sanchit-gandhi HF staff commited on
Commit
8c71811
1 Parent(s): a53bfaa

End of training

Browse files
all_results.json CHANGED
@@ -1,14 +1,14 @@
1
  {
2
  "epoch": 1.0,
3
- "eval_loss": 4.515435218811035,
4
- "eval_runtime": 931.871,
5
  "eval_samples": 2642,
6
- "eval_samples_per_second": 2.835,
7
- "eval_steps_per_second": 0.237,
8
- "eval_wer": 2.3907179690598968,
9
- "train_loss": 4.478918995520081,
10
- "train_runtime": 5719.3723,
11
  "train_samples": 28538,
12
- "train_samples_per_second": 4.99,
13
- "train_steps_per_second": 0.104
14
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "eval_loss": 4.729526996612549,
4
+ "eval_runtime": 946.0728,
5
  "eval_samples": 2642,
6
+ "eval_samples_per_second": 2.793,
7
+ "eval_steps_per_second": 0.234,
8
+ "eval_wer": 2.3516065053550177,
9
+ "train_loss": 4.303745459225844,
10
+ "train_runtime": 4795.9642,
11
  "train_samples": 28538,
12
+ "train_samples_per_second": 5.95,
13
+ "train_steps_per_second": 0.062
14
  }
eval_results.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
  "epoch": 1.0,
3
- "eval_loss": 4.515435218811035,
4
- "eval_runtime": 931.871,
5
  "eval_samples": 2642,
6
- "eval_samples_per_second": 2.835,
7
- "eval_steps_per_second": 0.237,
8
- "eval_wer": 2.3907179690598968
9
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "eval_loss": 4.729526996612549,
4
+ "eval_runtime": 946.0728,
5
  "eval_samples": 2642,
6
+ "eval_samples_per_second": 2.793,
7
+ "eval_steps_per_second": 0.234,
8
+ "eval_wer": 2.3516065053550177
9
  }
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 1.0,
3
- "train_loss": 4.478918995520081,
4
- "train_runtime": 5719.3723,
5
  "train_samples": 28538,
6
- "train_samples_per_second": 4.99,
7
- "train_steps_per_second": 0.104
8
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "train_loss": 4.303745459225844,
4
+ "train_runtime": 4795.9642,
5
  "train_samples": 28538,
6
+ "train_samples_per_second": 5.95,
7
+ "train_steps_per_second": 0.062
8
  }
trainer_state.json CHANGED
@@ -2,3595 +2,1804 @@
2
  "best_metric": null,
3
  "best_model_checkpoint": null,
4
  "epoch": 0.9987389659520807,
5
- "global_step": 594,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
9
  "log_history": [
10
  {
11
  "epoch": 0.0,
12
- "learning_rate": 0.0,
13
- "loss": 4.8182,
14
  "step": 1
15
  },
16
  {
17
- "epoch": 0.0,
18
- "learning_rate": 6e-07,
19
- "loss": 5.0028,
20
  "step": 2
21
  },
22
  {
23
  "epoch": 0.01,
24
- "learning_rate": 1.2e-06,
25
- "loss": 4.8689,
26
  "step": 3
27
  },
28
  {
29
  "epoch": 0.01,
30
- "learning_rate": 1.8e-06,
31
- "loss": 4.8344,
32
  "step": 4
33
  },
34
  {
35
- "epoch": 0.01,
36
- "learning_rate": 2.4e-06,
37
- "loss": 4.7712,
38
  "step": 5
39
  },
40
  {
41
- "epoch": 0.01,
42
  "learning_rate": 2.9999999999999997e-06,
43
- "loss": 4.7584,
44
  "step": 6
45
  },
46
  {
47
- "epoch": 0.01,
48
  "learning_rate": 3.6e-06,
49
- "loss": 4.8249,
50
  "step": 7
51
  },
52
  {
53
- "epoch": 0.01,
54
  "learning_rate": 4.2e-06,
55
- "loss": 4.6967,
56
  "step": 8
57
  },
58
  {
59
- "epoch": 0.02,
60
  "learning_rate": 4.8e-06,
61
- "loss": 4.7905,
62
  "step": 9
63
  },
64
  {
65
- "epoch": 0.02,
66
  "learning_rate": 5.399999999999999e-06,
67
- "loss": 4.6558,
68
  "step": 10
69
  },
70
  {
71
- "epoch": 0.02,
72
  "learning_rate": 5.999999999999999e-06,
73
- "loss": 4.6367,
74
  "step": 11
75
  },
76
  {
77
- "epoch": 0.02,
78
  "learning_rate": 6.599999999999999e-06,
79
- "loss": 4.646,
80
  "step": 12
81
  },
82
  {
83
- "epoch": 0.02,
84
  "learning_rate": 7.2e-06,
85
- "loss": 4.543,
86
  "step": 13
87
  },
88
  {
89
- "epoch": 0.02,
90
  "learning_rate": 7.799999999999998e-06,
91
- "loss": 4.4923,
92
  "step": 14
93
  },
94
  {
95
- "epoch": 0.03,
96
  "learning_rate": 8.4e-06,
97
- "loss": 4.3786,
98
  "step": 15
99
  },
100
  {
101
- "epoch": 0.03,
102
  "learning_rate": 8.999999999999999e-06,
103
- "loss": 4.5164,
104
  "step": 16
105
  },
106
  {
107
- "epoch": 0.03,
108
  "learning_rate": 9.6e-06,
109
- "loss": 4.5063,
110
  "step": 17
111
  },
112
  {
113
- "epoch": 0.03,
114
  "learning_rate": 1.02e-05,
115
- "loss": 4.5819,
116
  "step": 18
117
  },
118
  {
119
- "epoch": 0.03,
120
  "learning_rate": 1.0799999999999998e-05,
121
- "loss": 4.3233,
122
  "step": 19
123
  },
124
  {
125
- "epoch": 0.03,
126
  "learning_rate": 1.14e-05,
127
- "loss": 4.5417,
128
  "step": 20
129
  },
130
  {
131
- "epoch": 0.04,
132
  "learning_rate": 1.1999999999999999e-05,
133
- "loss": 4.4638,
134
  "step": 21
135
  },
136
  {
137
- "epoch": 0.04,
138
  "learning_rate": 1.26e-05,
139
- "loss": 4.4333,
140
  "step": 22
141
  },
142
  {
143
- "epoch": 0.04,
144
  "learning_rate": 1.3199999999999997e-05,
145
- "loss": 4.4354,
146
  "step": 23
147
  },
148
  {
149
- "epoch": 0.04,
150
  "learning_rate": 1.3799999999999998e-05,
151
- "loss": 4.4696,
152
  "step": 24
153
  },
154
  {
155
- "epoch": 0.04,
156
  "learning_rate": 1.44e-05,
157
- "loss": 4.2463,
158
  "step": 25
159
  },
160
  {
161
- "epoch": 0.04,
162
  "learning_rate": 1.4999999999999999e-05,
163
- "loss": 4.6625,
164
  "step": 26
165
  },
166
  {
167
- "epoch": 0.05,
168
  "learning_rate": 1.5599999999999996e-05,
169
- "loss": 4.3226,
170
  "step": 27
171
  },
172
  {
173
- "epoch": 0.05,
174
  "learning_rate": 1.6199999999999997e-05,
175
- "loss": 4.2108,
176
  "step": 28
177
  },
178
  {
179
- "epoch": 0.05,
180
  "learning_rate": 1.68e-05,
181
- "loss": 4.2346,
182
  "step": 29
183
  },
184
  {
185
- "epoch": 0.05,
186
  "learning_rate": 1.74e-05,
187
- "loss": 4.3936,
188
  "step": 30
189
  },
190
  {
191
- "epoch": 0.05,
192
  "learning_rate": 1.7999999999999997e-05,
193
- "loss": 4.4266,
194
  "step": 31
195
  },
196
  {
197
- "epoch": 0.05,
198
  "learning_rate": 1.8599999999999998e-05,
199
- "loss": 4.2831,
200
  "step": 32
201
  },
202
  {
203
- "epoch": 0.06,
204
  "learning_rate": 1.92e-05,
205
- "loss": 4.2859,
206
  "step": 33
207
  },
208
  {
209
- "epoch": 0.06,
210
  "learning_rate": 1.98e-05,
211
- "loss": 4.2428,
212
  "step": 34
213
  },
214
  {
215
- "epoch": 0.06,
216
  "learning_rate": 2.04e-05,
217
- "loss": 4.295,
218
  "step": 35
219
  },
220
  {
221
- "epoch": 0.06,
222
  "learning_rate": 2.1e-05,
223
- "loss": 4.452,
224
  "step": 36
225
  },
226
  {
227
- "epoch": 0.06,
228
  "learning_rate": 2.1599999999999996e-05,
229
- "loss": 4.2423,
230
  "step": 37
231
  },
232
  {
233
- "epoch": 0.06,
234
  "learning_rate": 2.2199999999999998e-05,
235
- "loss": 4.4074,
236
  "step": 38
237
  },
238
  {
239
- "epoch": 0.07,
240
  "learning_rate": 2.28e-05,
241
- "loss": 4.556,
242
  "step": 39
243
  },
244
  {
245
- "epoch": 0.07,
246
  "learning_rate": 2.34e-05,
247
- "loss": 4.2127,
248
  "step": 40
249
  },
250
  {
251
- "epoch": 0.07,
252
  "learning_rate": 2.3999999999999997e-05,
253
- "loss": 4.2372,
254
  "step": 41
255
  },
256
  {
257
- "epoch": 0.07,
258
  "learning_rate": 2.4599999999999998e-05,
259
- "loss": 4.4704,
260
  "step": 42
261
  },
262
  {
263
- "epoch": 0.07,
264
  "learning_rate": 2.52e-05,
265
- "loss": 4.1838,
266
  "step": 43
267
  },
268
  {
269
- "epoch": 0.07,
270
  "learning_rate": 2.5799999999999997e-05,
271
- "loss": 4.3398,
272
  "step": 44
273
  },
274
  {
275
- "epoch": 0.08,
276
  "learning_rate": 2.6399999999999995e-05,
277
- "loss": 4.4231,
278
  "step": 45
279
  },
280
  {
281
- "epoch": 0.08,
282
  "learning_rate": 2.6999999999999996e-05,
283
- "loss": 4.4815,
284
  "step": 46
285
  },
286
  {
287
- "epoch": 0.08,
288
  "learning_rate": 2.7599999999999997e-05,
289
- "loss": 4.3865,
290
  "step": 47
291
  },
292
  {
293
- "epoch": 0.08,
294
  "learning_rate": 2.8199999999999998e-05,
295
- "loss": 4.7313,
296
  "step": 48
297
  },
298
  {
299
- "epoch": 0.08,
300
- "learning_rate": 2.8199999999999998e-05,
301
- "loss": 4.7684,
302
  "step": 49
303
  },
304
  {
305
- "epoch": 0.08,
306
- "learning_rate": 2.88e-05,
307
- "loss": 4.4833,
308
  "step": 50
309
  },
310
  {
311
- "epoch": 0.09,
312
- "learning_rate": 2.94e-05,
313
- "loss": 4.193,
314
  "step": 51
315
  },
316
  {
317
- "epoch": 0.09,
318
- "learning_rate": 2.9999999999999997e-05,
319
- "loss": 4.2105,
320
  "step": 52
321
  },
322
  {
323
- "epoch": 0.09,
324
- "learning_rate": 3.06e-05,
325
- "loss": 4.2301,
326
  "step": 53
327
  },
328
  {
329
- "epoch": 0.09,
330
- "learning_rate": 3.119999999999999e-05,
331
- "loss": 4.3011,
332
  "step": 54
333
  },
334
  {
335
- "epoch": 0.09,
336
- "learning_rate": 3.1799999999999994e-05,
337
- "loss": 4.1692,
338
  "step": 55
339
  },
340
  {
341
- "epoch": 0.09,
342
- "learning_rate": 3.2399999999999995e-05,
343
- "loss": 4.3675,
344
  "step": 56
345
  },
346
  {
347
- "epoch": 0.1,
348
- "learning_rate": 3.2999999999999996e-05,
349
- "loss": 4.1814,
350
  "step": 57
351
  },
352
  {
353
- "epoch": 0.1,
354
- "learning_rate": 3.36e-05,
355
- "loss": 4.4746,
356
  "step": 58
357
  },
358
  {
359
- "epoch": 0.1,
360
- "learning_rate": 3.42e-05,
361
- "loss": 4.2425,
362
  "step": 59
363
  },
364
  {
365
- "epoch": 0.1,
366
- "learning_rate": 3.48e-05,
367
- "loss": 4.0318,
368
  "step": 60
369
  },
370
  {
371
- "epoch": 0.1,
372
- "learning_rate": 3.539999999999999e-05,
373
- "loss": 4.2939,
374
  "step": 61
375
  },
376
  {
377
- "epoch": 0.1,
378
- "learning_rate": 3.5999999999999994e-05,
379
- "loss": 4.3447,
380
  "step": 62
381
  },
382
  {
383
- "epoch": 0.11,
384
- "learning_rate": 3.6599999999999995e-05,
385
- "loss": 4.2286,
386
  "step": 63
387
  },
388
  {
389
- "epoch": 0.11,
390
- "learning_rate": 3.7199999999999996e-05,
391
- "loss": 4.2042,
392
  "step": 64
393
  },
394
  {
395
- "epoch": 0.11,
396
- "learning_rate": 3.78e-05,
397
- "loss": 4.236,
398
  "step": 65
399
  },
400
  {
401
- "epoch": 0.11,
402
- "learning_rate": 3.84e-05,
403
- "loss": 4.1772,
404
  "step": 66
405
  },
406
  {
407
- "epoch": 0.11,
408
- "learning_rate": 3.9e-05,
409
- "loss": 4.142,
410
  "step": 67
411
  },
412
  {
413
- "epoch": 0.11,
414
- "learning_rate": 3.96e-05,
415
- "loss": 4.1345,
416
  "step": 68
417
  },
418
  {
419
- "epoch": 0.12,
420
- "learning_rate": 4.02e-05,
421
- "loss": 4.2305,
422
  "step": 69
423
  },
424
  {
425
- "epoch": 0.12,
426
- "learning_rate": 4.08e-05,
427
- "loss": 4.2083,
428
  "step": 70
429
  },
430
  {
431
- "epoch": 0.12,
432
- "learning_rate": 4.14e-05,
433
- "loss": 4.3085,
434
  "step": 71
435
  },
436
  {
437
- "epoch": 0.12,
438
- "learning_rate": 4.2e-05,
439
- "loss": 4.2821,
440
  "step": 72
441
  },
442
  {
443
- "epoch": 0.12,
444
- "learning_rate": 4.259999999999999e-05,
445
- "loss": 4.0847,
446
  "step": 73
447
  },
448
  {
449
- "epoch": 0.12,
450
- "learning_rate": 4.319999999999999e-05,
451
- "loss": 4.2869,
452
  "step": 74
453
  },
454
  {
455
- "epoch": 0.13,
456
- "learning_rate": 4.3799999999999994e-05,
457
- "loss": 4.2665,
458
  "step": 75
459
  },
460
  {
461
- "epoch": 0.13,
462
- "learning_rate": 4.4399999999999995e-05,
463
- "loss": 4.24,
464
  "step": 76
465
  },
466
  {
467
- "epoch": 0.13,
468
- "learning_rate": 4.4999999999999996e-05,
469
- "loss": 4.219,
470
  "step": 77
471
  },
472
  {
473
- "epoch": 0.13,
474
- "learning_rate": 4.56e-05,
475
- "loss": 4.1975,
476
  "step": 78
477
  },
478
  {
479
- "epoch": 0.13,
480
- "learning_rate": 4.62e-05,
481
- "loss": 4.1354,
482
  "step": 79
483
  },
484
  {
485
- "epoch": 0.13,
486
- "learning_rate": 4.68e-05,
487
- "loss": 4.3206,
488
  "step": 80
489
  },
490
  {
491
- "epoch": 0.14,
492
- "learning_rate": 4.7399999999999993e-05,
493
- "loss": 4.2772,
494
  "step": 81
495
  },
496
  {
497
- "epoch": 0.14,
498
- "learning_rate": 4.7999999999999994e-05,
499
- "loss": 4.2269,
500
  "step": 82
501
  },
502
  {
503
- "epoch": 0.14,
504
- "learning_rate": 4.8599999999999995e-05,
505
- "loss": 4.3281,
506
  "step": 83
507
  },
508
  {
509
- "epoch": 0.14,
510
- "learning_rate": 4.9199999999999997e-05,
511
- "loss": 4.291,
512
  "step": 84
513
  },
514
  {
515
- "epoch": 0.14,
516
- "learning_rate": 4.98e-05,
517
- "loss": 4.2093,
518
  "step": 85
519
  },
520
  {
521
- "epoch": 0.14,
522
- "learning_rate": 5.04e-05,
523
- "loss": 4.1592,
524
  "step": 86
525
  },
526
  {
527
- "epoch": 0.15,
528
- "learning_rate": 5.1e-05,
529
- "loss": 4.4026,
530
  "step": 87
531
  },
532
  {
533
- "epoch": 0.15,
534
- "learning_rate": 5.1599999999999994e-05,
535
- "loss": 4.2414,
536
  "step": 88
537
  },
538
  {
539
- "epoch": 0.15,
540
- "learning_rate": 5.2199999999999995e-05,
541
- "loss": 4.245,
542
  "step": 89
543
  },
544
  {
545
- "epoch": 0.15,
546
- "learning_rate": 5.279999999999999e-05,
547
- "loss": 4.2845,
548
  "step": 90
549
  },
550
  {
551
- "epoch": 0.15,
552
- "learning_rate": 5.339999999999999e-05,
553
- "loss": 4.247,
554
  "step": 91
555
  },
556
  {
557
- "epoch": 0.15,
558
- "learning_rate": 5.399999999999999e-05,
559
- "loss": 4.3707,
560
  "step": 92
561
  },
562
  {
563
- "epoch": 0.16,
564
- "learning_rate": 5.459999999999999e-05,
565
- "loss": 4.2318,
566
  "step": 93
567
  },
568
  {
569
- "epoch": 0.16,
570
- "learning_rate": 5.519999999999999e-05,
571
- "loss": 4.4576,
572
  "step": 94
573
  },
574
  {
575
- "epoch": 0.16,
576
- "learning_rate": 5.5799999999999994e-05,
577
- "loss": 4.3531,
578
  "step": 95
579
  },
580
  {
581
- "epoch": 0.16,
582
- "learning_rate": 5.6399999999999995e-05,
583
- "loss": 4.409,
584
  "step": 96
585
  },
586
  {
587
- "epoch": 0.16,
588
- "learning_rate": 5.6999999999999996e-05,
589
- "loss": 4.3879,
590
  "step": 97
591
  },
592
  {
593
- "epoch": 0.16,
594
- "learning_rate": 5.76e-05,
595
- "loss": 4.309,
596
  "step": 98
597
  },
598
  {
599
- "epoch": 0.17,
600
- "learning_rate": 5.82e-05,
601
- "loss": 4.5432,
602
  "step": 99
603
  },
604
  {
605
- "epoch": 0.17,
606
- "learning_rate": 5.88e-05,
607
- "loss": 4.205,
608
  "step": 100
609
  },
610
  {
611
- "epoch": 0.17,
612
- "learning_rate": 5.94e-05,
613
- "loss": 4.3588,
614
  "step": 101
615
  },
616
  {
617
- "epoch": 0.17,
618
- "learning_rate": 5.9999999999999995e-05,
619
- "loss": 4.1904,
620
  "step": 102
621
  },
622
  {
623
- "epoch": 0.17,
624
- "learning_rate": 6.0599999999999996e-05,
625
- "loss": 4.2077,
626
  "step": 103
627
  },
628
  {
629
- "epoch": 0.17,
630
- "learning_rate": 6.12e-05,
631
- "loss": 4.13,
632
  "step": 104
633
  },
634
  {
635
- "epoch": 0.18,
636
- "learning_rate": 6.18e-05,
637
- "loss": 4.1796,
638
  "step": 105
639
  },
640
  {
641
- "epoch": 0.18,
642
- "learning_rate": 6.239999999999999e-05,
643
- "loss": 4.1452,
644
  "step": 106
645
  },
646
  {
647
- "epoch": 0.18,
648
- "learning_rate": 6.299999999999999e-05,
649
- "loss": 4.2699,
650
  "step": 107
651
  },
652
  {
653
- "epoch": 0.18,
654
- "learning_rate": 6.359999999999999e-05,
655
- "loss": 4.3016,
656
  "step": 108
657
  },
658
  {
659
- "epoch": 0.18,
660
- "learning_rate": 6.419999999999999e-05,
661
- "loss": 4.4238,
662
  "step": 109
663
  },
664
  {
665
- "epoch": 0.18,
666
- "learning_rate": 6.479999999999999e-05,
667
- "loss": 4.2019,
668
  "step": 110
669
  },
670
  {
671
- "epoch": 0.19,
672
- "learning_rate": 6.539999999999999e-05,
673
- "loss": 4.2325,
674
  "step": 111
675
  },
676
  {
677
- "epoch": 0.19,
678
- "learning_rate": 6.599999999999999e-05,
679
- "loss": 4.2623,
680
  "step": 112
681
  },
682
  {
683
- "epoch": 0.19,
684
- "learning_rate": 6.659999999999999e-05,
685
- "loss": 4.2221,
686
  "step": 113
687
  },
688
  {
689
- "epoch": 0.19,
690
- "learning_rate": 6.72e-05,
691
- "loss": 4.2787,
692
  "step": 114
693
  },
694
  {
695
- "epoch": 0.19,
696
- "learning_rate": 6.78e-05,
697
- "loss": 4.2717,
698
  "step": 115
699
  },
700
  {
701
- "epoch": 0.2,
702
- "learning_rate": 6.84e-05,
703
- "loss": 4.2513,
704
  "step": 116
705
  },
706
  {
707
- "epoch": 0.2,
708
- "learning_rate": 6.9e-05,
709
- "loss": 4.3379,
710
  "step": 117
711
  },
712
  {
713
- "epoch": 0.2,
714
- "learning_rate": 6.96e-05,
715
- "loss": 4.1303,
716
  "step": 118
717
  },
718
  {
719
- "epoch": 0.2,
720
- "learning_rate": 7.02e-05,
721
- "loss": 4.1263,
722
  "step": 119
723
  },
724
  {
725
- "epoch": 0.2,
726
- "learning_rate": 7.079999999999999e-05,
727
- "loss": 4.1342,
728
  "step": 120
729
  },
730
  {
731
- "epoch": 0.2,
732
- "learning_rate": 7.139999999999999e-05,
733
- "loss": 4.3001,
734
  "step": 121
735
  },
736
  {
737
- "epoch": 0.21,
738
- "learning_rate": 7.199999999999999e-05,
739
- "loss": 4.4657,
740
  "step": 122
741
  },
742
  {
743
- "epoch": 0.21,
744
- "learning_rate": 7.259999999999999e-05,
745
- "loss": 4.194,
746
  "step": 123
747
  },
748
  {
749
- "epoch": 0.21,
750
- "learning_rate": 7.319999999999999e-05,
751
- "loss": 4.2687,
752
  "step": 124
753
  },
754
  {
755
- "epoch": 0.21,
756
- "learning_rate": 7.379999999999999e-05,
757
- "loss": 4.2877,
758
  "step": 125
759
  },
760
  {
761
- "epoch": 0.21,
762
- "learning_rate": 7.439999999999999e-05,
763
- "loss": 4.1317,
764
  "step": 126
765
  },
766
  {
767
- "epoch": 0.21,
768
- "learning_rate": 7.5e-05,
769
- "loss": 4.145,
770
  "step": 127
771
  },
772
  {
773
- "epoch": 0.22,
774
- "learning_rate": 7.56e-05,
775
- "loss": 4.1317,
776
  "step": 128
777
  },
778
  {
779
- "epoch": 0.22,
780
- "learning_rate": 7.62e-05,
781
- "loss": 4.2327,
782
  "step": 129
783
  },
784
  {
785
- "epoch": 0.22,
786
- "learning_rate": 7.68e-05,
787
- "loss": 4.2944,
788
  "step": 130
789
  },
790
  {
791
- "epoch": 0.22,
792
- "learning_rate": 7.74e-05,
793
- "loss": 4.254,
794
  "step": 131
795
  },
796
  {
797
- "epoch": 0.22,
798
- "learning_rate": 7.8e-05,
799
- "loss": 4.2138,
800
  "step": 132
801
  },
802
  {
803
- "epoch": 0.22,
804
- "learning_rate": 7.86e-05,
805
- "loss": 4.2627,
806
  "step": 133
807
  },
808
  {
809
- "epoch": 0.23,
810
- "learning_rate": 7.92e-05,
811
- "loss": 4.3398,
812
  "step": 134
813
  },
814
  {
815
- "epoch": 0.23,
816
- "learning_rate": 7.98e-05,
817
- "loss": 4.3485,
818
  "step": 135
819
  },
820
  {
821
- "epoch": 0.23,
822
- "learning_rate": 8.04e-05,
823
- "loss": 4.2115,
824
  "step": 136
825
  },
826
  {
827
- "epoch": 0.23,
828
- "learning_rate": 8.1e-05,
829
- "loss": 4.1802,
830
  "step": 137
831
  },
832
  {
833
- "epoch": 0.23,
834
- "learning_rate": 8.16e-05,
835
- "loss": 4.24,
836
  "step": 138
837
  },
838
  {
839
- "epoch": 0.23,
840
- "learning_rate": 8.22e-05,
841
- "loss": 4.2233,
842
  "step": 139
843
  },
844
  {
845
- "epoch": 0.24,
846
- "learning_rate": 8.28e-05,
847
- "loss": 4.3909,
848
  "step": 140
849
  },
850
  {
851
- "epoch": 0.24,
852
- "learning_rate": 8.34e-05,
853
- "loss": 4.2786,
854
  "step": 141
855
  },
856
  {
857
- "epoch": 0.24,
858
- "learning_rate": 8.4e-05,
859
- "loss": 4.3874,
860
  "step": 142
861
  },
862
  {
863
- "epoch": 0.24,
864
- "learning_rate": 8.459999999999998e-05,
865
- "loss": 4.4902,
866
  "step": 143
867
  },
868
  {
869
- "epoch": 0.24,
870
- "learning_rate": 8.519999999999998e-05,
871
- "loss": 4.3232,
872
  "step": 144
873
  },
874
  {
875
- "epoch": 0.24,
876
- "learning_rate": 8.579999999999998e-05,
877
- "loss": 4.3679,
878
  "step": 145
879
  },
880
  {
881
- "epoch": 0.25,
882
- "learning_rate": 8.639999999999999e-05,
883
- "loss": 4.6866,
884
  "step": 146
885
  },
886
  {
887
- "epoch": 0.25,
888
- "learning_rate": 8.699999999999999e-05,
889
- "loss": 4.6394,
890
  "step": 147
891
  },
892
  {
893
- "epoch": 0.25,
894
- "learning_rate": 8.759999999999999e-05,
895
- "loss": 4.5798,
896
  "step": 148
897
  },
898
  {
899
- "epoch": 0.25,
900
- "learning_rate": 8.819999999999999e-05,
901
- "loss": 4.6793,
902
  "step": 149
903
  },
904
  {
905
- "epoch": 0.25,
906
- "learning_rate": 8.879999999999999e-05,
907
- "loss": 4.5903,
908
  "step": 150
909
  },
910
  {
911
- "epoch": 0.25,
912
- "learning_rate": 8.939999999999999e-05,
913
- "loss": 4.2962,
914
  "step": 151
915
  },
916
  {
917
- "epoch": 0.26,
918
- "learning_rate": 8.999999999999999e-05,
919
- "loss": 4.2278,
920
  "step": 152
921
  },
922
  {
923
- "epoch": 0.26,
924
- "learning_rate": 9.059999999999999e-05,
925
- "loss": 4.2305,
926
  "step": 153
927
  },
928
  {
929
- "epoch": 0.26,
930
- "learning_rate": 9.12e-05,
931
- "loss": 4.145,
932
  "step": 154
933
  },
934
  {
935
- "epoch": 0.26,
936
- "learning_rate": 9.18e-05,
937
- "loss": 4.1925,
938
  "step": 155
939
  },
940
  {
941
- "epoch": 0.26,
942
- "learning_rate": 9.24e-05,
943
- "loss": 4.1712,
944
  "step": 156
945
  },
946
  {
947
- "epoch": 0.26,
948
- "learning_rate": 9.3e-05,
949
- "loss": 4.301,
950
  "step": 157
951
  },
952
  {
953
- "epoch": 0.27,
954
- "learning_rate": 9.36e-05,
955
- "loss": 4.1469,
956
  "step": 158
957
  },
958
  {
959
- "epoch": 0.27,
960
- "learning_rate": 9.419999999999999e-05,
961
- "loss": 4.0775,
962
  "step": 159
963
  },
964
  {
965
- "epoch": 0.27,
966
- "learning_rate": 9.479999999999999e-05,
967
- "loss": 4.2078,
968
  "step": 160
969
  },
970
  {
971
- "epoch": 0.27,
972
- "learning_rate": 9.539999999999999e-05,
973
- "loss": 4.2729,
974
  "step": 161
975
  },
976
  {
977
- "epoch": 0.27,
978
- "learning_rate": 9.599999999999999e-05,
979
- "loss": 4.1514,
980
  "step": 162
981
  },
982
  {
983
- "epoch": 0.27,
984
- "learning_rate": 9.659999999999999e-05,
985
- "loss": 4.1282,
986
  "step": 163
987
  },
988
  {
989
- "epoch": 0.28,
990
- "learning_rate": 9.719999999999999e-05,
991
- "loss": 4.1626,
992
  "step": 164
993
  },
994
  {
995
- "epoch": 0.28,
996
- "learning_rate": 9.779999999999999e-05,
997
- "loss": 4.3451,
998
  "step": 165
999
  },
1000
  {
1001
- "epoch": 0.28,
1002
- "learning_rate": 9.839999999999999e-05,
1003
- "loss": 4.3345,
1004
  "step": 166
1005
  },
1006
  {
1007
- "epoch": 0.28,
1008
- "learning_rate": 9.9e-05,
1009
- "loss": 4.1537,
1010
  "step": 167
1011
  },
1012
  {
1013
- "epoch": 0.28,
1014
- "learning_rate": 9.96e-05,
1015
- "loss": 4.3872,
1016
  "step": 168
1017
  },
1018
  {
1019
- "epoch": 0.28,
1020
- "learning_rate": 0.0001002,
1021
- "loss": 4.2078,
1022
  "step": 169
1023
  },
1024
  {
1025
- "epoch": 0.29,
1026
- "learning_rate": 0.0001008,
1027
- "loss": 4.2435,
1028
  "step": 170
1029
  },
1030
  {
1031
- "epoch": 0.29,
1032
- "learning_rate": 0.0001014,
1033
- "loss": 4.1005,
1034
  "step": 171
1035
  },
1036
  {
1037
- "epoch": 0.29,
1038
- "learning_rate": 0.000102,
1039
- "loss": 4.1444,
1040
  "step": 172
1041
  },
1042
  {
1043
- "epoch": 0.29,
1044
- "learning_rate": 0.0001026,
1045
- "loss": 4.1539,
1046
  "step": 173
1047
  },
1048
  {
1049
- "epoch": 0.29,
1050
- "learning_rate": 0.00010319999999999999,
1051
- "loss": 4.2127,
1052
  "step": 174
1053
  },
1054
  {
1055
- "epoch": 0.29,
1056
- "learning_rate": 0.00010379999999999999,
1057
- "loss": 4.3975,
1058
  "step": 175
1059
  },
1060
  {
1061
- "epoch": 0.3,
1062
- "learning_rate": 0.00010439999999999999,
1063
- "loss": 4.3712,
1064
  "step": 176
1065
  },
1066
  {
1067
- "epoch": 0.3,
1068
- "learning_rate": 0.00010499999999999999,
1069
- "loss": 4.3348,
1070
  "step": 177
1071
  },
1072
  {
1073
- "epoch": 0.3,
1074
- "learning_rate": 0.00010559999999999998,
1075
- "loss": 4.1479,
1076
  "step": 178
1077
  },
1078
  {
1079
- "epoch": 0.3,
1080
- "learning_rate": 0.00010619999999999998,
1081
- "loss": 4.2947,
1082
  "step": 179
1083
  },
1084
  {
1085
- "epoch": 0.3,
1086
- "learning_rate": 0.00010679999999999998,
1087
- "loss": 4.1779,
1088
  "step": 180
1089
  },
1090
  {
1091
- "epoch": 0.3,
1092
- "learning_rate": 0.00010739999999999998,
1093
- "loss": 4.3328,
1094
  "step": 181
1095
  },
1096
  {
1097
- "epoch": 0.31,
1098
- "learning_rate": 0.00010799999999999998,
1099
- "loss": 4.2,
1100
  "step": 182
1101
  },
1102
  {
1103
- "epoch": 0.31,
1104
- "learning_rate": 0.00010859999999999998,
1105
- "loss": 4.2252,
1106
  "step": 183
1107
  },
1108
  {
1109
- "epoch": 0.31,
1110
- "learning_rate": 0.00010919999999999998,
1111
- "loss": 4.1258,
1112
  "step": 184
1113
  },
1114
  {
1115
- "epoch": 0.31,
1116
- "learning_rate": 0.00010979999999999999,
1117
- "loss": 4.4858,
1118
  "step": 185
1119
  },
1120
  {
1121
- "epoch": 0.31,
1122
- "learning_rate": 0.00011039999999999999,
1123
- "loss": 4.2418,
1124
  "step": 186
1125
  },
1126
  {
1127
- "epoch": 0.31,
1128
- "learning_rate": 0.00011099999999999999,
1129
- "loss": 4.2566,
1130
  "step": 187
1131
  },
1132
  {
1133
- "epoch": 0.32,
1134
- "learning_rate": 0.00011159999999999999,
1135
- "loss": 4.3974,
1136
  "step": 188
1137
  },
1138
  {
1139
- "epoch": 0.32,
1140
- "learning_rate": 0.00011219999999999999,
1141
- "loss": 4.4813,
1142
  "step": 189
1143
  },
1144
  {
1145
- "epoch": 0.32,
1146
- "learning_rate": 0.00011279999999999999,
1147
- "loss": 4.1671,
1148
  "step": 190
1149
  },
1150
  {
1151
- "epoch": 0.32,
1152
- "learning_rate": 0.00011339999999999999,
1153
- "loss": 4.4851,
1154
  "step": 191
1155
  },
1156
  {
1157
- "epoch": 0.32,
1158
- "learning_rate": 0.00011399999999999999,
1159
- "loss": 4.3957,
1160
  "step": 192
1161
  },
1162
  {
1163
- "epoch": 0.32,
1164
- "learning_rate": 0.0001146,
1165
- "loss": 4.6878,
1166
  "step": 193
1167
  },
1168
  {
1169
- "epoch": 0.33,
1170
- "learning_rate": 0.0001152,
1171
- "loss": 4.3941,
1172
  "step": 194
1173
  },
1174
  {
1175
- "epoch": 0.33,
1176
- "learning_rate": 0.0001158,
1177
- "loss": 4.3453,
1178
  "step": 195
1179
  },
1180
  {
1181
- "epoch": 0.33,
1182
- "learning_rate": 0.0001164,
1183
- "loss": 4.4555,
1184
  "step": 196
1185
  },
1186
  {
1187
- "epoch": 0.33,
1188
- "learning_rate": 0.000117,
1189
- "loss": 4.7167,
1190
  "step": 197
1191
  },
1192
  {
1193
- "epoch": 0.33,
1194
- "learning_rate": 0.0001176,
1195
- "loss": 4.538,
1196
  "step": 198
1197
  },
1198
  {
1199
- "epoch": 0.33,
1200
- "learning_rate": 0.0001182,
1201
- "loss": 4.5329,
1202
  "step": 199
1203
  },
1204
  {
1205
- "epoch": 0.34,
1206
- "learning_rate": 0.0001188,
1207
- "loss": 4.5335,
1208
  "step": 200
1209
  },
1210
  {
1211
- "epoch": 0.34,
1212
- "learning_rate": 0.0001194,
1213
- "loss": 4.1187,
1214
  "step": 201
1215
  },
1216
  {
1217
- "epoch": 0.34,
1218
- "learning_rate": 0.00011999999999999999,
1219
- "loss": 4.2866,
1220
  "step": 202
1221
  },
1222
  {
1223
- "epoch": 0.34,
1224
- "learning_rate": 0.00012059999999999999,
1225
- "loss": 4.3993,
1226
  "step": 203
1227
  },
1228
  {
1229
- "epoch": 0.34,
1230
- "learning_rate": 0.00012119999999999999,
1231
- "loss": 4.1615,
1232
  "step": 204
1233
  },
1234
  {
1235
- "epoch": 0.34,
1236
- "learning_rate": 0.00012179999999999999,
1237
- "loss": 4.226,
1238
  "step": 205
1239
  },
1240
  {
1241
- "epoch": 0.35,
1242
- "learning_rate": 0.0001224,
1243
- "loss": 4.1952,
1244
  "step": 206
1245
  },
1246
  {
1247
- "epoch": 0.35,
1248
- "learning_rate": 0.00012299999999999998,
1249
- "loss": 4.2312,
1250
  "step": 207
1251
  },
1252
  {
1253
- "epoch": 0.35,
1254
- "learning_rate": 0.0001236,
1255
- "loss": 4.3538,
1256
  "step": 208
1257
  },
1258
  {
1259
- "epoch": 0.35,
1260
- "learning_rate": 0.00012419999999999998,
1261
- "loss": 4.265,
1262
  "step": 209
1263
  },
1264
  {
1265
- "epoch": 0.35,
1266
- "learning_rate": 0.00012479999999999997,
1267
- "loss": 4.3468,
1268
  "step": 210
1269
  },
1270
  {
1271
- "epoch": 0.35,
1272
- "learning_rate": 0.00012539999999999999,
1273
- "loss": 4.3613,
1274
  "step": 211
1275
  },
1276
  {
1277
- "epoch": 0.36,
1278
- "learning_rate": 0.00012599999999999997,
1279
- "loss": 4.2659,
1280
  "step": 212
1281
  },
1282
  {
1283
- "epoch": 0.36,
1284
- "learning_rate": 0.0001266,
1285
- "loss": 4.2808,
1286
  "step": 213
1287
  },
1288
  {
1289
- "epoch": 0.36,
1290
- "learning_rate": 0.00012719999999999997,
1291
- "loss": 4.3405,
1292
  "step": 214
1293
  },
1294
  {
1295
- "epoch": 0.36,
1296
- "learning_rate": 0.0001278,
1297
- "loss": 4.2305,
1298
  "step": 215
1299
  },
1300
  {
1301
- "epoch": 0.36,
1302
- "learning_rate": 0.00012839999999999998,
1303
- "loss": 4.277,
1304
  "step": 216
1305
  },
1306
  {
1307
- "epoch": 0.36,
1308
- "learning_rate": 0.000129,
1309
- "loss": 4.1606,
1310
  "step": 217
1311
  },
1312
  {
1313
- "epoch": 0.37,
1314
- "learning_rate": 0.00012959999999999998,
1315
- "loss": 4.3264,
1316
  "step": 218
1317
  },
1318
  {
1319
- "epoch": 0.37,
1320
- "learning_rate": 0.0001302,
1321
- "loss": 4.4251,
1322
  "step": 219
1323
  },
1324
  {
1325
- "epoch": 0.37,
1326
- "learning_rate": 0.00013079999999999998,
1327
- "loss": 4.4401,
1328
  "step": 220
1329
  },
1330
  {
1331
- "epoch": 0.37,
1332
- "learning_rate": 0.0001314,
1333
- "loss": 4.2936,
1334
  "step": 221
1335
  },
1336
  {
1337
- "epoch": 0.37,
1338
- "learning_rate": 0.00013199999999999998,
1339
- "loss": 4.3649,
1340
  "step": 222
1341
  },
1342
  {
1343
- "epoch": 0.37,
1344
- "learning_rate": 0.0001326,
1345
- "loss": 4.3694,
1346
  "step": 223
1347
  },
1348
  {
1349
- "epoch": 0.38,
1350
- "learning_rate": 0.00013319999999999999,
1351
- "loss": 4.3328,
1352
  "step": 224
1353
  },
1354
  {
1355
- "epoch": 0.38,
1356
- "learning_rate": 0.0001338,
1357
- "loss": 4.5794,
1358
  "step": 225
1359
  },
1360
  {
1361
- "epoch": 0.38,
1362
- "learning_rate": 0.0001344,
1363
- "loss": 4.2256,
1364
  "step": 226
1365
  },
1366
  {
1367
- "epoch": 0.38,
1368
- "learning_rate": 0.000135,
1369
- "loss": 4.1422,
1370
  "step": 227
1371
  },
1372
  {
1373
- "epoch": 0.38,
1374
- "learning_rate": 0.0001356,
1375
- "loss": 4.2117,
1376
  "step": 228
1377
  },
1378
  {
1379
- "epoch": 0.39,
1380
- "learning_rate": 0.0001362,
1381
- "loss": 4.4223,
1382
  "step": 229
1383
  },
1384
  {
1385
- "epoch": 0.39,
1386
- "learning_rate": 0.0001368,
1387
- "loss": 4.384,
1388
  "step": 230
1389
  },
1390
  {
1391
- "epoch": 0.39,
1392
- "learning_rate": 0.0001374,
1393
- "loss": 4.1981,
1394
  "step": 231
1395
  },
1396
  {
1397
- "epoch": 0.39,
1398
- "learning_rate": 0.000138,
1399
- "loss": 4.2797,
1400
  "step": 232
1401
  },
1402
  {
1403
- "epoch": 0.39,
1404
- "learning_rate": 0.0001386,
1405
- "loss": 4.5837,
1406
  "step": 233
1407
  },
1408
  {
1409
- "epoch": 0.39,
1410
- "learning_rate": 0.0001392,
1411
- "loss": 4.2845,
1412
  "step": 234
1413
  },
1414
  {
1415
- "epoch": 0.4,
1416
- "learning_rate": 0.00013979999999999998,
1417
- "loss": 4.4128,
1418
  "step": 235
1419
  },
1420
  {
1421
- "epoch": 0.4,
1422
- "learning_rate": 0.0001404,
1423
- "loss": 4.1982,
1424
  "step": 236
1425
  },
1426
  {
1427
- "epoch": 0.4,
1428
- "learning_rate": 0.00014099999999999998,
1429
- "loss": 4.158,
1430
  "step": 237
1431
  },
1432
  {
1433
- "epoch": 0.4,
1434
- "learning_rate": 0.00014159999999999997,
1435
- "loss": 4.2719,
1436
  "step": 238
1437
  },
1438
  {
1439
- "epoch": 0.4,
1440
- "learning_rate": 0.0001422,
1441
- "loss": 4.3127,
1442
  "step": 239
1443
  },
1444
  {
1445
- "epoch": 0.4,
1446
- "learning_rate": 0.00014279999999999997,
1447
- "loss": 4.3545,
1448
  "step": 240
1449
  },
1450
  {
1451
- "epoch": 0.41,
1452
- "learning_rate": 0.0001434,
1453
- "loss": 4.5494,
1454
  "step": 241
1455
  },
1456
  {
1457
- "epoch": 0.41,
1458
- "learning_rate": 0.00014399999999999998,
1459
- "loss": 4.4465,
1460
  "step": 242
1461
  },
1462
  {
1463
- "epoch": 0.41,
1464
- "learning_rate": 0.0001446,
1465
- "loss": 4.2999,
1466
  "step": 243
1467
  },
1468
  {
1469
- "epoch": 0.41,
1470
- "learning_rate": 0.00014519999999999998,
1471
- "loss": 4.4911,
1472
  "step": 244
1473
  },
1474
  {
1475
- "epoch": 0.41,
1476
- "learning_rate": 0.0001458,
1477
- "loss": 4.7491,
1478
  "step": 245
1479
  },
1480
  {
1481
- "epoch": 0.41,
1482
- "learning_rate": 0.00014639999999999998,
1483
- "loss": 4.5465,
1484
  "step": 246
1485
  },
1486
  {
1487
- "epoch": 0.42,
1488
- "learning_rate": 0.000147,
1489
- "loss": 4.9226,
1490
  "step": 247
1491
  },
1492
  {
1493
- "epoch": 0.42,
1494
- "learning_rate": 0.00014759999999999998,
1495
- "loss": 4.895,
1496
  "step": 248
1497
  },
1498
  {
1499
- "epoch": 0.42,
1500
- "learning_rate": 0.0001482,
1501
- "loss": 4.4854,
1502
  "step": 249
1503
  },
1504
  {
1505
- "epoch": 0.42,
1506
- "learning_rate": 0.00014879999999999998,
1507
- "loss": 4.6152,
1508
  "step": 250
1509
  },
1510
  {
1511
- "epoch": 0.42,
1512
- "learning_rate": 0.0001494,
1513
- "loss": 4.4272,
1514
  "step": 251
1515
  },
1516
  {
1517
- "epoch": 0.42,
1518
- "learning_rate": 0.00015,
1519
- "loss": 4.582,
1520
  "step": 252
1521
  },
1522
  {
1523
- "epoch": 0.43,
1524
- "learning_rate": 0.00015059999999999997,
1525
- "loss": 4.7457,
1526
  "step": 253
1527
  },
1528
  {
1529
- "epoch": 0.43,
1530
- "learning_rate": 0.0001512,
1531
- "loss": 4.4816,
1532
  "step": 254
1533
  },
1534
  {
1535
- "epoch": 0.43,
1536
- "learning_rate": 0.00015179999999999998,
1537
- "loss": 4.8845,
1538
  "step": 255
1539
  },
1540
  {
1541
- "epoch": 0.43,
1542
- "learning_rate": 0.0001524,
1543
- "loss": 4.8351,
1544
  "step": 256
1545
  },
1546
  {
1547
- "epoch": 0.43,
1548
- "learning_rate": 0.00015299999999999998,
1549
- "loss": 4.5857,
1550
  "step": 257
1551
  },
1552
  {
1553
- "epoch": 0.43,
1554
- "learning_rate": 0.0001536,
1555
- "loss": 4.389,
1556
  "step": 258
1557
  },
1558
  {
1559
- "epoch": 0.44,
1560
- "learning_rate": 0.00015419999999999998,
1561
- "loss": 4.4282,
1562
  "step": 259
1563
  },
1564
  {
1565
- "epoch": 0.44,
1566
- "learning_rate": 0.0001548,
1567
- "loss": 4.647,
1568
  "step": 260
1569
  },
1570
  {
1571
- "epoch": 0.44,
1572
- "learning_rate": 0.00015539999999999998,
1573
- "loss": 4.6697,
1574
  "step": 261
1575
  },
1576
  {
1577
- "epoch": 0.44,
1578
- "learning_rate": 0.000156,
1579
- "loss": 4.5029,
1580
  "step": 262
1581
  },
1582
  {
1583
- "epoch": 0.44,
1584
- "learning_rate": 0.00015659999999999998,
1585
- "loss": 4.8514,
1586
  "step": 263
1587
  },
1588
  {
1589
- "epoch": 0.44,
1590
- "learning_rate": 0.0001572,
1591
- "loss": 4.1349,
1592
  "step": 264
1593
  },
1594
  {
1595
- "epoch": 0.45,
1596
- "learning_rate": 0.0001578,
1597
- "loss": 4.3635,
1598
  "step": 265
1599
  },
1600
  {
1601
- "epoch": 0.45,
1602
- "learning_rate": 0.0001584,
1603
- "loss": 4.5442,
1604
  "step": 266
1605
  },
1606
  {
1607
- "epoch": 0.45,
1608
- "learning_rate": 0.000159,
1609
- "loss": 5.1162,
1610
  "step": 267
1611
  },
1612
  {
1613
- "epoch": 0.45,
1614
- "learning_rate": 0.0001596,
1615
- "loss": 4.4451,
1616
  "step": 268
1617
  },
1618
  {
1619
- "epoch": 0.45,
1620
- "learning_rate": 0.0001602,
1621
- "loss": 4.5503,
1622
  "step": 269
1623
  },
1624
  {
1625
- "epoch": 0.45,
1626
- "learning_rate": 0.0001608,
1627
- "loss": 4.4475,
1628
  "step": 270
1629
  },
1630
  {
1631
- "epoch": 0.46,
1632
- "learning_rate": 0.0001614,
1633
- "loss": 4.3478,
1634
  "step": 271
1635
  },
1636
  {
1637
- "epoch": 0.46,
1638
- "learning_rate": 0.000162,
1639
- "loss": 4.6113,
1640
  "step": 272
1641
  },
1642
  {
1643
- "epoch": 0.46,
1644
- "learning_rate": 0.0001626,
1645
- "loss": 4.7242,
1646
  "step": 273
1647
  },
1648
- {
1649
- "epoch": 0.46,
1650
- "learning_rate": 0.0001632,
1651
- "loss": 4.6371,
1652
- "step": 274
1653
- },
1654
- {
1655
- "epoch": 0.46,
1656
- "learning_rate": 0.0001638,
1657
- "loss": 4.4794,
1658
- "step": 275
1659
- },
1660
- {
1661
- "epoch": 0.46,
1662
- "learning_rate": 0.0001644,
1663
- "loss": 4.3902,
1664
- "step": 276
1665
- },
1666
- {
1667
- "epoch": 0.47,
1668
- "learning_rate": 0.000165,
1669
- "loss": 4.5252,
1670
- "step": 277
1671
- },
1672
- {
1673
- "epoch": 0.47,
1674
- "learning_rate": 0.0001656,
1675
- "loss": 4.5436,
1676
- "step": 278
1677
- },
1678
- {
1679
- "epoch": 0.47,
1680
- "learning_rate": 0.0001662,
1681
- "loss": 4.4553,
1682
- "step": 279
1683
- },
1684
- {
1685
- "epoch": 0.47,
1686
- "learning_rate": 0.0001668,
1687
- "loss": 4.5571,
1688
- "step": 280
1689
- },
1690
- {
1691
- "epoch": 0.47,
1692
- "learning_rate": 0.0001674,
1693
- "loss": 4.5108,
1694
- "step": 281
1695
- },
1696
- {
1697
- "epoch": 0.47,
1698
- "learning_rate": 0.000168,
1699
- "loss": 4.5576,
1700
- "step": 282
1701
- },
1702
- {
1703
- "epoch": 0.48,
1704
- "learning_rate": 0.0001686,
1705
- "loss": 4.4762,
1706
- "step": 283
1707
- },
1708
- {
1709
- "epoch": 0.48,
1710
- "learning_rate": 0.00016919999999999997,
1711
- "loss": 4.2964,
1712
- "step": 284
1713
- },
1714
- {
1715
- "epoch": 0.48,
1716
- "learning_rate": 0.00016979999999999998,
1717
- "loss": 4.5742,
1718
- "step": 285
1719
- },
1720
- {
1721
- "epoch": 0.48,
1722
- "learning_rate": 0.00017039999999999997,
1723
- "loss": 4.2314,
1724
- "step": 286
1725
- },
1726
- {
1727
- "epoch": 0.48,
1728
- "learning_rate": 0.00017099999999999998,
1729
- "loss": 4.7449,
1730
- "step": 287
1731
- },
1732
- {
1733
- "epoch": 0.48,
1734
- "learning_rate": 0.00017159999999999997,
1735
- "loss": 4.3179,
1736
- "step": 288
1737
- },
1738
- {
1739
- "epoch": 0.49,
1740
- "learning_rate": 0.00017219999999999998,
1741
- "loss": 4.3885,
1742
- "step": 289
1743
- },
1744
- {
1745
- "epoch": 0.49,
1746
- "learning_rate": 0.00017279999999999997,
1747
- "loss": 4.6569,
1748
- "step": 290
1749
- },
1750
- {
1751
- "epoch": 0.49,
1752
- "learning_rate": 0.00017339999999999996,
1753
- "loss": 4.5262,
1754
- "step": 291
1755
- },
1756
- {
1757
- "epoch": 0.49,
1758
- "learning_rate": 0.00017399999999999997,
1759
- "loss": 4.3064,
1760
- "step": 292
1761
- },
1762
- {
1763
- "epoch": 0.49,
1764
- "learning_rate": 0.00017459999999999996,
1765
- "loss": 4.6379,
1766
- "step": 293
1767
- },
1768
- {
1769
- "epoch": 0.49,
1770
- "learning_rate": 0.00017519999999999998,
1771
- "loss": 4.6143,
1772
- "step": 294
1773
- },
1774
- {
1775
- "epoch": 0.5,
1776
- "learning_rate": 0.00017579999999999996,
1777
- "loss": 4.4993,
1778
- "step": 295
1779
- },
1780
- {
1781
- "epoch": 0.5,
1782
- "learning_rate": 0.00017639999999999998,
1783
- "loss": 4.348,
1784
- "step": 296
1785
- },
1786
- {
1787
- "epoch": 0.5,
1788
- "learning_rate": 0.00017699999999999997,
1789
- "loss": 5.2729,
1790
- "step": 297
1791
- },
1792
- {
1793
- "epoch": 0.5,
1794
- "learning_rate": 0.00017759999999999998,
1795
- "loss": 5.2359,
1796
- "step": 298
1797
- },
1798
- {
1799
- "epoch": 0.5,
1800
- "learning_rate": 0.00017819999999999997,
1801
- "loss": 4.8578,
1802
- "step": 299
1803
- },
1804
- {
1805
- "epoch": 0.5,
1806
- "learning_rate": 0.00017879999999999998,
1807
- "loss": 5.5686,
1808
- "step": 300
1809
- },
1810
- {
1811
- "epoch": 0.51,
1812
- "learning_rate": 0.00017939999999999997,
1813
- "loss": 4.6844,
1814
- "step": 301
1815
- },
1816
- {
1817
- "epoch": 0.51,
1818
- "learning_rate": 0.00017999999999999998,
1819
- "loss": 4.5929,
1820
- "step": 302
1821
- },
1822
- {
1823
- "epoch": 0.51,
1824
- "learning_rate": 0.00018059999999999997,
1825
- "loss": 4.7074,
1826
- "step": 303
1827
- },
1828
- {
1829
- "epoch": 0.51,
1830
- "learning_rate": 0.00018119999999999999,
1831
- "loss": 4.4101,
1832
- "step": 304
1833
- },
1834
- {
1835
- "epoch": 0.51,
1836
- "learning_rate": 0.00018179999999999997,
1837
- "loss": 4.5093,
1838
- "step": 305
1839
- },
1840
- {
1841
- "epoch": 0.51,
1842
- "learning_rate": 0.0001824,
1843
- "loss": 4.725,
1844
- "step": 306
1845
- },
1846
- {
1847
- "epoch": 0.52,
1848
- "learning_rate": 0.00018299999999999998,
1849
- "loss": 4.6516,
1850
- "step": 307
1851
- },
1852
- {
1853
- "epoch": 0.52,
1854
- "learning_rate": 0.0001836,
1855
- "loss": 4.7071,
1856
- "step": 308
1857
- },
1858
- {
1859
- "epoch": 0.52,
1860
- "learning_rate": 0.00018419999999999998,
1861
- "loss": 4.4795,
1862
- "step": 309
1863
- },
1864
- {
1865
- "epoch": 0.52,
1866
- "learning_rate": 0.0001848,
1867
- "loss": 4.4157,
1868
- "step": 310
1869
- },
1870
- {
1871
- "epoch": 0.52,
1872
- "learning_rate": 0.00018539999999999998,
1873
- "loss": 4.2875,
1874
- "step": 311
1875
- },
1876
- {
1877
- "epoch": 0.52,
1878
- "learning_rate": 0.000186,
1879
- "loss": 4.5176,
1880
- "step": 312
1881
- },
1882
- {
1883
- "epoch": 0.53,
1884
- "learning_rate": 0.00018659999999999998,
1885
- "loss": 4.4679,
1886
- "step": 313
1887
- },
1888
- {
1889
- "epoch": 0.53,
1890
- "learning_rate": 0.0001872,
1891
- "loss": 4.6015,
1892
- "step": 314
1893
- },
1894
- {
1895
- "epoch": 0.53,
1896
- "learning_rate": 0.00018779999999999998,
1897
- "loss": 4.338,
1898
- "step": 315
1899
- },
1900
- {
1901
- "epoch": 0.53,
1902
- "learning_rate": 0.00018839999999999997,
1903
- "loss": 4.6336,
1904
- "step": 316
1905
- },
1906
- {
1907
- "epoch": 0.53,
1908
- "learning_rate": 0.00018899999999999999,
1909
- "loss": 4.4357,
1910
- "step": 317
1911
- },
1912
- {
1913
- "epoch": 0.53,
1914
- "learning_rate": 0.00018959999999999997,
1915
- "loss": 4.8962,
1916
- "step": 318
1917
- },
1918
- {
1919
- "epoch": 0.54,
1920
- "learning_rate": 0.0001902,
1921
- "loss": 4.5626,
1922
- "step": 319
1923
- },
1924
- {
1925
- "epoch": 0.54,
1926
- "learning_rate": 0.00019079999999999998,
1927
- "loss": 4.5467,
1928
- "step": 320
1929
- },
1930
- {
1931
- "epoch": 0.54,
1932
- "learning_rate": 0.0001914,
1933
- "loss": 4.2766,
1934
- "step": 321
1935
- },
1936
- {
1937
- "epoch": 0.54,
1938
- "learning_rate": 0.00019199999999999998,
1939
- "loss": 4.3254,
1940
- "step": 322
1941
- },
1942
- {
1943
- "epoch": 0.54,
1944
- "learning_rate": 0.0001926,
1945
- "loss": 4.8041,
1946
- "step": 323
1947
- },
1948
- {
1949
- "epoch": 0.54,
1950
- "learning_rate": 0.00019319999999999998,
1951
- "loss": 4.3947,
1952
- "step": 324
1953
- },
1954
- {
1955
- "epoch": 0.55,
1956
- "learning_rate": 0.0001938,
1957
- "loss": 4.7643,
1958
- "step": 325
1959
- },
1960
- {
1961
- "epoch": 0.55,
1962
- "learning_rate": 0.00019439999999999998,
1963
- "loss": 4.4117,
1964
- "step": 326
1965
- },
1966
- {
1967
- "epoch": 0.55,
1968
- "learning_rate": 0.000195,
1969
- "loss": 4.7301,
1970
- "step": 327
1971
- },
1972
- {
1973
- "epoch": 0.55,
1974
- "learning_rate": 0.00019559999999999998,
1975
- "loss": 4.4717,
1976
- "step": 328
1977
- },
1978
- {
1979
- "epoch": 0.55,
1980
- "learning_rate": 0.0001962,
1981
- "loss": 4.4449,
1982
- "step": 329
1983
- },
1984
- {
1985
- "epoch": 0.55,
1986
- "learning_rate": 0.00019679999999999999,
1987
- "loss": 4.4055,
1988
- "step": 330
1989
- },
1990
- {
1991
- "epoch": 0.56,
1992
- "learning_rate": 0.0001974,
1993
- "loss": 4.5839,
1994
- "step": 331
1995
- },
1996
- {
1997
- "epoch": 0.56,
1998
- "learning_rate": 0.000198,
1999
- "loss": 4.2595,
2000
- "step": 332
2001
- },
2002
- {
2003
- "epoch": 0.56,
2004
- "learning_rate": 0.0001986,
2005
- "loss": 4.4989,
2006
- "step": 333
2007
- },
2008
- {
2009
- "epoch": 0.56,
2010
- "learning_rate": 0.0001992,
2011
- "loss": 4.5707,
2012
- "step": 334
2013
- },
2014
- {
2015
- "epoch": 0.56,
2016
- "learning_rate": 0.0001998,
2017
- "loss": 4.4243,
2018
- "step": 335
2019
- },
2020
- {
2021
- "epoch": 0.56,
2022
- "learning_rate": 0.0002004,
2023
- "loss": 4.4127,
2024
- "step": 336
2025
- },
2026
- {
2027
- "epoch": 0.57,
2028
- "learning_rate": 0.000201,
2029
- "loss": 4.4754,
2030
- "step": 337
2031
- },
2032
- {
2033
- "epoch": 0.57,
2034
- "learning_rate": 0.0002016,
2035
- "loss": 4.3751,
2036
- "step": 338
2037
- },
2038
- {
2039
- "epoch": 0.57,
2040
- "learning_rate": 0.0002022,
2041
- "loss": 4.4543,
2042
- "step": 339
2043
- },
2044
- {
2045
- "epoch": 0.57,
2046
- "learning_rate": 0.0002028,
2047
- "loss": 4.6159,
2048
- "step": 340
2049
- },
2050
- {
2051
- "epoch": 0.57,
2052
- "learning_rate": 0.00020339999999999998,
2053
- "loss": 4.4517,
2054
- "step": 341
2055
- },
2056
- {
2057
- "epoch": 0.58,
2058
- "learning_rate": 0.000204,
2059
- "loss": 4.7063,
2060
- "step": 342
2061
- },
2062
- {
2063
- "epoch": 0.58,
2064
- "learning_rate": 0.00020459999999999999,
2065
- "loss": 4.7711,
2066
- "step": 343
2067
- },
2068
- {
2069
- "epoch": 0.58,
2070
- "learning_rate": 0.0002052,
2071
- "loss": 4.6853,
2072
- "step": 344
2073
- },
2074
- {
2075
- "epoch": 0.58,
2076
- "learning_rate": 0.0002058,
2077
- "loss": 4.8209,
2078
- "step": 345
2079
- },
2080
- {
2081
- "epoch": 0.58,
2082
- "learning_rate": 0.00020639999999999998,
2083
- "loss": 4.3988,
2084
- "step": 346
2085
- },
2086
- {
2087
- "epoch": 0.58,
2088
- "learning_rate": 0.00020699999999999996,
2089
- "loss": 4.5112,
2090
- "step": 347
2091
- },
2092
- {
2093
- "epoch": 0.59,
2094
- "learning_rate": 0.00020759999999999998,
2095
- "loss": 4.6472,
2096
- "step": 348
2097
- },
2098
- {
2099
- "epoch": 0.59,
2100
- "learning_rate": 0.00020819999999999996,
2101
- "loss": 4.4794,
2102
- "step": 349
2103
- },
2104
- {
2105
- "epoch": 0.59,
2106
- "learning_rate": 0.00020879999999999998,
2107
- "loss": 5.2373,
2108
- "step": 350
2109
- },
2110
- {
2111
- "epoch": 0.59,
2112
- "learning_rate": 0.00020939999999999997,
2113
- "loss": 4.5732,
2114
- "step": 351
2115
- },
2116
- {
2117
- "epoch": 0.59,
2118
- "learning_rate": 0.00020999999999999998,
2119
- "loss": 4.5354,
2120
- "step": 352
2121
- },
2122
- {
2123
- "epoch": 0.59,
2124
- "learning_rate": 0.00021059999999999997,
2125
- "loss": 4.5812,
2126
- "step": 353
2127
- },
2128
- {
2129
- "epoch": 0.6,
2130
- "learning_rate": 0.00021119999999999996,
2131
- "loss": 4.4975,
2132
- "step": 354
2133
- },
2134
- {
2135
- "epoch": 0.6,
2136
- "learning_rate": 0.00021179999999999997,
2137
- "loss": 4.4032,
2138
- "step": 355
2139
- },
2140
- {
2141
- "epoch": 0.6,
2142
- "learning_rate": 0.00021239999999999996,
2143
- "loss": 4.3613,
2144
- "step": 356
2145
- },
2146
- {
2147
- "epoch": 0.6,
2148
- "learning_rate": 0.00021299999999999997,
2149
- "loss": 4.3984,
2150
- "step": 357
2151
- },
2152
- {
2153
- "epoch": 0.6,
2154
- "learning_rate": 0.00021359999999999996,
2155
- "loss": 4.7058,
2156
- "step": 358
2157
- },
2158
- {
2159
- "epoch": 0.6,
2160
- "learning_rate": 0.00021419999999999998,
2161
- "loss": 4.3473,
2162
- "step": 359
2163
- },
2164
- {
2165
- "epoch": 0.61,
2166
- "learning_rate": 0.00021479999999999996,
2167
- "loss": 4.6245,
2168
- "step": 360
2169
- },
2170
- {
2171
- "epoch": 0.61,
2172
- "learning_rate": 0.00021539999999999998,
2173
- "loss": 4.6861,
2174
- "step": 361
2175
- },
2176
- {
2177
- "epoch": 0.61,
2178
- "learning_rate": 0.00021599999999999996,
2179
- "loss": 4.4029,
2180
- "step": 362
2181
- },
2182
- {
2183
- "epoch": 0.61,
2184
- "learning_rate": 0.00021659999999999998,
2185
- "loss": 4.7367,
2186
- "step": 363
2187
- },
2188
- {
2189
- "epoch": 0.61,
2190
- "learning_rate": 0.00021719999999999997,
2191
- "loss": 4.4922,
2192
- "step": 364
2193
- },
2194
- {
2195
- "epoch": 0.61,
2196
- "learning_rate": 0.00021779999999999998,
2197
- "loss": 4.7612,
2198
- "step": 365
2199
- },
2200
- {
2201
- "epoch": 0.62,
2202
- "learning_rate": 0.00021839999999999997,
2203
- "loss": 4.447,
2204
- "step": 366
2205
- },
2206
- {
2207
- "epoch": 0.62,
2208
- "learning_rate": 0.00021899999999999998,
2209
- "loss": 4.5431,
2210
- "step": 367
2211
- },
2212
- {
2213
- "epoch": 0.62,
2214
- "learning_rate": 0.00021959999999999997,
2215
- "loss": 4.3665,
2216
- "step": 368
2217
- },
2218
- {
2219
- "epoch": 0.62,
2220
- "learning_rate": 0.00022019999999999999,
2221
- "loss": 4.4973,
2222
- "step": 369
2223
- },
2224
- {
2225
- "epoch": 0.62,
2226
- "learning_rate": 0.00022079999999999997,
2227
- "loss": 4.8256,
2228
- "step": 370
2229
- },
2230
- {
2231
- "epoch": 0.62,
2232
- "learning_rate": 0.0002214,
2233
- "loss": 4.2738,
2234
- "step": 371
2235
- },
2236
- {
2237
- "epoch": 0.63,
2238
- "learning_rate": 0.00022199999999999998,
2239
- "loss": 4.4241,
2240
- "step": 372
2241
- },
2242
- {
2243
- "epoch": 0.63,
2244
- "learning_rate": 0.0002226,
2245
- "loss": 4.3496,
2246
- "step": 373
2247
- },
2248
- {
2249
- "epoch": 0.63,
2250
- "learning_rate": 0.00022319999999999998,
2251
- "loss": 4.6338,
2252
- "step": 374
2253
- },
2254
- {
2255
- "epoch": 0.63,
2256
- "learning_rate": 0.0002238,
2257
- "loss": 4.5995,
2258
- "step": 375
2259
- },
2260
- {
2261
- "epoch": 0.63,
2262
- "learning_rate": 0.00022439999999999998,
2263
- "loss": 4.5976,
2264
- "step": 376
2265
- },
2266
- {
2267
- "epoch": 0.63,
2268
- "learning_rate": 0.000225,
2269
- "loss": 4.572,
2270
- "step": 377
2271
- },
2272
- {
2273
- "epoch": 0.64,
2274
- "learning_rate": 0.00022559999999999998,
2275
- "loss": 4.6357,
2276
- "step": 378
2277
- },
2278
- {
2279
- "epoch": 0.64,
2280
- "learning_rate": 0.00022619999999999997,
2281
- "loss": 4.5382,
2282
- "step": 379
2283
- },
2284
- {
2285
- "epoch": 0.64,
2286
- "learning_rate": 0.00022679999999999998,
2287
- "loss": 4.5339,
2288
- "step": 380
2289
- },
2290
- {
2291
- "epoch": 0.64,
2292
- "learning_rate": 0.00022739999999999997,
2293
- "loss": 4.6268,
2294
- "step": 381
2295
- },
2296
- {
2297
- "epoch": 0.64,
2298
- "learning_rate": 0.00022799999999999999,
2299
- "loss": 4.7323,
2300
- "step": 382
2301
- },
2302
- {
2303
- "epoch": 0.64,
2304
- "learning_rate": 0.00022859999999999997,
2305
- "loss": 4.7925,
2306
- "step": 383
2307
- },
2308
- {
2309
- "epoch": 0.65,
2310
- "learning_rate": 0.0002292,
2311
- "loss": 4.5124,
2312
- "step": 384
2313
- },
2314
- {
2315
- "epoch": 0.65,
2316
- "learning_rate": 0.00022979999999999997,
2317
- "loss": 4.706,
2318
- "step": 385
2319
- },
2320
- {
2321
- "epoch": 0.65,
2322
- "learning_rate": 0.0002304,
2323
- "loss": 4.8178,
2324
- "step": 386
2325
- },
2326
- {
2327
- "epoch": 0.65,
2328
- "learning_rate": 0.00023099999999999998,
2329
- "loss": 4.981,
2330
- "step": 387
2331
- },
2332
- {
2333
- "epoch": 0.65,
2334
- "learning_rate": 0.0002316,
2335
- "loss": 4.4796,
2336
- "step": 388
2337
- },
2338
- {
2339
- "epoch": 0.65,
2340
- "learning_rate": 0.00023219999999999998,
2341
- "loss": 4.4747,
2342
- "step": 389
2343
- },
2344
- {
2345
- "epoch": 0.66,
2346
- "learning_rate": 0.0002328,
2347
- "loss": 4.5041,
2348
- "step": 390
2349
- },
2350
- {
2351
- "epoch": 0.66,
2352
- "learning_rate": 0.00023339999999999998,
2353
- "loss": 4.5166,
2354
- "step": 391
2355
- },
2356
- {
2357
- "epoch": 0.66,
2358
- "learning_rate": 0.000234,
2359
- "loss": 4.7084,
2360
- "step": 392
2361
- },
2362
- {
2363
- "epoch": 0.66,
2364
- "learning_rate": 0.00023459999999999998,
2365
- "loss": 4.9243,
2366
- "step": 393
2367
- },
2368
- {
2369
- "epoch": 0.66,
2370
- "learning_rate": 0.0002352,
2371
- "loss": 4.4673,
2372
- "step": 394
2373
- },
2374
- {
2375
- "epoch": 0.66,
2376
- "learning_rate": 0.00023579999999999999,
2377
- "loss": 4.9729,
2378
- "step": 395
2379
- },
2380
- {
2381
- "epoch": 0.67,
2382
- "learning_rate": 0.0002364,
2383
- "loss": 4.6977,
2384
- "step": 396
2385
- },
2386
- {
2387
- "epoch": 0.67,
2388
- "learning_rate": 0.000237,
2389
- "loss": 4.9387,
2390
- "step": 397
2391
- },
2392
- {
2393
- "epoch": 0.67,
2394
- "learning_rate": 0.0002376,
2395
- "loss": 5.7888,
2396
- "step": 398
2397
- },
2398
- {
2399
- "epoch": 0.67,
2400
- "learning_rate": 0.0002382,
2401
- "loss": 4.9183,
2402
- "step": 399
2403
- },
2404
- {
2405
- "epoch": 0.67,
2406
- "learning_rate": 0.0002388,
2407
- "loss": 5.0151,
2408
- "step": 400
2409
- },
2410
- {
2411
- "epoch": 0.67,
2412
- "learning_rate": 0.0002394,
2413
- "loss": 4.7088,
2414
- "step": 401
2415
- },
2416
- {
2417
- "epoch": 0.68,
2418
- "learning_rate": 0.00023999999999999998,
2419
- "loss": 4.644,
2420
- "step": 402
2421
- },
2422
- {
2423
- "epoch": 0.68,
2424
- "learning_rate": 0.0002406,
2425
- "loss": 4.5208,
2426
- "step": 403
2427
- },
2428
- {
2429
- "epoch": 0.68,
2430
- "learning_rate": 0.00024119999999999998,
2431
- "loss": 4.4707,
2432
- "step": 404
2433
- },
2434
- {
2435
- "epoch": 0.68,
2436
- "learning_rate": 0.0002418,
2437
- "loss": 4.4193,
2438
- "step": 405
2439
- },
2440
- {
2441
- "epoch": 0.68,
2442
- "learning_rate": 0.00024239999999999998,
2443
- "loss": 4.3914,
2444
- "step": 406
2445
- },
2446
- {
2447
- "epoch": 0.68,
2448
- "learning_rate": 0.000243,
2449
- "loss": 4.6832,
2450
- "step": 407
2451
- },
2452
- {
2453
- "epoch": 0.69,
2454
- "learning_rate": 0.00024359999999999999,
2455
- "loss": 4.3622,
2456
- "step": 408
2457
- },
2458
- {
2459
- "epoch": 0.69,
2460
- "learning_rate": 0.00024419999999999997,
2461
- "loss": 4.5366,
2462
- "step": 409
2463
- },
2464
- {
2465
- "epoch": 0.69,
2466
- "learning_rate": 0.0002448,
2467
- "loss": 4.4287,
2468
- "step": 410
2469
- },
2470
- {
2471
- "epoch": 0.69,
2472
- "learning_rate": 0.00024539999999999995,
2473
- "loss": 4.3764,
2474
- "step": 411
2475
- },
2476
- {
2477
- "epoch": 0.69,
2478
- "learning_rate": 0.00024599999999999996,
2479
- "loss": 4.6033,
2480
- "step": 412
2481
- },
2482
- {
2483
- "epoch": 0.69,
2484
- "learning_rate": 0.0002466,
2485
- "loss": 4.4157,
2486
- "step": 413
2487
- },
2488
- {
2489
- "epoch": 0.7,
2490
- "learning_rate": 0.0002472,
2491
- "loss": 4.5498,
2492
- "step": 414
2493
- },
2494
- {
2495
- "epoch": 0.7,
2496
- "learning_rate": 0.00024779999999999995,
2497
- "loss": 4.4146,
2498
- "step": 415
2499
- },
2500
- {
2501
- "epoch": 0.7,
2502
- "learning_rate": 0.00024839999999999997,
2503
- "loss": 4.4638,
2504
- "step": 416
2505
- },
2506
- {
2507
- "epoch": 0.7,
2508
- "learning_rate": 0.000249,
2509
- "loss": 4.4486,
2510
- "step": 417
2511
- },
2512
- {
2513
- "epoch": 0.7,
2514
- "learning_rate": 0.00024959999999999994,
2515
- "loss": 4.1511,
2516
- "step": 418
2517
- },
2518
- {
2519
- "epoch": 0.7,
2520
- "learning_rate": 0.00025019999999999996,
2521
- "loss": 4.6831,
2522
- "step": 419
2523
- },
2524
- {
2525
- "epoch": 0.71,
2526
- "learning_rate": 0.00025079999999999997,
2527
- "loss": 4.502,
2528
- "step": 420
2529
- },
2530
- {
2531
- "epoch": 0.71,
2532
- "learning_rate": 0.0002514,
2533
- "loss": 4.196,
2534
- "step": 421
2535
- },
2536
- {
2537
- "epoch": 0.71,
2538
- "learning_rate": 0.00025199999999999995,
2539
- "loss": 4.4275,
2540
- "step": 422
2541
- },
2542
- {
2543
- "epoch": 0.71,
2544
- "learning_rate": 0.00025259999999999996,
2545
- "loss": 4.5887,
2546
- "step": 423
2547
- },
2548
- {
2549
- "epoch": 0.71,
2550
- "learning_rate": 0.0002532,
2551
- "loss": 4.4421,
2552
- "step": 424
2553
- },
2554
- {
2555
- "epoch": 0.71,
2556
- "learning_rate": 0.0002538,
2557
- "loss": 4.4905,
2558
- "step": 425
2559
- },
2560
- {
2561
- "epoch": 0.72,
2562
- "learning_rate": 0.00025439999999999995,
2563
- "loss": 4.4904,
2564
- "step": 426
2565
- },
2566
- {
2567
- "epoch": 0.72,
2568
- "learning_rate": 0.00025499999999999996,
2569
- "loss": 4.2661,
2570
- "step": 427
2571
- },
2572
- {
2573
- "epoch": 0.72,
2574
- "learning_rate": 0.0002556,
2575
- "loss": 4.2795,
2576
- "step": 428
2577
- },
2578
- {
2579
- "epoch": 0.72,
2580
- "learning_rate": 0.0002562,
2581
- "loss": 4.8814,
2582
- "step": 429
2583
- },
2584
- {
2585
- "epoch": 0.72,
2586
- "learning_rate": 0.00025679999999999995,
2587
- "loss": 4.5092,
2588
- "step": 430
2589
- },
2590
- {
2591
- "epoch": 0.72,
2592
- "learning_rate": 0.00025739999999999997,
2593
- "loss": 4.7066,
2594
- "step": 431
2595
- },
2596
- {
2597
- "epoch": 0.73,
2598
- "learning_rate": 0.000258,
2599
- "loss": 4.22,
2600
- "step": 432
2601
- },
2602
- {
2603
- "epoch": 0.73,
2604
- "learning_rate": 0.0002586,
2605
- "loss": 4.4888,
2606
- "step": 433
2607
- },
2608
- {
2609
- "epoch": 0.73,
2610
- "learning_rate": 0.00025919999999999996,
2611
- "loss": 4.5165,
2612
- "step": 434
2613
- },
2614
- {
2615
- "epoch": 0.73,
2616
- "learning_rate": 0.00025979999999999997,
2617
- "loss": 4.6135,
2618
- "step": 435
2619
- },
2620
- {
2621
- "epoch": 0.73,
2622
- "learning_rate": 0.0002604,
2623
- "loss": 4.4378,
2624
- "step": 436
2625
- },
2626
- {
2627
- "epoch": 0.73,
2628
- "learning_rate": 0.000261,
2629
- "loss": 4.6876,
2630
- "step": 437
2631
- },
2632
- {
2633
- "epoch": 0.74,
2634
- "learning_rate": 0.00026159999999999996,
2635
- "loss": 4.5232,
2636
- "step": 438
2637
- },
2638
- {
2639
- "epoch": 0.74,
2640
- "learning_rate": 0.0002622,
2641
- "loss": 4.6054,
2642
- "step": 439
2643
- },
2644
- {
2645
- "epoch": 0.74,
2646
- "learning_rate": 0.0002628,
2647
- "loss": 4.6991,
2648
- "step": 440
2649
- },
2650
- {
2651
- "epoch": 0.74,
2652
- "learning_rate": 0.00026339999999999995,
2653
- "loss": 4.7554,
2654
- "step": 441
2655
- },
2656
- {
2657
- "epoch": 0.74,
2658
- "learning_rate": 0.00026399999999999997,
2659
- "loss": 4.8751,
2660
- "step": 442
2661
- },
2662
- {
2663
- "epoch": 0.74,
2664
- "learning_rate": 0.0002646,
2665
- "loss": 4.8587,
2666
- "step": 443
2667
- },
2668
- {
2669
- "epoch": 0.75,
2670
- "learning_rate": 0.0002652,
2671
- "loss": 4.5289,
2672
- "step": 444
2673
- },
2674
- {
2675
- "epoch": 0.75,
2676
- "learning_rate": 0.00026579999999999996,
2677
- "loss": 4.3795,
2678
- "step": 445
2679
- },
2680
- {
2681
- "epoch": 0.75,
2682
- "learning_rate": 0.00026639999999999997,
2683
- "loss": 4.7699,
2684
- "step": 446
2685
- },
2686
- {
2687
- "epoch": 0.75,
2688
- "learning_rate": 0.000267,
2689
- "loss": 4.5368,
2690
- "step": 447
2691
- },
2692
- {
2693
- "epoch": 0.75,
2694
- "learning_rate": 0.0002676,
2695
- "loss": 4.6414,
2696
- "step": 448
2697
- },
2698
- {
2699
- "epoch": 0.75,
2700
- "learning_rate": 0.00026819999999999996,
2701
- "loss": 5.1838,
2702
- "step": 449
2703
- },
2704
- {
2705
- "epoch": 0.76,
2706
- "learning_rate": 0.0002688,
2707
- "loss": 4.3724,
2708
- "step": 450
2709
- },
2710
- {
2711
- "epoch": 0.76,
2712
- "learning_rate": 0.0002694,
2713
- "loss": 4.4935,
2714
- "step": 451
2715
- },
2716
- {
2717
- "epoch": 0.76,
2718
- "learning_rate": 0.00027,
2719
- "loss": 4.6275,
2720
- "step": 452
2721
- },
2722
- {
2723
- "epoch": 0.76,
2724
- "learning_rate": 0.00027059999999999996,
2725
- "loss": 4.4756,
2726
- "step": 453
2727
- },
2728
- {
2729
- "epoch": 0.76,
2730
- "learning_rate": 0.0002712,
2731
- "loss": 4.7369,
2732
- "step": 454
2733
- },
2734
- {
2735
- "epoch": 0.77,
2736
- "learning_rate": 0.0002718,
2737
- "loss": 4.6098,
2738
- "step": 455
2739
- },
2740
- {
2741
- "epoch": 0.77,
2742
- "learning_rate": 0.0002724,
2743
- "loss": 4.6067,
2744
- "step": 456
2745
- },
2746
- {
2747
- "epoch": 0.77,
2748
- "learning_rate": 0.00027299999999999997,
2749
- "loss": 4.8474,
2750
- "step": 457
2751
- },
2752
- {
2753
- "epoch": 0.77,
2754
- "learning_rate": 0.0002736,
2755
- "loss": 4.6679,
2756
- "step": 458
2757
- },
2758
- {
2759
- "epoch": 0.77,
2760
- "learning_rate": 0.0002742,
2761
- "loss": 4.7952,
2762
- "step": 459
2763
- },
2764
- {
2765
- "epoch": 0.77,
2766
- "learning_rate": 0.0002748,
2767
- "loss": 4.5183,
2768
- "step": 460
2769
- },
2770
- {
2771
- "epoch": 0.78,
2772
- "learning_rate": 0.00027539999999999997,
2773
- "loss": 4.8535,
2774
- "step": 461
2775
- },
2776
- {
2777
- "epoch": 0.78,
2778
- "learning_rate": 0.000276,
2779
- "loss": 4.4351,
2780
- "step": 462
2781
- },
2782
- {
2783
- "epoch": 0.78,
2784
- "learning_rate": 0.0002766,
2785
- "loss": 4.7839,
2786
- "step": 463
2787
- },
2788
- {
2789
- "epoch": 0.78,
2790
- "learning_rate": 0.0002772,
2791
- "loss": 4.6477,
2792
- "step": 464
2793
- },
2794
- {
2795
- "epoch": 0.78,
2796
- "learning_rate": 0.0002778,
2797
- "loss": 4.8635,
2798
- "step": 465
2799
- },
2800
- {
2801
- "epoch": 0.78,
2802
- "learning_rate": 0.0002784,
2803
- "loss": 4.4758,
2804
- "step": 466
2805
- },
2806
- {
2807
- "epoch": 0.79,
2808
- "learning_rate": 0.000279,
2809
- "loss": 4.5836,
2810
- "step": 467
2811
- },
2812
- {
2813
- "epoch": 0.79,
2814
- "learning_rate": 0.00027959999999999997,
2815
- "loss": 4.4608,
2816
- "step": 468
2817
- },
2818
- {
2819
- "epoch": 0.79,
2820
- "learning_rate": 0.0002802,
2821
- "loss": 4.7147,
2822
- "step": 469
2823
- },
2824
- {
2825
- "epoch": 0.79,
2826
- "learning_rate": 0.0002808,
2827
- "loss": 4.3855,
2828
- "step": 470
2829
- },
2830
- {
2831
- "epoch": 0.79,
2832
- "learning_rate": 0.00028139999999999996,
2833
- "loss": 4.5603,
2834
- "step": 471
2835
- },
2836
- {
2837
- "epoch": 0.79,
2838
- "learning_rate": 0.00028199999999999997,
2839
- "loss": 4.3725,
2840
- "step": 472
2841
- },
2842
- {
2843
- "epoch": 0.8,
2844
- "learning_rate": 0.0002826,
2845
- "loss": 4.4598,
2846
- "step": 473
2847
- },
2848
- {
2849
- "epoch": 0.8,
2850
- "learning_rate": 0.00028319999999999994,
2851
- "loss": 4.6407,
2852
- "step": 474
2853
- },
2854
- {
2855
- "epoch": 0.8,
2856
- "learning_rate": 0.00028379999999999996,
2857
- "loss": 4.4258,
2858
- "step": 475
2859
- },
2860
- {
2861
- "epoch": 0.8,
2862
- "learning_rate": 0.0002844,
2863
- "loss": 4.6098,
2864
- "step": 476
2865
- },
2866
- {
2867
- "epoch": 0.8,
2868
- "learning_rate": 0.000285,
2869
- "loss": 4.5606,
2870
- "step": 477
2871
- },
2872
- {
2873
- "epoch": 0.8,
2874
- "learning_rate": 0.00028559999999999995,
2875
- "loss": 4.5687,
2876
- "step": 478
2877
- },
2878
- {
2879
- "epoch": 0.81,
2880
- "learning_rate": 0.00028619999999999996,
2881
- "loss": 4.3645,
2882
- "step": 479
2883
- },
2884
- {
2885
- "epoch": 0.81,
2886
- "learning_rate": 0.0002868,
2887
- "loss": 4.3326,
2888
- "step": 480
2889
- },
2890
- {
2891
- "epoch": 0.81,
2892
- "learning_rate": 0.00028739999999999994,
2893
- "loss": 4.9138,
2894
- "step": 481
2895
- },
2896
- {
2897
- "epoch": 0.81,
2898
- "learning_rate": 0.00028799999999999995,
2899
- "loss": 4.992,
2900
- "step": 482
2901
- },
2902
- {
2903
- "epoch": 0.81,
2904
- "learning_rate": 0.00028859999999999997,
2905
- "loss": 4.7403,
2906
- "step": 483
2907
- },
2908
- {
2909
- "epoch": 0.81,
2910
- "learning_rate": 0.0002892,
2911
- "loss": 4.7293,
2912
- "step": 484
2913
- },
2914
- {
2915
- "epoch": 0.82,
2916
- "learning_rate": 0.00028979999999999994,
2917
- "loss": 4.5899,
2918
- "step": 485
2919
- },
2920
- {
2921
- "epoch": 0.82,
2922
- "learning_rate": 0.00029039999999999996,
2923
- "loss": 4.4421,
2924
- "step": 486
2925
- },
2926
- {
2927
- "epoch": 0.82,
2928
- "learning_rate": 0.00029099999999999997,
2929
- "loss": 4.6241,
2930
- "step": 487
2931
- },
2932
- {
2933
- "epoch": 0.82,
2934
- "learning_rate": 0.0002916,
2935
- "loss": 4.4847,
2936
- "step": 488
2937
- },
2938
- {
2939
- "epoch": 0.82,
2940
- "learning_rate": 0.00029219999999999995,
2941
- "loss": 4.5834,
2942
- "step": 489
2943
- },
2944
- {
2945
- "epoch": 0.82,
2946
- "learning_rate": 0.00029279999999999996,
2947
- "loss": 4.4479,
2948
- "step": 490
2949
- },
2950
- {
2951
- "epoch": 0.83,
2952
- "learning_rate": 0.0002934,
2953
- "loss": 4.7391,
2954
- "step": 491
2955
- },
2956
- {
2957
- "epoch": 0.83,
2958
- "learning_rate": 0.000294,
2959
- "loss": 4.7008,
2960
- "step": 492
2961
- },
2962
- {
2963
- "epoch": 0.83,
2964
- "learning_rate": 0.00029459999999999995,
2965
- "loss": 4.3374,
2966
- "step": 493
2967
- },
2968
- {
2969
- "epoch": 0.83,
2970
- "learning_rate": 0.00029519999999999997,
2971
- "loss": 4.378,
2972
- "step": 494
2973
- },
2974
- {
2975
- "epoch": 0.83,
2976
- "learning_rate": 0.0002958,
2977
- "loss": 4.636,
2978
- "step": 495
2979
- },
2980
- {
2981
- "epoch": 0.83,
2982
- "learning_rate": 0.0002964,
2983
- "loss": 4.8833,
2984
- "step": 496
2985
- },
2986
- {
2987
- "epoch": 0.84,
2988
- "learning_rate": 0.00029699999999999996,
2989
- "loss": 4.5708,
2990
- "step": 497
2991
- },
2992
- {
2993
- "epoch": 0.84,
2994
- "learning_rate": 0.00029759999999999997,
2995
- "loss": 4.6291,
2996
- "step": 498
2997
- },
2998
- {
2999
- "epoch": 0.84,
3000
- "learning_rate": 0.0002982,
3001
- "loss": 5.1499,
3002
- "step": 499
3003
- },
3004
- {
3005
- "epoch": 0.84,
3006
- "learning_rate": 0.0002988,
3007
- "loss": 4.7669,
3008
- "step": 500
3009
- },
3010
- {
3011
- "epoch": 0.84,
3012
- "eval_loss": 4.873856067657471,
3013
- "eval_runtime": 784.3492,
3014
- "eval_samples_per_second": 3.368,
3015
- "eval_steps_per_second": 0.282,
3016
- "eval_wer": 1.9278262594208648,
3017
- "step": 500
3018
- },
3019
- {
3020
- "epoch": 0.84,
3021
- "learning_rate": 0.00029939999999999996,
3022
- "loss": 4.5736,
3023
- "step": 501
3024
- },
3025
- {
3026
- "epoch": 0.84,
3027
- "learning_rate": 0.0003,
3028
- "loss": 4.6392,
3029
- "step": 502
3030
- },
3031
- {
3032
- "epoch": 0.85,
3033
- "learning_rate": 0.00029680851063829784,
3034
- "loss": 4.7407,
3035
- "step": 503
3036
- },
3037
- {
3038
- "epoch": 0.85,
3039
- "learning_rate": 0.0002936170212765957,
3040
- "loss": 4.9224,
3041
- "step": 504
3042
- },
3043
- {
3044
- "epoch": 0.85,
3045
- "learning_rate": 0.0002904255319148936,
3046
- "loss": 4.7926,
3047
- "step": 505
3048
- },
3049
- {
3050
- "epoch": 0.85,
3051
- "learning_rate": 0.0002872340425531915,
3052
- "loss": 5.0733,
3053
- "step": 506
3054
- },
3055
- {
3056
- "epoch": 0.85,
3057
- "learning_rate": 0.00028404255319148934,
3058
- "loss": 4.7467,
3059
- "step": 507
3060
- },
3061
- {
3062
- "epoch": 0.85,
3063
- "learning_rate": 0.0002808510638297872,
3064
- "loss": 4.7175,
3065
- "step": 508
3066
- },
3067
- {
3068
- "epoch": 0.86,
3069
- "learning_rate": 0.00027765957446808506,
3070
- "loss": 4.6484,
3071
- "step": 509
3072
- },
3073
- {
3074
- "epoch": 0.86,
3075
- "learning_rate": 0.000274468085106383,
3076
- "loss": 4.7577,
3077
- "step": 510
3078
- },
3079
- {
3080
- "epoch": 0.86,
3081
- "learning_rate": 0.00027127659574468084,
3082
- "loss": 4.7992,
3083
- "step": 511
3084
- },
3085
- {
3086
- "epoch": 0.86,
3087
- "learning_rate": 0.0002680851063829787,
3088
- "loss": 4.5604,
3089
- "step": 512
3090
- },
3091
- {
3092
- "epoch": 0.86,
3093
- "learning_rate": 0.00026489361702127656,
3094
- "loss": 4.5546,
3095
- "step": 513
3096
- },
3097
- {
3098
- "epoch": 0.86,
3099
- "learning_rate": 0.0002617021276595745,
3100
- "loss": 4.9298,
3101
- "step": 514
3102
- },
3103
- {
3104
- "epoch": 0.87,
3105
- "learning_rate": 0.0002585106382978723,
3106
- "loss": 4.6264,
3107
- "step": 515
3108
- },
3109
- {
3110
- "epoch": 0.87,
3111
- "learning_rate": 0.0002553191489361702,
3112
- "loss": 4.6125,
3113
- "step": 516
3114
- },
3115
- {
3116
- "epoch": 0.87,
3117
- "learning_rate": 0.00025212765957446806,
3118
- "loss": 4.5963,
3119
- "step": 517
3120
- },
3121
- {
3122
- "epoch": 0.87,
3123
- "learning_rate": 0.0002489361702127659,
3124
- "loss": 4.7157,
3125
- "step": 518
3126
- },
3127
- {
3128
- "epoch": 0.87,
3129
- "learning_rate": 0.00024574468085106384,
3130
- "loss": 4.5033,
3131
- "step": 519
3132
- },
3133
- {
3134
- "epoch": 0.87,
3135
- "learning_rate": 0.00024255319148936167,
3136
- "loss": 4.5918,
3137
- "step": 520
3138
- },
3139
- {
3140
- "epoch": 0.88,
3141
- "learning_rate": 0.00023936170212765956,
3142
- "loss": 4.6696,
3143
- "step": 521
3144
- },
3145
- {
3146
- "epoch": 0.88,
3147
- "learning_rate": 0.00023617021276595742,
3148
- "loss": 4.8345,
3149
- "step": 522
3150
- },
3151
- {
3152
- "epoch": 0.88,
3153
- "learning_rate": 0.00023297872340425529,
3154
- "loss": 4.3613,
3155
- "step": 523
3156
- },
3157
- {
3158
- "epoch": 0.88,
3159
- "learning_rate": 0.00022978723404255317,
3160
- "loss": 4.1583,
3161
- "step": 524
3162
- },
3163
- {
3164
- "epoch": 0.88,
3165
- "learning_rate": 0.00022659574468085106,
3166
- "loss": 4.5778,
3167
- "step": 525
3168
- },
3169
- {
3170
- "epoch": 0.88,
3171
- "learning_rate": 0.0002234042553191489,
3172
- "loss": 4.7744,
3173
- "step": 526
3174
- },
3175
- {
3176
- "epoch": 0.89,
3177
- "learning_rate": 0.00022021276595744679,
3178
- "loss": 4.6864,
3179
- "step": 527
3180
- },
3181
- {
3182
- "epoch": 0.89,
3183
- "learning_rate": 0.00021702127659574468,
3184
- "loss": 4.4407,
3185
- "step": 528
3186
- },
3187
- {
3188
- "epoch": 0.89,
3189
- "learning_rate": 0.00021382978723404254,
3190
- "loss": 4.6176,
3191
- "step": 529
3192
- },
3193
- {
3194
- "epoch": 0.89,
3195
- "learning_rate": 0.0002106382978723404,
3196
- "loss": 4.3369,
3197
- "step": 530
3198
- },
3199
- {
3200
- "epoch": 0.89,
3201
- "learning_rate": 0.0002074468085106383,
3202
- "loss": 4.4068,
3203
- "step": 531
3204
- },
3205
- {
3206
- "epoch": 0.89,
3207
- "learning_rate": 0.00020425531914893615,
3208
- "loss": 4.7584,
3209
- "step": 532
3210
- },
3211
- {
3212
- "epoch": 0.9,
3213
- "learning_rate": 0.00020106382978723404,
3214
- "loss": 4.6745,
3215
- "step": 533
3216
- },
3217
- {
3218
- "epoch": 0.9,
3219
- "learning_rate": 0.00019787234042553187,
3220
- "loss": 4.4945,
3221
- "step": 534
3222
- },
3223
- {
3224
- "epoch": 0.9,
3225
- "learning_rate": 0.00019468085106382976,
3226
- "loss": 4.3341,
3227
- "step": 535
3228
- },
3229
- {
3230
- "epoch": 0.9,
3231
- "learning_rate": 0.00019148936170212765,
3232
- "loss": 4.5124,
3233
- "step": 536
3234
- },
3235
- {
3236
- "epoch": 0.9,
3237
- "learning_rate": 0.0001882978723404255,
3238
- "loss": 4.5719,
3239
- "step": 537
3240
- },
3241
- {
3242
- "epoch": 0.9,
3243
- "learning_rate": 0.0001851063829787234,
3244
- "loss": 4.3806,
3245
- "step": 538
3246
- },
3247
- {
3248
- "epoch": 0.91,
3249
- "learning_rate": 0.00018191489361702126,
3250
- "loss": 4.623,
3251
- "step": 539
3252
- },
3253
- {
3254
- "epoch": 0.91,
3255
- "learning_rate": 0.00017872340425531912,
3256
- "loss": 4.5661,
3257
- "step": 540
3258
- },
3259
- {
3260
- "epoch": 0.91,
3261
- "learning_rate": 0.000175531914893617,
3262
- "loss": 4.6673,
3263
- "step": 541
3264
- },
3265
- {
3266
- "epoch": 0.91,
3267
- "learning_rate": 0.0001723404255319149,
3268
- "loss": 4.7293,
3269
- "step": 542
3270
- },
3271
- {
3272
- "epoch": 0.91,
3273
- "learning_rate": 0.00016914893617021274,
3274
- "loss": 4.4276,
3275
- "step": 543
3276
- },
3277
- {
3278
- "epoch": 0.91,
3279
- "learning_rate": 0.00016595744680851062,
3280
- "loss": 4.4851,
3281
- "step": 544
3282
- },
3283
- {
3284
- "epoch": 0.92,
3285
- "learning_rate": 0.00016276595744680849,
3286
- "loss": 4.5706,
3287
- "step": 545
3288
- },
3289
- {
3290
- "epoch": 0.92,
3291
- "learning_rate": 0.00015957446808510637,
3292
- "loss": 4.7982,
3293
- "step": 546
3294
- },
3295
- {
3296
- "epoch": 0.92,
3297
- "learning_rate": 0.00015638297872340426,
3298
- "loss": 4.4439,
3299
- "step": 547
3300
- },
3301
  {
3302
  "epoch": 0.92,
3303
- "learning_rate": 0.0001531914893617021,
3304
- "loss": 5.0839,
3305
- "step": 548
3306
- },
3307
- {
3308
- "epoch": 0.92,
3309
- "learning_rate": 0.00015,
3310
- "loss": 5.1486,
3311
- "step": 549
3312
  },
3313
  {
3314
  "epoch": 0.92,
3315
- "learning_rate": 0.00014680851063829785,
3316
- "loss": 4.481,
3317
- "step": 550
3318
- },
3319
- {
3320
- "epoch": 0.93,
3321
- "learning_rate": 0.00014361702127659574,
3322
- "loss": 4.5966,
3323
- "step": 551
3324
- },
3325
- {
3326
- "epoch": 0.93,
3327
- "learning_rate": 0.0001404255319148936,
3328
- "loss": 4.7382,
3329
- "step": 552
3330
- },
3331
- {
3332
- "epoch": 0.93,
3333
- "learning_rate": 0.0001372340425531915,
3334
- "loss": 4.5965,
3335
- "step": 553
3336
  },
3337
  {
3338
  "epoch": 0.93,
3339
- "learning_rate": 0.00013404255319148935,
3340
- "loss": 4.3645,
3341
- "step": 554
3342
  },
3343
  {
3344
  "epoch": 0.93,
3345
- "learning_rate": 0.00013085106382978724,
3346
- "loss": 4.4921,
3347
- "step": 555
3348
  },
3349
  {
3350
  "epoch": 0.93,
3351
- "learning_rate": 0.0001276595744680851,
3352
- "loss": 4.547,
3353
- "step": 556
3354
- },
3355
- {
3356
- "epoch": 0.94,
3357
- "learning_rate": 0.00012446808510638296,
3358
- "loss": 4.6281,
3359
- "step": 557
3360
- },
3361
- {
3362
- "epoch": 0.94,
3363
- "learning_rate": 0.00012127659574468084,
3364
- "loss": 4.5267,
3365
- "step": 558
3366
- },
3367
- {
3368
- "epoch": 0.94,
3369
- "learning_rate": 0.00011808510638297871,
3370
- "loss": 4.3407,
3371
- "step": 559
3372
  },
3373
  {
3374
  "epoch": 0.94,
3375
- "learning_rate": 0.00011489361702127659,
3376
- "loss": 4.2849,
3377
- "step": 560
3378
  },
3379
  {
3380
  "epoch": 0.94,
3381
- "learning_rate": 0.00011170212765957445,
3382
- "loss": 4.7138,
3383
- "step": 561
3384
  },
3385
  {
3386
  "epoch": 0.94,
3387
- "learning_rate": 0.00010851063829787234,
3388
- "loss": 4.3811,
3389
- "step": 562
3390
- },
3391
- {
3392
- "epoch": 0.95,
3393
- "learning_rate": 0.0001053191489361702,
3394
- "loss": 4.4681,
3395
- "step": 563
3396
- },
3397
- {
3398
- "epoch": 0.95,
3399
- "learning_rate": 0.00010212765957446807,
3400
- "loss": 4.7141,
3401
- "step": 564
3402
- },
3403
- {
3404
- "epoch": 0.95,
3405
- "learning_rate": 9.893617021276594e-05,
3406
- "loss": 4.5566,
3407
- "step": 565
3408
  },
3409
  {
3410
  "epoch": 0.95,
3411
- "learning_rate": 9.574468085106382e-05,
3412
- "loss": 4.5957,
3413
- "step": 566
3414
  },
3415
  {
3416
  "epoch": 0.95,
3417
- "learning_rate": 9.25531914893617e-05,
3418
- "loss": 4.4626,
3419
- "step": 567
3420
- },
3421
- {
3422
- "epoch": 0.96,
3423
- "learning_rate": 8.936170212765956e-05,
3424
- "loss": 4.2414,
3425
- "step": 568
3426
- },
3427
- {
3428
- "epoch": 0.96,
3429
- "learning_rate": 8.617021276595745e-05,
3430
- "loss": 4.3564,
3431
- "step": 569
3432
- },
3433
- {
3434
- "epoch": 0.96,
3435
- "learning_rate": 8.297872340425531e-05,
3436
- "loss": 4.4139,
3437
- "step": 570
3438
  },
3439
  {
3440
  "epoch": 0.96,
3441
- "learning_rate": 7.978723404255319e-05,
3442
- "loss": 4.5387,
3443
- "step": 571
3444
  },
3445
  {
3446
  "epoch": 0.96,
3447
- "learning_rate": 7.659574468085105e-05,
3448
- "loss": 4.4192,
3449
- "step": 572
3450
  },
3451
  {
3452
  "epoch": 0.96,
3453
- "learning_rate": 7.340425531914892e-05,
3454
- "loss": 4.5384,
3455
- "step": 573
3456
- },
3457
- {
3458
- "epoch": 0.97,
3459
- "learning_rate": 7.02127659574468e-05,
3460
- "loss": 4.3597,
3461
- "step": 574
3462
- },
3463
- {
3464
- "epoch": 0.97,
3465
- "learning_rate": 6.702127659574467e-05,
3466
- "loss": 4.3714,
3467
- "step": 575
3468
- },
3469
- {
3470
- "epoch": 0.97,
3471
- "learning_rate": 6.382978723404255e-05,
3472
- "loss": 4.4526,
3473
- "step": 576
3474
  },
3475
  {
3476
  "epoch": 0.97,
3477
- "learning_rate": 6.063829787234042e-05,
3478
- "loss": 4.4474,
3479
- "step": 577
3480
  },
3481
  {
3482
  "epoch": 0.97,
3483
- "learning_rate": 5.7446808510638294e-05,
3484
- "loss": 4.188,
3485
- "step": 578
3486
  },
3487
  {
3488
  "epoch": 0.97,
3489
- "learning_rate": 5.425531914893617e-05,
3490
- "loss": 4.3285,
3491
- "step": 579
3492
- },
3493
- {
3494
- "epoch": 0.98,
3495
- "learning_rate": 5.106382978723404e-05,
3496
- "loss": 4.3842,
3497
- "step": 580
3498
- },
3499
- {
3500
- "epoch": 0.98,
3501
- "learning_rate": 4.787234042553191e-05,
3502
- "loss": 4.5096,
3503
- "step": 581
3504
- },
3505
- {
3506
- "epoch": 0.98,
3507
- "learning_rate": 4.468085106382978e-05,
3508
- "loss": 4.3595,
3509
- "step": 582
3510
  },
3511
  {
3512
  "epoch": 0.98,
3513
- "learning_rate": 4.1489361702127656e-05,
3514
- "loss": 4.2159,
3515
- "step": 583
3516
  },
3517
  {
3518
  "epoch": 0.98,
3519
- "learning_rate": 3.8297872340425525e-05,
3520
- "loss": 4.4105,
3521
- "step": 584
3522
  },
3523
  {
3524
  "epoch": 0.98,
3525
- "learning_rate": 3.51063829787234e-05,
3526
- "loss": 4.3474,
3527
- "step": 585
3528
- },
3529
- {
3530
- "epoch": 0.99,
3531
- "learning_rate": 3.1914893617021275e-05,
3532
- "loss": 4.2935,
3533
- "step": 586
3534
- },
3535
- {
3536
- "epoch": 0.99,
3537
- "learning_rate": 2.8723404255319147e-05,
3538
- "loss": 4.5936,
3539
- "step": 587
3540
- },
3541
- {
3542
- "epoch": 0.99,
3543
- "learning_rate": 2.553191489361702e-05,
3544
- "loss": 4.5503,
3545
- "step": 588
3546
  },
3547
  {
3548
  "epoch": 0.99,
3549
- "learning_rate": 2.234042553191489e-05,
3550
- "loss": 4.4791,
3551
- "step": 589
3552
  },
3553
  {
3554
  "epoch": 0.99,
3555
- "learning_rate": 1.9148936170212762e-05,
3556
- "loss": 4.3574,
3557
- "step": 590
3558
  },
3559
  {
3560
  "epoch": 0.99,
3561
- "learning_rate": 1.5957446808510637e-05,
3562
- "loss": 4.5734,
3563
- "step": 591
3564
- },
3565
- {
3566
- "epoch": 1.0,
3567
- "learning_rate": 1.276595744680851e-05,
3568
- "loss": 5.0126,
3569
- "step": 592
3570
  },
3571
  {
3572
  "epoch": 1.0,
3573
- "learning_rate": 9.574468085106381e-06,
3574
- "loss": 4.6574,
3575
- "step": 593
3576
  },
3577
  {
3578
  "epoch": 1.0,
3579
- "learning_rate": 6.382978723404255e-06,
3580
- "loss": 4.4898,
3581
- "step": 594
3582
  },
3583
  {
3584
  "epoch": 1.0,
3585
- "step": 594,
3586
  "total_flos": 0.0,
3587
- "train_loss": 4.478918995520081,
3588
- "train_runtime": 5719.3723,
3589
- "train_samples_per_second": 4.99,
3590
- "train_steps_per_second": 0.104
3591
  }
3592
  ],
3593
- "max_steps": 594,
3594
  "num_train_epochs": 1,
3595
  "total_flos": 0.0,
3596
  "trial_name": null,
 
2
  "best_metric": null,
3
  "best_model_checkpoint": null,
4
  "epoch": 0.9987389659520807,
5
+ "global_step": 297,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
9
  "log_history": [
10
  {
11
  "epoch": 0.0,
12
+ "learning_rate": 6e-07,
13
+ "loss": 4.7698,
14
  "step": 1
15
  },
16
  {
17
+ "epoch": 0.01,
18
+ "learning_rate": 1.2e-06,
19
+ "loss": 4.8469,
20
  "step": 2
21
  },
22
  {
23
  "epoch": 0.01,
24
+ "learning_rate": 1.8e-06,
25
+ "loss": 4.9086,
26
  "step": 3
27
  },
28
  {
29
  "epoch": 0.01,
30
+ "learning_rate": 2.4e-06,
31
+ "loss": 4.8415,
32
  "step": 4
33
  },
34
  {
35
+ "epoch": 0.02,
36
+ "learning_rate": 2.9999999999999997e-06,
37
+ "loss": 4.7132,
38
  "step": 5
39
  },
40
  {
41
+ "epoch": 0.02,
42
  "learning_rate": 2.9999999999999997e-06,
43
+ "loss": 4.8118,
44
  "step": 6
45
  },
46
  {
47
+ "epoch": 0.02,
48
  "learning_rate": 3.6e-06,
49
+ "loss": 4.7937,
50
  "step": 7
51
  },
52
  {
53
+ "epoch": 0.03,
54
  "learning_rate": 4.2e-06,
55
+ "loss": 4.8404,
56
  "step": 8
57
  },
58
  {
59
+ "epoch": 0.03,
60
  "learning_rate": 4.8e-06,
61
+ "loss": 4.7309,
62
  "step": 9
63
  },
64
  {
65
+ "epoch": 0.03,
66
  "learning_rate": 5.399999999999999e-06,
67
+ "loss": 4.5415,
68
  "step": 10
69
  },
70
  {
71
+ "epoch": 0.04,
72
  "learning_rate": 5.999999999999999e-06,
73
+ "loss": 4.6528,
74
  "step": 11
75
  },
76
  {
77
+ "epoch": 0.04,
78
  "learning_rate": 6.599999999999999e-06,
79
+ "loss": 4.662,
80
  "step": 12
81
  },
82
  {
83
+ "epoch": 0.04,
84
  "learning_rate": 7.2e-06,
85
+ "loss": 4.5529,
86
  "step": 13
87
  },
88
  {
89
+ "epoch": 0.05,
90
  "learning_rate": 7.799999999999998e-06,
91
+ "loss": 4.474,
92
  "step": 14
93
  },
94
  {
95
+ "epoch": 0.05,
96
  "learning_rate": 8.4e-06,
97
+ "loss": 4.4883,
98
  "step": 15
99
  },
100
  {
101
+ "epoch": 0.05,
102
  "learning_rate": 8.999999999999999e-06,
103
+ "loss": 4.4004,
104
  "step": 16
105
  },
106
  {
107
+ "epoch": 0.06,
108
  "learning_rate": 9.6e-06,
109
+ "loss": 4.4632,
110
  "step": 17
111
  },
112
  {
113
+ "epoch": 0.06,
114
  "learning_rate": 1.02e-05,
115
+ "loss": 4.4696,
116
  "step": 18
117
  },
118
  {
119
+ "epoch": 0.06,
120
  "learning_rate": 1.0799999999999998e-05,
121
+ "loss": 4.4129,
122
  "step": 19
123
  },
124
  {
125
+ "epoch": 0.07,
126
  "learning_rate": 1.14e-05,
127
+ "loss": 4.4643,
128
  "step": 20
129
  },
130
  {
131
+ "epoch": 0.07,
132
  "learning_rate": 1.1999999999999999e-05,
133
+ "loss": 4.4646,
134
  "step": 21
135
  },
136
  {
137
+ "epoch": 0.07,
138
  "learning_rate": 1.26e-05,
139
+ "loss": 4.4311,
140
  "step": 22
141
  },
142
  {
143
+ "epoch": 0.08,
144
  "learning_rate": 1.3199999999999997e-05,
145
+ "loss": 4.306,
146
  "step": 23
147
  },
148
  {
149
+ "epoch": 0.08,
150
  "learning_rate": 1.3799999999999998e-05,
151
+ "loss": 4.4385,
152
  "step": 24
153
  },
154
  {
155
+ "epoch": 0.08,
156
  "learning_rate": 1.44e-05,
157
+ "loss": 4.276,
158
  "step": 25
159
  },
160
  {
161
+ "epoch": 0.09,
162
  "learning_rate": 1.4999999999999999e-05,
163
+ "loss": 4.5521,
164
  "step": 26
165
  },
166
  {
167
+ "epoch": 0.09,
168
  "learning_rate": 1.5599999999999996e-05,
169
+ "loss": 4.3171,
170
  "step": 27
171
  },
172
  {
173
+ "epoch": 0.09,
174
  "learning_rate": 1.6199999999999997e-05,
175
+ "loss": 4.2827,
176
  "step": 28
177
  },
178
  {
179
+ "epoch": 0.1,
180
  "learning_rate": 1.68e-05,
181
+ "loss": 4.1956,
182
  "step": 29
183
  },
184
  {
185
+ "epoch": 0.1,
186
  "learning_rate": 1.74e-05,
187
+ "loss": 4.3854,
188
  "step": 30
189
  },
190
  {
191
+ "epoch": 0.1,
192
  "learning_rate": 1.7999999999999997e-05,
193
+ "loss": 4.3831,
194
  "step": 31
195
  },
196
  {
197
+ "epoch": 0.11,
198
  "learning_rate": 1.8599999999999998e-05,
199
+ "loss": 4.2882,
200
  "step": 32
201
  },
202
  {
203
+ "epoch": 0.11,
204
  "learning_rate": 1.92e-05,
205
+ "loss": 4.3278,
206
  "step": 33
207
  },
208
  {
209
+ "epoch": 0.11,
210
  "learning_rate": 1.98e-05,
211
+ "loss": 4.2851,
212
  "step": 34
213
  },
214
  {
215
+ "epoch": 0.12,
216
  "learning_rate": 2.04e-05,
217
+ "loss": 4.2957,
218
  "step": 35
219
  },
220
  {
221
+ "epoch": 0.12,
222
  "learning_rate": 2.1e-05,
223
+ "loss": 4.2894,
224
  "step": 36
225
  },
226
  {
227
+ "epoch": 0.12,
228
  "learning_rate": 2.1599999999999996e-05,
229
+ "loss": 4.3836,
230
  "step": 37
231
  },
232
  {
233
+ "epoch": 0.13,
234
  "learning_rate": 2.2199999999999998e-05,
235
+ "loss": 4.3498,
236
  "step": 38
237
  },
238
  {
239
+ "epoch": 0.13,
240
  "learning_rate": 2.28e-05,
241
+ "loss": 4.3839,
242
  "step": 39
243
  },
244
  {
245
+ "epoch": 0.13,
246
  "learning_rate": 2.34e-05,
247
+ "loss": 4.2473,
248
  "step": 40
249
  },
250
  {
251
+ "epoch": 0.14,
252
  "learning_rate": 2.3999999999999997e-05,
253
+ "loss": 4.3261,
254
  "step": 41
255
  },
256
  {
257
+ "epoch": 0.14,
258
  "learning_rate": 2.4599999999999998e-05,
259
+ "loss": 4.3655,
260
  "step": 42
261
  },
262
  {
263
+ "epoch": 0.14,
264
  "learning_rate": 2.52e-05,
265
+ "loss": 4.1703,
266
  "step": 43
267
  },
268
  {
269
+ "epoch": 0.15,
270
  "learning_rate": 2.5799999999999997e-05,
271
+ "loss": 4.468,
272
  "step": 44
273
  },
274
  {
275
+ "epoch": 0.15,
276
  "learning_rate": 2.6399999999999995e-05,
277
+ "loss": 4.4049,
278
  "step": 45
279
  },
280
  {
281
+ "epoch": 0.15,
282
  "learning_rate": 2.6999999999999996e-05,
283
+ "loss": 4.4439,
284
  "step": 46
285
  },
286
  {
287
+ "epoch": 0.16,
288
  "learning_rate": 2.7599999999999997e-05,
289
+ "loss": 4.4014,
290
  "step": 47
291
  },
292
  {
293
+ "epoch": 0.16,
294
  "learning_rate": 2.8199999999999998e-05,
295
+ "loss": 4.5049,
296
  "step": 48
297
  },
298
  {
299
+ "epoch": 0.16,
300
+ "learning_rate": 2.88e-05,
301
+ "loss": 4.5962,
302
  "step": 49
303
  },
304
  {
305
+ "epoch": 0.17,
306
+ "learning_rate": 2.94e-05,
307
+ "loss": 4.3296,
308
  "step": 50
309
  },
310
  {
311
+ "epoch": 0.17,
312
+ "learning_rate": 2.9999999999999997e-05,
313
+ "loss": 4.2871,
314
  "step": 51
315
  },
316
  {
317
+ "epoch": 0.17,
318
+ "learning_rate": 3.06e-05,
319
+ "loss": 4.1845,
320
  "step": 52
321
  },
322
  {
323
+ "epoch": 0.18,
324
+ "learning_rate": 3.119999999999999e-05,
325
+ "loss": 4.163,
326
  "step": 53
327
  },
328
  {
329
+ "epoch": 0.18,
330
+ "learning_rate": 3.1799999999999994e-05,
331
+ "loss": 4.1515,
332
  "step": 54
333
  },
334
  {
335
+ "epoch": 0.18,
336
+ "learning_rate": 3.2399999999999995e-05,
337
+ "loss": 4.1247,
338
  "step": 55
339
  },
340
  {
341
+ "epoch": 0.19,
342
+ "learning_rate": 3.2999999999999996e-05,
343
+ "loss": 4.1402,
344
  "step": 56
345
  },
346
  {
347
+ "epoch": 0.19,
348
+ "learning_rate": 3.36e-05,
349
+ "loss": 4.2412,
350
  "step": 57
351
  },
352
  {
353
+ "epoch": 0.2,
354
+ "learning_rate": 3.42e-05,
355
+ "loss": 4.2467,
356
  "step": 58
357
  },
358
  {
359
+ "epoch": 0.2,
360
+ "learning_rate": 3.48e-05,
361
+ "loss": 4.2223,
362
  "step": 59
363
  },
364
  {
365
+ "epoch": 0.2,
366
+ "learning_rate": 3.539999999999999e-05,
367
+ "loss": 4.1693,
368
  "step": 60
369
  },
370
  {
371
+ "epoch": 0.21,
372
+ "learning_rate": 3.5999999999999994e-05,
373
+ "loss": 4.2163,
374
  "step": 61
375
  },
376
  {
377
+ "epoch": 0.21,
378
+ "learning_rate": 3.6599999999999995e-05,
379
+ "loss": 4.1301,
380
  "step": 62
381
  },
382
  {
383
+ "epoch": 0.21,
384
+ "learning_rate": 3.7199999999999996e-05,
385
+ "loss": 4.2289,
386
  "step": 63
387
  },
388
  {
389
+ "epoch": 0.22,
390
+ "learning_rate": 3.78e-05,
391
+ "loss": 4.1746,
392
  "step": 64
393
  },
394
  {
395
+ "epoch": 0.22,
396
+ "learning_rate": 3.84e-05,
397
+ "loss": 4.2473,
398
  "step": 65
399
  },
400
  {
401
+ "epoch": 0.22,
402
+ "learning_rate": 3.9e-05,
403
+ "loss": 4.2665,
404
  "step": 66
405
  },
406
  {
407
+ "epoch": 0.23,
408
+ "learning_rate": 3.96e-05,
409
+ "loss": 4.2199,
410
  "step": 67
411
  },
412
  {
413
+ "epoch": 0.23,
414
+ "learning_rate": 4.02e-05,
415
+ "loss": 4.246,
416
  "step": 68
417
  },
418
  {
419
+ "epoch": 0.23,
420
+ "learning_rate": 4.08e-05,
421
+ "loss": 4.1884,
422
  "step": 69
423
  },
424
  {
425
+ "epoch": 0.24,
426
+ "learning_rate": 4.14e-05,
427
+ "loss": 4.1402,
428
  "step": 70
429
  },
430
  {
431
+ "epoch": 0.24,
432
+ "learning_rate": 4.2e-05,
433
+ "loss": 4.1451,
434
  "step": 71
435
  },
436
  {
437
+ "epoch": 0.24,
438
+ "learning_rate": 4.259999999999999e-05,
439
+ "loss": 4.2652,
440
  "step": 72
441
  },
442
  {
443
+ "epoch": 0.25,
444
+ "learning_rate": 4.319999999999999e-05,
445
+ "loss": 4.1772,
446
  "step": 73
447
  },
448
  {
449
+ "epoch": 0.25,
450
+ "learning_rate": 4.3799999999999994e-05,
451
+ "loss": 4.1788,
452
  "step": 74
453
  },
454
  {
455
+ "epoch": 0.25,
456
+ "learning_rate": 4.4399999999999995e-05,
457
+ "loss": 4.2676,
458
  "step": 75
459
  },
460
  {
461
+ "epoch": 0.26,
462
+ "learning_rate": 4.4999999999999996e-05,
463
+ "loss": 4.1542,
464
  "step": 76
465
  },
466
  {
467
+ "epoch": 0.26,
468
+ "learning_rate": 4.56e-05,
469
+ "loss": 4.1175,
470
  "step": 77
471
  },
472
  {
473
+ "epoch": 0.26,
474
+ "learning_rate": 4.62e-05,
475
+ "loss": 4.1771,
476
  "step": 78
477
  },
478
  {
479
+ "epoch": 0.27,
480
+ "learning_rate": 4.68e-05,
481
+ "loss": 4.0612,
482
  "step": 79
483
  },
484
  {
485
+ "epoch": 0.27,
486
+ "learning_rate": 4.7399999999999993e-05,
487
+ "loss": 4.2617,
488
  "step": 80
489
  },
490
  {
491
+ "epoch": 0.27,
492
+ "learning_rate": 4.7999999999999994e-05,
493
+ "loss": 4.1813,
494
  "step": 81
495
  },
496
  {
497
+ "epoch": 0.28,
498
+ "learning_rate": 4.8599999999999995e-05,
499
+ "loss": 4.1929,
500
  "step": 82
501
  },
502
  {
503
+ "epoch": 0.28,
504
+ "learning_rate": 4.9199999999999997e-05,
505
+ "loss": 4.1566,
506
  "step": 83
507
  },
508
  {
509
+ "epoch": 0.28,
510
+ "learning_rate": 4.98e-05,
511
+ "loss": 4.2066,
512
  "step": 84
513
  },
514
  {
515
+ "epoch": 0.29,
516
+ "learning_rate": 5.04e-05,
517
+ "loss": 4.297,
518
  "step": 85
519
  },
520
  {
521
+ "epoch": 0.29,
522
+ "learning_rate": 5.1e-05,
523
+ "loss": 4.1838,
524
  "step": 86
525
  },
526
  {
527
+ "epoch": 0.29,
528
+ "learning_rate": 5.1599999999999994e-05,
529
+ "loss": 4.2046,
530
  "step": 87
531
  },
532
  {
533
+ "epoch": 0.3,
534
+ "learning_rate": 5.2199999999999995e-05,
535
+ "loss": 4.2594,
536
  "step": 88
537
  },
538
  {
539
+ "epoch": 0.3,
540
+ "learning_rate": 5.279999999999999e-05,
541
+ "loss": 4.2559,
542
  "step": 89
543
  },
544
  {
545
+ "epoch": 0.3,
546
+ "learning_rate": 5.339999999999999e-05,
547
+ "loss": 4.2763,
548
  "step": 90
549
  },
550
  {
551
+ "epoch": 0.31,
552
+ "learning_rate": 5.399999999999999e-05,
553
+ "loss": 4.2853,
554
  "step": 91
555
  },
556
  {
557
+ "epoch": 0.31,
558
+ "learning_rate": 5.459999999999999e-05,
559
+ "loss": 4.3723,
560
  "step": 92
561
  },
562
  {
563
+ "epoch": 0.31,
564
+ "learning_rate": 5.519999999999999e-05,
565
+ "loss": 4.333,
566
  "step": 93
567
  },
568
  {
569
+ "epoch": 0.32,
570
+ "learning_rate": 5.5799999999999994e-05,
571
+ "loss": 4.2612,
572
  "step": 94
573
  },
574
  {
575
+ "epoch": 0.32,
576
+ "learning_rate": 5.6399999999999995e-05,
577
+ "loss": 4.2223,
578
  "step": 95
579
  },
580
  {
581
+ "epoch": 0.32,
582
+ "learning_rate": 5.6999999999999996e-05,
583
+ "loss": 4.3846,
584
  "step": 96
585
  },
586
  {
587
+ "epoch": 0.33,
588
+ "learning_rate": 5.76e-05,
589
+ "loss": 4.5903,
590
  "step": 97
591
  },
592
  {
593
+ "epoch": 0.33,
594
+ "learning_rate": 5.82e-05,
595
+ "loss": 4.4963,
596
  "step": 98
597
  },
598
  {
599
+ "epoch": 0.33,
600
+ "learning_rate": 5.88e-05,
601
+ "loss": 4.5083,
602
  "step": 99
603
  },
604
  {
605
+ "epoch": 0.34,
606
+ "learning_rate": 5.94e-05,
607
+ "loss": 4.5454,
608
  "step": 100
609
  },
610
  {
611
+ "epoch": 0.34,
612
+ "learning_rate": 5.9999999999999995e-05,
613
+ "loss": 4.146,
614
  "step": 101
615
  },
616
  {
617
+ "epoch": 0.34,
618
+ "learning_rate": 6.0599999999999996e-05,
619
+ "loss": 4.2164,
620
  "step": 102
621
  },
622
  {
623
+ "epoch": 0.35,
624
+ "learning_rate": 6.12e-05,
625
+ "loss": 4.2249,
626
  "step": 103
627
  },
628
  {
629
+ "epoch": 0.35,
630
+ "learning_rate": 6.18e-05,
631
+ "loss": 4.1094,
632
  "step": 104
633
  },
634
  {
635
+ "epoch": 0.35,
636
+ "learning_rate": 6.239999999999999e-05,
637
+ "loss": 4.2254,
638
  "step": 105
639
  },
640
  {
641
+ "epoch": 0.36,
642
+ "learning_rate": 6.299999999999999e-05,
643
+ "loss": 4.1199,
644
  "step": 106
645
  },
646
  {
647
+ "epoch": 0.36,
648
+ "learning_rate": 6.359999999999999e-05,
649
+ "loss": 4.0813,
650
  "step": 107
651
  },
652
  {
653
+ "epoch": 0.36,
654
+ "learning_rate": 6.419999999999999e-05,
655
+ "loss": 4.1843,
656
  "step": 108
657
  },
658
  {
659
+ "epoch": 0.37,
660
+ "learning_rate": 6.479999999999999e-05,
661
+ "loss": 4.2144,
662
  "step": 109
663
  },
664
  {
665
+ "epoch": 0.37,
666
+ "learning_rate": 6.539999999999999e-05,
667
+ "loss": 4.1853,
668
  "step": 110
669
  },
670
  {
671
+ "epoch": 0.37,
672
+ "learning_rate": 6.599999999999999e-05,
673
+ "loss": 4.256,
674
  "step": 111
675
  },
676
  {
677
+ "epoch": 0.38,
678
+ "learning_rate": 6.659999999999999e-05,
679
+ "loss": 4.1764,
680
  "step": 112
681
  },
682
  {
683
+ "epoch": 0.38,
684
+ "learning_rate": 6.72e-05,
685
+ "loss": 4.1245,
686
  "step": 113
687
  },
688
  {
689
+ "epoch": 0.38,
690
+ "learning_rate": 6.78e-05,
691
+ "loss": 4.1798,
692
  "step": 114
693
  },
694
  {
695
+ "epoch": 0.39,
696
+ "learning_rate": 6.84e-05,
697
+ "loss": 4.1431,
698
  "step": 115
699
  },
700
  {
701
+ "epoch": 0.39,
702
+ "learning_rate": 6.9e-05,
703
+ "loss": 4.1608,
704
  "step": 116
705
  },
706
  {
707
+ "epoch": 0.39,
708
+ "learning_rate": 6.96e-05,
709
+ "loss": 4.09,
710
  "step": 117
711
  },
712
  {
713
+ "epoch": 0.4,
714
+ "learning_rate": 7.02e-05,
715
+ "loss": 4.1146,
716
  "step": 118
717
  },
718
  {
719
+ "epoch": 0.4,
720
+ "learning_rate": 7.079999999999999e-05,
721
+ "loss": 4.1764,
722
  "step": 119
723
  },
724
  {
725
+ "epoch": 0.4,
726
+ "learning_rate": 7.139999999999999e-05,
727
+ "loss": 4.181,
728
  "step": 120
729
  },
730
  {
731
+ "epoch": 0.41,
732
+ "learning_rate": 7.199999999999999e-05,
733
+ "loss": 4.1816,
734
  "step": 121
735
  },
736
  {
737
+ "epoch": 0.41,
738
+ "learning_rate": 7.259999999999999e-05,
739
+ "loss": 4.2696,
740
  "step": 122
741
  },
742
  {
743
+ "epoch": 0.41,
744
+ "learning_rate": 7.319999999999999e-05,
745
+ "loss": 4.1079,
746
  "step": 123
747
  },
748
  {
749
+ "epoch": 0.42,
750
+ "learning_rate": 7.379999999999999e-05,
751
+ "loss": 4.2193,
752
  "step": 124
753
  },
754
  {
755
+ "epoch": 0.42,
756
+ "learning_rate": 7.439999999999999e-05,
757
+ "loss": 4.2735,
758
  "step": 125
759
  },
760
  {
761
+ "epoch": 0.42,
762
+ "learning_rate": 7.5e-05,
763
+ "loss": 4.1796,
764
  "step": 126
765
  },
766
  {
767
+ "epoch": 0.43,
768
+ "learning_rate": 7.56e-05,
769
+ "loss": 4.2039,
770
  "step": 127
771
  },
772
  {
773
+ "epoch": 0.43,
774
+ "learning_rate": 7.62e-05,
775
+ "loss": 4.1144,
776
  "step": 128
777
  },
778
  {
779
+ "epoch": 0.43,
780
+ "learning_rate": 7.68e-05,
781
+ "loss": 4.2619,
782
  "step": 129
783
  },
784
  {
785
+ "epoch": 0.44,
786
+ "learning_rate": 7.74e-05,
787
+ "loss": 4.3677,
788
  "step": 130
789
  },
790
  {
791
+ "epoch": 0.44,
792
+ "learning_rate": 7.8e-05,
793
+ "loss": 4.2429,
794
  "step": 131
795
  },
796
  {
797
+ "epoch": 0.44,
798
+ "learning_rate": 7.86e-05,
799
+ "loss": 4.1074,
800
  "step": 132
801
  },
802
  {
803
+ "epoch": 0.45,
804
+ "learning_rate": 7.92e-05,
805
+ "loss": 4.2802,
806
  "step": 133
807
  },
808
  {
809
+ "epoch": 0.45,
810
+ "learning_rate": 7.98e-05,
811
+ "loss": 4.1459,
812
  "step": 134
813
  },
814
  {
815
+ "epoch": 0.45,
816
+ "learning_rate": 8.04e-05,
817
+ "loss": 4.3063,
818
  "step": 135
819
  },
820
  {
821
+ "epoch": 0.46,
822
+ "learning_rate": 8.1e-05,
823
+ "loss": 4.2316,
824
  "step": 136
825
  },
826
  {
827
+ "epoch": 0.46,
828
+ "learning_rate": 8.16e-05,
829
+ "loss": 4.1519,
830
  "step": 137
831
  },
832
  {
833
+ "epoch": 0.46,
834
+ "learning_rate": 8.22e-05,
835
+ "loss": 4.0775,
836
  "step": 138
837
  },
838
  {
839
+ "epoch": 0.47,
840
+ "learning_rate": 8.28e-05,
841
+ "loss": 4.1537,
842
  "step": 139
843
  },
844
  {
845
+ "epoch": 0.47,
846
+ "learning_rate": 8.34e-05,
847
+ "loss": 4.1356,
848
  "step": 140
849
  },
850
  {
851
+ "epoch": 0.47,
852
+ "learning_rate": 8.4e-05,
853
+ "loss": 4.3637,
854
  "step": 141
855
  },
856
  {
857
+ "epoch": 0.48,
858
+ "learning_rate": 8.459999999999998e-05,
859
+ "loss": 4.156,
860
  "step": 142
861
  },
862
  {
863
+ "epoch": 0.48,
864
+ "learning_rate": 8.519999999999998e-05,
865
+ "loss": 4.2406,
866
  "step": 143
867
  },
868
  {
869
+ "epoch": 0.48,
870
+ "learning_rate": 8.579999999999998e-05,
871
+ "loss": 4.3063,
872
  "step": 144
873
  },
874
  {
875
+ "epoch": 0.49,
876
+ "learning_rate": 8.639999999999999e-05,
877
+ "loss": 4.4181,
878
  "step": 145
879
  },
880
  {
881
+ "epoch": 0.49,
882
+ "learning_rate": 8.699999999999999e-05,
883
+ "loss": 4.173,
884
  "step": 146
885
  },
886
  {
887
+ "epoch": 0.49,
888
+ "learning_rate": 8.759999999999999e-05,
889
+ "loss": 4.4079,
890
  "step": 147
891
  },
892
  {
893
+ "epoch": 0.5,
894
+ "learning_rate": 8.819999999999999e-05,
895
+ "loss": 4.4766,
896
  "step": 148
897
  },
898
  {
899
+ "epoch": 0.5,
900
+ "learning_rate": 8.879999999999999e-05,
901
+ "loss": 4.4345,
902
  "step": 149
903
  },
904
  {
905
+ "epoch": 0.5,
906
+ "learning_rate": 8.939999999999999e-05,
907
+ "loss": 4.5567,
908
  "step": 150
909
  },
910
  {
911
+ "epoch": 0.51,
912
+ "learning_rate": 8.999999999999999e-05,
913
+ "loss": 4.2932,
914
  "step": 151
915
  },
916
  {
917
+ "epoch": 0.51,
918
+ "learning_rate": 9.059999999999999e-05,
919
+ "loss": 4.2495,
920
  "step": 152
921
  },
922
  {
923
+ "epoch": 0.51,
924
+ "learning_rate": 9.12e-05,
925
+ "loss": 4.2386,
926
  "step": 153
927
  },
928
  {
929
+ "epoch": 0.52,
930
+ "learning_rate": 9.18e-05,
931
+ "loss": 4.1652,
932
  "step": 154
933
  },
934
  {
935
+ "epoch": 0.52,
936
+ "learning_rate": 9.24e-05,
937
+ "loss": 4.1926,
938
  "step": 155
939
  },
940
  {
941
+ "epoch": 0.52,
942
+ "learning_rate": 9.3e-05,
943
+ "loss": 4.2151,
944
  "step": 156
945
  },
946
  {
947
+ "epoch": 0.53,
948
+ "learning_rate": 9.36e-05,
949
+ "loss": 4.1392,
950
  "step": 157
951
  },
952
  {
953
+ "epoch": 0.53,
954
+ "learning_rate": 9.419999999999999e-05,
955
+ "loss": 4.165,
956
  "step": 158
957
  },
958
  {
959
+ "epoch": 0.53,
960
+ "learning_rate": 9.479999999999999e-05,
961
+ "loss": 4.2136,
962
  "step": 159
963
  },
964
  {
965
+ "epoch": 0.54,
966
+ "learning_rate": 9.539999999999999e-05,
967
+ "loss": 4.1931,
968
  "step": 160
969
  },
970
  {
971
+ "epoch": 0.54,
972
+ "learning_rate": 9.599999999999999e-05,
973
+ "loss": 4.2008,
974
  "step": 161
975
  },
976
  {
977
+ "epoch": 0.54,
978
+ "learning_rate": 9.659999999999999e-05,
979
+ "loss": 4.1198,
980
  "step": 162
981
  },
982
  {
983
+ "epoch": 0.55,
984
+ "learning_rate": 9.719999999999999e-05,
985
+ "loss": 4.2464,
986
  "step": 163
987
  },
988
  {
989
+ "epoch": 0.55,
990
+ "learning_rate": 9.779999999999999e-05,
991
+ "loss": 4.1406,
992
  "step": 164
993
  },
994
  {
995
+ "epoch": 0.55,
996
+ "learning_rate": 9.839999999999999e-05,
997
+ "loss": 4.2095,
998
  "step": 165
999
  },
1000
  {
1001
+ "epoch": 0.56,
1002
+ "learning_rate": 9.9e-05,
1003
+ "loss": 4.1875,
1004
  "step": 166
1005
  },
1006
  {
1007
+ "epoch": 0.56,
1008
+ "learning_rate": 9.96e-05,
1009
+ "loss": 4.152,
1010
  "step": 167
1011
  },
1012
  {
1013
+ "epoch": 0.56,
1014
+ "learning_rate": 0.0001002,
1015
+ "loss": 4.1881,
1016
  "step": 168
1017
  },
1018
  {
1019
+ "epoch": 0.57,
1020
+ "learning_rate": 0.0001008,
1021
+ "loss": 4.2114,
1022
  "step": 169
1023
  },
1024
  {
1025
+ "epoch": 0.57,
1026
+ "learning_rate": 0.0001014,
1027
+ "loss": 4.1061,
1028
  "step": 170
1029
  },
1030
  {
1031
+ "epoch": 0.58,
1032
+ "learning_rate": 0.000102,
1033
+ "loss": 4.1188,
1034
  "step": 171
1035
  },
1036
  {
1037
+ "epoch": 0.58,
1038
+ "learning_rate": 0.0001026,
1039
+ "loss": 4.1931,
1040
  "step": 172
1041
  },
1042
  {
1043
+ "epoch": 0.58,
1044
+ "learning_rate": 0.00010319999999999999,
1045
+ "loss": 4.0806,
1046
  "step": 173
1047
  },
1048
  {
1049
+ "epoch": 0.59,
1050
+ "learning_rate": 0.00010379999999999999,
1051
+ "loss": 4.1139,
1052
  "step": 174
1053
  },
1054
  {
1055
+ "epoch": 0.59,
1056
+ "learning_rate": 0.00010439999999999999,
1057
+ "loss": 4.3033,
1058
  "step": 175
1059
  },
1060
  {
1061
+ "epoch": 0.59,
1062
+ "learning_rate": 0.00010499999999999999,
1063
+ "loss": 4.1447,
1064
  "step": 176
1065
  },
1066
  {
1067
+ "epoch": 0.6,
1068
+ "learning_rate": 0.00010559999999999998,
1069
+ "loss": 4.2387,
1070
  "step": 177
1071
  },
1072
  {
1073
+ "epoch": 0.6,
1074
+ "learning_rate": 0.00010619999999999998,
1075
+ "loss": 4.1149,
1076
  "step": 178
1077
  },
1078
  {
1079
+ "epoch": 0.6,
1080
+ "learning_rate": 0.00010679999999999998,
1081
+ "loss": 4.1095,
1082
  "step": 179
1083
  },
1084
  {
1085
+ "epoch": 0.61,
1086
+ "learning_rate": 0.00010739999999999998,
1087
+ "loss": 4.3137,
1088
  "step": 180
1089
  },
1090
  {
1091
+ "epoch": 0.61,
1092
+ "learning_rate": 0.00010799999999999998,
1093
+ "loss": 4.1593,
1094
  "step": 181
1095
  },
1096
  {
1097
+ "epoch": 0.61,
1098
+ "learning_rate": 0.00010859999999999998,
1099
+ "loss": 4.0903,
1100
  "step": 182
1101
  },
1102
  {
1103
+ "epoch": 0.62,
1104
+ "learning_rate": 0.00010919999999999998,
1105
+ "loss": 4.1328,
1106
  "step": 183
1107
  },
1108
  {
1109
+ "epoch": 0.62,
1110
+ "learning_rate": 0.00010979999999999999,
1111
+ "loss": 4.1959,
1112
  "step": 184
1113
  },
1114
  {
1115
+ "epoch": 0.62,
1116
+ "learning_rate": 0.00011039999999999999,
1117
+ "loss": 4.2055,
1118
  "step": 185
1119
  },
1120
  {
1121
+ "epoch": 0.63,
1122
+ "learning_rate": 0.00011099999999999999,
1123
+ "loss": 4.1536,
1124
  "step": 186
1125
  },
1126
  {
1127
+ "epoch": 0.63,
1128
+ "learning_rate": 0.00011159999999999999,
1129
+ "loss": 4.195,
1130
  "step": 187
1131
  },
1132
  {
1133
+ "epoch": 0.63,
1134
+ "learning_rate": 0.00011219999999999999,
1135
+ "loss": 4.1625,
1136
  "step": 188
1137
  },
1138
  {
1139
+ "epoch": 0.64,
1140
+ "learning_rate": 0.00011279999999999999,
1141
+ "loss": 4.1326,
1142
  "step": 189
1143
  },
1144
  {
1145
+ "epoch": 0.64,
1146
+ "learning_rate": 0.00011339999999999999,
1147
+ "loss": 4.2294,
1148
  "step": 190
1149
  },
1150
  {
1151
+ "epoch": 0.64,
1152
+ "learning_rate": 0.00011399999999999999,
1153
+ "loss": 4.3466,
1154
  "step": 191
1155
  },
1156
  {
1157
+ "epoch": 0.65,
1158
+ "learning_rate": 0.0001146,
1159
+ "loss": 4.3242,
1160
  "step": 192
1161
  },
1162
  {
1163
+ "epoch": 0.65,
1164
+ "learning_rate": 0.0001152,
1165
+ "loss": 4.3581,
1166
  "step": 193
1167
  },
1168
  {
1169
+ "epoch": 0.65,
1170
+ "learning_rate": 0.0001158,
1171
+ "loss": 4.3301,
1172
  "step": 194
1173
  },
1174
  {
1175
+ "epoch": 0.66,
1176
+ "learning_rate": 0.0001164,
1177
+ "loss": 4.2756,
1178
  "step": 195
1179
  },
1180
  {
1181
+ "epoch": 0.66,
1182
+ "learning_rate": 0.000117,
1183
+ "loss": 4.3115,
1184
  "step": 196
1185
  },
1186
  {
1187
+ "epoch": 0.66,
1188
+ "learning_rate": 0.0001176,
1189
+ "loss": 4.1626,
1190
  "step": 197
1191
  },
1192
  {
1193
+ "epoch": 0.67,
1194
+ "learning_rate": 0.0001182,
1195
+ "loss": 4.6198,
1196
  "step": 198
1197
  },
1198
  {
1199
+ "epoch": 0.67,
1200
+ "learning_rate": 0.0001188,
1201
+ "loss": 4.4561,
1202
  "step": 199
1203
  },
1204
  {
1205
+ "epoch": 0.67,
1206
+ "learning_rate": 0.0001194,
1207
+ "loss": 4.4628,
1208
  "step": 200
1209
  },
1210
  {
1211
+ "epoch": 0.68,
1212
+ "learning_rate": 0.00011999999999999999,
1213
+ "loss": 4.2635,
1214
  "step": 201
1215
  },
1216
  {
1217
+ "epoch": 0.68,
1218
+ "learning_rate": 0.00012059999999999999,
1219
+ "loss": 4.2424,
1220
  "step": 202
1221
  },
1222
  {
1223
+ "epoch": 0.68,
1224
+ "learning_rate": 0.00012119999999999999,
1225
+ "loss": 4.1895,
1226
  "step": 203
1227
  },
1228
  {
1229
+ "epoch": 0.69,
1230
+ "learning_rate": 0.00012179999999999999,
1231
+ "loss": 4.133,
1232
  "step": 204
1233
  },
1234
  {
1235
+ "epoch": 0.69,
1236
+ "learning_rate": 0.0001224,
1237
+ "loss": 4.1853,
1238
  "step": 205
1239
  },
1240
  {
1241
+ "epoch": 0.69,
1242
+ "learning_rate": 0.00012299999999999998,
1243
+ "loss": 4.1186,
1244
  "step": 206
1245
  },
1246
  {
1247
+ "epoch": 0.7,
1248
+ "learning_rate": 0.0001236,
1249
+ "loss": 4.3218,
1250
  "step": 207
1251
  },
1252
  {
1253
+ "epoch": 0.7,
1254
+ "learning_rate": 0.00012419999999999998,
1255
+ "loss": 4.19,
1256
  "step": 208
1257
  },
1258
  {
1259
+ "epoch": 0.7,
1260
+ "learning_rate": 0.00012479999999999997,
1261
+ "loss": 4.1701,
1262
  "step": 209
1263
  },
1264
  {
1265
+ "epoch": 0.71,
1266
+ "learning_rate": 0.00012539999999999999,
1267
+ "loss": 4.2564,
1268
  "step": 210
1269
  },
1270
  {
1271
+ "epoch": 0.71,
1272
+ "learning_rate": 0.00012599999999999997,
1273
+ "loss": 4.0469,
1274
  "step": 211
1275
  },
1276
  {
1277
+ "epoch": 0.71,
1278
+ "learning_rate": 0.0001266,
1279
+ "loss": 4.1559,
1280
  "step": 212
1281
  },
1282
  {
1283
+ "epoch": 0.72,
1284
+ "learning_rate": 0.00012719999999999997,
1285
+ "loss": 4.3674,
1286
  "step": 213
1287
  },
1288
  {
1289
+ "epoch": 0.72,
1290
+ "learning_rate": 0.0001278,
1291
+ "loss": 4.172,
1292
  "step": 214
1293
  },
1294
  {
1295
+ "epoch": 0.72,
1296
+ "learning_rate": 0.00012839999999999998,
1297
+ "loss": 4.2285,
1298
  "step": 215
1299
  },
1300
  {
1301
+ "epoch": 0.73,
1302
+ "learning_rate": 0.000129,
1303
+ "loss": 4.182,
1304
  "step": 216
1305
  },
1306
  {
1307
+ "epoch": 0.73,
1308
+ "learning_rate": 0.00012959999999999998,
1309
+ "loss": 4.1601,
1310
  "step": 217
1311
  },
1312
  {
1313
+ "epoch": 0.73,
1314
+ "learning_rate": 0.0001302,
1315
+ "loss": 4.0804,
1316
  "step": 218
1317
  },
1318
  {
1319
+ "epoch": 0.74,
1320
+ "learning_rate": 0.00013079999999999998,
1321
+ "loss": 4.0997,
1322
  "step": 219
1323
  },
1324
  {
1325
+ "epoch": 0.74,
1326
+ "learning_rate": 0.0001314,
1327
+ "loss": 4.1931,
1328
  "step": 220
1329
  },
1330
  {
1331
+ "epoch": 0.74,
1332
+ "learning_rate": 0.00013199999999999998,
1333
+ "loss": 4.1306,
1334
  "step": 221
1335
  },
1336
  {
1337
+ "epoch": 0.75,
1338
+ "learning_rate": 0.0001326,
1339
+ "loss": 4.0679,
1340
  "step": 222
1341
  },
1342
  {
1343
+ "epoch": 0.75,
1344
+ "learning_rate": 0.00013319999999999999,
1345
+ "loss": 4.1208,
1346
  "step": 223
1347
  },
1348
  {
1349
+ "epoch": 0.75,
1350
+ "learning_rate": 0.0001338,
1351
+ "loss": 4.1506,
1352
  "step": 224
1353
  },
1354
  {
1355
+ "epoch": 0.76,
1356
+ "learning_rate": 0.0001344,
1357
+ "loss": 4.1266,
1358
  "step": 225
1359
  },
1360
  {
1361
+ "epoch": 0.76,
1362
+ "learning_rate": 0.000135,
1363
+ "loss": 4.2373,
1364
  "step": 226
1365
  },
1366
  {
1367
+ "epoch": 0.76,
1368
+ "learning_rate": 0.0001356,
1369
+ "loss": 4.1706,
1370
  "step": 227
1371
  },
1372
  {
1373
+ "epoch": 0.77,
1374
+ "learning_rate": 0.0001362,
1375
+ "loss": 4.063,
1376
  "step": 228
1377
  },
1378
  {
1379
+ "epoch": 0.77,
1380
+ "learning_rate": 0.0001368,
1381
+ "loss": 4.1355,
1382
  "step": 229
1383
  },
1384
  {
1385
+ "epoch": 0.77,
1386
+ "learning_rate": 0.0001374,
1387
+ "loss": 4.1105,
1388
  "step": 230
1389
  },
1390
  {
1391
+ "epoch": 0.78,
1392
+ "learning_rate": 0.000138,
1393
+ "loss": 4.2918,
1394
  "step": 231
1395
  },
1396
  {
1397
+ "epoch": 0.78,
1398
+ "learning_rate": 0.0001386,
1399
+ "loss": 4.1313,
1400
  "step": 232
1401
  },
1402
  {
1403
+ "epoch": 0.78,
1404
+ "learning_rate": 0.0001392,
1405
+ "loss": 4.2142,
1406
  "step": 233
1407
  },
1408
  {
1409
+ "epoch": 0.79,
1410
+ "learning_rate": 0.00013979999999999998,
1411
+ "loss": 4.2723,
1412
  "step": 234
1413
  },
1414
  {
1415
+ "epoch": 0.79,
1416
+ "learning_rate": 0.0001404,
1417
+ "loss": 4.24,
1418
  "step": 235
1419
  },
1420
  {
1421
+ "epoch": 0.79,
1422
+ "learning_rate": 0.00014099999999999998,
1423
+ "loss": 4.0821,
1424
  "step": 236
1425
  },
1426
  {
1427
+ "epoch": 0.8,
1428
+ "learning_rate": 0.00014159999999999997,
1429
+ "loss": 4.177,
1430
  "step": 237
1431
  },
1432
  {
1433
+ "epoch": 0.8,
1434
+ "learning_rate": 0.0001422,
1435
+ "loss": 4.2761,
1436
  "step": 238
1437
  },
1438
  {
1439
+ "epoch": 0.8,
1440
+ "learning_rate": 0.00014279999999999997,
1441
+ "loss": 4.1533,
1442
  "step": 239
1443
  },
1444
  {
1445
+ "epoch": 0.81,
1446
+ "learning_rate": 0.0001434,
1447
+ "loss": 4.1522,
1448
  "step": 240
1449
  },
1450
  {
1451
+ "epoch": 0.81,
1452
+ "learning_rate": 0.00014399999999999998,
1453
+ "loss": 4.3344,
1454
  "step": 241
1455
  },
1456
  {
1457
+ "epoch": 0.81,
1458
+ "learning_rate": 0.0001446,
1459
+ "loss": 4.3554,
1460
  "step": 242
1461
  },
1462
  {
1463
+ "epoch": 0.82,
1464
+ "learning_rate": 0.00014519999999999998,
1465
+ "loss": 4.1236,
1466
  "step": 243
1467
  },
1468
  {
1469
+ "epoch": 0.82,
1470
+ "learning_rate": 0.0001458,
1471
+ "loss": 4.1599,
1472
  "step": 244
1473
  },
1474
  {
1475
+ "epoch": 0.82,
1476
+ "learning_rate": 0.00014639999999999998,
1477
+ "loss": 4.1697,
1478
  "step": 245
1479
  },
1480
  {
1481
+ "epoch": 0.83,
1482
+ "learning_rate": 0.000147,
1483
+ "loss": 4.2965,
1484
  "step": 246
1485
  },
1486
  {
1487
+ "epoch": 0.83,
1488
+ "learning_rate": 0.00014759999999999998,
1489
+ "loss": 4.3114,
1490
  "step": 247
1491
  },
1492
  {
1493
+ "epoch": 0.83,
1494
+ "learning_rate": 0.0001482,
1495
+ "loss": 4.2331,
1496
  "step": 248
1497
  },
1498
  {
1499
+ "epoch": 0.84,
1500
+ "learning_rate": 0.00014879999999999998,
1501
+ "loss": 4.4326,
1502
  "step": 249
1503
  },
1504
  {
1505
+ "epoch": 0.84,
1506
+ "learning_rate": 0.0001494,
1507
+ "loss": 4.1969,
1508
  "step": 250
1509
  },
1510
  {
1511
+ "epoch": 0.84,
1512
+ "learning_rate": 0.00015,
1513
+ "loss": 4.3411,
1514
  "step": 251
1515
  },
1516
  {
1517
+ "epoch": 0.85,
1518
+ "learning_rate": 0.00015059999999999997,
1519
+ "loss": 4.2298,
1520
  "step": 252
1521
  },
1522
  {
1523
+ "epoch": 0.85,
1524
+ "learning_rate": 0.0001512,
1525
+ "loss": 4.2068,
1526
  "step": 253
1527
  },
1528
  {
1529
+ "epoch": 0.85,
1530
+ "learning_rate": 0.00015179999999999998,
1531
+ "loss": 4.1757,
1532
  "step": 254
1533
  },
1534
  {
1535
+ "epoch": 0.86,
1536
+ "learning_rate": 0.0001524,
1537
+ "loss": 4.1387,
1538
  "step": 255
1539
  },
1540
  {
1541
+ "epoch": 0.86,
1542
+ "learning_rate": 0.00015299999999999998,
1543
+ "loss": 4.1108,
1544
  "step": 256
1545
  },
1546
  {
1547
+ "epoch": 0.86,
1548
+ "learning_rate": 0.0001536,
1549
+ "loss": 4.3753,
1550
  "step": 257
1551
  },
1552
  {
1553
+ "epoch": 0.87,
1554
+ "learning_rate": 0.00015419999999999998,
1555
+ "loss": 4.1422,
1556
  "step": 258
1557
  },
1558
  {
1559
+ "epoch": 0.87,
1560
+ "learning_rate": 0.0001548,
1561
+ "loss": 4.2031,
1562
  "step": 259
1563
  },
1564
  {
1565
+ "epoch": 0.87,
1566
+ "learning_rate": 0.00015539999999999998,
1567
+ "loss": 4.1135,
1568
  "step": 260
1569
  },
1570
  {
1571
+ "epoch": 0.88,
1572
+ "learning_rate": 0.000156,
1573
+ "loss": 4.2013,
1574
  "step": 261
1575
  },
1576
  {
1577
+ "epoch": 0.88,
1578
+ "learning_rate": 0.00015659999999999998,
1579
+ "loss": 4.2562,
1580
  "step": 262
1581
  },
1582
  {
1583
+ "epoch": 0.88,
1584
+ "learning_rate": 0.0001572,
1585
+ "loss": 4.1273,
1586
  "step": 263
1587
  },
1588
  {
1589
+ "epoch": 0.89,
1590
+ "learning_rate": 0.0001578,
1591
+ "loss": 4.1866,
1592
  "step": 264
1593
  },
1594
  {
1595
+ "epoch": 0.89,
1596
+ "learning_rate": 0.0001584,
1597
+ "loss": 4.1688,
1598
  "step": 265
1599
  },
1600
  {
1601
+ "epoch": 0.89,
1602
+ "learning_rate": 0.000159,
1603
+ "loss": 4.3492,
1604
  "step": 266
1605
  },
1606
  {
1607
+ "epoch": 0.9,
1608
+ "learning_rate": 0.0001596,
1609
+ "loss": 4.1599,
1610
  "step": 267
1611
  },
1612
  {
1613
+ "epoch": 0.9,
1614
+ "learning_rate": 0.0001602,
1615
+ "loss": 4.1327,
1616
  "step": 268
1617
  },
1618
  {
1619
+ "epoch": 0.9,
1620
+ "learning_rate": 0.0001608,
1621
+ "loss": 4.0783,
1622
  "step": 269
1623
  },
1624
  {
1625
+ "epoch": 0.91,
1626
+ "learning_rate": 0.0001614,
1627
+ "loss": 4.2492,
1628
  "step": 270
1629
  },
1630
  {
1631
+ "epoch": 0.91,
1632
+ "learning_rate": 0.000162,
1633
+ "loss": 4.0783,
1634
  "step": 271
1635
  },
1636
  {
1637
+ "epoch": 0.91,
1638
+ "learning_rate": 0.0001626,
1639
+ "loss": 4.0342,
1640
  "step": 272
1641
  },
1642
  {
1643
+ "epoch": 0.92,
1644
+ "learning_rate": 0.0001632,
1645
+ "loss": 4.1312,
1646
  "step": 273
1647
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1648
  {
1649
  "epoch": 0.92,
1650
+ "learning_rate": 0.0001638,
1651
+ "loss": 4.1231,
1652
+ "step": 274
 
 
 
 
 
 
1653
  },
1654
  {
1655
  "epoch": 0.92,
1656
+ "learning_rate": 0.0001644,
1657
+ "loss": 4.1969,
1658
+ "step": 275
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1659
  },
1660
  {
1661
  "epoch": 0.93,
1662
+ "learning_rate": 0.000165,
1663
+ "loss": 4.1583,
1664
+ "step": 276
1665
  },
1666
  {
1667
  "epoch": 0.93,
1668
+ "learning_rate": 0.0001656,
1669
+ "loss": 4.161,
1670
+ "step": 277
1671
  },
1672
  {
1673
  "epoch": 0.93,
1674
+ "learning_rate": 0.0001662,
1675
+ "loss": 4.0859,
1676
+ "step": 278
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1677
  },
1678
  {
1679
  "epoch": 0.94,
1680
+ "learning_rate": 0.0001668,
1681
+ "loss": 4.1179,
1682
+ "step": 279
1683
  },
1684
  {
1685
  "epoch": 0.94,
1686
+ "learning_rate": 0.0001674,
1687
+ "loss": 4.1002,
1688
+ "step": 280
1689
  },
1690
  {
1691
  "epoch": 0.94,
1692
+ "learning_rate": 0.000168,
1693
+ "loss": 4.2117,
1694
+ "step": 281
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1695
  },
1696
  {
1697
  "epoch": 0.95,
1698
+ "learning_rate": 0.0001686,
1699
+ "loss": 4.0938,
1700
+ "step": 282
1701
  },
1702
  {
1703
  "epoch": 0.95,
1704
+ "learning_rate": 0.00016919999999999997,
1705
+ "loss": 4.2846,
1706
+ "step": 283
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1707
  },
1708
  {
1709
  "epoch": 0.96,
1710
+ "learning_rate": 0.00016979999999999998,
1711
+ "loss": 4.1019,
1712
+ "step": 284
1713
  },
1714
  {
1715
  "epoch": 0.96,
1716
+ "learning_rate": 0.00017039999999999997,
1717
+ "loss": 4.2145,
1718
+ "step": 285
1719
  },
1720
  {
1721
  "epoch": 0.96,
1722
+ "learning_rate": 0.00017099999999999998,
1723
+ "loss": 4.1841,
1724
+ "step": 286
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1725
  },
1726
  {
1727
  "epoch": 0.97,
1728
+ "learning_rate": 0.00017159999999999997,
1729
+ "loss": 4.1436,
1730
+ "step": 287
1731
  },
1732
  {
1733
  "epoch": 0.97,
1734
+ "learning_rate": 0.00017219999999999998,
1735
+ "loss": 4.178,
1736
+ "step": 288
1737
  },
1738
  {
1739
  "epoch": 0.97,
1740
+ "learning_rate": 0.00017279999999999997,
1741
+ "loss": 4.1402,
1742
+ "step": 289
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1743
  },
1744
  {
1745
  "epoch": 0.98,
1746
+ "learning_rate": 0.00017339999999999996,
1747
+ "loss": 4.2526,
1748
+ "step": 290
1749
  },
1750
  {
1751
  "epoch": 0.98,
1752
+ "learning_rate": 0.00017399999999999997,
1753
+ "loss": 4.2094,
1754
+ "step": 291
1755
  },
1756
  {
1757
  "epoch": 0.98,
1758
+ "learning_rate": 0.00017459999999999996,
1759
+ "loss": 4.3642,
1760
+ "step": 292
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1761
  },
1762
  {
1763
  "epoch": 0.99,
1764
+ "learning_rate": 0.00017519999999999998,
1765
+ "loss": 15.0924,
1766
+ "step": 293
1767
  },
1768
  {
1769
  "epoch": 0.99,
1770
+ "learning_rate": 0.00017579999999999996,
1771
+ "loss": 6.5611,
1772
+ "step": 294
1773
  },
1774
  {
1775
  "epoch": 0.99,
1776
+ "learning_rate": 0.00017639999999999998,
1777
+ "loss": 4.5139,
1778
+ "step": 295
 
 
 
 
 
 
1779
  },
1780
  {
1781
  "epoch": 1.0,
1782
+ "learning_rate": 0.00017699999999999997,
1783
+ "loss": 4.8705,
1784
+ "step": 296
1785
  },
1786
  {
1787
  "epoch": 1.0,
1788
+ "learning_rate": 0.00017759999999999998,
1789
+ "loss": 4.5291,
1790
+ "step": 297
1791
  },
1792
  {
1793
  "epoch": 1.0,
1794
+ "step": 297,
1795
  "total_flos": 0.0,
1796
+ "train_loss": 4.303745459225844,
1797
+ "train_runtime": 4795.9642,
1798
+ "train_samples_per_second": 5.95,
1799
+ "train_steps_per_second": 0.062
1800
  }
1801
  ],
1802
+ "max_steps": 297,
1803
  "num_train_epochs": 1,
1804
  "total_flos": 0.0,
1805
  "trial_name": null,
wandb/run-20220302_200036-31e4k99c/files/config.yaml CHANGED
@@ -10646,6 +10646,26 @@ _wandb:
10646
  5: 1
10647
  6:
10648
  - 1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10649
  python_version: 3.9.5
10650
  start_time: 1646251236
10651
  t:
 
10646
  5: 1
10647
  6:
10648
  - 1
10649
+ - 1: eval/loss
10650
+ 5: 1
10651
+ 6:
10652
+ - 1
10653
+ - 1: eval/wer
10654
+ 5: 1
10655
+ 6:
10656
+ - 1
10657
+ - 1: eval/runtime
10658
+ 5: 1
10659
+ 6:
10660
+ - 1
10661
+ - 1: eval/samples_per_second
10662
+ 5: 1
10663
+ 6:
10664
+ - 1
10665
+ - 1: eval/steps_per_second
10666
+ 5: 1
10667
+ 6:
10668
+ - 1
10669
  python_version: 3.9.5
10670
  start_time: 1646251236
10671
  t:
wandb/run-20220302_200036-31e4k99c/files/output.log CHANGED
@@ -2548,3 +2548,247 @@ Upload file wandb/run-20220302_200036-31e4k99c/run-31e4k99c.wandb: 100%|██
2548
  Upload file wandb/run-20220302_200036-31e4k99c/run-31e4k99c.wandb: 100%|████████████| 34.6M/34.6M [03:29<00:00, 100kB/s]if the files are large.2022-03-02 21:20:32,141 >> 7,819 >> Could not estimate the number of tokens of the input, floating-point operations will not be computed
2549
  03/02/2022 21:26:15 - WARNING - huggingface_hub.repository - To https://huggingface.co/sanchit-gandhi/wav2vec2-gpt2-wandb-grid-search
2550
  [INFO|modelcard.py:460] 2022-03-02 21:26:18,447 >> Dropping the following result as it does not have all the necessary fields: files are large.2022-03-02 21:20:32,141 >> 7,819 >> Could not estimate the number of tokens of the input, floating-point operations will not be computed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2548
  Upload file wandb/run-20220302_200036-31e4k99c/run-31e4k99c.wandb: 100%|████████████| 34.6M/34.6M [03:29<00:00, 100kB/s]if the files are large.2022-03-02 21:20:32,141 >> 7,819 >> Could not estimate the number of tokens of the input, floating-point operations will not be computed
2549
  03/02/2022 21:26:15 - WARNING - huggingface_hub.repository - To https://huggingface.co/sanchit-gandhi/wav2vec2-gpt2-wandb-grid-search
2550
  [INFO|modelcard.py:460] 2022-03-02 21:26:18,447 >> Dropping the following result as it does not have all the necessary fields: files are large.2022-03-02 21:20:32,141 >> 7,819 >> Could not estimate the number of tokens of the input, floating-point operations will not be computed
2551
+ Upload file wandb/run-20220302_200036-31e4k99c/run-31e4k99c.wandb: 43%|████▋ | 14.9M/34.6M [00:01<00:01, 15.6MB/s]To https://huggingface.co/sanchit-gandhi/wav2vec2-gpt2-wandb-grid-searchimate the number of tokens of the input, floating-point operations will not be computed
2552
+ Upload file wandb/run-20220302_200036-31e4k99c/run-31e4k99c.wandb: 43%|████▋ | 14.9M/34.6M [00:01<00:01, 15.6MB/s]To https://huggingface.co/sanchit-gandhi/wav2vec2-gpt2-wandb-grid-searchimate the number of tokens of the input, floating-point operations will not be computed
2553
+ 03/02/2022 21:26:24 - WARNING - huggingface_hub.repository - To https://huggingface.co/sanchit-gandhi/wav2vec2-gpt2-wandb-grid-search
2554
+ 70b21bb..a53bfaa main -> main
2555
+ [INFO|trainer.py:2366] 2022-03-02 21:26:27,097 >> Num examples = 2642in the evaluation set don't have a corresponding argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2556
+ [INFO|trainer.py:2366] 2022-03-02 21:26:27,097 >> Num examples = 2642in the evaluation set don't have a corresponding argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2557
+ ***** train metrics *****
2558
+ epoch = 1.0
2559
+ train_loss = 4.3037
2560
+ train_runtime = 1:19:55.96
2561
+ train_samples = 28538
2562
+ train_samples_per_second = 5.95
2563
+ train_steps_per_second = 0.062
2564
+ 0%| | 0/221 [00:00<?, ?it/s] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2565
+ 1%|▊ | 2/221 [00:03<06:18, 1.73s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2566
+ 1%|█▏ | 3/221 [00:07<09:51, 2.71s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2567
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2568
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2569
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2570
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2571
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2572
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2573
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2574
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2575
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2576
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2577
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2578
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2579
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2580
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2581
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2582
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2583
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2584
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2585
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2586
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2587
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2588
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2589
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2590
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2591
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2592
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2593
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2594
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2595
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2596
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2597
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2598
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2599
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2600
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2601
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2602
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2603
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2604
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2605
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2606
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2607
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2608
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2609
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2610
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2611
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2612
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2613
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2614
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2615
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2616
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2617
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2618
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2619
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2620
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2621
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2622
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2623
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2624
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2625
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2626
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2627
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2628
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2629
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2630
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2631
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2632
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2633
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2634
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2635
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2636
+ 2%|█�� | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2637
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2638
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2639
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2640
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2641
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2642
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2643
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2644
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2645
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2646
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2647
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2648
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2649
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2650
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2651
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2652
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2653
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2654
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2655
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2656
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2657
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2658
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2659
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2660
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2661
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2662
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2663
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2664
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2665
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2666
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2667
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2668
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2669
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2670
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2671
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2672
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2673
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2674
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2675
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2676
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2677
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2678
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2679
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2680
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2681
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2682
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2683
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2684
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2685
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2686
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2687
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2688
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2689
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2690
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2691
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2692
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2693
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2694
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2695
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2696
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2697
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2698
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2699
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2700
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2701
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2702
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2703
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2704
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2705
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2706
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2707
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2708
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2709
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2710
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2711
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2712
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2713
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2714
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2715
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2716
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2717
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2718
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2719
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2720
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2721
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2722
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2723
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2724
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2725
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2726
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2727
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2728
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2729
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2730
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2731
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2732
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2733
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2734
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2735
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2736
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2737
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2738
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2739
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2740
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2741
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2742
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2743
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2744
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2745
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2746
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2747
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2748
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2749
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2750
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2751
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2752
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2753
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2754
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2755
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2756
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2757
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2758
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2759
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2760
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2761
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2762
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2763
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2764
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2765
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2766
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2767
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2768
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2769
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2770
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2771
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2772
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2773
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2774
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2775
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2776
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2777
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2778
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2779
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2780
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2781
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2782
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2783
+ 2%|█▌ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2784
+ [INFO|trainer.py:2114] 2022-03-02 21:42:13,173 >> Saving model checkpoint to ./ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2785
+ [INFO|trainer.py:2114] 2022-03-02 21:42:13,173 >> Saving model checkpoint to ./ | 4/221 [00:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
2786
+ 03/02/2022 21:42:13 - INFO - datasets.metric - Removing /home/sanchit_huggingface_co/.cache/huggingface/metrics/wer/default/default_experiment-1-0.arrow
2787
+ ***** eval metrics *****
2788
+ epoch = 1.0
2789
+ eval_loss = 4.7295
2790
+ eval_runtime = 0:15:46.07
2791
+ eval_samples = 2642
2792
+ eval_samples_per_second = 2.793
2793
+ eval_steps_per_second = 0.234
2794
+ [INFO|modeling_utils.py:1081] 2022-03-02 21:42:29,550 >> Model weights saved in ./pytorch_model.bin:11<11:47, 3.26s/it] argument in `SpeechEncoderDecoderModel.forward` and have been ignored: input_length. If input_length are not expected by `SpeechEncoderDecoderModel.forward`, you can safely ignore this message.
wandb/run-20220302_200036-31e4k99c/files/wandb-summary.json CHANGED
The diff for this file is too large to render. See raw diff
 
wandb/run-20220302_200036-31e4k99c/logs/debug-internal.log CHANGED
@@ -7122,3 +7122,403 @@
7122
  2022-03-02 21:26:14,800 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7123
  2022-03-02 21:26:16,800 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7124
  2022-03-02 21:26:18,801 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7122
  2022-03-02 21:26:14,800 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7123
  2022-03-02 21:26:16,800 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7124
  2022-03-02 21:26:18,801 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7125
+ 2022-03-02 21:26:20,057 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7126
+ 2022-03-02 21:26:22,990 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7127
+ 2022-03-02 21:26:22,990 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7128
+ 2022-03-02 21:26:24,803 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7129
+ 2022-03-02 21:26:26,804 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7130
+ 2022-03-02 21:26:28,804 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7131
+ 2022-03-02 21:26:32,806 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7132
+ 2022-03-02 21:26:36,807 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7133
+ 2022-03-02 21:26:38,032 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7134
+ 2022-03-02 21:26:38,033 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7135
+ 2022-03-02 21:26:38,808 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7136
+ 2022-03-02 21:26:42,809 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7137
+ 2022-03-02 21:26:47,811 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7138
+ 2022-03-02 21:26:50,445 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7139
+ 2022-03-02 21:26:51,812 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7140
+ 2022-03-02 21:26:53,078 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7141
+ 2022-03-02 21:26:53,078 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7142
+ 2022-03-02 21:26:57,814 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7143
+ 2022-03-02 21:27:01,816 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7144
+ 2022-03-02 21:27:03,817 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7145
+ 2022-03-02 21:27:08,139 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7146
+ 2022-03-02 21:27:08,139 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7147
+ 2022-03-02 21:27:09,819 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7148
+ 2022-03-02 21:27:13,820 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7149
+ 2022-03-02 21:27:17,821 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7150
+ 2022-03-02 21:27:20,832 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7151
+ 2022-03-02 21:27:21,823 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7152
+ 2022-03-02 21:27:23,193 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7153
+ 2022-03-02 21:27:23,193 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7154
+ 2022-03-02 21:27:25,824 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7155
+ 2022-03-02 21:27:31,826 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7156
+ 2022-03-02 21:27:37,829 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7157
+ 2022-03-02 21:27:38,239 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7158
+ 2022-03-02 21:27:38,240 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7159
+ 2022-03-02 21:27:41,830 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7160
+ 2022-03-02 21:27:45,831 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7161
+ 2022-03-02 21:27:49,833 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7162
+ 2022-03-02 21:27:51,223 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7163
+ 2022-03-02 21:27:53,287 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7164
+ 2022-03-02 21:27:53,287 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7165
+ 2022-03-02 21:27:53,834 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7166
+ 2022-03-02 21:27:57,836 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7167
+ 2022-03-02 21:28:01,837 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7168
+ 2022-03-02 21:28:05,839 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7169
+ 2022-03-02 21:28:08,381 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7170
+ 2022-03-02 21:28:08,381 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7171
+ 2022-03-02 21:28:09,840 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7172
+ 2022-03-02 21:28:13,842 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7173
+ 2022-03-02 21:28:19,844 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7174
+ 2022-03-02 21:28:21,606 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7175
+ 2022-03-02 21:28:21,844 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7176
+ 2022-03-02 21:28:23,480 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7177
+ 2022-03-02 21:28:23,480 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7178
+ 2022-03-02 21:28:27,846 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7179
+ 2022-03-02 21:28:31,848 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7180
+ 2022-03-02 21:28:35,849 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7181
+ 2022-03-02 21:28:38,541 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7182
+ 2022-03-02 21:28:38,541 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7183
+ 2022-03-02 21:28:39,850 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7184
+ 2022-03-02 21:28:43,852 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7185
+ 2022-03-02 21:28:47,853 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7186
+ 2022-03-02 21:28:51,855 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7187
+ 2022-03-02 21:28:51,979 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7188
+ 2022-03-02 21:28:53,584 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7189
+ 2022-03-02 21:28:53,584 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7190
+ 2022-03-02 21:28:55,856 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7191
+ 2022-03-02 21:28:59,858 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7192
+ 2022-03-02 21:29:05,860 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7193
+ 2022-03-02 21:29:07,860 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7194
+ 2022-03-02 21:29:08,651 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7195
+ 2022-03-02 21:29:08,652 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7196
+ 2022-03-02 21:29:13,862 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7197
+ 2022-03-02 21:29:15,863 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7198
+ 2022-03-02 21:29:21,865 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7199
+ 2022-03-02 21:29:22,348 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7200
+ 2022-03-02 21:29:23,718 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7201
+ 2022-03-02 21:29:23,718 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7202
+ 2022-03-02 21:29:25,866 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7203
+ 2022-03-02 21:29:29,868 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7204
+ 2022-03-02 21:29:35,870 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7205
+ 2022-03-02 21:29:38,782 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7206
+ 2022-03-02 21:29:38,782 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7207
+ 2022-03-02 21:29:41,872 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7208
+ 2022-03-02 21:29:45,873 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7209
+ 2022-03-02 21:29:51,875 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7210
+ 2022-03-02 21:29:52,723 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7211
+ 2022-03-02 21:29:53,822 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7212
+ 2022-03-02 21:29:53,822 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7213
+ 2022-03-02 21:29:55,877 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7214
+ 2022-03-02 21:29:59,878 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7215
+ 2022-03-02 21:30:05,880 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7216
+ 2022-03-02 21:30:08,862 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7217
+ 2022-03-02 21:30:08,862 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7218
+ 2022-03-02 21:30:09,882 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7219
+ 2022-03-02 21:30:11,883 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7220
+ 2022-03-02 21:30:15,884 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7221
+ 2022-03-02 21:30:19,885 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7222
+ 2022-03-02 21:30:23,112 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7223
+ 2022-03-02 21:30:24,036 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7224
+ 2022-03-02 21:30:24,037 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7225
+ 2022-03-02 21:30:25,888 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7226
+ 2022-03-02 21:30:29,889 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7227
+ 2022-03-02 21:30:33,890 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7228
+ 2022-03-02 21:30:39,132 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7229
+ 2022-03-02 21:30:39,132 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7230
+ 2022-03-02 21:30:39,892 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7231
+ 2022-03-02 21:30:41,893 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7232
+ 2022-03-02 21:30:45,894 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7233
+ 2022-03-02 21:30:49,896 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7234
+ 2022-03-02 21:30:53,484 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7235
+ 2022-03-02 21:30:53,897 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7236
+ 2022-03-02 21:30:54,176 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7237
+ 2022-03-02 21:30:54,176 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7238
+ 2022-03-02 21:30:57,898 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7239
+ 2022-03-02 21:31:01,900 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7240
+ 2022-03-02 21:31:05,901 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7241
+ 2022-03-02 21:31:09,288 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7242
+ 2022-03-02 21:31:09,288 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7243
+ 2022-03-02 21:31:09,902 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7244
+ 2022-03-02 21:31:13,903 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7245
+ 2022-03-02 21:31:17,905 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7246
+ 2022-03-02 21:31:21,906 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7247
+ 2022-03-02 21:31:23,877 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7248
+ 2022-03-02 21:31:24,409 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7249
+ 2022-03-02 21:31:24,409 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7250
+ 2022-03-02 21:31:25,907 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7251
+ 2022-03-02 21:31:29,909 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7252
+ 2022-03-02 21:31:33,910 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7253
+ 2022-03-02 21:31:37,911 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7254
+ 2022-03-02 21:31:39,525 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7255
+ 2022-03-02 21:31:39,526 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7256
+ 2022-03-02 21:31:42,913 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7257
+ 2022-03-02 21:31:46,914 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7258
+ 2022-03-02 21:31:50,916 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7259
+ 2022-03-02 21:31:54,256 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7260
+ 2022-03-02 21:31:54,582 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7261
+ 2022-03-02 21:31:54,583 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7262
+ 2022-03-02 21:31:54,917 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7263
+ 2022-03-02 21:31:58,919 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7264
+ 2022-03-02 21:32:02,920 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7265
+ 2022-03-02 21:32:06,921 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7266
+ 2022-03-02 21:32:09,658 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7267
+ 2022-03-02 21:32:09,659 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7268
+ 2022-03-02 21:32:10,923 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7269
+ 2022-03-02 21:32:14,924 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7270
+ 2022-03-02 21:32:20,926 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7271
+ 2022-03-02 21:32:24,652 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7272
+ 2022-03-02 21:32:24,718 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7273
+ 2022-03-02 21:32:24,718 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7274
+ 2022-03-02 21:32:24,928 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7275
+ 2022-03-02 21:32:30,930 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7276
+ 2022-03-02 21:32:34,931 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7277
+ 2022-03-02 21:32:39,760 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7278
+ 2022-03-02 21:32:39,761 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7279
+ 2022-03-02 21:32:40,933 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7280
+ 2022-03-02 21:32:44,934 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7281
+ 2022-03-02 21:32:48,936 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7282
+ 2022-03-02 21:32:52,937 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7283
+ 2022-03-02 21:32:54,826 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7284
+ 2022-03-02 21:32:54,826 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7285
+ 2022-03-02 21:32:55,071 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7286
+ 2022-03-02 21:32:56,939 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7287
+ 2022-03-02 21:33:02,941 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7288
+ 2022-03-02 21:33:06,942 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7289
+ 2022-03-02 21:33:09,883 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7290
+ 2022-03-02 21:33:09,883 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7291
+ 2022-03-02 21:33:12,944 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7292
+ 2022-03-02 21:33:16,945 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7293
+ 2022-03-02 21:33:20,947 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7294
+ 2022-03-02 21:33:24,988 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7295
+ 2022-03-02 21:33:24,988 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7296
+ 2022-03-02 21:33:25,447 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7297
+ 2022-03-02 21:33:26,949 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7298
+ 2022-03-02 21:33:30,950 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7299
+ 2022-03-02 21:33:34,952 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7300
+ 2022-03-02 21:33:38,953 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7301
+ 2022-03-02 21:33:40,078 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7302
+ 2022-03-02 21:33:40,078 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7303
+ 2022-03-02 21:33:42,954 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7304
+ 2022-03-02 21:33:46,956 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7305
+ 2022-03-02 21:33:50,957 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7306
+ 2022-03-02 21:33:54,959 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7307
+ 2022-03-02 21:33:55,125 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7308
+ 2022-03-02 21:33:55,125 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7309
+ 2022-03-02 21:33:55,849 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7310
+ 2022-03-02 21:34:00,961 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7311
+ 2022-03-02 21:34:04,962 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7312
+ 2022-03-02 21:34:08,964 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7313
+ 2022-03-02 21:34:10,184 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7314
+ 2022-03-02 21:34:10,184 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7315
+ 2022-03-02 21:34:14,966 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7316
+ 2022-03-02 21:34:18,967 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7317
+ 2022-03-02 21:34:22,968 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7318
+ 2022-03-02 21:34:25,250 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7319
+ 2022-03-02 21:34:25,250 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7320
+ 2022-03-02 21:34:26,274 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7321
+ 2022-03-02 21:34:26,970 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7322
+ 2022-03-02 21:34:30,971 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7323
+ 2022-03-02 21:34:36,973 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7324
+ 2022-03-02 21:34:38,974 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7325
+ 2022-03-02 21:34:40,390 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7326
+ 2022-03-02 21:34:40,390 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7327
+ 2022-03-02 21:34:42,975 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7328
+ 2022-03-02 21:34:46,976 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7329
+ 2022-03-02 21:34:50,978 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7330
+ 2022-03-02 21:34:55,460 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7331
+ 2022-03-02 21:34:55,460 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7332
+ 2022-03-02 21:34:56,657 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7333
+ 2022-03-02 21:34:56,980 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7334
+ 2022-03-02 21:35:00,981 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7335
+ 2022-03-02 21:35:06,983 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7336
+ 2022-03-02 21:35:10,508 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7337
+ 2022-03-02 21:35:10,509 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7338
+ 2022-03-02 21:35:10,984 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7339
+ 2022-03-02 21:35:12,985 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7340
+ 2022-03-02 21:35:16,987 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7341
+ 2022-03-02 21:35:20,988 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7342
+ 2022-03-02 21:35:24,989 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7343
+ 2022-03-02 21:35:25,948 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7344
+ 2022-03-02 21:35:25,948 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7345
+ 2022-03-02 21:35:27,034 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7346
+ 2022-03-02 21:35:28,990 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7347
+ 2022-03-02 21:35:30,991 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7348
+ 2022-03-02 21:35:34,993 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7349
+ 2022-03-02 21:35:38,994 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7350
+ 2022-03-02 21:35:41,003 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7351
+ 2022-03-02 21:35:41,003 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7352
+ 2022-03-02 21:35:42,995 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7353
+ 2022-03-02 21:35:46,997 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7354
+ 2022-03-02 21:35:50,998 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7355
+ 2022-03-02 21:35:54,999 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7356
+ 2022-03-02 21:35:56,131 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7357
+ 2022-03-02 21:35:56,131 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7358
+ 2022-03-02 21:35:57,000 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7359
+ 2022-03-02 21:35:57,435 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7360
+ 2022-03-02 21:36:01,001 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7361
+ 2022-03-02 21:36:05,003 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7362
+ 2022-03-02 21:36:11,005 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7363
+ 2022-03-02 21:36:11,242 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7364
+ 2022-03-02 21:36:11,242 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7365
+ 2022-03-02 21:36:15,006 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7366
+ 2022-03-02 21:36:19,007 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7367
+ 2022-03-02 21:36:23,009 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7368
+ 2022-03-02 21:36:25,009 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7369
+ 2022-03-02 21:36:26,302 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7370
+ 2022-03-02 21:36:26,302 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7371
+ 2022-03-02 21:36:27,807 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7372
+ 2022-03-02 21:36:29,011 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7373
+ 2022-03-02 21:36:33,012 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7374
+ 2022-03-02 21:36:37,013 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7375
+ 2022-03-02 21:36:41,015 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7376
+ 2022-03-02 21:36:41,361 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7377
+ 2022-03-02 21:36:41,362 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7378
+ 2022-03-02 21:36:45,016 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7379
+ 2022-03-02 21:36:49,018 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7380
+ 2022-03-02 21:36:53,019 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7381
+ 2022-03-02 21:36:56,423 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7382
+ 2022-03-02 21:36:56,424 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7383
+ 2022-03-02 21:36:57,020 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7384
+ 2022-03-02 21:36:58,185 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7385
+ 2022-03-02 21:37:01,022 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7386
+ 2022-03-02 21:37:05,023 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7387
+ 2022-03-02 21:37:09,024 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7388
+ 2022-03-02 21:37:11,503 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7389
+ 2022-03-02 21:37:11,504 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7390
+ 2022-03-02 21:37:13,026 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7391
+ 2022-03-02 21:37:17,027 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7392
+ 2022-03-02 21:37:21,029 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7393
+ 2022-03-02 21:37:25,030 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7394
+ 2022-03-02 21:37:26,580 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7395
+ 2022-03-02 21:37:26,580 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7396
+ 2022-03-02 21:37:28,561 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7397
+ 2022-03-02 21:37:29,031 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7398
+ 2022-03-02 21:37:33,033 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7399
+ 2022-03-02 21:37:40,035 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7400
+ 2022-03-02 21:37:41,628 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7401
+ 2022-03-02 21:37:41,628 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7402
+ 2022-03-02 21:37:44,037 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7403
+ 2022-03-02 21:37:50,039 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7404
+ 2022-03-02 21:37:54,040 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7405
+ 2022-03-02 21:37:56,888 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7406
+ 2022-03-02 21:37:56,888 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7407
+ 2022-03-02 21:37:58,042 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7408
+ 2022-03-02 21:37:58,960 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7409
+ 2022-03-02 21:38:04,043 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7410
+ 2022-03-02 21:38:08,045 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7411
+ 2022-03-02 21:38:12,001 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7412
+ 2022-03-02 21:38:12,001 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7413
+ 2022-03-02 21:38:12,047 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7414
+ 2022-03-02 21:38:16,048 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7415
+ 2022-03-02 21:38:20,049 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7416
+ 2022-03-02 21:38:24,051 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7417
+ 2022-03-02 21:38:27,168 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7418
+ 2022-03-02 21:38:27,169 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7419
+ 2022-03-02 21:38:28,052 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7420
+ 2022-03-02 21:38:29,363 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7421
+ 2022-03-02 21:38:34,054 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7422
+ 2022-03-02 21:38:38,056 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7423
+ 2022-03-02 21:38:42,057 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7424
+ 2022-03-02 21:38:42,279 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7425
+ 2022-03-02 21:38:42,280 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7426
+ 2022-03-02 21:38:44,058 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7427
+ 2022-03-02 21:38:48,060 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7428
+ 2022-03-02 21:38:54,061 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7429
+ 2022-03-02 21:38:56,062 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7430
+ 2022-03-02 21:38:57,333 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7431
+ 2022-03-02 21:38:57,333 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7432
+ 2022-03-02 21:38:59,734 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7433
+ 2022-03-02 21:39:02,064 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7434
+ 2022-03-02 21:39:06,066 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7435
+ 2022-03-02 21:39:10,067 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7436
+ 2022-03-02 21:39:12,405 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7437
+ 2022-03-02 21:39:12,405 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7438
+ 2022-03-02 21:39:14,068 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7439
+ 2022-03-02 21:39:20,070 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7440
+ 2022-03-02 21:39:24,072 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7441
+ 2022-03-02 21:39:27,534 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7442
+ 2022-03-02 21:39:27,534 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7443
+ 2022-03-02 21:39:28,073 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7444
+ 2022-03-02 21:39:30,111 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7445
+ 2022-03-02 21:39:32,075 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7446
+ 2022-03-02 21:39:38,077 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7447
+ 2022-03-02 21:39:42,078 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7448
+ 2022-03-02 21:39:42,584 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7449
+ 2022-03-02 21:39:42,584 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7450
+ 2022-03-02 21:39:46,079 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7451
+ 2022-03-02 21:39:52,081 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7452
+ 2022-03-02 21:39:56,083 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7453
+ 2022-03-02 21:39:57,647 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7454
+ 2022-03-02 21:39:57,647 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7455
+ 2022-03-02 21:40:00,482 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7456
+ 2022-03-02 21:40:02,085 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7457
+ 2022-03-02 21:40:06,086 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7458
+ 2022-03-02 21:40:10,087 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7459
+ 2022-03-02 21:40:12,695 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7460
+ 2022-03-02 21:40:12,695 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7461
+ 2022-03-02 21:40:14,089 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7462
+ 2022-03-02 21:40:18,090 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7463
+ 2022-03-02 21:40:22,092 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7464
+ 2022-03-02 21:40:26,093 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7465
+ 2022-03-02 21:40:27,747 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7466
+ 2022-03-02 21:40:27,748 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7467
+ 2022-03-02 21:40:30,095 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7468
+ 2022-03-02 21:40:30,856 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7469
+ 2022-03-02 21:40:36,097 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7470
+ 2022-03-02 21:40:40,098 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7471
+ 2022-03-02 21:40:42,807 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7472
+ 2022-03-02 21:40:42,807 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7473
+ 2022-03-02 21:40:44,099 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7474
+ 2022-03-02 21:40:48,101 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7475
+ 2022-03-02 21:40:52,102 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7476
+ 2022-03-02 21:40:56,103 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7477
+ 2022-03-02 21:40:57,852 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7478
+ 2022-03-02 21:40:57,852 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7479
+ 2022-03-02 21:41:01,242 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7480
+ 2022-03-02 21:41:02,105 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7481
+ 2022-03-02 21:41:08,107 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7482
+ 2022-03-02 21:41:12,109 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7483
+ 2022-03-02 21:41:12,906 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7484
+ 2022-03-02 21:41:12,907 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7485
+ 2022-03-02 21:41:16,110 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7486
+ 2022-03-02 21:41:20,111 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7487
+ 2022-03-02 21:41:24,113 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7488
+ 2022-03-02 21:41:28,024 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7489
+ 2022-03-02 21:41:28,024 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7490
+ 2022-03-02 21:41:30,115 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7491
+ 2022-03-02 21:41:31,618 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7492
+ 2022-03-02 21:41:34,116 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7493
+ 2022-03-02 21:41:38,117 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7494
+ 2022-03-02 21:41:42,119 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7495
+ 2022-03-02 21:41:43,075 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7496
+ 2022-03-02 21:41:43,076 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7497
+ 2022-03-02 21:41:49,123 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7498
+ 2022-03-02 21:41:53,124 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7499
+ 2022-03-02 21:41:57,126 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7500
+ 2022-03-02 21:41:58,122 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7501
+ 2022-03-02 21:41:58,122 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7502
+ 2022-03-02 21:42:02,001 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7503
+ 2022-03-02 21:42:03,127 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7504
+ 2022-03-02 21:42:07,129 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7505
+ 2022-03-02 21:42:11,130 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7506
+ 2022-03-02 21:42:13,177 DEBUG SenderThread:264383 [sender.py:send():235] send: metric
7507
+ 2022-03-02 21:42:13,182 DEBUG SenderThread:264383 [sender.py:send():235] send: metric
7508
+ 2022-03-02 21:42:13,187 DEBUG SenderThread:264383 [sender.py:send():235] send: metric
7509
+ 2022-03-02 21:42:13,193 DEBUG SenderThread:264383 [sender.py:send():235] send: metric
7510
+ 2022-03-02 21:42:13,193 DEBUG SenderThread:264383 [sender.py:send():235] send: metric
7511
+ 2022-03-02 21:42:13,198 DEBUG SenderThread:264383 [sender.py:send():235] send: history
7512
+ 2022-03-02 21:42:13,230 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7513
+ 2022-03-02 21:42:13,230 DEBUG SenderThread:264383 [sender.py:send():235] send: summary
7514
+ 2022-03-02 21:42:13,312 INFO SenderThread:264383 [sender.py:_save_file():944] saving file wandb-summary.json with policy end
7515
+ 2022-03-02 21:42:13,313 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7516
+ 2022-03-02 21:42:14,131 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/wandb-summary.json
7517
+ 2022-03-02 21:42:15,132 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7518
+ 2022-03-02 21:42:28,370 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7519
+ 2022-03-02 21:42:28,371 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7520
+ 2022-03-02 21:42:31,137 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/output.log
7521
+ 2022-03-02 21:42:32,380 DEBUG SenderThread:264383 [sender.py:send():235] send: stats
7522
+ 2022-03-02 21:42:43,423 DEBUG HandlerThread:264383 [handler.py:handle_request():131] handle_request: stop_status
7523
+ 2022-03-02 21:42:43,569 DEBUG SenderThread:264383 [sender.py:send_request():249] send_request: stop_status
7524
+ 2022-03-02 21:42:44,190 INFO Thread-8 :264383 [dir_watcher.py:_on_file_modified():230] file/dir modified: /home/sanchit_huggingface_co/wav2vec2-gpt2-wandb-grid-search/wandb/run-20220302_200036-31e4k99c/files/config.yaml
wandb/run-20220302_200036-31e4k99c/run-31e4k99c.wandb CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cb78d10d06d334bb905cdcdee7d207ec83967984a3cd82ca5bd14b2c6a1762ee
3
- size 36302358
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa51a0a26ce5e172e476442812ae6212e91addd95b58f5626c318ed39b664595
3
+ size 36405850