
365
- "loss": 0.3448,
366
- "step": 510
367
- },
368
- {
369
- "epoch": 0.06261288380493679,
370
- "grad_norm": 0.7767229676246643,
371
- "learning_rate": 6.257521058965102e-07,
372
- "loss": 0.3751,
373
- "step": 520
374
- },
375
- {
376
- "epoch": 0.0638169777242625,
377
- "grad_norm": 0.6051218509674072,
378
- "learning_rate": 6.377858002406738e-07,
379
- "loss": 0.3502,
380
- "step": 530
381
- },
382
- {
383
- "epoch": 0.0650210716435882,
384
- "grad_norm": 0.7111226916313171,
385
- "learning_rate": 6.498194945848375e-07,
386
- "loss": 0.3625,
387
- "step": 540
388
- },
389
- {
390
- "epoch": 0.06622516556291391,
391
- "grad_norm": 0.7441733479499817,
392
- "learning_rate": 6.618531889290013e-07,
393
- "loss": 0.3269,
394
- "step": 550
395
- },
396
- {
397
- "epoch": 0.06742925948223961,
398
- "grad_norm": 0.6909326910972595,
399
- "learning_rate": 6.738868832731648e-07,
400
- "loss": 0.3302,
401
- "step": 560
402
- },
403
- {
404
- "epoch": 0.06863335340156532,
405
- "grad_norm": 0.7504749298095703,
406
- "learning_rate": 6.859205776173285e-07,
407
- "loss": 0.3425,
408
- "step": 570
409
- },
410
- {
411
- "epoch": 0.06983744732089103,
412
- "grad_norm": 0.5878099799156189,
413
- "learning_rate": 6.979542719614921e-07,
414
- "loss": 0.3504,
415
- "step": 580
416
- },
417
- {
418
- "epoch": 0.07104154124021674,
419
- "grad_norm": 0.5515761971473694,
420
- "learning_rate": 7.099879663056558e-07,
421
- "loss": 0.3409,
422
- "step": 590
423
- },
424
- {
425
- "epoch": 0.07224563515954245,
426
- "grad_norm": 0.57797771692276,
427
- "learning_rate": 7.220216606498195e-07,
428
- "loss": 0.3416,
429
- "step": 600
430
- },
431
- {
432
- "epoch": 0.07344972907886815,
433
- "grad_norm": 0.4524708390235901,
434
- "learning_rate": 7.34055354993983e-07,
435
- "loss": 0.3581,
436
- "step": 610
437
- },
438
- {
439
- "epoch": 0.07465382299819386,
440
- "grad_norm": 0.718927800655365,
441
- "learning_rate": 7.460890493381468e-07,
442
- "loss": 0.3609,
443
- "step": 620
444
- },
445
- {
446
- "epoch": 0.07585791691751957,
447
- "grad_norm": 0.5666077733039856,
448
- "learning_rate": 7.581227436823105e-07,
449
- "loss": 0.335,
450
- "step": 630
451
- },
452
- {
453
- "epoch": 0.07706201083684527,
454
- "grad_norm": 0.5896601676940918,
455
- "learning_rate": 7.701564380264741e-07,
456
- "loss": 0.3274,
457
- "step": 640
458
- },
459
- {
460
- "epoch": 0.07826610475617098,
461
- "grad_norm": 0.6044319868087769,
462
- "learning_rate": 7.821901323706378e-07,
463
- "loss": 0.3407,
464
- "step": 650
465
- },
466
- {
467
- "epoch": 0.07947019867549669,
468
- "grad_norm": 0.6831541061401367,
469
- "learning_rate": 7.942238267148013e-07,
470
- "loss": 0.3333,
471
- "step": 660
472
- },
473
- {
474
- "epoch": 0.0806742925948224,
475
- "grad_norm": 0.7124572396278381,
476
- "learning_rate": 8.06257521058965e-07,
477
- "loss": 0.3326,
478
- "step": 670
479
- },
480
- {
481
- "epoch": 0.08187838651414811,
482
- "grad_norm": 0.732711136341095,
483
- "learning_rate": 8.182912154031288e-07,
484
- "loss": 0.3487,
485
- "step": 680
486
- },
487
- {
488
- "epoch": 0.08308248043347381,
489
- "grad_norm": 0.7555579543113708,
490
- "learning_rate": 8.303249097472924e-07,
491
- "loss": 0.3218,
492
- "step": 690
493
- },
494
- {
495
- "epoch": 0.08428657435279951,
496
- "grad_norm": 0.7618419528007507,
497
- "learning_rate": 8.423586040914561e-07,
498
- "loss": 0.3231,
499
- "step": 700
500
- },
501
- {
502
- "epoch": 0.08549066827212523,
503
- "grad_norm": 0.7383216023445129,
504
- "learning_rate": 8.543922984356197e-07,
505
- "loss": 0.3218,
506
- "step": 710
507
- },
508
- {
509
- "epoch": 0.08669476219145093,
510
- "grad_norm": 0.5902182459831238,
511
- "learning_rate": 8.664259927797833e-07,
512
- "loss": 0.3367,
513
- "step": 720
514
- },
515
- {
516
- "epoch": 0.08789885611077664,
517
- "grad_norm": 0.6107906103134155,
518
- "learning_rate": 8.78459687123947e-07,
519
- "loss": 0.3331,
520
- "step": 730
521
- },
522
- {
523
- "epoch": 0.08910295003010235,
524
- "grad_norm": 0.7179387211799622,
525
- "learning_rate": 8.904933814681108e-07,
526
- "loss": 0.3347,
527
- "step": 740
528
- },
529
- {
530
- "epoch": 0.09030704394942805,
531
- "grad_norm": 0.8263080716133118,
532
- "learning_rate": 9.025270758122743e-07,
533
- "loss": 0.3247,
534
- "step": 750
535
- },
536
- {
537
- "epoch": 0.09151113786875377,
538
- "grad_norm": 0.8549688458442688,
539
- "learning_rate": 9.14560770156438e-07,
540
- "loss": 0.3239,
541
- "step": 760
542
- },
543
- {
544
- "epoch": 0.09271523178807947,
545
- "grad_norm": 0.6674267053604126,
546
- "learning_rate": 9.265944645006016e-07,
547
- "loss": 0.333,
548
- "step": 770
549
- },
550
- {
551
- "epoch": 0.09391932570740517,
552
- "grad_norm": 0.5892189741134644,
553
- "learning_rate": 9.386281588447653e-07,
554
- "loss": 0.322,
555
- "step": 780
556
- },
557
- {
558
- "epoch": 0.09512341962673089,
559
- "grad_norm": 0.7087513208389282,
560
- "learning_rate": 9.50661853188929e-07,
561
- "loss": 0.327,
562
- "step": 790
563
- },
564
- {
565
- "epoch": 0.09632751354605659,
566
- "grad_norm": 0.6016402840614319,
567
- "learning_rate": 9.626955475330926e-07,
568
- "loss": 0.3255,
569
- "step": 800
570
- },
571
- {
572
- "epoch": 0.0975316074653823,
573
- "grad_norm": 0.5783524513244629,
574
- "learning_rate": 9.747292418772562e-07,
575
- "loss": 0.3128,
576
- "step": 810
577
- },
578
- {
579
- "epoch": 0.09873570138470801,
580
- "grad_norm": 0.6049711108207703,
581
- "learning_rate": 9.8676293622142e-07,
582
- "loss": 0.3257,
583
- "step": 820
584
- },
585
- {
586
- "epoch": 0.09993979530403371,
587
- "grad_norm": 0.6259274482727051,
588
- "learning_rate": 9.987966305655835e-07,
589
- "loss": 0.3318,
590
- "step": 830
591
- },
592
- {
593
- "epoch": 0.10114388922335943,
594
- "grad_norm": 0.5331777930259705,
595
- "learning_rate": 9.999964221834556e-07,
596
- "loss": 0.3133,
597
- "step": 840
598
- },
599
- {
600
- "epoch": 0.10234798314268513,
601
- "grad_norm": 0.5190764665603638,
602
- "learning_rate": 9.999840544882987e-07,
603
- "loss": 0.3349,
604
- "step": 850
605
- },
606
- {
607
- "epoch": 0.10355207706201083,
608
- "grad_norm": 0.5867928862571716,
609
- "learning_rate": 9.99962852962418e-07,
610
- "loss": 0.3252,
611
- "step": 860
612
- },
613
- {
614
- "epoch": 0.10475617098133655,
615
- "grad_norm": 0.7667666673660278,
616
- "learning_rate": 9.999328179804064e-07,
617
- "loss": 0.3269,
618
- "step": 870
619
- },
620
- {
621
- "epoch": 0.10596026490066225,
622
- "grad_norm": 0.5684708952903748,
623
- "learning_rate": 9.998939500729291e-07,
624
- "loss": 0.3204,
625
- "step": 880
626
- },
627
- {
628
- "epoch": 0.10716435881998795,
629
- "grad_norm": 0.5369793772697449,
630
- "learning_rate": 9.99846249926713e-07,
631
- "loss": 0.2997,
632
- "step": 890
633
- },
634
- {
635
- "epoch": 0.10836845273931367,
636
- "grad_norm": 0.5773791074752808,
637
- "learning_rate": 9.997897183845347e-07,
638
- "loss": 0.3147,
639
- "step": 900
640
- },
641
- {
642
- "epoch": 0.10957254665863937,
643
- "grad_norm": 0.571826159954071,
644
- "learning_rate": 9.997243564452064e-07,
645
- "loss": 0.32,
646
- "step": 910
647
- },
648
- {
649
- "epoch": 0.11077664057796509,
650
- "grad_norm": 0.420244961977005,
651
- "learning_rate": 9.996501652635578e-07,
652
- "loss": 0.3141,
653
- "step": 920
654
- },
655
- {
656
- "epoch": 0.11198073449729079,
657
- "grad_norm": 0.5253920555114746,
658
- "learning_rate": 9.99567146150415e-07,
659
- "loss": 0.3201,
660
- "step": 930
661
- },
662
- {
663
- "epoch": 0.11318482841661649,
664
- "grad_norm": 0.49279969930648804,
665
- "learning_rate": 9.994753005725785e-07,
666
- "loss": 0.3076,
667
- "step": 940
668
- },
669
- {
670
- "epoch": 0.11438892233594221,
671
- "grad_norm": 0.6114805936813354,
672
- "learning_rate": 9.993746301527965e-07,
673
- "loss": 0.3209,
674
- "step": 950
675
- },
676
- {
677
- "epoch": 0.11559301625526791,
678
- "grad_norm": 1.6514418125152588,
679
- "learning_rate": 9.99265136669737e-07,
680
- "loss": 0.319,
681
- "step": 960
682
- },
683
- {
684
- "epoch": 0.11679711017459361,
685
- "grad_norm": 0.6415925621986389,
686
- "learning_rate": 9.99146822057955e-07,
687
- "loss": 0.3268,
688
- "step": 970
689
- },
690
- {
691
- "epoch": 0.11800120409391933,
692
- "grad_norm": 0.5680079460144043,
693
- "learning_rate": 9.990196884078599e-07,
694
- "loss": 0.3139,
695
- "step": 980
696
- },
697
- {
698
- "epoch": 0.11920529801324503,
699
- "grad_norm": 0.715497612953186,
700
- "learning_rate": 9.988837379656778e-07,
701
- "loss": 0.3143,
702
- "step": 990
703
- },
704
- {
705
- "epoch": 0.12040939193257075,
706
- "grad_norm": 0.6379466652870178,
707
- "learning_rate": 9.987389731334112e-07,
708
- "loss": 0.3037,
709
- "step": 1000
710
- },
711
- {
712
- "epoch": 0.12161348585189645,
713
- "grad_norm": 0.5227240920066833,
714
- "learning_rate": 9.985853964687985e-07,
715
- "loss": 0.3202,
716
- "step": 1010
717
- },
718
- {
719
- "epoch": 0.12281757977122215,
720
- "grad_norm": 0.5148226022720337,
721
- "learning_rate": 9.984230106852658e-07,
722
- "loss": 0.3089,
723
- "step": 1020
724
- },
725
- {
726
- "epoch": 0.12402167369054787,
727
- "grad_norm": 0.8337252140045166,
728
- "learning_rate": 9.982518186518824e-07,
729
- "loss": 0.3093,
730
- "step": 1030
731
- },
732
- {
733
- "epoch": 0.12522576760987358,
734
- "grad_norm": 0.5874176621437073,
735
- "learning_rate": 9.980718233933072e-07,
736
- "loss": 0.3257,
737
- "step": 1040
738
- },
739
- {
740
- "epoch": 0.12642986152919927,
741
- "grad_norm": 0.6203235983848572,
742
- "learning_rate": 9.978830280897373e-07,
743
- "loss": 0.3094,
744
- "step": 1050
745
- },
746
- {
747
- "epoch": 0.127633955448525,
748
- "grad_norm": 0.7386701107025146,
749
- "learning_rate": 9.976854360768501e-07,
750
- "loss": 0.3283,
751
- "step": 1060
752
- },
753
- {
754
- "epoch": 0.1288380493678507,
755
- "grad_norm": 0.7480394244194031,
756
- "learning_rate": 9.97479050845746e-07,
757
- "loss": 0.322,
758
- "step": 1070
759
- },
760
- {
761
- "epoch": 0.1300421432871764,
762
- "grad_norm": 0.6779530048370361,
763
- "learning_rate": 9.97263876042886e-07,
764
- "loss": 0.3263,
765
- "step": 1080
766
- },
767
- {
768
- "epoch": 0.1312462372065021,
769
- "grad_norm": 1.0457607507705688,
770
- "learning_rate": 9.970399154700262e-07,
771
- "loss": 0.324,
772
- "step": 1090
773
- },
774
- {
775
- "epoch": 0.13245033112582782,
776
- "grad_norm": 0.4574492871761322,
777
- "learning_rate": 9.96807173084153e-07,
778
- "loss": 0.3033,
779
- "step": 1100
780
- },
781
- {
782
- "epoch": 0.1336544250451535,
783
- "grad_norm": 0.4800940454006195,
784
- "learning_rate": 9.965656529974108e-07,
785
- "loss": 0.3076,
786
- "step": 1110
787
- },
788
- {
789
- "epoch": 0.13485851896447923,
790
- "grad_norm": 0.5336936116218567,
791
- "learning_rate": 9.96315359477031e-07,
792
- "loss": 0.3029,
793
- "step": 1120
794
- },
795
- {
796
- "epoch": 0.13606261288380495,
797
- "grad_norm": 0.9403670430183411,
798
- "learning_rate": 9.960562969452559e-07,
799
- "loss": 0.3019,
800
- "step": 1130
801
- },
802
- {
803
- "epoch": 0.13726670680313063,
804
- "grad_norm": 0.6152085661888123,
805
- "learning_rate": 9.957884699792604e-07,
806
- "loss": 0.3051,
807
- "step": 1140
808
- },
809
- {
810
- "epoch": 0.13847080072245635,
811
- "grad_norm": 0.7313536405563354,
812
- "learning_rate": 9.955118833110716e-07,
813
- "loss": 0.3137,
814
- "step": 1150
815
- },
816
- {
817
- "epoch": 0.13967489464178207,
818
- "grad_norm": 0.47397103905677795,
819
- "learning_rate": 9.95226541827485e-07,
820
- "loss": 0.3214,
821
- "step": 1160
822
- },
823
- {
824
- "epoch": 0.14087898856110775,
825
- "grad_norm": 0.4812333881855011,
826
- "learning_rate": 9.949324505699782e-07,
827
- "loss": 0.3164,
828
- "step": 1170
829
- },
830
- {
831
- "epoch": 0.14208308248043347,
832
- "grad_norm": 0.6729305386543274,
833
- "learning_rate": 9.946296147346215e-07,
834
- "loss": 0.2946,
835
- "step": 1180
836
- },
837
- {
838
- "epoch": 0.1432871763997592,
839
- "grad_norm": 0.6568790078163147,
840
- "learning_rate": 9.943180396719867e-07,
841
- "loss": 0.2929,
842
- "step": 1190
843
- },
844
- {
845
- "epoch": 0.1444912703190849,
846
- "grad_norm": 0.5633556842803955,
847
- "learning_rate": 9.939977308870518e-07,
848
- "loss": 0.3073,
849
- "step": 1200
850
- },
851
- {
852
- "epoch": 0.1456953642384106,
853
- "grad_norm": 1.1128957271575928,
854
- "learning_rate": 9.936686940391048e-07,
855
- "loss": 0.3264,
856
- "step": 1210
857
- },
858
- {
859
- "epoch": 0.1468994581577363,
860
- "grad_norm": 0.5192599892616272,
861
- "learning_rate": 9.933309349416428e-07,
862
- "loss": 0.3064,
863
- "step": 1220
864
- },
865
- {
866
- "epoch": 0.14810355207706202,
867
- "grad_norm": 0.49194392561912537,
868
- "learning_rate": 9.92984459562269e-07,
869
- "loss": 0.302,
870
- "step": 1230
871
- },
872
- {
873
- "epoch": 0.1493076459963877,
874
- "grad_norm": 0.5606468915939331,
875
- "learning_rate": 9.926292740225888e-07,
876
- "loss": 0.3037,
877
- "step": 1240
878
- },
879
- {
880
- "epoch": 0.15051173991571343,
881
- "grad_norm": 0.544266939163208,
882
- "learning_rate": 9.922653845981e-07,
883
- "loss": 0.3025,
884
- "step": 1250
885
- },
886
- {
887
- "epoch": 0.15171583383503914,
888
- "grad_norm": 1.0137197971343994,
889
- "learning_rate": 9.918927977180826e-07,
890
- "loss": 0.2998,
891
- "step": 1260
892
- },
893
- {
894
- "epoch": 0.15291992775436483,
895
- "grad_norm": 0.4881134629249573,
896
- "learning_rate": 9.91511519965486e-07,
897
- "loss": 0.2975,
898
- "step": 1270
899
- },
900
- {
901
- "epoch": 0.15412402167369055,
902
- "grad_norm": 0.4854426383972168,
903
- "learning_rate": 9.911215580768106e-07,
904
- "loss": 0.3109,
905
- "step": 1280
906
- },
907
- {
908
- "epoch": 0.15532811559301626,
909
- "grad_norm": 0.5056730508804321,
910
- "learning_rate": 9.90722918941991e-07,
911
- "loss": 0.3121,
912
- "step": 1290
913
- },
914
- {
915
- "epoch": 0.15653220951234195,
916
- "grad_norm": 0.5286668539047241,
917
- "learning_rate": 9.903156096042734e-07,
918
- "loss": 0.2982,
919
- "step": 1300
920
- },
921
- {
922
- "epoch": 0.15773630343166767,
923
- "grad_norm": 0.5490984916687012,
924
- "learning_rate": 9.898996372600903e-07,
925
- "loss": 0.3115,
926
- "step": 1310
927
- },
928
- {
929
- "epoch": 0.15894039735099338,
930
- "grad_norm": 0.614521861076355,
931
- "learning_rate": 9.894750092589349e-07,
932
- "loss": 0.2985,
933
- "step": 1320
934
- },
935
- {
936
- "epoch": 0.16014449127031907,
937
- "grad_norm": 0.5678403973579407,
938
- "learning_rate": 9.8904173310323e-07,
939
- "loss": 0.3046,
940
- "step": 1330
941
- },
942
- {
943
- "epoch": 0.1613485851896448,
944
- "grad_norm": 0.5179656147956848,
945
- "learning_rate": 9.885998164481966e-07,
946
- "loss": 0.3053,
947
- "step": 1340
948
- },
949
- {
950
- "epoch": 0.1625526791089705,
951
- "grad_norm": 0.526849091053009,
952
- "learning_rate": 9.881492671017172e-07,
953
- "loss": 0.3143,
954
- "step": 1350
955
- },
956
- {
957
- "epoch": 0.16375677302829622,
958
- "grad_norm": 0.5683344006538391,
959
- "learning_rate": 9.876900930241991e-07,
960
- "loss": 0.3031,
961
- "step": 1360
962
- },
963
- {
964
- "epoch": 0.1649608669476219,
965
- "grad_norm": 0.5243839621543884,
966
- "learning_rate": 9.872223023284333e-07,
967
- "loss": 0.312,
968
- "step": 1370
969
- },
970
- {
971
- "epoch": 0.16616496086694763,
972
- "grad_norm": 0.5260365605354309,
973
- "learning_rate": 9.867459032794508e-07,
974
- "loss": 0.3037,
975
- "step": 1380
976
- },
977
- {
978
- "epoch": 0.16736905478627334,
979
- "grad_norm": 0.4755154252052307,
980
- "learning_rate": 9.86260904294377e-07,
981
- "loss": 0.2916,
982
- "step": 1390
983
- },
984
- {
985
- "epoch": 0.16857314870559903,
986
- "grad_norm": 0.5555715560913086,
987
- "learning_rate": 9.857673139422833e-07,
988
- "loss": 0.3135,
989
- "step": 1400
990
- },
991
- {
992
- "epoch": 0.16977724262492475,
993
- "grad_norm": 0.5810279250144958,
994
- "learning_rate": 9.85265140944035e-07,
995
- "loss": 0.3104,
996
- "step": 1410
997
- },
998
- {
999
- "epoch": 0.17098133654425046,
1000
- "grad_norm": 0.48022618889808655,
1001
- "learning_rate": 9.847543941721379e-07,
1002
- "loss": 0.3022,
1003
- "step": 1420
1004
- },
1005
- {
1006
- "epoch": 0.17218543046357615,
1007
- "grad_norm": 0.5191965103149414,
1008
- "learning_rate": 9.842350826505802e-07,
1009
- "loss": 0.3018,
1010
- "step": 1430
1011
- },
1012
- {
1013
- "epoch": 0.17338952438290187,
1014
- "grad_norm": 1.2972302436828613,
1015
- "learning_rate": 9.837072155546753e-07,
1016
- "loss": 0.3026,
1017
- "step": 1440
1018
- },
1019
- {
1020
- "epoch": 0.17459361830222758,
1021
- "grad_norm": 0.47315987944602966,
1022
- "learning_rate": 9.831708022108972e-07,
1023
- "loss": 0.311,
1024
- "step": 1450
1025
- },
1026
- {
1027
- "epoch": 0.17579771222155327,
1028
- "grad_norm": 0.5953189134597778,
1029
- "learning_rate": 9.826258520967177e-07,
1030
- "loss": 0.3071,
1031
- "step": 1460
1032
- },
1033
- {
1034
- "epoch": 0.177001806140879,
1035
- "grad_norm": 0.5407562851905823,
1036
- "learning_rate": 9.820723748404382e-07,
1037
- "loss": 0.31,
1038
- "step": 1470
1039
- },
1040
- {
1041
- "epoch": 0.1782059000602047,
1042
- "grad_norm": 0.5249618291854858,
1043
- "learning_rate": 9.815103802210193e-07,
1044
- "loss": 0.2898,
1045
- "step": 1480
1046
- },
1047
- {
1048
- "epoch": 0.1794099939795304,
1049
- "grad_norm": 0.5347439646720886,
1050
- "learning_rate": 9.80939878167908e-07,
1051
- "loss": 0.2944,
1052
- "step": 1490
1053
- },
1054
- {
1055
- "epoch": 0.1806140878988561,
1056
- "grad_norm": 0.49509304761886597,
1057
- "learning_rate": 9.80360878760863e-07,
1058
- "loss": 0.3073,
1059
- "step": 1500
1060
- },
1061
- {
1062
- "epoch": 0.18181818181818182,
1063
- "grad_norm": 0.5182557106018066,
1064
- "learning_rate": 9.79773392229776e-07,
1065
- "loss": 0.3092,
1066
- "step": 1510
1067
- },
1068
- {
1069
- "epoch": 0.18302227573750754,
1070
- "grad_norm": 0.5343918204307556,
1071
- "learning_rate": 9.79177428954492e-07,
1072
- "loss": 0.3058,
1073
- "step": 1520
1074
- },
1075
- {
1076
- "epoch": 0.18422636965683323,
1077
- "grad_norm": 0.42448320984840393,
1078
- "learning_rate": 9.785729994646228e-07,
1079
- "loss": 0.2966,
1080
- "step": 1530
1081
- },
1082
- {
1083
- "epoch": 0.18543046357615894,
1084
- "grad_norm": 0.514305055141449,
1085
- "learning_rate": 9.779601144393655e-07,
1086
- "loss": 0.3063,
1087
- "step": 1540
1088
- },
1089
- {
1090
- "epoch": 0.18663455749548466,
1091
- "grad_norm": 0.559808075428009,
1092
- "learning_rate": 9.773387847073102e-07,
1093
- "loss": 0.3103,
1094
- "step": 1550
1095
- },
1096
- {
1097
- "epoch": 0.18783865141481035,
1098
- "grad_norm": 0.5099034905433655,
1099
- "learning_rate": 9.767090212462506e-07,
1100
- "loss": 0.3045,
1101
- "step": 1560
1102
- },
1103
- {
1104
- "epoch": 0.18904274533413606,
1105
- "grad_norm": 0.5309582352638245,
1106
- "learning_rate": 9.76070835182989e-07,
1107
- "loss": 0.3198,
1108
- "step": 1570
1109
- },
1110
- {
1111
- "epoch": 0.19024683925346178,
1112
- "grad_norm": 0.5174340605735779,
1113
- "learning_rate": 9.754242377931402e-07,
1114
- "loss": 0.3019,
1115
- "step": 1580
1116
- },
1117
- {
1118
- "epoch": 0.19145093317278747,
1119
- "grad_norm": 0.47818174958229065,
1120
- "learning_rate": 9.747692405009327e-07,
1121
- "loss": 0.2885,
1122
- "step": 1590
1123
- },
1124
- {
1125
- "epoch": 0.19265502709211318,
1126
- "grad_norm": 0.4435511529445648,
1127
- "learning_rate": 9.741058548790055e-07,
1128
- "loss": 0.2716,
1129
- "step": 1600
1130
- },
1131
- {
1132
- "epoch": 0.1938591210114389,
1133
- "grad_norm": 0.47226864099502563,
1134
- "learning_rate": 9.734340926482052e-07,
1135
- "loss": 0.2911,
1136
- "step": 1610
1137
- },
1138
- {
1139
- "epoch": 0.1950632149307646,
1140
- "grad_norm": 0.4990203082561493,
1141
- "learning_rate": 9.72753965677378e-07,
1142
- "loss": 0.3119,
1143
- "step": 1620
1144
- },
1145
- {
1146
- "epoch": 0.1962673088500903,
1147
- "grad_norm": 0.6255252957344055,
1148
- "learning_rate": 9.7206548598316e-07,
1149
- "loss": 0.2902,
1150
- "step": 1630
1151
- },
1152
- {
1153
- "epoch": 0.19747140276941602,
1154
- "grad_norm": 0.5827116370201111,
1155
- "learning_rate": 9.713686657297655e-07,
1156
- "loss": 0.3079,
1157
- "step": 1640
1158
- },
1159
- {
1160
- "epoch": 0.1986754966887417,
1161
- "grad_norm": 0.5475650429725647,
1162
- "learning_rate": 9.706635172287715e-07,
1163
- "loss": 0.3095,
1164
- "step": 1650
1165
- },
1166
- {
1167
- "epoch": 0.19987959060806743,
1168
- "grad_norm": 0.674460768699646,
1169
- "learning_rate": 9.699500529389001e-07,
1170
- "loss": 0.2953,
1171
- "step": 1660
1172
- },
1173
- {
1174
- "epoch": 0.20108368452739314,
1175
- "grad_norm": 0.5000407695770264,
1176
- "learning_rate": 9.692282854657989e-07,
1177
- "loss": 0.3055,
1178
- "step": 1670
1179
- },
1180
- {
1181
- "epoch": 0.20228777844671886,
1182
- "grad_norm": 0.5063086748123169,
1183
- "learning_rate": 9.684982275618178e-07,
1184
- "loss": 0.2952,
1185
- "step": 1680
1186
- },
1187
- {
1188
- "epoch": 0.20349187236604455,
1189
- "grad_norm": 0.6266674399375916,
1190
- "learning_rate": 9.677598921257842e-07,
1191
- "loss": 0.3028,
1192
- "step": 1690
1193
- },
1194
- {
1195
- "epoch": 0.20469596628537026,
1196
- "grad_norm": 1.3428351879119873,
1197
- "learning_rate": 9.67013292202775e-07,
1198
- "loss": 0.3165,
1199
- "step": 1700
1200
- },
1201
- {
1202
- "epoch": 0.20590006020469598,
1203
- "grad_norm": 0.6307231187820435,
1204
- "learning_rate": 9.66258440983885e-07,
1205
- "loss": 0.3112,
1206
- "step": 1710
1207
- },
1208
- {
1209
- "epoch": 0.20710415412402167,
1210
- "grad_norm": 0.5176913738250732,
1211
- "learning_rate": 9.654953518059953e-07,
1212
- "loss": 0.3042,
1213
- "step": 1720
1214
- },
1215
- {
1216
- "epoch": 0.20830824804334738,
1217
- "grad_norm": 0.4618211090564728,
1218
- "learning_rate": 9.647240381515376e-07,
1219
- "loss": 0.3107,
1220
- "step": 1730
1221
- },
1222
- {
1223
- "epoch": 0.2095123419626731,
1224
- "grad_norm": 0.4354129135608673,
1225
- "learning_rate": 9.639445136482546e-07,
1226
- "loss": 0.2932,
1227
- "step": 1740
1228
- },
1229
- {
1230
- "epoch": 0.2107164358819988,
1231
- "grad_norm": 0.6150096654891968,
1232
- "learning_rate": 9.631567920689607e-07,
1233
- "loss": 0.2898,
1234
- "step": 1750
1235
- },
1236
- {
1237
- "epoch": 0.2119205298013245,
1238
- "grad_norm": 0.4629852771759033,
1239
- "learning_rate": 9.623608873312979e-07,
1240
- "loss": 0.2969,
1241
- "step": 1760
1242
- },
1243
- {
1244
- "epoch": 0.21312462372065022,
1245
- "grad_norm": 0.4912186563014984,
1246
- "learning_rate": 9.615568134974902e-07,
1247
- "loss": 0.3037,
1248
- "step": 1770
1249
- },
1250
- {
1251
- "epoch": 0.2143287176399759,
1252
- "grad_norm": 0.5452593564987183,
1253
- "learning_rate": 9.607445847740946e-07,
1254
- "loss": 0.3011,
1255
- "step": 1780
1256
- },
1257
- {
1258
- "epoch": 0.21553281155930162,
1259
- "grad_norm": 0.5524305701255798,
1260
- "learning_rate": 9.599242155117514e-07,
1261
- "loss": 0.3056,
1262
- "step": 1790
1263
- },
1264
- {
1265
- "epoch": 0.21673690547862734,
1266
- "grad_norm": 0.4734737277030945,
1267
- "learning_rate": 9.590957202049288e-07,
1268
- "loss": 0.2937,
1269
- "step": 1800
1270
- },
1271
- {
1272
- "epoch": 0.21794099939795303,
1273
- "grad_norm": 0.5050627589225769,
1274
- "learning_rate": 9.582591134916683e-07,
1275
- "loss": 0.2964,
1276
- "step": 1810
1277
- },
1278
- {
1279
- "epoch": 0.21914509331727874,
1280
- "grad_norm": 0.5784972310066223,
1281
- "learning_rate": 9.574144101533258e-07,
1282
- "loss": 0.3126,
1283
- "step": 1820
1284
- },
1285
- {
1286
- "epoch": 0.22034918723660446,
1287
- "grad_norm": 0.67679762840271,
1288
- "learning_rate": 9.565616251143093e-07,
1289
- "loss": 0.2997,
1290
- "step": 1830
1291
- },
1292
- {
1293
- "epoch": 0.22155328115593018,
1294
- "grad_norm": 0.730844259262085,
1295
- "learning_rate": 9.55700773441817e-07,
1296
- "loss": 0.2992,
1297
- "step": 1840
1298
- },
1299
- {
1300
- "epoch": 0.22275737507525586,
1301
- "grad_norm": 0.511701226234436,
1302
- "learning_rate": 9.5483187034557e-07,
1303
- "loss": 0.2843,
1304
- "step": 1850
1305
- },
1306
- {
1307
- "epoch": 0.22396146899458158,
1308
- "grad_norm": 0.49653661251068115,
1309
- "learning_rate": 9.539549311775434e-07,
1310
- "loss": 0.3003,
1311
- "step": 1860
1312
- },
1313
- {
1314
- "epoch": 0.2251655629139073,
1315
- "grad_norm": 0.479397714138031,
1316
- "learning_rate": 9.530699714316955e-07,
1317
- "loss": 0.3007,
1318
- "step": 1870
1319
- },
1320
- {
1321
- "epoch": 0.22636965683323299,
1322
- "grad_norm": 0.5917854905128479,
1323
- "learning_rate": 9.521770067436944e-07,
1324
- "loss": 0.2818,
1325
- "step": 1880
1326
- },
1327
- {
1328
- "epoch": 0.2275737507525587,
1329
- "grad_norm": 0.4750485420227051,
1330
- "learning_rate": 9.512760528906409e-07,
1331
- "loss": 0.3107,
1332
- "step": 1890
1333
- },
1334
- {
1335
- "epoch": 0.22877784467188442,
1336
- "grad_norm": 0.5081465244293213,
1337
- "learning_rate": 9.503671257907905e-07,
1338
- "loss": 0.3003,
1339
- "step": 1900
1340
- },
1341
- {
1342
- "epoch": 0.2299819385912101,
1343
- "grad_norm": 0.7816819548606873,
1344
- "learning_rate": 9.494502415032714e-07,
1345
- "loss": 0.2898,
1346
- "step": 1910
1347
- },
1348
- {
1349
- "epoch": 0.23118603251053582,
1350
- "grad_norm": 0.600690484046936,
1351
- "learning_rate": 9.485254162278013e-07,
1352
- "loss": 0.2975,
1353
- "step": 1920
1354
- },
1355
- {
1356
- "epoch": 0.23239012642986154,
1357
- "grad_norm": 0.6016291379928589,
1358
- "learning_rate": 9.475926663044016e-07,
1359
- "loss": 0.2895,
1360
- "step": 1930
1361
- },
1362
- {
1363
- "epoch": 0.23359422034918723,
1364
- "grad_norm": 0.5959491729736328,
1365
- "learning_rate": 9.466520082131074e-07,
1366
- "loss": 0.293,
1367
- "step": 1940
1368
- },
1369
- {
1370
- "epoch": 0.23479831426851294,
1371
- "grad_norm": 0.5337576270103455,
1372
- "learning_rate": 9.457034585736776e-07,
1373
- "loss": 0.2954,
1374
- "step": 1950
1375
- },
1376
- {
1377
- "epoch": 0.23600240818783866,
1378
- "grad_norm": 0.5701966881752014,
1379
- "learning_rate": 9.447470341453003e-07,
1380
- "loss": 0.3016,
1381
- "step": 1960
1382
- },
1383
- {
1384
- "epoch": 0.23720650210716435,
1385
- "grad_norm": 0.48122677206993103,
1386
- "learning_rate": 9.437827518262976e-07,
1387
- "loss": 0.2834,
1388
- "step": 1970
1389
- },
1390
- {
1391
- "epoch": 0.23841059602649006,
1392
- "grad_norm": 0.6107509732246399,
1393
- "learning_rate": 9.428106286538263e-07,
1394
- "loss": 0.2865,
1395
- "step": 1980
1396
- },
1397
- {
1398
- "epoch": 0.23961468994581578,
1399
- "grad_norm": 0.4537561237812042,
1400
- "learning_rate": 9.418306818035773e-07,
1401
- "loss": 0.2981,
1402
- "step": 1990
1403
- },
1404
- {
1405
- "epoch": 0.2408187838651415,
1406
- "grad_norm": 0.6205712556838989,
1407
- "learning_rate": 9.408429285894721e-07,
1408
- "loss": 0.3099,
1409
- "step": 2000
1410
- },
1411
- {
1412
- "epoch": 0.24202287778446718,
1413
- "grad_norm": 0.4940670132637024,
1414
- "learning_rate": 9.398473864633564e-07,
1415
- "loss": 0.2942,
1416
- "step": 2010
1417
- },
1418
- {
1419
- "epoch": 0.2432269717037929,
1420
- "grad_norm": 0.45464888215065,
1421
- "learning_rate": 9.388440730146923e-07,
1422
- "loss": 0.2875,
1423
- "step": 2020
1424
- },
1425
- {
1426
- "epoch": 0.24443106562311862,
1427
- "grad_norm": 0.4339371919631958,
1428
- "learning_rate": 9.378330059702479e-07,
1429
- "loss": 0.284,
1430
- "step": 2030
1431
- },
1432
- {
1433
- "epoch": 0.2456351595424443,
1434
- "grad_norm": 0.6798887848854065,
1435
- "learning_rate": 9.368142031937826e-07,
1436
- "loss": 0.3079,
1437
- "step": 2040
1438
- },
1439
- {
1440
- "epoch": 0.24683925346177002,
1441
- "grad_norm": 0.504805326461792,
1442
- "learning_rate": 9.357876826857334e-07,
1443
- "loss": 0.2942,
1444
- "step": 2050
1445
- },
1446
- {
1447
- "epoch": 0.24804334738109574,
1448
- "grad_norm": 1.0256134271621704,
1449
- "learning_rate": 9.347534625828955e-07,
1450
- "loss": 0.2958,
1451
- "step": 2060
1452
- },
1453
- {
1454
- "epoch": 0.24924744130042142,
1455
- "grad_norm": 0.7034043073654175,
1456
- "learning_rate": 9.337115611581019e-07,
1457
- "loss": 0.2977,
1458
- "step": 2070
1459
- },
1460
- {
1461
- "epoch": 0.25045153521974717,
1462
- "grad_norm": 0.6767880916595459,
1463
- "learning_rate": 9.326619968199016e-07,
1464
- "loss": 0.2843,
1465
- "step": 2080
1466
- },
1467
- {
1468
- "epoch": 0.25165562913907286,
1469
- "grad_norm": 0.5257042050361633,
1470
- "learning_rate": 9.316047881122334e-07,
1471
- "loss": 0.2869,
1472
- "step": 2090
1473
- },
1474
- {
1475
- "epoch": 0.25285972305839854,
1476
- "grad_norm": 0.5919986963272095,
1477
- "learning_rate": 9.305399537140983e-07,
1478
- "loss": 0.3009,
1479
- "step": 2100
1480
- },
1481
- {
1482
- "epoch": 0.2540638169777243,
1483
- "grad_norm": 0.5936114192008972,
1484
- "learning_rate": 9.294675124392302e-07,
1485
- "loss": 0.2863,
1486
- "step": 2110
1487
- },
1488
- {
1489
- "epoch": 0.25526791089705,
1490
- "grad_norm": 1.1754176616668701,
1491
- "learning_rate": 9.283874832357625e-07,
1492
- "loss": 0.2808,
1493
- "step": 2120
1494
- },
1495
- {
1496
- "epoch": 0.25647200481637566,
1497
- "grad_norm": 0.6144666075706482,
1498
- "learning_rate": 9.272998851858943e-07,
1499
- "loss": 0.2854,
1500
- "step": 2130
1501
- },
1502
- {
1503
- "epoch": 0.2576760987357014,
1504
- "grad_norm": 0.47984328866004944,
1505
- "learning_rate": 9.262047375055524e-07,
1506
- "loss": 0.2978,
1507
- "step": 2140
1508
- },
1509
- {
1510
- "epoch": 0.2588801926550271,
1511
- "grad_norm": 0.6158226728439331,
1512
- "learning_rate": 9.251020595440524e-07,
1513
- "loss": 0.3072,
1514
- "step": 2150
1515
- },
1516
- {
1517
- "epoch": 0.2600842865743528,
1518
- "grad_norm": 0.6357386708259583,
1519
- "learning_rate": 9.239918707837564e-07,
1520
- "loss": 0.2927,
1521
- "step": 2160
1522
- },
1523
- {
1524
- "epoch": 0.26128838049367853,
1525
- "grad_norm": 0.6893799901008606,
1526
- "learning_rate": 9.228741908397293e-07,
1527
- "loss": 0.2988,
1528
- "step": 2170
1529
- },
1530
- {
1531
- "epoch": 0.2624924744130042,
1532
- "grad_norm": 0.5763195157051086,
1533
- "learning_rate": 9.217490394593914e-07,
1534
- "loss": 0.3049,
1535
- "step": 2180
1536
- },
1537
- {
1538
- "epoch": 0.2636965683323299,
1539
- "grad_norm": 0.5649781823158264,
1540
- "learning_rate": 9.206164365221706e-07,
1541
- "loss": 0.3083,
1542
- "step": 2190
1543
- },
1544
- {
1545
- "epoch": 0.26490066225165565,
1546
- "grad_norm": 0.4519605040550232,
1547
- "learning_rate": 9.194764020391506e-07,
1548
- "loss": 0.274,
1549
- "step": 2200
1550
- },
1551
- {
1552
- "epoch": 0.26610475617098134,
1553
- "grad_norm": 0.5203403830528259,
1554
- "learning_rate": 9.183289561527164e-07,
1555
- "loss": 0.2823,
1556
- "step": 2210
1557
- },
1558
- {
1559
- "epoch": 0.267308850090307,
1560
- "grad_norm": 0.525934100151062,
1561
- "learning_rate": 9.171741191362005e-07,
1562
- "loss": 0.2928,
1563
- "step": 2220
1564
- },
1565
- {
1566
- "epoch": 0.26851294400963277,
1567
- "grad_norm": 0.5151864290237427,
1568
- "learning_rate": 9.160119113935227e-07,
1569
- "loss": 0.2914,
1570
- "step": 2230
1571
- },
1572
- {
1573
- "epoch": 0.26971703792895846,
1574
- "grad_norm": 0.663339376449585,
1575
- "learning_rate": 9.14842353458831e-07,
1576
- "loss": 0.301,
1577
- "step": 2240
1578
- },
1579
- {
1580
- "epoch": 0.27092113184828415,
1581
- "grad_norm": 0.5526972413063049,
1582
- "learning_rate": 9.136654659961381e-07,
1583
- "loss": 0.2931,
1584
- "step": 2250
1585
- },
1586
- {
1587
- "epoch": 0.2721252257676099,
1588
- "grad_norm": 0.6518740057945251,
1589
- "learning_rate": 9.12481269798956e-07,
1590
- "loss": 0.2772,
1591
- "step": 2260
1592
- },
1593
- {
1594
- "epoch": 0.2733293196869356,
1595
- "grad_norm": 0.5191295742988586,
1596
- "learning_rate": 9.112897857899298e-07,
1597
- "loss": 0.2933,
1598
- "step": 2270
1599
- },
1600
- {
1601
- "epoch": 0.27453341360626127,
1602
- "grad_norm": 1.087936282157898,
1603
- "learning_rate": 9.100910350204669e-07,
1604
- "loss": 0.2956,
1605
- "step": 2280
1606
- },
1607
- {
1608
- "epoch": 0.275737507525587,
1609
- "grad_norm": 0.5870952010154724,
1610
- "learning_rate": 9.088850386703653e-07,
1611
- "loss": 0.2857,
1612
- "step": 2290
1613
- },
1614
- {
1615
- "epoch": 0.2769416014449127,
1616
- "grad_norm": 0.5123207569122314,
1617
- "learning_rate": 9.076718180474399e-07,
1618
- "loss": 0.3005,
1619
- "step": 2300
1620
- },
1621
- {
1622
- "epoch": 0.2781456953642384,
1623
- "grad_norm": 0.47658002376556396,
1624
- "learning_rate": 9.064513945871457e-07,
1625
- "loss": 0.2889,
1626
- "step": 2310
1627
- },
1628
- {
1629
- "epoch": 0.27934978928356413,
1630
- "grad_norm": 0.564738929271698,
1631
- "learning_rate": 9.052237898521984e-07,
1632
- "loss": 0.2929,
1633
- "step": 2320
1634
- },
1635
- {
1636
- "epoch": 0.2805538832028898,
1637
- "grad_norm": 0.47116583585739136,
1638
- "learning_rate": 9.03989025532195e-07,
1639
- "loss": 0.2942,
1640
- "step": 2330
1641
- },
1642
- {
1643
- "epoch": 0.2817579771222155,
1644
- "grad_norm": 0.5838178396224976,
1645
- "learning_rate": 9.027471234432292e-07,
1646
- "loss": 0.2883,
1647
- "step": 2340
1648
- },
1649
- {
1650
- "epoch": 0.28296207104154125,
1651
- "grad_norm": 0.48679229617118835,
1652
- "learning_rate": 9.014981055275059e-07,
1653
- "loss": 0.29,
1654
- "step": 2350
1655
- },
1656
- {
1657
- "epoch": 0.28416616496086694,
1658
- "grad_norm": 0.5863898992538452,
1659
- "learning_rate": 9.00241993852955e-07,
1660
- "loss": 0.2871,
1661
- "step": 2360
1662
- },
1663
- {
1664
- "epoch": 0.28537025888019263,
1665
- "grad_norm": 0.5949921607971191,
1666
- "learning_rate": 8.989788106128402e-07,
1667
- "loss": 0.2927,
1668
- "step": 2370
1669
- },
1670
- {
1671
- "epoch": 0.2865743527995184,
1672
- "grad_norm": 0.42538484930992126,
1673
- "learning_rate": 8.977085781253668e-07,
1674
- "loss": 0.2825,
1675
- "step": 2380
1676
- },
1677
- {
1678
- "epoch": 0.28777844671884406,
1679
- "grad_norm": 0.5678000450134277,
1680
- "learning_rate": 8.964313188332881e-07,
1681
- "loss": 0.294,
1682
- "step": 2390
1683
- },
1684
- {
1685
- "epoch": 0.2889825406381698,
1686
- "grad_norm": 0.5283777713775635,
1687
- "learning_rate": 8.951470553035086e-07,
1688
- "loss": 0.286,
1689
- "step": 2400
1690
- },
1691
- {
1692
- "epoch": 0.2901866345574955,
1693
- "grad_norm": 0.8639681935310364,
1694
- "learning_rate": 8.938558102266851e-07,
1695
- "loss": 0.2971,
1696
- "step": 2410
1697
- },
1698
- {
1699
- "epoch": 0.2913907284768212,
1700
- "grad_norm": 0.5353107452392578,
1701
- "learning_rate": 8.925576064168261e-07,
1702
- "loss": 0.3038,
1703
- "step": 2420
1704
- },
1705
- {
1706
- "epoch": 0.2925948223961469,
1707
- "grad_norm": 0.5691916346549988,
1708
- "learning_rate": 8.912524668108885e-07,
1709
- "loss": 0.2901,
1710
- "step": 2430
1711
- },
1712
- {
1713
- "epoch": 0.2937989163154726,
1714
- "grad_norm": 0.5999578833580017,
1715
- "learning_rate": 8.899404144683724e-07,
1716
- "loss": 0.2864,
1717
- "step": 2440
1718
- },
1719
- {
1720
- "epoch": 0.2950030102347983,
1721
- "grad_norm": 0.6660271883010864,
1722
- "learning_rate": 8.886214725709136e-07,
1723
- "loss": 0.2866,
1724
- "step": 2450
1725
- },
1726
- {
1727
- "epoch": 0.29620710415412405,
1728
- "grad_norm": 0.5501262545585632,
1729
- "learning_rate": 8.872956644218742e-07,
1730
- "loss": 0.2909,
1731
- "step": 2460
1732
- },
1733
- {
1734
- "epoch": 0.29741119807344973,
1735
- "grad_norm": 0.44489532709121704,
1736
- "learning_rate": 8.859630134459308e-07,
1737
- "loss": 0.2869,
1738
- "step": 2470
1739
- },
1740
- {
1741
- "epoch": 0.2986152919927754,
1742
- "grad_norm": 0.619097113609314,
1743
- "learning_rate": 8.846235431886604e-07,
1744
- "loss": 0.2782,
1745
- "step": 2480
1746
- },
1747
- {
1748
- "epoch": 0.29981938591210117,
1749
- "grad_norm": 0.49712878465652466,
1750
- "learning_rate": 8.832772773161251e-07,
1751
- "loss": 0.2848,
1752
- "step": 2490
1753
- },
1754
- {
1755
- "epoch": 0.30102347983142685,
1756
- "grad_norm": 0.46963346004486084,
1757
- "learning_rate": 8.819242396144529e-07,
1758
- "loss": 0.2915,
1759
- "step": 2500
1760
- },
1761
- {
1762
- "epoch": 0.30222757375075254,
1763
- "grad_norm": 0.5881354212760925,
1764
- "learning_rate": 8.805644539894181e-07,
1765
- "loss": 0.2969,
1766
- "step": 2510
1767
- },
1768
- {
1769
- "epoch": 0.3034316676700783,
1770
- "grad_norm": 0.5345028042793274,
1771
- "learning_rate": 8.791979444660193e-07,
1772
- "loss": 0.2985,
1773
- "step": 2520
1774
- },
1775
- {
1776
- "epoch": 0.304635761589404,
1777
- "grad_norm": 0.5038124322891235,
1778
- "learning_rate": 8.778247351880536e-07,
1779
- "loss": 0.2931,
1780
- "step": 2530
1781
- },
1782
- {
1783
- "epoch": 0.30583985550872966,
1784
- "grad_norm": 0.6723479628562927,
1785
- "learning_rate": 8.764448504176919e-07,
1786
- "loss": 0.2885,
1787
- "step": 2540
1788
- },
1789
- {
1790
- "epoch": 0.3070439494280554,
1791
- "grad_norm": 0.474516361951828,
1792
- "learning_rate": 8.750583145350483e-07,
1793
- "loss": 0.2906,
1794
- "step": 2550
1795
- },
1796
- {
1797
- "epoch": 0.3082480433473811,
1798
- "grad_norm": 0.509379506111145,
1799
- "learning_rate": 8.736651520377507e-07,
1800
- "loss": 0.2874,
1801
- "step": 2560
1802
- },
1803
- {
1804
- "epoch": 0.3094521372667068,
1805
- "grad_norm": 0.9317507743835449,
1806
- "learning_rate": 8.722653875405075e-07,
1807
- "loss": 0.2891,
1808
- "step": 2570
1809
- },
1810
- {
1811
- "epoch": 0.3106562311860325,
1812
- "grad_norm": 0.4634588360786438,
1813
- "learning_rate": 8.708590457746727e-07,
1814
- "loss": 0.284,
1815
- "step": 2580
1816
- },
1817
- {
1818
- "epoch": 0.3118603251053582,
1819
- "grad_norm": 0.4674171209335327,
1820
- "learning_rate": 8.694461515878088e-07,
1821
- "loss": 0.2851,
1822
- "step": 2590
1823
- },
1824
- {
1825
- "epoch": 0.3130644190246839,
1826
- "grad_norm": 0.4606451988220215,
1827
- "learning_rate": 8.68026729943248e-07,
1828
- "loss": 0.282,
1829
- "step": 2600
1830
- },
1831
- {
1832
- "epoch": 0.31426851294400965,
1833
- "grad_norm": 0.5793256163597107,
1834
- "learning_rate": 8.666008059196513e-07,
1835
- "loss": 0.2852,
1836
- "step": 2610
1837
- },
1838
- {
1839
- "epoch": 0.31547260686333534,
1840
- "grad_norm": 0.742026686668396,
1841
- "learning_rate": 8.65168404710565e-07,
1842
- "loss": 0.2909,
1843
- "step": 2620
1844
- },
1845
- {
1846
- "epoch": 0.316676700782661,
1847
- "grad_norm": 0.469868928194046,
1848
- "learning_rate": 8.637295516239757e-07,
1849
- "loss": 0.2784,
1850
- "step": 2630
1851
- },
1852
- {
1853
- "epoch": 0.31788079470198677,
1854
- "grad_norm": 0.6895257234573364,
1855
- "learning_rate": 8.622842720818635e-07,
1856
- "loss": 0.2849,
1857
- "step": 2640
1858
- },
1859
- {
1860
- "epoch": 0.31908488862131246,
1861
- "grad_norm": 0.6843047142028809,
1862
- "learning_rate": 8.608325916197524e-07,
1863
- "loss": 0.2969,
1864
- "step": 2650
1865
- },
1866
- {
1867
- "epoch": 0.32028898254063815,
1868
- "grad_norm": 2.822052240371704,
1869
- "learning_rate": 8.593745358862592e-07,
1870
- "loss": 0.2954,
1871
- "step": 2660
1872
- },
1873
- {
1874
- "epoch": 0.3214930764599639,
1875
- "grad_norm": 0.5745678544044495,
1876
- "learning_rate": 8.579101306426406e-07,
1877
- "loss": 0.3005,
1878
- "step": 2670
1879
- },
1880
- {
1881
- "epoch": 0.3226971703792896,
1882
- "grad_norm": 0.4625186026096344,
1883
- "learning_rate": 8.564394017623378e-07,
1884
- "loss": 0.2889,
1885
- "step": 2680
1886
- },
1887
- {
1888
- "epoch": 0.32390126429861527,
1889
- "grad_norm": 0.5813141465187073,
1890
- "learning_rate": 8.549623752305192e-07,
1891
- "loss": 0.2926,
1892
- "step": 2690
1893
- },
1894
- {
1895
- "epoch": 0.325105358217941,
1896
- "grad_norm": 0.49706658720970154,
1897
- "learning_rate": 8.534790771436222e-07,
1898
- "loss": 0.2884,
1899
- "step": 2700
1900
- },
1901
- {
1902
- "epoch": 0.3263094521372667,
1903
- "grad_norm": 0.5477120280265808,
1904
- "learning_rate": 8.519895337088907e-07,
1905
- "loss": 0.2922,
1906
- "step": 2710
1907
- },
1908
- {
1909
- "epoch": 0.32751354605659244,
1910
- "grad_norm": 1.157457709312439,
1911
- "learning_rate": 8.504937712439131e-07,
1912
- "loss": 0.2699,
1913
- "step": 2720
1914
- },
1915
- {
1916
- "epoch": 0.32871763997591813,
1917
- "grad_norm": 0.5263344049453735,
1918
- "learning_rate": 8.48991816176157e-07,
1919
- "loss": 0.2888,
1920
- "step": 2730
1921
- },
1922
- {
1923
- "epoch": 0.3299217338952438,
1924
- "grad_norm": 0.764481782913208,
1925
- "learning_rate": 8.474836950425026e-07,
1926
- "loss": 0.292,
1927
- "step": 2740
1928
- },
1929
- {
1930
- "epoch": 0.33112582781456956,
1931
- "grad_norm": 0.5704035758972168,
1932
- "learning_rate": 8.459694344887731e-07,
1933
- "loss": 0.2928,
1934
- "step": 2750
1935
- },
1936
- {
1937
- "epoch": 0.33232992173389525,
1938
- "grad_norm": 0.46473219990730286,
1939
- "learning_rate": 8.444490612692645e-07,
1940
- "loss": 0.2816,
1941
- "step": 2760
1942
- },
1943
- {
1944
- "epoch": 0.33353401565322094,
1945
- "grad_norm": 0.5250662565231323,
1946
- "learning_rate": 8.429226022462728e-07,
1947
- "loss": 0.2881,
1948
- "step": 2770
1949
- },
1950
- {
1951
- "epoch": 0.3347381095725467,
1952
- "grad_norm": 0.6085227727890015,
1953
- "learning_rate": 8.413900843896193e-07,
1954
- "loss": 0.3122,
1955
- "step": 2780
1956
- },
1957
- {
1958
- "epoch": 0.33594220349187237,
1959
- "grad_norm": 0.7203246355056763,
1960
- "learning_rate": 8.398515347761745e-07,
1961
- "loss": 0.2911,
1962
- "step": 2790
1963
- },
1964
- {
1965
- "epoch": 0.33714629741119806,
1966
- "grad_norm": 0.5305497050285339,
1967
- "learning_rate": 8.383069805893784e-07,
1968
- "loss": 0.2888,
1969
- "step": 2800
1970
- },
1971
- {
1972
- "epoch": 0.3383503913305238,
1973
- "grad_norm": 0.5452449917793274,
1974
- "learning_rate": 8.367564491187622e-07,
1975
- "loss": 0.2866,
1976
- "step": 2810
1977
- },
1978
- {
1979
- "epoch": 0.3395544852498495,
1980
- "grad_norm": 0.4815659523010254,
1981
- "learning_rate": 8.351999677594645e-07,
1982
- "loss": 0.2863,
1983
- "step": 2820
1984
- },
1985
- {
1986
- "epoch": 0.3407585791691752,
1987
- "grad_norm": 0.5499128103256226,
1988
- "learning_rate": 8.336375640117481e-07,
1989
- "loss": 0.2865,
1990
- "step": 2830
1991
- },
1992
- {
1993
- "epoch": 0.3419626730885009,
1994
- "grad_norm": 0.559804379940033,
1995
- "learning_rate": 8.320692654805136e-07,
1996
- "loss": 0.2833,
1997
- "step": 2840
1998
- },
1999
- {
2000
- "epoch": 0.3431667670078266,
2001
- "grad_norm": 0.5070551633834839,
2002
- "learning_rate": 8.304950998748124e-07,
2003
- "loss": 0.2969,
2004
- "step": 2850
2005
- },
2006
- {
2007
- "epoch": 0.3443708609271523,
2008
- "grad_norm": 0.5566725730895996,
2009
- "learning_rate": 8.289150950073564e-07,
2010
- "loss": 0.2814,
2011
- "step": 2860
2012
- },
2013
- {
2014
- "epoch": 0.34557495484647804,
2015
- "grad_norm": 0.5421969890594482,
2016
- "learning_rate": 8.273292787940268e-07,
2017
- "loss": 0.2805,
2018
- "step": 2870
2019
- },
2020
- {
2021
- "epoch": 0.34677904876580373,
2022
- "grad_norm": 0.49686506390571594,
2023
- "learning_rate": 8.257376792533813e-07,
2024
- "loss": 0.2872,
2025
- "step": 2880
2026
- },
2027
- {
2028
- "epoch": 0.3479831426851294,
2029
- "grad_norm": 0.4665164649486542,
2030
- "learning_rate": 8.241403245061584e-07,
2031
- "loss": 0.2816,
2032
- "step": 2890
2033
- },
2034
- {
2035
- "epoch": 0.34918723660445516,
2036
- "grad_norm": 0.4437556266784668,
2037
- "learning_rate": 8.225372427747813e-07,
2038
- "loss": 0.286,
2039
- "step": 2900
2040
- },
2041
- {
2042
- "epoch": 0.35039133052378085,
2043
- "grad_norm": 0.5280335545539856,
2044
- "learning_rate": 8.209284623828583e-07,
2045
- "loss": 0.2895,
2046
- "step": 2910
2047
- },
2048
- {
2049
- "epoch": 0.35159542444310654,
2050
- "grad_norm": 0.5298367142677307,
2051
- "learning_rate": 8.193140117546832e-07,
2052
- "loss": 0.282,
2053
- "step": 2920
2054
- },
2055
- {
2056
- "epoch": 0.3527995183624323,
2057
- "grad_norm": 0.7123149633407593,
2058
- "learning_rate": 8.176939194147329e-07,
2059
- "loss": 0.2841,
2060
- "step": 2930
2061
- },
2062
- {
2063
- "epoch": 0.354003612281758,
2064
- "grad_norm": 0.6565315127372742,
2065
- "learning_rate": 8.160682139871632e-07,
2066
- "loss": 0.2793,
2067
- "step": 2940
2068
- },
2069
- {
2070
- "epoch": 0.35520770620108366,
2071
- "grad_norm": 0.7005172967910767,
2072
- "learning_rate": 8.144369241953032e-07,
2073
- "loss": 0.2854,
2074
- "step": 2950
2075
- },
2076
- {
2077
- "epoch": 0.3564118001204094,
2078
- "grad_norm": 0.7468757033348083,
2079
- "learning_rate": 8.128000788611478e-07,
2080
- "loss": 0.2992,
2081
- "step": 2960
2082
- },
2083
- {
2084
- "epoch": 0.3576158940397351,
2085
- "grad_norm": 0.5055456161499023,
2086
- "learning_rate": 8.111577069048487e-07,
2087
- "loss": 0.2979,
2088
- "step": 2970
2089
- },
2090
- {
2091
- "epoch": 0.3588199879590608,
2092
- "grad_norm": 0.576806366443634,
2093
- "learning_rate": 8.095098373442027e-07,
2094
- "loss": 0.2915,
2095
- "step": 2980
2096
- },
2097
- {
2098
- "epoch": 0.3600240818783865,
2099
- "grad_norm": 0.5598990321159363,
2100
- "learning_rate": 8.078564992941401e-07,
2101
- "loss": 0.2741,
2102
- "step": 2990
2103
- },
2104
- {
2105
- "epoch": 0.3612281757977122,
2106
- "grad_norm": 0.5614596009254456,
2107
- "learning_rate": 8.061977219662092e-07,
2108
- "loss": 0.2913,
2109
- "step": 3000
2110
- },
2111
- {
2112
- "epoch": 0.3624322697170379,
2113
- "grad_norm": 0.37974095344543457,
2114
- "learning_rate": 8.045335346680611e-07,
2115
- "loss": 0.2787,
2116
- "step": 3010
2117
- },
2118
- {
2119
- "epoch": 0.36363636363636365,
2120
- "grad_norm": 0.6439441442489624,
2121
- "learning_rate": 8.028639668029309e-07,
2122
- "loss": 0.2868,
2123
- "step": 3020
2124
- },
2125
- {
2126
- "epoch": 0.36484045755568933,
2127
- "grad_norm": 0.46323299407958984,
2128
- "learning_rate": 8.011890478691196e-07,
2129
- "loss": 0.2831,
2130
- "step": 3030
2131
- },
2132
- {
2133
- "epoch": 0.3660445514750151,
2134
- "grad_norm": 0.4963575005531311,
2135
- "learning_rate": 7.995088074594713e-07,
2136
- "loss": 0.2782,
2137
- "step": 3040
2138
- },
2139
- {
2140
- "epoch": 0.36724864539434077,
2141
- "grad_norm": 0.6179429888725281,
2142
- "learning_rate": 7.978232752608516e-07,
2143
- "loss": 0.2703,
2144
- "step": 3050
2145
- },
2146
- {
2147
- "epoch": 0.36845273931366646,
2148
- "grad_norm": 0.5127160549163818,
2149
- "learning_rate": 7.961324810536223e-07,
2150
- "loss": 0.3007,
2151
- "step": 3060
2152
- },
2153
- {
2154
- "epoch": 0.3696568332329922,
2155
- "grad_norm": 0.45177775621414185,
2156
- "learning_rate": 7.94436454711116e-07,
2157
- "loss": 0.288,
2158
- "step": 3070
2159
- },
2160
- {
2161
- "epoch": 0.3708609271523179,
2162
- "grad_norm": 0.47144508361816406,
2163
- "learning_rate": 7.927352261991074e-07,
2164
- "loss": 0.2901,
2165
- "step": 3080
2166
- },
2167
- {
2168
- "epoch": 0.3720650210716436,
2169
- "grad_norm": 0.5511527061462402,
2170
- "learning_rate": 7.910288255752844e-07,
2171
- "loss": 0.2754,
2172
- "step": 3090
2173
- },
2174
- {
2175
- "epoch": 0.3732691149909693,
2176
- "grad_norm": 0.5164305567741394,
2177
- "learning_rate": 7.893172829887171e-07,
2178
- "loss": 0.2847,
2179
- "step": 3100
2180
- },
2181
- {
2182
- "epoch": 0.374473208910295,
2183
- "grad_norm": 0.5629504919052124,
2184
- "learning_rate": 7.876006286793251e-07,
2185
- "loss": 0.2953,
2186
- "step": 3110
2187
- },
2188
- {
2189
- "epoch": 0.3756773028296207,
2190
- "grad_norm": 0.513200044631958,
2191
- "learning_rate": 7.858788929773422e-07,
2192
- "loss": 0.2702,
2193
- "step": 3120
2194
- },
2195
- {
2196
- "epoch": 0.37688139674894644,
2197
- "grad_norm": 0.504371166229248,
2198
- "learning_rate": 7.841521063027825e-07,
2199
- "loss": 0.2873,
2200
- "step": 3130
2201
- },
2202
- {
2203
- "epoch": 0.37808549066827213,
2204
- "grad_norm": 0.613593578338623,
2205
- "learning_rate": 7.824202991649013e-07,
2206
- "loss": 0.27,
2207
- "step": 3140
2208
- },
2209
- {
2210
- "epoch": 0.3792895845875978,
2211
- "grad_norm": 0.7345304489135742,
2212
- "learning_rate": 7.806835021616564e-07,
2213
- "loss": 0.2895,
2214
- "step": 3150
2215
- },
2216
- {
2217
- "epoch": 0.38049367850692356,
2218
- "grad_norm": 0.48514464497566223,
2219
- "learning_rate": 7.789417459791681e-07,
2220
- "loss": 0.2809,
2221
- "step": 3160
2222
- },
2223
- {
2224
- "epoch": 0.38169777242624925,
2225
- "grad_norm": 0.4638960063457489,
2226
- "learning_rate": 7.77195061391176e-07,
2227
- "loss": 0.2839,
2228
- "step": 3170
2229
- },
2230
- {
2231
- "epoch": 0.38290186634557494,
2232
- "grad_norm": 0.5008341073989868,
2233
- "learning_rate": 7.754434792584968e-07,
2234
- "loss": 0.2701,
2235
- "step": 3180
2236
- },
2237
- {
2238
- "epoch": 0.3841059602649007,
2239
- "grad_norm": 0.5258957743644714,
2240
- "learning_rate": 7.73687030528477e-07,
2241
- "loss": 0.2709,
2242
- "step": 3190
2243
- },
2244
- {
2245
- "epoch": 0.38531005418422637,
2246
- "grad_norm": 0.5781968832015991,
2247
- "learning_rate": 7.719257462344481e-07,
2248
- "loss": 0.2994,
2249
- "step": 3200
2250
- },
2251
- {
2252
- "epoch": 0.38651414810355206,
2253
- "grad_norm": 0.5485130548477173,
2254
- "learning_rate": 7.701596574951771e-07,
2255
- "loss": 0.3001,
2256
- "step": 3210
2257
- },
2258
- {
2259
- "epoch": 0.3877182420228778,
2260
- "grad_norm": 0.4708418846130371,
2261
- "learning_rate": 7.683887955143169e-07,
2262
- "loss": 0.2736,
2263
- "step": 3220
2264
- },
2265
- {
2266
- "epoch": 0.3889223359422035,
2267
- "grad_norm": 0.5321612358093262,
2268
- "learning_rate": 7.666131915798556e-07,
2269
- "loss": 0.2892,
2270
- "step": 3230
2271
- },
2272
- {
2273
- "epoch": 0.3901264298615292,
2274
- "grad_norm": 0.524898111820221,
2275
- "learning_rate": 7.648328770635623e-07,
2276
- "loss": 0.2897,
2277
- "step": 3240
2278
- },
2279
- {
2280
- "epoch": 0.3913305237808549,
2281
- "grad_norm": 0.4973953664302826,
2282
- "learning_rate": 7.630478834204351e-07,
2283
- "loss": 0.2804,
2284
- "step": 3250
2285
- },
2286
- {
2287
- "epoch": 0.3925346177001806,
2288
- "grad_norm": 0.5439997315406799,
2289
- "learning_rate": 7.612582421881423e-07,
2290
- "loss": 0.2824,
2291
- "step": 3260
2292
- },
2293
- {
2294
- "epoch": 0.3937387116195063,
2295
- "grad_norm": 0.5040695667266846,
2296
- "learning_rate": 7.594639849864681e-07,
2297
- "loss": 0.2806,
2298
- "step": 3270
2299
- },
2300
- {
2301
- "epoch": 0.39494280553883204,
2302
- "grad_norm": 0.57867830991745,
2303
- "learning_rate": 7.576651435167523e-07,
2304
- "loss": 0.2788,
2305
- "step": 3280
2306
- },
2307
- {
2308
- "epoch": 0.39614689945815773,
2309
- "grad_norm": 0.43785402178764343,
2310
- "learning_rate": 7.558617495613304e-07,
2311
- "loss": 0.272,
2312
- "step": 3290
2313
- },
2314
- {
2315
- "epoch": 0.3973509933774834,
2316
- "grad_norm": 0.6042655110359192,
2317
- "learning_rate": 7.540538349829725e-07,
2318
- "loss": 0.2918,
2319
- "step": 3300
2320
- },
2321
- {
2322
- "epoch": 0.39855508729680916,
2323
- "grad_norm": 0.6529451012611389,
2324
- "learning_rate": 7.522414317243198e-07,
2325
- "loss": 0.2882,
2326
- "step": 3310
2327
- },
2328
- {
2329
- "epoch": 0.39975918121613485,
2330
- "grad_norm": 0.5043284296989441,
2331
- "learning_rate": 7.50424571807321e-07,
2332
- "loss": 0.2859,
2333
- "step": 3320
2334
- },
2335
- {
2336
- "epoch": 0.40096327513546054,
2337
- "grad_norm": 0.44874584674835205,
2338
- "learning_rate": 7.486032873326656e-07,
2339
- "loss": 0.2912,
2340
- "step": 3330
2341
- },
2342
- {
2343
- "epoch": 0.4021673690547863,
2344
- "grad_norm": 0.515211284160614,
2345
- "learning_rate": 7.467776104792171e-07,
2346
- "loss": 0.2747,
2347
- "step": 3340
2348
- },
2349
- {
2350
- "epoch": 0.40337146297411197,
2351
- "grad_norm": 0.5425666570663452,
2352
- "learning_rate": 7.449475735034453e-07,
2353
- "loss": 0.2964,
2354
- "step": 3350
2355
- },
2356
- {
2357
- "epoch": 0.4045755568934377,
2358
- "grad_norm": 0.5557084083557129,
2359
- "learning_rate": 7.431132087388546e-07,
2360
- "loss": 0.2809,
2361
- "step": 3360
2362
- },
2363
- {
2364
- "epoch": 0.4057796508127634,
2365
- "grad_norm": 0.4438600540161133,
2366
- "learning_rate": 7.412745485954144e-07,
2367
- "loss": 0.269,
2368
- "step": 3370
2369
- },
2370
- {
2371
- "epoch": 0.4069837447320891,
2372
- "grad_norm": 0.586608350276947,
2373
- "learning_rate": 7.394316255589854e-07,
2374
- "loss": 0.2848,
2375
- "step": 3380
2376
- },
2377
- {
2378
- "epoch": 0.40818783865141484,
2379
- "grad_norm": 0.6429834961891174,
2380
- "learning_rate": 7.375844721907466e-07,
2381
- "loss": 0.2917,
2382
- "step": 3390
2383
- },
2384
- {
2385
- "epoch": 0.4093919325707405,
2386
- "grad_norm": 0.5150188207626343,
2387
- "learning_rate": 7.35733121126619e-07,
2388
- "loss": 0.2772,
2389
- "step": 3400
2390
- },
2391
- {
2392
- "epoch": 0.4105960264900662,
2393
- "grad_norm": 0.5537393093109131,
2394
- "learning_rate": 7.338776050766896e-07,
2395
- "loss": 0.2819,
2396
- "step": 3410
2397
- },
2398
- {
2399
- "epoch": 0.41180012040939196,
2400
- "grad_norm": 0.4834784269332886,
2401
- "learning_rate": 7.320179568246333e-07,
2402
- "loss": 0.2851,
2403
- "step": 3420
2404
- },
2405
- {
2406
- "epoch": 0.41300421432871764,
2407
- "grad_norm": 0.6806831955909729,
2408
- "learning_rate": 7.301542092271337e-07,
2409
- "loss": 0.2841,
2410
- "step": 3430
2411
- },
2412
- {
2413
- "epoch": 0.41420830824804333,
2414
- "grad_norm": 0.5081019997596741,
2415
- "learning_rate": 7.282863952133022e-07,
2416
- "loss": 0.2763,
2417
- "step": 3440
2418
- },
2419
- {
2420
- "epoch": 0.4154124021673691,
2421
- "grad_norm": 0.5681424140930176,
2422
- "learning_rate": 7.264145477840974e-07,
2423
- "loss": 0.2719,
2424
- "step": 3450
2425
- },
2426
- {
2427
- "epoch": 0.41661649608669477,
2428
- "grad_norm": 0.6257504820823669,
2429
- "learning_rate": 7.245387000117404e-07,
2430
- "loss": 0.2813,
2431
- "step": 3460
2432
- },
2433
- {
2434
- "epoch": 0.41782059000602045,
2435
- "grad_norm": 0.5195356607437134,
2436
- "learning_rate": 7.226588850391317e-07,
2437
- "loss": 0.2761,
2438
- "step": 3470
2439
- },
2440
- {
2441
- "epoch": 0.4190246839253462,
2442
- "grad_norm": 0.5490323305130005,
2443
- "learning_rate": 7.207751360792647e-07,
2444
- "loss": 0.291,
2445
- "step": 3480
2446
- },
2447
- {
2448
- "epoch": 0.4202287778446719,
2449
- "grad_norm": 0.6458017230033875,
2450
- "learning_rate": 7.188874864146397e-07,
2451
- "loss": 0.2919,
2452
- "step": 3490
2453
- },
2454
- {
2455
- "epoch": 0.4214328717639976,
2456
- "grad_norm": 0.5081551671028137,
2457
- "learning_rate": 7.16995969396676e-07,
2458
- "loss": 0.2762,
2459
- "step": 3500
2460
- },
2461
- {
2462
- "epoch": 0.4226369656833233,
2463
- "grad_norm": 0.6496263742446899,
2464
- "learning_rate": 7.151006184451212e-07,
2465
- "loss": 0.2766,
2466
- "step": 3510
2467
- },
2468
- {
2469
- "epoch": 0.423841059602649,
2470
- "grad_norm": 0.6383594870567322,
2471
- "learning_rate": 7.132014670474625e-07,
2472
- "loss": 0.2829,
2473
- "step": 3520
2474
- },
2475
- {
2476
- "epoch": 0.4250451535219747,
2477
- "grad_norm": 0.6374247074127197,
2478
- "learning_rate": 7.112985487583333e-07,
2479
- "loss": 0.2776,
2480
- "step": 3530
2481
- },
2482
- {
2483
- "epoch": 0.42624924744130044,
2484
- "grad_norm": 0.48250874876976013,
2485
- "learning_rate": 7.093918971989229e-07,
2486
- "loss": 0.2794,
2487
- "step": 3540
2488
- },
2489
- {
2490
- "epoch": 0.4274533413606261,
2491
- "grad_norm": 0.5055521726608276,
2492
- "learning_rate": 7.07481546056379e-07,
2493
- "loss": 0.2818,
2494
- "step": 3550
2495
- },
2496
- {
2497
- "epoch": 0.4286574352799518,
2498
- "grad_norm": 0.558320164680481,
2499
- "learning_rate": 7.055675290832157e-07,
2500
- "loss": 0.29,
2501
- "step": 3560
2502
- },
2503
- {
2504
- "epoch": 0.42986152919927756,
2505
- "grad_norm": 0.54196697473526,
2506
- "learning_rate": 7.036498800967153e-07,
2507
- "loss": 0.2819,
2508
- "step": 3570
2509
- },
2510
- {
2511
- "epoch": 0.43106562311860325,
2512
- "grad_norm": 0.5442371368408203,
2513
- "learning_rate": 7.017286329783314e-07,
2514
- "loss": 0.3044,
2515
- "step": 3580
2516
- },
2517
- {
2518
- "epoch": 0.43226971703792894,
2519
- "grad_norm": 0.531579315662384,
2520
- "learning_rate": 6.9980382167309e-07,
2521
- "loss": 0.2875,
2522
- "step": 3590
2523
- },
2524
- {
2525
- "epoch": 0.4334738109572547,
2526
- "grad_norm": 0.6069034934043884,
2527
- "learning_rate": 6.978754801889902e-07,
2528
- "loss": 0.2915,
2529
- "step": 3600
2530
- },
2531
- {
2532
- "epoch": 0.43467790487658037,
2533
- "grad_norm": 0.5376235246658325,
2534
- "learning_rate": 6.959436425964033e-07,
2535
- "loss": 0.2768,
2536
- "step": 3610
2537
- },
2538
- {
2539
- "epoch": 0.43588199879590606,
2540
- "grad_norm": 0.5438763499259949,
2541
- "learning_rate": 6.9400834302747e-07,
2542
- "loss": 0.2911,
2543
- "step": 3620
2544
- },
2545
- {
2546
- "epoch": 0.4370860927152318,
2547
- "grad_norm": 0.4325105547904968,
2548
- "learning_rate": 6.920696156754985e-07,
2549
- "loss": 0.269,
2550
- "step": 3630
2551
- },
2552
- {
2553
- "epoch": 0.4382901866345575,
2554
- "grad_norm": 0.5107905864715576,
2555
- "learning_rate": 6.901274947943597e-07,
2556
- "loss": 0.2754,
2557
- "step": 3640
2558
- },
2559
- {
2560
- "epoch": 0.4394942805538832,
2561
- "grad_norm": 0.5302306413650513,
2562
- "learning_rate": 6.881820146978822e-07,
2563
- "loss": 0.2835,
2564
- "step": 3650
2565
- },
2566
- {
2567
- "epoch": 0.4406983744732089,
2568
- "grad_norm": 0.5489309430122375,
2569
- "learning_rate": 6.862332097592457e-07,
2570
- "loss": 0.2746,
2571
- "step": 3660
2572
- },
2573
- {
2574
- "epoch": 0.4419024683925346,
2575
- "grad_norm": 0.4515032172203064,
2576
- "learning_rate": 6.842811144103743e-07,
2577
- "loss": 0.2829,
2578
- "step": 3670
2579
- },
2580
- {
2581
- "epoch": 0.44310656231186035,
2582
- "grad_norm": 0.5359588861465454,
2583
- "learning_rate": 6.823257631413275e-07,
2584
- "loss": 0.2826,
2585
- "step": 3680
2586
- },
2587
- {
2588
- "epoch": 0.44431065623118604,
2589
- "grad_norm": 0.49561506509780884,
2590
- "learning_rate": 6.803671904996916e-07,
2591
- "loss": 0.2946,
2592
- "step": 3690
2593
- },
2594
- {
2595
- "epoch": 0.44551475015051173,
2596
- "grad_norm": 0.43841075897216797,
2597
- "learning_rate": 6.784054310899683e-07,
2598
- "loss": 0.2802,
2599
- "step": 3700
2600
- },
2601
- {
2602
- "epoch": 0.4467188440698375,
2603
- "grad_norm": 0.7528261542320251,
2604
- "learning_rate": 6.764405195729639e-07,
2605
- "loss": 0.2829,
2606
- "step": 3710
2607
- },
2608
- {
2609
- "epoch": 0.44792293798916316,
2610
- "grad_norm": 1.1440777778625488,
2611
- "learning_rate": 6.744724906651774e-07,
2612
- "loss": 0.2665,
2613
- "step": 3720
2614
- },
2615
- {
2616
- "epoch": 0.44912703190848885,
2617
- "grad_norm": 0.5153807997703552,
2618
- "learning_rate": 6.72501379138186e-07,
2619
- "loss": 0.2754,
2620
- "step": 3730
2621
- },
2622
- {
2623
- "epoch": 0.4503311258278146,
2624
- "grad_norm": 0.582036554813385,
2625
- "learning_rate": 6.705272198180312e-07,
2626
- "loss": 0.2818,
2627
- "step": 3740
2628
- },
2629
- {
2630
- "epoch": 0.4515352197471403,
2631
- "grad_norm": 0.7196856737136841,
2632
- "learning_rate": 6.685500475846044e-07,
2633
- "loss": 0.2744,
2634
- "step": 3750
2635
- },
2636
- {
2637
- "epoch": 0.45273931366646597,
2638
- "grad_norm": 1.0595272779464722,
2639
- "learning_rate": 6.665698973710288e-07,
2640
- "loss": 0.2602,
2641
- "step": 3760
2642
- },
2643
- {
2644
- "epoch": 0.4539434075857917,
2645
- "grad_norm": 0.4910378158092499,
2646
- "learning_rate": 6.645868041630439e-07,
2647
- "loss": 0.2887,
2648
- "step": 3770
2649
- },
2650
- {
2651
- "epoch": 0.4551475015051174,
2652
- "grad_norm": 0.4395122230052948,
2653
- "learning_rate": 6.626008029983867e-07,
2654
- "loss": 0.2771,
2655
- "step": 3780
2656
- },
2657
- {
2658
- "epoch": 0.4563515954244431,
2659
- "grad_norm": 0.5630185008049011,
2660
- "learning_rate": 6.606119289661721e-07,
2661
- "loss": 0.2976,
2662
- "step": 3790
2663
- },
2664
- {
2665
- "epoch": 0.45755568934376883,
2666
- "grad_norm": 0.6062456965446472,
2667
- "learning_rate": 6.58620217206274e-07,
2668
- "loss": 0.2707,
2669
- "step": 3800
2670
- },
2671
- {
2672
- "epoch": 0.4587597832630945,
2673
- "grad_norm": 0.6882142424583435,
2674
- "learning_rate": 6.566257029087039e-07,
2675
- "loss": 0.2732,
2676
- "step": 3810
2677
- },
2678
- {
2679
- "epoch": 0.4599638771824202,
2680
- "grad_norm": 0.4631926417350769,
2681
- "learning_rate": 6.546284213129885e-07,
2682
- "loss": 0.2794,
2683
- "step": 3820
2684
- },
2685
- {
2686
- "epoch": 0.46116797110174595,
2687
- "grad_norm": 0.4465793967247009,
2688
- "learning_rate": 6.526284077075488e-07,
2689
- "loss": 0.2809,
2690
- "step": 3830
2691
- },
2692
- {
2693
- "epoch": 0.46237206502107164,
2694
- "grad_norm": 0.5073222517967224,
2695
- "learning_rate": 6.506256974290747e-07,
2696
- "loss": 0.2908,
2697
- "step": 3840
2698
- },
2699
- {
2700
- "epoch": 0.46357615894039733,
2701
- "grad_norm": 0.5717306137084961,
2702
- "learning_rate": 6.486203258619016e-07,
2703
- "loss": 0.282,
2704
- "step": 3850
2705
- },
2706
- {
2707
- "epoch": 0.4647802528597231,
2708
- "grad_norm": 0.5614638924598694,
2709
- "learning_rate": 6.466123284373858e-07,
2710
- "loss": 0.2764,
2711
- "step": 3860
2712
- },
2713
- {
2714
- "epoch": 0.46598434677904876,
2715
- "grad_norm": 0.626006007194519,
2716
- "learning_rate": 6.446017406332772e-07,
2717
- "loss": 0.277,
2718
- "step": 3870
2719
- },
2720
- {
2721
- "epoch": 0.46718844069837445,
2722
- "grad_norm": 0.47509709000587463,
2723
- "learning_rate": 6.425885979730933e-07,
2724
- "loss": 0.2828,
2725
- "step": 3880
2726
- },
2727
- {
2728
- "epoch": 0.4683925346177002,
2729
- "grad_norm": 0.5545176267623901,
2730
- "learning_rate": 6.405729360254914e-07,
2731
- "loss": 0.2893,
2732
- "step": 3890
2733
- },
2734
- {
2735
- "epoch": 0.4695966285370259,
2736
- "grad_norm": 0.4888879060745239,
2737
- "learning_rate": 6.3855479040364e-07,
2738
- "loss": 0.2811,
2739
- "step": 3900
2740
- },
2741
- {
2742
- "epoch": 0.4708007224563516,
2743
- "grad_norm": 0.44063079357147217,
2744
- "learning_rate": 6.365341967645902e-07,
2745
- "loss": 0.2782,
2746
- "step": 3910
2747
- },
2748
- {
2749
- "epoch": 0.4720048163756773,
2750
- "grad_norm": 0.5356207489967346,
2751
- "learning_rate": 6.345111908086444e-07,
2752
- "loss": 0.2658,
2753
- "step": 3920
2754
- },
2755
- {
2756
- "epoch": 0.473208910295003,
2757
- "grad_norm": 0.5134460926055908,
2758
- "learning_rate": 6.324858082787275e-07,
2759
- "loss": 0.2782,
2760
- "step": 3930
2761
- },
2762
- {
2763
- "epoch": 0.4744130042143287,
2764
- "grad_norm": 0.5685980916023254,
2765
- "learning_rate": 6.304580849597527e-07,
2766
- "loss": 0.2704,
2767
- "step": 3940
2768
- },
2769
- {
2770
- "epoch": 0.47561709813365444,
2771
- "grad_norm": 0.8610411286354065,
2772
- "learning_rate": 6.284280566779923e-07,
2773
- "loss": 0.29,
2774
- "step": 3950
2775
- },
2776
- {
2777
- "epoch": 0.4768211920529801,
2778
- "grad_norm": 0.5496920943260193,
2779
- "learning_rate": 6.263957593004421e-07,
2780
- "loss": 0.2704,
2781
- "step": 3960
2782
- },
2783
- {
2784
- "epoch": 0.4780252859723058,
2785
- "grad_norm": 0.4593532383441925,
2786
- "learning_rate": 6.243612287341896e-07,
2787
- "loss": 0.2806,
2788
- "step": 3970
2789
- },
2790
- {
2791
- "epoch": 0.47922937989163156,
2792
- "grad_norm": 0.5178139805793762,
2793
- "learning_rate": 6.223245009257783e-07,
2794
- "loss": 0.2683,
2795
- "step": 3980
2796
- },
2797
- {
2798
- "epoch": 0.48043347381095725,
2799
- "grad_norm": 0.6350088119506836,
2800
- "learning_rate": 6.20285611860573e-07,
2801
- "loss": 0.2796,
2802
- "step": 3990
2803
- },
2804
- {
2805
- "epoch": 0.481637567730283,
2806
- "grad_norm": 0.4848230183124542,
2807
- "learning_rate": 6.182445975621246e-07,
2808
- "loss": 0.2727,
2809
- "step": 4000
2810
- },
2811
- {
2812
- "epoch": 0.4828416616496087,
2813
- "grad_norm": 0.6039783358573914,
2814
- "learning_rate": 6.162014940915323e-07,
2815
- "loss": 0.295,
2816
- "step": 4010
2817
- },
2818
- {
2819
- "epoch": 0.48404575556893437,
2820
- "grad_norm": 0.5623034834861755,
2821
- "learning_rate": 6.141563375468082e-07,
2822
- "loss": 0.2843,
2823
- "step": 4020
2824
- },
2825
- {
2826
- "epoch": 0.4852498494882601,
2827
- "grad_norm": 0.5298231244087219,
2828
- "learning_rate": 6.12109164062238e-07,
2829
- "loss": 0.2685,
2830
- "step": 4030
2831
- },
2832
- {
2833
- "epoch": 0.4864539434075858,
2834
- "grad_norm": 0.49439486861228943,
2835
- "learning_rate": 6.100600098077431e-07,
2836
- "loss": 0.2588,
2837
- "step": 4040
2838
- },
2839
- {
2840
- "epoch": 0.4876580373269115,
2841
- "grad_norm": 0.4667768180370331,
2842
- "learning_rate": 6.080089109882418e-07,
2843
- "loss": 0.275,
2844
- "step": 4050
2845
- },
2846
- {
2847
- "epoch": 0.48886213124623723,
2848
- "grad_norm": 0.5490863919258118,
2849
- "learning_rate": 6.059559038430094e-07,
2850
- "loss": 0.2837,
2851
- "step": 4060
2852
- },
2853
- {
2854
- "epoch": 0.4900662251655629,
2855
- "grad_norm": 0.467192143201828,
2856
- "learning_rate": 6.039010246450376e-07,
2857
- "loss": 0.2733,
2858
- "step": 4070
2859
- },
2860
- {
2861
- "epoch": 0.4912703190848886,
2862
- "grad_norm": 0.49663642048835754,
2863
- "learning_rate": 6.018443097003945e-07,
2864
- "loss": 0.2738,
2865
- "step": 4080
2866
- },
2867
- {
2868
- "epoch": 0.49247441300421435,
2869
- "grad_norm": 0.501777708530426,
2870
- "learning_rate": 5.997857953475823e-07,
2871
- "loss": 0.2743,
2872
- "step": 4090
2873
- },
2874
- {
2875
- "epoch": 0.49367850692354004,
2876
- "grad_norm": 0.5064652562141418,
2877
- "learning_rate": 5.977255179568955e-07,
2878
- "loss": 0.2748,
2879
- "step": 4100
2880
- },
2881
- {
2882
- "epoch": 0.4948826008428657,
2883
- "grad_norm": 0.6248656511306763,
2884
- "learning_rate": 5.956635139297783e-07,
2885
- "loss": 0.2765,
2886
- "step": 4110
2887
- },
2888
- {
2889
- "epoch": 0.49608669476219147,
2890
- "grad_norm": 0.45688706636428833,
2891
- "learning_rate": 5.935998196981817e-07,
2892
- "loss": 0.271,
2893
- "step": 4120
2894
- },
2895
- {
2896
- "epoch": 0.49729078868151716,
2897
- "grad_norm": 0.7225250601768494,
2898
- "learning_rate": 5.915344717239197e-07,
2899
- "loss": 0.2853,
2900
- "step": 4130
2901
- },
2902
- {
2903
- "epoch": 0.49849488260084285,
2904
- "grad_norm": 0.5863081812858582,
2905
- "learning_rate": 5.894675064980246e-07,
2906
- "loss": 0.2685,
2907
- "step": 4140
2908
- },
2909
- {
2910
- "epoch": 0.4996989765201686,
2911
- "grad_norm": 0.5770187973976135,
2912
- "learning_rate": 5.87398960540103e-07,
2913
- "loss": 0.2774,
2914
- "step": 4150
2915
- },
2916
- {
2917
- "epoch": 0.5009030704394943,
2918
- "grad_norm": 0.41943806409835815,
2919
- "learning_rate": 5.8532887039769e-07,
2920
- "loss": 0.2622,
2921
- "step": 4160
2922
- },
2923
- {
2924
- "epoch": 0.50210716435882,
2925
- "grad_norm": 0.6374907493591309,
2926
- "learning_rate": 5.832572726456039e-07,
2927
- "loss": 0.2858,
2928
- "step": 4170
2929
- },
2930
- {
2931
- "epoch": 0.5033112582781457,
2932
- "grad_norm": 0.5210843086242676,
2933
- "learning_rate": 5.811842038852996e-07,
2934
- "loss": 0.2706,
2935
- "step": 4180
2936
- },
2937
- {
2938
- "epoch": 0.5045153521974715,
2939
- "grad_norm": 0.596387505531311,
2940
- "learning_rate": 5.791097007442222e-07,
2941
- "loss": 0.2823,
2942
- "step": 4190
2943
- },
2944
- {
2945
- "epoch": 0.5057194461167971,
2946
- "grad_norm": 0.6676878929138184,
2947
- "learning_rate": 5.7703379987516e-07,
2948
- "loss": 0.2848,
2949
- "step": 4200
2950
- },
2951
- {
2952
- "epoch": 0.5069235400361228,
2953
- "grad_norm": 0.6097555160522461,
2954
- "learning_rate": 5.749565379555961e-07,
2955
- "loss": 0.2766,
2956
- "step": 4210
2957
- },
2958
- {
2959
- "epoch": 0.5081276339554486,
2960
- "grad_norm": 0.6043739318847656,
2961
- "learning_rate": 5.728779516870615e-07,
2962
- "loss": 0.2885,
2963
- "step": 4220
2964
- },
2965
- {
2966
- "epoch": 0.5093317278747742,
2967
- "grad_norm": 0.5565124750137329,
2968
- "learning_rate": 5.707980777944859e-07,
2969
- "loss": 0.2643,
2970
- "step": 4230
2971
- },
2972
- {
2973
- "epoch": 0.5105358217941,
2974
- "grad_norm": 0.49649959802627563,
2975
- "learning_rate": 5.687169530255487e-07,
2976
- "loss": 0.2672,
2977
- "step": 4240
2978
- },
2979
- {
2980
- "epoch": 0.5117399157134257,
2981
- "grad_norm": 0.49968451261520386,
2982
- "learning_rate": 5.666346141500307e-07,
2983
- "loss": 0.2754,
2984
- "step": 4250
2985
- },
2986
- {
2987
- "epoch": 0.5129440096327513,
2988
- "grad_norm": 0.4982677698135376,
2989
- "learning_rate": 5.645510979591634e-07,
2990
- "loss": 0.2785,
2991
- "step": 4260
2992
- },
2993
- {
2994
- "epoch": 0.5141481035520771,
2995
- "grad_norm": 0.904083251953125,
2996
- "learning_rate": 5.624664412649797e-07,
2997
- "loss": 0.2833,
2998
- "step": 4270
2999
- },
3000
- {
3001
- "epoch": 0.5153521974714028,
3002
- "grad_norm": 0.5038682222366333,
3003
- "learning_rate": 5.603806808996625e-07,
3004
- "loss": 0.2746,
3005
- "step": 4280
3006
- },
3007
- {
3008
- "epoch": 0.5165562913907285,
3009
- "grad_norm": 0.7115175724029541,
3010
- "learning_rate": 5.58293853714895e-07,
3011
- "loss": 0.2712,
3012
- "step": 4290
3013
- },
3014
- {
3015
- "epoch": 0.5177603853100542,
3016
- "grad_norm": 0.5522176027297974,
3017
- "learning_rate": 5.562059965812097e-07,
3018
- "loss": 0.2869,
3019
- "step": 4300
3020
- },
3021
- {
3022
- "epoch": 0.5189644792293799,
3023
- "grad_norm": 0.6081178784370422,
3024
- "learning_rate": 5.541171463873357e-07,
3025
- "loss": 0.2751,
3026
- "step": 4310
3027
- },
3028
- {
3029
- "epoch": 0.5201685731487056,
3030
- "grad_norm": 0.5689599514007568,
3031
- "learning_rate": 5.52027340039548e-07,
3032
- "loss": 0.2875,
3033
- "step": 4320
3034
- },
3035
- {
3036
- "epoch": 0.5213726670680313,
3037
- "grad_norm": 0.43370601534843445,
3038
- "learning_rate": 5.499366144610153e-07,
3039
- "loss": 0.2673,
3040
- "step": 4330
3041
- },
3042
- {
3043
- "epoch": 0.5225767609873571,
3044
- "grad_norm": 0.5115625262260437,
3045
- "learning_rate": 5.478450065911473e-07,
3046
- "loss": 0.2791,
3047
- "step": 4340
3048
- },
3049
- {
3050
- "epoch": 0.5237808549066827,
3051
- "grad_norm": 0.518798291683197,
3052
- "learning_rate": 5.45752553384942e-07,
3053
- "loss": 0.277,
3054
- "step": 4350
3055
- },
3056
- {
3057
- "epoch": 0.5249849488260084,
3058
- "grad_norm": 0.5628324151039124,
3059
- "learning_rate": 5.436592918123337e-07,
3060
- "loss": 0.2884,
3061
- "step": 4360
3062
- },
3063
- {
3064
- "epoch": 0.5261890427453342,
3065
- "grad_norm": 0.47458890080451965,
3066
- "learning_rate": 5.415652588575385e-07,
3067
- "loss": 0.27,
3068
- "step": 4370
3069
- },
3070
- {
3071
- "epoch": 0.5273931366646598,
3072
- "grad_norm": 0.6163709759712219,
3073
- "learning_rate": 5.394704915184014e-07,
3074
- "loss": 0.2643,
3075
- "step": 4380
3076
- },
3077
- {
3078
- "epoch": 0.5285972305839856,
3079
- "grad_norm": 0.44985631108283997,
3080
- "learning_rate": 5.373750268057431e-07,
3081
- "loss": 0.2774,
3082
- "step": 4390
3083
- },
3084
- {
3085
- "epoch": 0.5298013245033113,
3086
- "grad_norm": 0.47572416067123413,
3087
- "learning_rate": 5.352789017427052e-07,
3088
- "loss": 0.278,
3089
- "step": 4400
3090
- },
3091
- {
3092
- "epoch": 0.5310054184226369,
3093
- "grad_norm": 0.5311432480812073,
3094
- "learning_rate": 5.33182153364097e-07,
3095
- "loss": 0.283,
3096
- "step": 4410
3097
- },
3098
- {
3099
- "epoch": 0.5322095123419627,
3100
- "grad_norm": 0.5810163617134094,
3101
- "learning_rate": 5.310848187157403e-07,
3102
- "loss": 0.257,
3103
- "step": 4420
3104
- },
3105
- {
3106
- "epoch": 0.5334136062612884,
3107
- "grad_norm": 0.8989514708518982,
3108
- "learning_rate": 5.289869348538153e-07,
3109
- "loss": 0.2846,
3110
- "step": 4430
3111
- },
3112
- {
3113
- "epoch": 0.534617700180614,
3114
- "grad_norm": 0.4534051716327667,
3115
- "learning_rate": 5.26888538844206e-07,
3116
- "loss": 0.2836,
3117
- "step": 4440
3118
- },
3119
- {
3120
- "epoch": 0.5358217940999398,
3121
- "grad_norm": 0.4670819938182831,
3122
- "learning_rate": 5.247896677618452e-07,
3123
- "loss": 0.2614,
3124
- "step": 4450
3125
- },
3126
- {
3127
- "epoch": 0.5370258880192655,
3128
- "grad_norm": 0.5935913324356079,
3129
- "learning_rate": 5.226903586900587e-07,
3130
- "loss": 0.2826,
3131
- "step": 4460
3132
- },
3133
- {
3134
- "epoch": 0.5382299819385912,
3135
- "grad_norm": 0.45839351415634155,
3136
- "learning_rate": 5.205906487199119e-07,
3137
- "loss": 0.2514,
3138
- "step": 4470
3139
- },
3140
- {
3141
- "epoch": 0.5394340758579169,
3142
- "grad_norm": 0.4929831624031067,
3143
- "learning_rate": 5.184905749495525e-07,
3144
- "loss": 0.2815,
3145
- "step": 4480
3146
- },
3147
- {
3148
- "epoch": 0.5406381697772427,
3149
- "grad_norm": 0.529437780380249,
3150
- "learning_rate": 5.163901744835564e-07,
3151
- "loss": 0.2744,
3152
- "step": 4490
3153
- },
3154
- {
3155
- "epoch": 0.5418422636965683,
3156
- "grad_norm": 0.44370970129966736,
3157
- "learning_rate": 5.14289484432271e-07,
3158
- "loss": 0.2837,
3159
- "step": 4500
3160
- },
3161
- {
3162
- "epoch": 0.543046357615894,
3163
- "grad_norm": 0.46680358052253723,
3164
- "learning_rate": 5.121885419111611e-07,
3165
- "loss": 0.2833,
3166
- "step": 4510
3167
- },
3168
- {
3169
- "epoch": 0.5442504515352198,
3170
- "grad_norm": 0.5581067204475403,
3171
- "learning_rate": 5.100873840401513e-07,
3172
- "loss": 0.2846,
3173
- "step": 4520
3174
- },
3175
- {
3176
- "epoch": 0.5454545454545454,
3177
- "grad_norm": 0.4683559238910675,
3178
- "learning_rate": 5.079860479429718e-07,
3179
- "loss": 0.2666,
3180
- "step": 4530
3181
- },
3182
- {
3183
- "epoch": 0.5466586393738712,
3184
- "grad_norm": 0.464067280292511,
3185
- "learning_rate": 5.058845707465009e-07,
3186
- "loss": 0.2693,
3187
- "step": 4540
3188
- },
3189
- {
3190
- "epoch": 0.5478627332931969,
3191
- "grad_norm": 0.5715063214302063,
3192
- "learning_rate": 5.037829895801106e-07,
3193
- "loss": 0.2746,
3194
- "step": 4550
3195
- },
3196
- {
3197
- "epoch": 0.5490668272125225,
3198
- "grad_norm": 0.585356593132019,
3199
- "learning_rate": 5.016813415750097e-07,
3200
- "loss": 0.281,
3201
- "step": 4560
3202
- },
3203
- {
3204
- "epoch": 0.5502709211318483,
3205
- "grad_norm": 0.4893047511577606,
3206
- "learning_rate": 4.995796638635875e-07,
3207
- "loss": 0.2799,
3208
- "step": 4570
3209
- },
3210
- {
3211
- "epoch": 0.551475015051174,
3212
- "grad_norm": 1.0689632892608643,
3213
- "learning_rate": 4.974779935787589e-07,
3214
- "loss": 0.2574,
3215
- "step": 4580
3216
- },
3217
- {
3218
- "epoch": 0.5526791089704997,
3219
- "grad_norm": 0.6054455637931824,
3220
- "learning_rate": 4.953763678533068e-07,
3221
- "loss": 0.2635,
3222
- "step": 4590
3223
- },
3224
- {
3225
- "epoch": 0.5538832028898254,
3226
- "grad_norm": 0.46325477957725525,
3227
- "learning_rate": 4.932748238192273e-07,
3228
- "loss": 0.2769,
3229
- "step": 4600
3230
- },
3231
- {
3232
- "epoch": 0.5550872968091511,
3233
- "grad_norm": 0.5770764350891113,
3234
- "learning_rate": 4.911733986070735e-07,
3235
- "loss": 0.2671,
3236
- "step": 4610
3237
- },
3238
- {
3239
- "epoch": 0.5562913907284768,
3240
- "grad_norm": 0.5715611577033997,
3241
- "learning_rate": 4.890721293452979e-07,
3242
- "loss": 0.2917,
3243
- "step": 4620
3244
- },
3245
- {
3246
- "epoch": 0.5574954846478025,
3247
- "grad_norm": 0.5384266972541809,
3248
- "learning_rate": 4.869710531595988e-07,
3249
- "loss": 0.2771,
3250
- "step": 4630
3251
- },
3252
- {
3253
- "epoch": 0.5586995785671283,
3254
- "grad_norm": 0.4611688256263733,
3255
- "learning_rate": 4.848702071722629e-07,
3256
- "loss": 0.2828,
3257
- "step": 4640
3258
- },
3259
- {
3260
- "epoch": 0.5599036724864539,
3261
- "grad_norm": 0.6118834018707275,
3262
- "learning_rate": 4.827696285015094e-07,
3263
- "loss": 0.2832,
3264
- "step": 4650
3265
- },
3266
- {
3267
- "epoch": 0.5611077664057796,
3268
- "grad_norm": 0.5026919841766357,
3269
- "learning_rate": 4.806693542608348e-07,
3270
- "loss": 0.2735,
3271
- "step": 4660
3272
- },
3273
- {
3274
- "epoch": 0.5623118603251054,
3275
- "grad_norm": 0.548273503780365,
3276
- "learning_rate": 4.785694215583566e-07,
3277
- "loss": 0.2742,
3278
- "step": 4670
3279
- },
3280
- {
3281
- "epoch": 0.563515954244431,
3282
- "grad_norm": 0.6186013221740723,
3283
- "learning_rate": 4.764698674961581e-07,
3284
- "loss": 0.2784,
3285
- "step": 4680
3286
- },
3287
- {
3288
- "epoch": 0.5647200481637568,
3289
- "grad_norm": 0.45300328731536865,
3290
- "learning_rate": 4.743707291696329e-07,
3291
- "loss": 0.2786,
3292
- "step": 4690
3293
- },
3294
- {
3295
- "epoch": 0.5659241420830825,
3296
- "grad_norm": 0.49064886569976807,
3297
- "learning_rate": 4.7227204366682873e-07,
3298
- "loss": 0.2747,
3299
- "step": 4700
3300
- },
3301
- {
3302
- "epoch": 0.5671282360024081,
3303
- "grad_norm": 0.5186241865158081,
3304
- "learning_rate": 4.7017384806779336e-07,
3305
- "loss": 0.2788,
3306
- "step": 4710
3307
- },
3308
- {
3309
- "epoch": 0.5683323299217339,
3310
- "grad_norm": 0.5284368395805359,
3311
- "learning_rate": 4.6807617944391843e-07,
3312
- "loss": 0.264,
3313
- "step": 4720
3314
- },
3315
- {
3316
- "epoch": 0.5695364238410596,
3317
- "grad_norm": 0.5770208239555359,
3318
- "learning_rate": 4.6597907485728477e-07,
3319
- "loss": 0.2759,
3320
- "step": 4730
3321
- },
3322
- {
3323
- "epoch": 0.5707405177603853,
3324
- "grad_norm": 0.5039085149765015,
3325
- "learning_rate": 4.6388257136000807e-07,
3326
- "loss": 0.2807,
3327
- "step": 4740
3328
- },
3329
- {
3330
- "epoch": 0.571944611679711,
3331
- "grad_norm": 1.2547776699066162,
3332
- "learning_rate": 4.617867059935838e-07,
3333
- "loss": 0.2651,
3334
- "step": 4750
3335
- },
3336
- {
3337
- "epoch": 0.5731487055990367,
3338
- "grad_norm": 0.5457895398139954,
3339
- "learning_rate": 4.5969151578823224e-07,
3340
- "loss": 0.27,
3341
- "step": 4760
3342
- },
3343
- {
3344
- "epoch": 0.5743527995183624,
3345
- "grad_norm": 0.4974658787250519,
3346
- "learning_rate": 4.5759703776224555e-07,
3347
- "loss": 0.2794,
3348
- "step": 4770
3349
- },
3350
- {
3351
- "epoch": 0.5755568934376881,
3352
- "grad_norm": 0.5161871314048767,
3353
- "learning_rate": 4.555033089213321e-07,
3354
- "loss": 0.2816,
3355
- "step": 4780
3356
- },
3357
- {
3358
- "epoch": 0.5767609873570139,
3359
- "grad_norm": 0.43015995621681213,
3360
- "learning_rate": 4.534103662579642e-07,
3361
- "loss": 0.267,
3362
- "step": 4790
3363
- },
3364
- {
3365
- "epoch": 0.5779650812763396,
3366
- "grad_norm": 0.4864785969257355,
3367
- "learning_rate": 4.5131824675072364e-07,
3368
- "loss": 0.2793,
3369
- "step": 4800
3370
- },
3371
- {
3372
- "epoch": 0.5791691751956652,
3373
- "grad_norm": 0.6006112694740295,
3374
- "learning_rate": 4.492269873636482e-07,
3375
- "loss": 0.2689,
3376
- "step": 4810
3377
- },
3378
- {
3379
- "epoch": 0.580373269114991,
3380
- "grad_norm": 0.4434204399585724,
3381
- "learning_rate": 4.4713662504557927e-07,
3382
- "loss": 0.2876,
3383
- "step": 4820
3384
- },
3385
- {
3386
- "epoch": 0.5815773630343167,
3387
- "grad_norm": 0.565077543258667,
3388
- "learning_rate": 4.450471967295083e-07,
3389
- "loss": 0.2658,
3390
- "step": 4830
3391
- },
3392
- {
3393
- "epoch": 0.5827814569536424,
3394
- "grad_norm": 0.5381281971931458,
3395
- "learning_rate": 4.429587393319246e-07,
3396
- "loss": 0.2715,
3397
- "step": 4840
3398
- },
3399
- {
3400
- "epoch": 0.5839855508729681,
3401
- "grad_norm": 0.49021026492118835,
3402
- "learning_rate": 4.408712897521633e-07,
3403
- "loss": 0.2688,
3404
- "step": 4850
3405
- },
3406
- {
3407
- "epoch": 0.5851896447922939,
3408
- "grad_norm": 0.5293102264404297,
3409
- "learning_rate": 4.3878488487175323e-07,
3410
- "loss": 0.2604,
3411
- "step": 4860
3412
- },
3413
- {
3414
- "epoch": 0.5863937387116195,
3415
- "grad_norm": 0.6353856921195984,
3416
- "learning_rate": 4.3669956155376476e-07,
3417
- "loss": 0.2586,
3418
- "step": 4870
3419
- },
3420
- {
3421
- "epoch": 0.5875978326309452,
3422
- "grad_norm": 0.5306446552276611,
3423
- "learning_rate": 4.3461535664215923e-07,
3424
- "loss": 0.2624,
3425
- "step": 4880
3426
- },
3427
- {
3428
- "epoch": 0.588801926550271,
3429
- "grad_norm": 0.5957462191581726,
3430
- "learning_rate": 4.325323069611383e-07,
3431
- "loss": 0.2731,
3432
- "step": 4890
3433
- },
3434
- {
3435
- "epoch": 0.5900060204695966,
3436
- "grad_norm": 0.6803829073905945,
3437
- "learning_rate": 4.3045044931449156e-07,
3438
- "loss": 0.2779,
3439
- "step": 4900
3440
- },
3441
- {
3442
- "epoch": 0.5912101143889223,
3443
- "grad_norm": 0.5501326322555542,
3444
- "learning_rate": 4.2836982048494854e-07,
3445
- "loss": 0.2675,
3446
- "step": 4910
3447
- },
3448
- {
3449
- "epoch": 0.5924142083082481,
3450
- "grad_norm": 0.49481987953186035,
3451
- "learning_rate": 4.262904572335272e-07,
3452
- "loss": 0.2725,
3453
- "step": 4920
3454
- },
3455
- {
3456
- "epoch": 0.5936183022275737,
3457
- "grad_norm": 0.5254814028739929,
3458
- "learning_rate": 4.242123962988851e-07,
3459
- "loss": 0.2804,
3460
- "step": 4930
3461
- },
3462
- {
3463
- "epoch": 0.5948223961468995,
3464
- "grad_norm": 0.5598310232162476,
3465
- "learning_rate": 4.2213567439667037e-07,
3466
- "loss": 0.2703,
3467
- "step": 4940
3468
- },
3469
- {
3470
- "epoch": 0.5960264900662252,
3471
- "grad_norm": 0.5715354681015015,
3472
- "learning_rate": 4.200603282188724e-07,
3473
- "loss": 0.2799,
3474
- "step": 4950
3475
- },
3476
- {
3477
- "epoch": 0.5972305839855508,
3478
- "grad_norm": 0.6474336981773376,
3479
- "learning_rate": 4.179863944331743e-07,
3480
- "loss": 0.2799,
3481
- "step": 4960
3482
- },
3483
- {
3484
- "epoch": 0.5984346779048766,
3485
- "grad_norm": 0.47116249799728394,
3486
- "learning_rate": 4.15913909682305e-07,
3487
- "loss": 0.2751,
3488
- "step": 4970
3489
- },
3490
- {
3491
- "epoch": 0.5996387718242023,
3492
- "grad_norm": 0.5750442147254944,
3493
- "learning_rate": 4.138429105833906e-07,
3494
- "loss": 0.2719,
3495
- "step": 4980
3496
- },
3497
- {
3498
- "epoch": 0.600842865743528,
3499
- "grad_norm": 0.5243822932243347,
3500
- "learning_rate": 4.1177343372730923e-07,
3501
- "loss": 0.2709,
3502
- "step": 4990
3503
- },
3504
- {
3505
- "epoch": 0.6020469596628537,
3506
- "grad_norm": 0.5334904789924622,
3507
- "learning_rate": 4.097055156780437e-07,
3508
- "loss": 0.272,
3509
- "step": 5000
3510
- }
3511
- ],
3512
- "logging_steps": 10,
3513
- "max_steps": 8305,
3514
- "num_input_tokens_seen": 0,
3515
- "num_train_epochs": 1,
3516
- "save_steps": 1000,
3517
- "stateful_callbacks": {
3518
- "TrainerControl": {
3519
- "args": {
3520
- "should_epoch_stop": false,
3521
- "should_evaluate": false,
3522
- "should_log": false,
3523
- "should_save": true,
3524
- "should_training_stop": false
3525
- },
3526
- "attributes": {}
3527
- }
3528
- },
3529
- "total_flos": 1967389652549632.0,
3530
- "train_batch_size": 1,
3531
- "trial_name": null,
3532
- "trial_params": null
3533
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
training_args.bin DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:1bf150d820bfae61b431f78524e1ada6e18847a6a0b58efeab889334baf2b6e5
3
- size 7672
 
 
 
 
zero_to_fp32.py DELETED
@@ -1,760 +0,0 @@
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)