tuanio commited on
Commit
e96123c
1 Parent(s): d485446

End of training

Browse files
Files changed (1) hide show
  1. trainer_state.json +1678 -0
trainer_state.json ADDED
@@ -0,0 +1,1678 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.1558224891558225,
3
+ "best_model_checkpoint": "w2v2_ablation_200epoch-with_ling_head-0drop-0load_best-best_on_tp0.025_tl10_fp0.001_fl16/checkpoint-8000",
4
+ "epoch": 200.0,
5
+ "eval_steps": 500,
6
+ "global_step": 21200,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.94,
13
+ "learning_rate": 6.698113207547171e-07,
14
+ "loss": 115.3958,
15
+ "step": 100
16
+ },
17
+ {
18
+ "epoch": 1.89,
19
+ "learning_rate": 1.6132075471698113e-06,
20
+ "loss": 94.1622,
21
+ "step": 200
22
+ },
23
+ {
24
+ "epoch": 2.83,
25
+ "learning_rate": 2.5566037735849057e-06,
26
+ "loss": 22.0241,
27
+ "step": 300
28
+ },
29
+ {
30
+ "epoch": 3.77,
31
+ "learning_rate": 3.5e-06,
32
+ "loss": 5.4965,
33
+ "step": 400
34
+ },
35
+ {
36
+ "epoch": 4.72,
37
+ "learning_rate": 4.443396226415095e-06,
38
+ "loss": 5.0922,
39
+ "step": 500
40
+ },
41
+ {
42
+ "epoch": 4.72,
43
+ "eval_loss": 5.229608058929443,
44
+ "eval_runtime": 2.1376,
45
+ "eval_samples_per_second": 335.427,
46
+ "eval_steps_per_second": 2.807,
47
+ "eval_wer": 1.0,
48
+ "step": 500
49
+ },
50
+ {
51
+ "epoch": 5.66,
52
+ "learning_rate": 5.38679245283019e-06,
53
+ "loss": 4.7302,
54
+ "step": 600
55
+ },
56
+ {
57
+ "epoch": 6.6,
58
+ "learning_rate": 6.330188679245284e-06,
59
+ "loss": 4.5277,
60
+ "step": 700
61
+ },
62
+ {
63
+ "epoch": 7.55,
64
+ "learning_rate": 7.273584905660378e-06,
65
+ "loss": 4.4156,
66
+ "step": 800
67
+ },
68
+ {
69
+ "epoch": 8.49,
70
+ "learning_rate": 8.216981132075471e-06,
71
+ "loss": 4.3728,
72
+ "step": 900
73
+ },
74
+ {
75
+ "epoch": 9.43,
76
+ "learning_rate": 9.160377358490566e-06,
77
+ "loss": 4.3263,
78
+ "step": 1000
79
+ },
80
+ {
81
+ "epoch": 9.43,
82
+ "eval_loss": 5.426015377044678,
83
+ "eval_runtime": 1.8299,
84
+ "eval_samples_per_second": 391.828,
85
+ "eval_steps_per_second": 3.279,
86
+ "eval_wer": 1.0,
87
+ "step": 1000
88
+ },
89
+ {
90
+ "epoch": 10.38,
91
+ "learning_rate": 1.010377358490566e-05,
92
+ "loss": 4.2404,
93
+ "step": 1100
94
+ },
95
+ {
96
+ "epoch": 11.32,
97
+ "learning_rate": 1.1047169811320754e-05,
98
+ "loss": 4.0915,
99
+ "step": 1200
100
+ },
101
+ {
102
+ "epoch": 12.26,
103
+ "learning_rate": 1.1990566037735848e-05,
104
+ "loss": 3.6665,
105
+ "step": 1300
106
+ },
107
+ {
108
+ "epoch": 13.21,
109
+ "learning_rate": 1.2933962264150946e-05,
110
+ "loss": 2.6838,
111
+ "step": 1400
112
+ },
113
+ {
114
+ "epoch": 14.15,
115
+ "learning_rate": 1.387735849056604e-05,
116
+ "loss": 1.9702,
117
+ "step": 1500
118
+ },
119
+ {
120
+ "epoch": 14.15,
121
+ "eval_loss": 1.6761473417282104,
122
+ "eval_runtime": 1.8334,
123
+ "eval_samples_per_second": 391.084,
124
+ "eval_steps_per_second": 3.273,
125
+ "eval_wer": 0.43685352018685353,
126
+ "step": 1500
127
+ },
128
+ {
129
+ "epoch": 15.09,
130
+ "learning_rate": 1.4820754716981134e-05,
131
+ "loss": 1.5243,
132
+ "step": 1600
133
+ },
134
+ {
135
+ "epoch": 16.04,
136
+ "learning_rate": 1.5764150943396227e-05,
137
+ "loss": 1.2131,
138
+ "step": 1700
139
+ },
140
+ {
141
+ "epoch": 16.98,
142
+ "learning_rate": 1.6707547169811323e-05,
143
+ "loss": 0.9532,
144
+ "step": 1800
145
+ },
146
+ {
147
+ "epoch": 17.92,
148
+ "learning_rate": 1.7650943396226415e-05,
149
+ "loss": 0.807,
150
+ "step": 1900
151
+ },
152
+ {
153
+ "epoch": 18.87,
154
+ "learning_rate": 1.859433962264151e-05,
155
+ "loss": 0.7013,
156
+ "step": 2000
157
+ },
158
+ {
159
+ "epoch": 18.87,
160
+ "eval_loss": 0.6799427270889282,
161
+ "eval_runtime": 1.6704,
162
+ "eval_samples_per_second": 429.238,
163
+ "eval_steps_per_second": 3.592,
164
+ "eval_wer": 0.235985985985986,
165
+ "step": 2000
166
+ },
167
+ {
168
+ "epoch": 19.81,
169
+ "learning_rate": 1.9537735849056604e-05,
170
+ "loss": 0.5817,
171
+ "step": 2100
172
+ },
173
+ {
174
+ "epoch": 20.75,
175
+ "learning_rate": 1.9999647425572696e-05,
176
+ "loss": 0.5499,
177
+ "step": 2200
178
+ },
179
+ {
180
+ "epoch": 21.7,
181
+ "learning_rate": 1.9996909387678377e-05,
182
+ "loss": 0.4612,
183
+ "step": 2300
184
+ },
185
+ {
186
+ "epoch": 22.64,
187
+ "learning_rate": 1.9991461164713644e-05,
188
+ "loss": 0.4711,
189
+ "step": 2400
190
+ },
191
+ {
192
+ "epoch": 23.58,
193
+ "learning_rate": 1.9983304233704244e-05,
194
+ "loss": 0.4391,
195
+ "step": 2500
196
+ },
197
+ {
198
+ "epoch": 23.58,
199
+ "eval_loss": 0.5237099528312683,
200
+ "eval_runtime": 1.6458,
201
+ "eval_samples_per_second": 435.648,
202
+ "eval_steps_per_second": 3.646,
203
+ "eval_wer": 0.19636302969636302,
204
+ "step": 2500
205
+ },
206
+ {
207
+ "epoch": 24.53,
208
+ "learning_rate": 1.997244080601288e-05,
209
+ "loss": 0.3742,
210
+ "step": 2600
211
+ },
212
+ {
213
+ "epoch": 25.47,
214
+ "learning_rate": 1.9958873826739734e-05,
215
+ "loss": 0.3537,
216
+ "step": 2700
217
+ },
218
+ {
219
+ "epoch": 26.42,
220
+ "learning_rate": 1.9942606973924003e-05,
221
+ "loss": 0.3391,
222
+ "step": 2800
223
+ },
224
+ {
225
+ "epoch": 27.36,
226
+ "learning_rate": 1.992364465754681e-05,
227
+ "loss": 0.3071,
228
+ "step": 2900
229
+ },
230
+ {
231
+ "epoch": 28.3,
232
+ "learning_rate": 1.990199201833563e-05,
233
+ "loss": 0.3015,
234
+ "step": 3000
235
+ },
236
+ {
237
+ "epoch": 28.3,
238
+ "eval_loss": 0.4436860978603363,
239
+ "eval_runtime": 1.7679,
240
+ "eval_samples_per_second": 405.574,
241
+ "eval_steps_per_second": 3.394,
242
+ "eval_wer": 0.18485151818485152,
243
+ "step": 3000
244
+ },
245
+ {
246
+ "epoch": 29.25,
247
+ "learning_rate": 1.9877654926370626e-05,
248
+ "loss": 0.2819,
249
+ "step": 3100
250
+ },
251
+ {
252
+ "epoch": 30.19,
253
+ "learning_rate": 1.9850639979493267e-05,
254
+ "loss": 0.2528,
255
+ "step": 3200
256
+ },
257
+ {
258
+ "epoch": 31.13,
259
+ "learning_rate": 1.9820954501517633e-05,
260
+ "loss": 0.2936,
261
+ "step": 3300
262
+ },
263
+ {
264
+ "epoch": 32.08,
265
+ "learning_rate": 1.9788606540244904e-05,
266
+ "loss": 0.2453,
267
+ "step": 3400
268
+ },
269
+ {
270
+ "epoch": 33.02,
271
+ "learning_rate": 1.97536048652816e-05,
272
+ "loss": 0.2416,
273
+ "step": 3500
274
+ },
275
+ {
276
+ "epoch": 33.02,
277
+ "eval_loss": 0.4311460554599762,
278
+ "eval_runtime": 1.665,
279
+ "eval_samples_per_second": 430.639,
280
+ "eval_steps_per_second": 3.604,
281
+ "eval_wer": 0.2081247914581248,
282
+ "step": 3500
283
+ },
284
+ {
285
+ "epoch": 33.96,
286
+ "learning_rate": 1.97159589656621e-05,
287
+ "loss": 0.2397,
288
+ "step": 3600
289
+ },
290
+ {
291
+ "epoch": 34.91,
292
+ "learning_rate": 1.967567904727617e-05,
293
+ "loss": 0.2325,
294
+ "step": 3700
295
+ },
296
+ {
297
+ "epoch": 35.85,
298
+ "learning_rate": 1.9632776030102088e-05,
299
+ "loss": 0.2252,
300
+ "step": 3800
301
+ },
302
+ {
303
+ "epoch": 36.79,
304
+ "learning_rate": 1.9587261545246246e-05,
305
+ "loss": 0.2268,
306
+ "step": 3900
307
+ },
308
+ {
309
+ "epoch": 37.74,
310
+ "learning_rate": 1.9539147931789902e-05,
311
+ "loss": 0.2057,
312
+ "step": 4000
313
+ },
314
+ {
315
+ "epoch": 37.74,
316
+ "eval_loss": 0.42023688554763794,
317
+ "eval_runtime": 1.6178,
318
+ "eval_samples_per_second": 443.182,
319
+ "eval_steps_per_second": 3.709,
320
+ "eval_wer": 0.16966966966966968,
321
+ "step": 4000
322
+ },
323
+ {
324
+ "epoch": 38.68,
325
+ "learning_rate": 1.9488448233444058e-05,
326
+ "loss": 0.1979,
327
+ "step": 4100
328
+ },
329
+ {
330
+ "epoch": 39.62,
331
+ "learning_rate": 1.943517619501325e-05,
332
+ "loss": 0.1867,
333
+ "step": 4200
334
+ },
335
+ {
336
+ "epoch": 40.57,
337
+ "learning_rate": 1.9379346258669315e-05,
338
+ "loss": 0.1842,
339
+ "step": 4300
340
+ },
341
+ {
342
+ "epoch": 41.51,
343
+ "learning_rate": 1.932156982230388e-05,
344
+ "loss": 0.1939,
345
+ "step": 4400
346
+ },
347
+ {
348
+ "epoch": 42.45,
349
+ "learning_rate": 1.926069537514039e-05,
350
+ "loss": 0.1714,
351
+ "step": 4500
352
+ },
353
+ {
354
+ "epoch": 42.45,
355
+ "eval_loss": 0.4269614815711975,
356
+ "eval_runtime": 1.7582,
357
+ "eval_samples_per_second": 407.806,
358
+ "eval_steps_per_second": 3.413,
359
+ "eval_wer": 0.1738405071738405,
360
+ "step": 4500
361
+ },
362
+ {
363
+ "epoch": 43.4,
364
+ "learning_rate": 1.919731033221434e-05,
365
+ "loss": 0.1687,
366
+ "step": 4600
367
+ },
368
+ {
369
+ "epoch": 44.34,
370
+ "learning_rate": 1.9131431877356292e-05,
371
+ "loss": 0.1673,
372
+ "step": 4700
373
+ },
374
+ {
375
+ "epoch": 45.28,
376
+ "learning_rate": 1.9063077870366504e-05,
377
+ "loss": 0.1627,
378
+ "step": 4800
379
+ },
380
+ {
381
+ "epoch": 46.23,
382
+ "learning_rate": 1.899226684217308e-05,
383
+ "loss": 0.1714,
384
+ "step": 4900
385
+ },
386
+ {
387
+ "epoch": 47.17,
388
+ "learning_rate": 1.8919017989808238e-05,
389
+ "loss": 0.1812,
390
+ "step": 5000
391
+ },
392
+ {
393
+ "epoch": 47.17,
394
+ "eval_loss": 0.44668373465538025,
395
+ "eval_runtime": 1.7171,
396
+ "eval_samples_per_second": 417.56,
397
+ "eval_steps_per_second": 3.494,
398
+ "eval_wer": 0.15999332665999333,
399
+ "step": 5000
400
+ },
401
+ {
402
+ "epoch": 48.11,
403
+ "learning_rate": 1.884335117120389e-05,
404
+ "loss": 0.1496,
405
+ "step": 5100
406
+ },
407
+ {
408
+ "epoch": 49.06,
409
+ "learning_rate": 1.8765286899808165e-05,
410
+ "loss": 0.1371,
411
+ "step": 5200
412
+ },
413
+ {
414
+ "epoch": 50.0,
415
+ "learning_rate": 1.8684846339024145e-05,
416
+ "loss": 0.1607,
417
+ "step": 5300
418
+ },
419
+ {
420
+ "epoch": 50.94,
421
+ "learning_rate": 1.8602051296472396e-05,
422
+ "loss": 0.1489,
423
+ "step": 5400
424
+ },
425
+ {
426
+ "epoch": 51.89,
427
+ "learning_rate": 1.851778695718426e-05,
428
+ "loss": 0.1498,
429
+ "step": 5500
430
+ },
431
+ {
432
+ "epoch": 51.89,
433
+ "eval_loss": 0.4322489798069,
434
+ "eval_runtime": 1.6167,
435
+ "eval_samples_per_second": 443.504,
436
+ "eval_steps_per_second": 3.711,
437
+ "eval_wer": 0.2197197197197197,
438
+ "step": 5500
439
+ },
440
+ {
441
+ "epoch": 52.83,
442
+ "learning_rate": 1.8430373894374803e-05,
443
+ "loss": 0.1347,
444
+ "step": 5600
445
+ },
446
+ {
447
+ "epoch": 53.77,
448
+ "learning_rate": 1.8340675337861192e-05,
449
+ "loss": 0.1343,
450
+ "step": 5700
451
+ },
452
+ {
453
+ "epoch": 54.72,
454
+ "learning_rate": 1.8248715605127888e-05,
455
+ "loss": 0.1279,
456
+ "step": 5800
457
+ },
458
+ {
459
+ "epoch": 55.66,
460
+ "learning_rate": 1.815451962666941e-05,
461
+ "loss": 0.1266,
462
+ "step": 5900
463
+ },
464
+ {
465
+ "epoch": 56.6,
466
+ "learning_rate": 1.8058112939231632e-05,
467
+ "loss": 0.1255,
468
+ "step": 6000
469
+ },
470
+ {
471
+ "epoch": 56.6,
472
+ "eval_loss": 0.44078171253204346,
473
+ "eval_runtime": 1.8002,
474
+ "eval_samples_per_second": 398.3,
475
+ "eval_steps_per_second": 3.333,
476
+ "eval_wer": 0.16958625291958626,
477
+ "step": 6000
478
+ },
479
+ {
480
+ "epoch": 57.55,
481
+ "learning_rate": 1.79595216788887e-05,
482
+ "loss": 0.1338,
483
+ "step": 6100
484
+ },
485
+ {
486
+ "epoch": 58.49,
487
+ "learning_rate": 1.785877257395748e-05,
488
+ "loss": 0.1325,
489
+ "step": 6200
490
+ },
491
+ {
492
+ "epoch": 59.43,
493
+ "learning_rate": 1.7755892937751472e-05,
494
+ "loss": 0.1307,
495
+ "step": 6300
496
+ },
497
+ {
498
+ "epoch": 60.38,
499
+ "learning_rate": 1.7650910661176075e-05,
500
+ "loss": 0.1133,
501
+ "step": 6400
502
+ },
503
+ {
504
+ "epoch": 61.32,
505
+ "learning_rate": 1.754385420516732e-05,
506
+ "loss": 0.1148,
507
+ "step": 6500
508
+ },
509
+ {
510
+ "epoch": 61.32,
511
+ "eval_loss": 0.45305711030960083,
512
+ "eval_runtime": 1.5973,
513
+ "eval_samples_per_second": 448.885,
514
+ "eval_steps_per_second": 3.756,
515
+ "eval_wer": 0.17650984317650983,
516
+ "step": 6500
517
+ },
518
+ {
519
+ "epoch": 62.26,
520
+ "learning_rate": 1.7434752592976063e-05,
521
+ "loss": 0.1185,
522
+ "step": 6600
523
+ },
524
+ {
525
+ "epoch": 63.21,
526
+ "learning_rate": 1.7323635402299657e-05,
527
+ "loss": 0.1205,
528
+ "step": 6700
529
+ },
530
+ {
531
+ "epoch": 64.15,
532
+ "learning_rate": 1.72116735114751e-05,
533
+ "loss": 0.1081,
534
+ "step": 6800
535
+ },
536
+ {
537
+ "epoch": 65.09,
538
+ "learning_rate": 1.7096635468859486e-05,
539
+ "loss": 0.1056,
540
+ "step": 6900
541
+ },
542
+ {
543
+ "epoch": 66.04,
544
+ "learning_rate": 1.69796735120868e-05,
545
+ "loss": 0.1112,
546
+ "step": 7000
547
+ },
548
+ {
549
+ "epoch": 66.04,
550
+ "eval_loss": 0.45722392201423645,
551
+ "eval_runtime": 1.7559,
552
+ "eval_samples_per_second": 408.338,
553
+ "eval_steps_per_second": 3.417,
554
+ "eval_wer": 0.21479813146479812,
555
+ "step": 7000
556
+ },
557
+ {
558
+ "epoch": 66.98,
559
+ "learning_rate": 1.6860819349811873e-05,
560
+ "loss": 0.1017,
561
+ "step": 7100
562
+ },
563
+ {
564
+ "epoch": 67.92,
565
+ "learning_rate": 1.674010520367077e-05,
566
+ "loss": 0.1073,
567
+ "step": 7200
568
+ },
569
+ {
570
+ "epoch": 68.87,
571
+ "learning_rate": 1.6617563799545462e-05,
572
+ "loss": 0.1074,
573
+ "step": 7300
574
+ },
575
+ {
576
+ "epoch": 69.81,
577
+ "learning_rate": 1.6493228358691748e-05,
578
+ "loss": 0.1067,
579
+ "step": 7400
580
+ },
581
+ {
582
+ "epoch": 70.75,
583
+ "learning_rate": 1.6367132588732895e-05,
584
+ "loss": 0.1038,
585
+ "step": 7500
586
+ },
587
+ {
588
+ "epoch": 70.75,
589
+ "eval_loss": 0.4648493826389313,
590
+ "eval_runtime": 1.7239,
591
+ "eval_samples_per_second": 415.929,
592
+ "eval_steps_per_second": 3.481,
593
+ "eval_wer": 0.18943943943943944,
594
+ "step": 7500
595
+ },
596
+ {
597
+ "epoch": 71.7,
598
+ "learning_rate": 1.6239310674521443e-05,
599
+ "loss": 0.1023,
600
+ "step": 7600
601
+ },
602
+ {
603
+ "epoch": 72.64,
604
+ "learning_rate": 1.6109797268871586e-05,
605
+ "loss": 0.0917,
606
+ "step": 7700
607
+ },
608
+ {
609
+ "epoch": 73.58,
610
+ "learning_rate": 1.597994726370795e-05,
611
+ "loss": 0.103,
612
+ "step": 7800
613
+ },
614
+ {
615
+ "epoch": 74.53,
616
+ "learning_rate": 1.5847172689095234e-05,
617
+ "loss": 0.1027,
618
+ "step": 7900
619
+ },
620
+ {
621
+ "epoch": 75.47,
622
+ "learning_rate": 1.571281293255163e-05,
623
+ "loss": 0.0923,
624
+ "step": 8000
625
+ },
626
+ {
627
+ "epoch": 75.47,
628
+ "eval_loss": 0.4811612367630005,
629
+ "eval_runtime": 1.6131,
630
+ "eval_samples_per_second": 444.487,
631
+ "eval_steps_per_second": 3.72,
632
+ "eval_wer": 0.1558224891558225,
633
+ "step": 8000
634
+ },
635
+ {
636
+ "epoch": 76.42,
637
+ "learning_rate": 1.557827105015563e-05,
638
+ "loss": 0.1087,
639
+ "step": 8100
640
+ },
641
+ {
642
+ "epoch": 77.36,
643
+ "learning_rate": 1.5440865560714836e-05,
644
+ "loss": 0.0964,
645
+ "step": 8200
646
+ },
647
+ {
648
+ "epoch": 78.3,
649
+ "learning_rate": 1.5301985040137824e-05,
650
+ "loss": 0.0786,
651
+ "step": 8300
652
+ },
653
+ {
654
+ "epoch": 79.25,
655
+ "learning_rate": 1.5161667139252354e-05,
656
+ "loss": 0.0875,
657
+ "step": 8400
658
+ },
659
+ {
660
+ "epoch": 80.19,
661
+ "learning_rate": 1.5019949898563271e-05,
662
+ "loss": 0.086,
663
+ "step": 8500
664
+ },
665
+ {
666
+ "epoch": 80.19,
667
+ "eval_loss": 0.48815199732780457,
668
+ "eval_runtime": 1.7194,
669
+ "eval_samples_per_second": 417.01,
670
+ "eval_steps_per_second": 3.49,
671
+ "eval_wer": 0.18943943943943944,
672
+ "step": 8500
673
+ },
674
+ {
675
+ "epoch": 81.13,
676
+ "learning_rate": 1.4876871737939665e-05,
677
+ "loss": 0.0858,
678
+ "step": 8600
679
+ },
680
+ {
681
+ "epoch": 82.08,
682
+ "learning_rate": 1.4732471446199137e-05,
683
+ "loss": 0.0798,
684
+ "step": 8700
685
+ },
686
+ {
687
+ "epoch": 83.02,
688
+ "learning_rate": 1.4586788170592067e-05,
689
+ "loss": 0.0856,
690
+ "step": 8800
691
+ },
692
+ {
693
+ "epoch": 83.96,
694
+ "learning_rate": 1.4439861406188711e-05,
695
+ "loss": 0.0814,
696
+ "step": 8900
697
+ },
698
+ {
699
+ "epoch": 84.91,
700
+ "learning_rate": 1.429173098517198e-05,
701
+ "loss": 0.0872,
702
+ "step": 9000
703
+ },
704
+ {
705
+ "epoch": 84.91,
706
+ "eval_loss": 0.4662412703037262,
707
+ "eval_runtime": 1.6673,
708
+ "eval_samples_per_second": 430.025,
709
+ "eval_steps_per_second": 3.599,
710
+ "eval_wer": 0.17442442442442443,
711
+ "step": 9000
712
+ },
713
+ {
714
+ "epoch": 85.85,
715
+ "learning_rate": 1.4142437066038866e-05,
716
+ "loss": 0.0852,
717
+ "step": 9100
718
+ },
719
+ {
720
+ "epoch": 86.79,
721
+ "learning_rate": 1.3992020122713358e-05,
722
+ "loss": 0.0852,
723
+ "step": 9200
724
+ },
725
+ {
726
+ "epoch": 87.74,
727
+ "learning_rate": 1.3840520933573889e-05,
728
+ "loss": 0.0893,
729
+ "step": 9300
730
+ },
731
+ {
732
+ "epoch": 88.68,
733
+ "learning_rate": 1.368798057039821e-05,
734
+ "loss": 0.073,
735
+ "step": 9400
736
+ },
737
+ {
738
+ "epoch": 89.62,
739
+ "learning_rate": 1.3534440387228745e-05,
740
+ "loss": 0.0778,
741
+ "step": 9500
742
+ },
743
+ {
744
+ "epoch": 89.62,
745
+ "eval_loss": 0.48002195358276367,
746
+ "eval_runtime": 1.7065,
747
+ "eval_samples_per_second": 420.152,
748
+ "eval_steps_per_second": 3.516,
749
+ "eval_wer": 0.17500834167500834,
750
+ "step": 9500
751
+ },
752
+ {
753
+ "epoch": 90.57,
754
+ "learning_rate": 1.3379942009161407e-05,
755
+ "loss": 0.0825,
756
+ "step": 9600
757
+ },
758
+ {
759
+ "epoch": 91.51,
760
+ "learning_rate": 1.3224527321060953e-05,
761
+ "loss": 0.0701,
762
+ "step": 9700
763
+ },
764
+ {
765
+ "epoch": 92.45,
766
+ "learning_rate": 1.3068238456205888e-05,
767
+ "loss": 0.0753,
768
+ "step": 9800
769
+ },
770
+ {
771
+ "epoch": 93.4,
772
+ "learning_rate": 1.2911117784866037e-05,
773
+ "loss": 0.071,
774
+ "step": 9900
775
+ },
776
+ {
777
+ "epoch": 94.34,
778
+ "learning_rate": 1.2753207902815873e-05,
779
+ "loss": 0.0709,
780
+ "step": 10000
781
+ },
782
+ {
783
+ "epoch": 94.34,
784
+ "eval_loss": 0.507671594619751,
785
+ "eval_runtime": 1.7723,
786
+ "eval_samples_per_second": 404.551,
787
+ "eval_steps_per_second": 3.385,
788
+ "eval_wer": 0.19602936269602936,
789
+ "step": 10000
790
+ },
791
+ {
792
+ "epoch": 95.28,
793
+ "learning_rate": 1.259455161978671e-05,
794
+ "loss": 0.0878,
795
+ "step": 10100
796
+ },
797
+ {
798
+ "epoch": 96.23,
799
+ "learning_rate": 1.2435191947860867e-05,
800
+ "loss": 0.0662,
801
+ "step": 10200
802
+ },
803
+ {
804
+ "epoch": 97.17,
805
+ "learning_rate": 1.2275172089811021e-05,
806
+ "loss": 0.0713,
807
+ "step": 10300
808
+ },
809
+ {
810
+ "epoch": 98.11,
811
+ "learning_rate": 1.2114535427387812e-05,
812
+ "loss": 0.0746,
813
+ "step": 10400
814
+ },
815
+ {
816
+ "epoch": 99.06,
817
+ "learning_rate": 1.1953325509558964e-05,
818
+ "loss": 0.0703,
819
+ "step": 10500
820
+ },
821
+ {
822
+ "epoch": 99.06,
823
+ "eval_loss": 0.5037663578987122,
824
+ "eval_runtime": 1.6202,
825
+ "eval_samples_per_second": 442.547,
826
+ "eval_steps_per_second": 3.703,
827
+ "eval_wer": 0.17400734067400733,
828
+ "step": 10500
829
+ },
830
+ {
831
+ "epoch": 100.0,
832
+ "learning_rate": 1.1791586040703056e-05,
833
+ "loss": 0.0708,
834
+ "step": 10600
835
+ },
836
+ {
837
+ "epoch": 100.94,
838
+ "learning_rate": 1.1630985380407055e-05,
839
+ "loss": 0.0648,
840
+ "step": 10700
841
+ },
842
+ {
843
+ "epoch": 101.89,
844
+ "learning_rate": 1.1468322684127968e-05,
845
+ "loss": 0.0858,
846
+ "step": 10800
847
+ },
848
+ {
849
+ "epoch": 102.83,
850
+ "learning_rate": 1.130526192220052e-05,
851
+ "loss": 0.0701,
852
+ "step": 10900
853
+ },
854
+ {
855
+ "epoch": 103.77,
856
+ "learning_rate": 1.1141847300771358e-05,
857
+ "loss": 0.0721,
858
+ "step": 11000
859
+ },
860
+ {
861
+ "epoch": 103.77,
862
+ "eval_loss": 0.5130674242973328,
863
+ "eval_runtime": 1.7166,
864
+ "eval_samples_per_second": 417.691,
865
+ "eval_steps_per_second": 3.495,
866
+ "eval_wer": 0.176343009676343,
867
+ "step": 11000
868
+ },
869
+ {
870
+ "epoch": 104.72,
871
+ "learning_rate": 1.0978123121919275e-05,
872
+ "loss": 0.0724,
873
+ "step": 11100
874
+ },
875
+ {
876
+ "epoch": 105.66,
877
+ "learning_rate": 1.0814133771644783e-05,
878
+ "loss": 0.0722,
879
+ "step": 11200
880
+ },
881
+ {
882
+ "epoch": 106.6,
883
+ "learning_rate": 1.064992370783697e-05,
884
+ "loss": 0.0772,
885
+ "step": 11300
886
+ },
887
+ {
888
+ "epoch": 107.55,
889
+ "learning_rate": 1.0485537448220873e-05,
890
+ "loss": 0.0791,
891
+ "step": 11400
892
+ },
893
+ {
894
+ "epoch": 108.49,
895
+ "learning_rate": 1.0321019558288621e-05,
896
+ "loss": 0.0717,
897
+ "step": 11500
898
+ },
899
+ {
900
+ "epoch": 108.49,
901
+ "eval_loss": 0.5090663433074951,
902
+ "eval_runtime": 1.6704,
903
+ "eval_samples_per_second": 429.242,
904
+ "eval_steps_per_second": 3.592,
905
+ "eval_wer": 0.18960627293960627,
906
+ "step": 11500
907
+ },
908
+ {
909
+ "epoch": 109.43,
910
+ "learning_rate": 1.0156414639217615e-05,
911
+ "loss": 0.0551,
912
+ "step": 11600
913
+ },
914
+ {
915
+ "epoch": 110.38,
916
+ "learning_rate": 9.991767315779063e-06,
917
+ "loss": 0.0686,
918
+ "step": 11700
919
+ },
920
+ {
921
+ "epoch": 111.32,
922
+ "learning_rate": 9.827122224240092e-06,
923
+ "loss": 0.0673,
924
+ "step": 11800
925
+ },
926
+ {
927
+ "epoch": 112.26,
928
+ "learning_rate": 9.662524000262758e-06,
929
+ "loss": 0.0597,
930
+ "step": 11900
931
+ },
932
+ {
933
+ "epoch": 113.21,
934
+ "learning_rate": 9.4980172668032e-06,
935
+ "loss": 0.0818,
936
+ "step": 12000
937
+ },
938
+ {
939
+ "epoch": 113.21,
940
+ "eval_loss": 0.5172753930091858,
941
+ "eval_runtime": 1.6672,
942
+ "eval_samples_per_second": 430.058,
943
+ "eval_steps_per_second": 3.599,
944
+ "eval_wer": 0.1907741074407741,
945
+ "step": 12000
946
+ },
947
+ {
948
+ "epoch": 114.15,
949
+ "learning_rate": 9.33364662201429e-06,
950
+ "loss": 0.0615,
951
+ "step": 12100
952
+ },
953
+ {
954
+ "epoch": 115.09,
955
+ "learning_rate": 9.169456627154936e-06,
956
+ "loss": 0.0671,
957
+ "step": 12200
958
+ },
959
+ {
960
+ "epoch": 116.04,
961
+ "learning_rate": 9.005491794509433e-06,
962
+ "loss": 0.0727,
963
+ "step": 12300
964
+ },
965
+ {
966
+ "epoch": 116.98,
967
+ "learning_rate": 8.841796575320085e-06,
968
+ "loss": 0.0654,
969
+ "step": 12400
970
+ },
971
+ {
972
+ "epoch": 117.92,
973
+ "learning_rate": 8.67841534773638e-06,
974
+ "loss": 0.0626,
975
+ "step": 12500
976
+ },
977
+ {
978
+ "epoch": 117.92,
979
+ "eval_loss": 0.5157958269119263,
980
+ "eval_runtime": 1.7198,
981
+ "eval_samples_per_second": 416.917,
982
+ "eval_steps_per_second": 3.489,
983
+ "eval_wer": 0.18651985318651984,
984
+ "step": 12500
985
+ },
986
+ {
987
+ "epoch": 118.87,
988
+ "learning_rate": 8.515392404783955e-06,
989
+ "loss": 0.0566,
990
+ "step": 12600
991
+ },
992
+ {
993
+ "epoch": 119.81,
994
+ "learning_rate": 8.352771942356647e-06,
995
+ "loss": 0.0677,
996
+ "step": 12700
997
+ },
998
+ {
999
+ "epoch": 120.75,
1000
+ "learning_rate": 8.190598047234921e-06,
1001
+ "loss": 0.0546,
1002
+ "step": 12800
1003
+ },
1004
+ {
1005
+ "epoch": 121.7,
1006
+ "learning_rate": 8.028914685133818e-06,
1007
+ "loss": 0.0496,
1008
+ "step": 12900
1009
+ },
1010
+ {
1011
+ "epoch": 122.64,
1012
+ "learning_rate": 7.867765688783732e-06,
1013
+ "loss": 0.0749,
1014
+ "step": 13000
1015
+ },
1016
+ {
1017
+ "epoch": 122.64,
1018
+ "eval_loss": 0.5207752585411072,
1019
+ "eval_runtime": 1.6773,
1020
+ "eval_samples_per_second": 427.477,
1021
+ "eval_steps_per_second": 3.577,
1022
+ "eval_wer": 0.18651985318651984,
1023
+ "step": 13000
1024
+ },
1025
+ {
1026
+ "epoch": 123.58,
1027
+ "learning_rate": 7.70879745103896e-06,
1028
+ "loss": 0.0556,
1029
+ "step": 13100
1030
+ },
1031
+ {
1032
+ "epoch": 124.53,
1033
+ "learning_rate": 7.548841662283848e-06,
1034
+ "loss": 0.0496,
1035
+ "step": 13200
1036
+ },
1037
+ {
1038
+ "epoch": 125.47,
1039
+ "learning_rate": 7.389550388177662e-06,
1040
+ "loss": 0.0644,
1041
+ "step": 13300
1042
+ },
1043
+ {
1044
+ "epoch": 126.42,
1045
+ "learning_rate": 7.230966812950273e-06,
1046
+ "loss": 0.0558,
1047
+ "step": 13400
1048
+ },
1049
+ {
1050
+ "epoch": 127.36,
1051
+ "learning_rate": 7.073133928972745e-06,
1052
+ "loss": 0.0592,
1053
+ "step": 13500
1054
+ },
1055
+ {
1056
+ "epoch": 127.36,
1057
+ "eval_loss": 0.5244377851486206,
1058
+ "eval_runtime": 1.6104,
1059
+ "eval_samples_per_second": 445.242,
1060
+ "eval_steps_per_second": 3.726,
1061
+ "eval_wer": 0.17809476142809477,
1062
+ "step": 13500
1063
+ },
1064
+ {
1065
+ "epoch": 128.3,
1066
+ "learning_rate": 6.916094525102035e-06,
1067
+ "loss": 0.0605,
1068
+ "step": 13600
1069
+ },
1070
+ {
1071
+ "epoch": 129.25,
1072
+ "learning_rate": 6.759891175080827e-06,
1073
+ "loss": 0.0577,
1074
+ "step": 13700
1075
+ },
1076
+ {
1077
+ "epoch": 130.19,
1078
+ "learning_rate": 6.604566225995672e-06,
1079
+ "loss": 0.0708,
1080
+ "step": 13800
1081
+ },
1082
+ {
1083
+ "epoch": 131.13,
1084
+ "learning_rate": 6.450161786796625e-06,
1085
+ "loss": 0.0506,
1086
+ "step": 13900
1087
+ },
1088
+ {
1089
+ "epoch": 132.08,
1090
+ "learning_rate": 6.296719716881401e-06,
1091
+ "loss": 0.055,
1092
+ "step": 14000
1093
+ },
1094
+ {
1095
+ "epoch": 132.08,
1096
+ "eval_loss": 0.5302874445915222,
1097
+ "eval_runtime": 1.7512,
1098
+ "eval_samples_per_second": 409.439,
1099
+ "eval_steps_per_second": 3.426,
1100
+ "eval_wer": 0.18101434768101435,
1101
+ "step": 14000
1102
+ },
1103
+ {
1104
+ "epoch": 133.02,
1105
+ "learning_rate": 6.14428161474718e-06,
1106
+ "loss": 0.0603,
1107
+ "step": 14100
1108
+ },
1109
+ {
1110
+ "epoch": 133.96,
1111
+ "learning_rate": 5.9928888067131815e-06,
1112
+ "loss": 0.0508,
1113
+ "step": 14200
1114
+ },
1115
+ {
1116
+ "epoch": 134.91,
1117
+ "learning_rate": 5.842582335717009e-06,
1118
+ "loss": 0.062,
1119
+ "step": 14300
1120
+ },
1121
+ {
1122
+ "epoch": 135.85,
1123
+ "learning_rate": 5.693402950187774e-06,
1124
+ "loss": 0.0457,
1125
+ "step": 14400
1126
+ },
1127
+ {
1128
+ "epoch": 136.79,
1129
+ "learning_rate": 5.546865300283049e-06,
1130
+ "loss": 0.0487,
1131
+ "step": 14500
1132
+ },
1133
+ {
1134
+ "epoch": 136.79,
1135
+ "eval_loss": 0.5263693332672119,
1136
+ "eval_runtime": 1.6777,
1137
+ "eval_samples_per_second": 427.383,
1138
+ "eval_steps_per_second": 3.576,
1139
+ "eval_wer": 0.17392392392392392,
1140
+ "step": 14500
1141
+ },
1142
+ {
1143
+ "epoch": 137.74,
1144
+ "learning_rate": 5.400048823682746e-06,
1145
+ "loss": 0.0534,
1146
+ "step": 14600
1147
+ },
1148
+ {
1149
+ "epoch": 138.68,
1150
+ "learning_rate": 5.254479404399784e-06,
1151
+ "loss": 0.0554,
1152
+ "step": 14700
1153
+ },
1154
+ {
1155
+ "epoch": 139.62,
1156
+ "learning_rate": 5.110196506637705e-06,
1157
+ "loss": 0.0547,
1158
+ "step": 14800
1159
+ },
1160
+ {
1161
+ "epoch": 140.57,
1162
+ "learning_rate": 4.967239245821106e-06,
1163
+ "loss": 0.0625,
1164
+ "step": 14900
1165
+ },
1166
+ {
1167
+ "epoch": 141.51,
1168
+ "learning_rate": 4.825646377991382e-06,
1169
+ "loss": 0.0486,
1170
+ "step": 15000
1171
+ },
1172
+ {
1173
+ "epoch": 141.51,
1174
+ "eval_loss": 0.5225411057472229,
1175
+ "eval_runtime": 1.6507,
1176
+ "eval_samples_per_second": 434.366,
1177
+ "eval_steps_per_second": 3.635,
1178
+ "eval_wer": 0.18143143143143142,
1179
+ "step": 15000
1180
+ },
1181
+ {
1182
+ "epoch": 142.45,
1183
+ "learning_rate": 4.685456289299853e-06,
1184
+ "loss": 0.0511,
1185
+ "step": 15100
1186
+ },
1187
+ {
1188
+ "epoch": 143.4,
1189
+ "learning_rate": 4.5467069856011844e-06,
1190
+ "loss": 0.0473,
1191
+ "step": 15200
1192
+ },
1193
+ {
1194
+ "epoch": 144.34,
1195
+ "learning_rate": 4.409436082149942e-06,
1196
+ "loss": 0.0479,
1197
+ "step": 15300
1198
+ },
1199
+ {
1200
+ "epoch": 145.28,
1201
+ "learning_rate": 4.273680793402986e-06,
1202
+ "loss": 0.0523,
1203
+ "step": 15400
1204
+ },
1205
+ {
1206
+ "epoch": 146.23,
1207
+ "learning_rate": 4.139477922930541e-06,
1208
+ "loss": 0.0478,
1209
+ "step": 15500
1210
+ },
1211
+ {
1212
+ "epoch": 146.23,
1213
+ "eval_loss": 0.5315635800361633,
1214
+ "eval_runtime": 1.7624,
1215
+ "eval_samples_per_second": 406.828,
1216
+ "eval_steps_per_second": 3.404,
1217
+ "eval_wer": 0.18702035368702036,
1218
+ "step": 15500
1219
+ },
1220
+ {
1221
+ "epoch": 147.17,
1222
+ "learning_rate": 4.006863853438702e-06,
1223
+ "loss": 0.0518,
1224
+ "step": 15600
1225
+ },
1226
+ {
1227
+ "epoch": 148.11,
1228
+ "learning_rate": 3.875874536905955e-06,
1229
+ "loss": 0.0469,
1230
+ "step": 15700
1231
+ },
1232
+ {
1233
+ "epoch": 149.06,
1234
+ "learning_rate": 3.7465454848365266e-06,
1235
+ "loss": 0.0565,
1236
+ "step": 15800
1237
+ },
1238
+ {
1239
+ "epoch": 150.0,
1240
+ "learning_rate": 3.618911758633151e-06,
1241
+ "loss": 0.0453,
1242
+ "step": 15900
1243
+ },
1244
+ {
1245
+ "epoch": 150.94,
1246
+ "learning_rate": 3.4942583224667994e-06,
1247
+ "loss": 0.0453,
1248
+ "step": 16000
1249
+ },
1250
+ {
1251
+ "epoch": 150.94,
1252
+ "eval_loss": 0.5269920229911804,
1253
+ "eval_runtime": 1.6131,
1254
+ "eval_samples_per_second": 444.479,
1255
+ "eval_steps_per_second": 3.719,
1256
+ "eval_wer": 0.17759426092759426,
1257
+ "step": 16000
1258
+ },
1259
+ {
1260
+ "epoch": 151.89,
1261
+ "learning_rate": 3.371333510514454e-06,
1262
+ "loss": 0.0611,
1263
+ "step": 16100
1264
+ },
1265
+ {
1266
+ "epoch": 152.83,
1267
+ "learning_rate": 3.2489552065041274e-06,
1268
+ "loss": 0.0573,
1269
+ "step": 16200
1270
+ },
1271
+ {
1272
+ "epoch": 153.77,
1273
+ "learning_rate": 3.1284071262259064e-06,
1274
+ "loss": 0.0622,
1275
+ "step": 16300
1276
+ },
1277
+ {
1278
+ "epoch": 154.72,
1279
+ "learning_rate": 3.009721950540925e-06,
1280
+ "loss": 0.0496,
1281
+ "step": 16400
1282
+ },
1283
+ {
1284
+ "epoch": 155.66,
1285
+ "learning_rate": 2.8929318552725984e-06,
1286
+ "loss": 0.0449,
1287
+ "step": 16500
1288
+ },
1289
+ {
1290
+ "epoch": 155.66,
1291
+ "eval_loss": 0.5318202972412109,
1292
+ "eval_runtime": 1.9396,
1293
+ "eval_samples_per_second": 369.673,
1294
+ "eval_steps_per_second": 3.093,
1295
+ "eval_wer": 0.18209876543209877,
1296
+ "step": 16500
1297
+ },
1298
+ {
1299
+ "epoch": 156.6,
1300
+ "learning_rate": 2.7792074959517755e-06,
1301
+ "loss": 0.0557,
1302
+ "step": 16600
1303
+ },
1304
+ {
1305
+ "epoch": 157.55,
1306
+ "learning_rate": 2.666282294133101e-06,
1307
+ "loss": 0.0543,
1308
+ "step": 16700
1309
+ },
1310
+ {
1311
+ "epoch": 158.49,
1312
+ "learning_rate": 2.555345280009872e-06,
1313
+ "loss": 0.0503,
1314
+ "step": 16800
1315
+ },
1316
+ {
1317
+ "epoch": 159.43,
1318
+ "learning_rate": 2.44642652886114e-06,
1319
+ "loss": 0.0515,
1320
+ "step": 16900
1321
+ },
1322
+ {
1323
+ "epoch": 160.38,
1324
+ "learning_rate": 2.339555568810221e-06,
1325
+ "loss": 0.0585,
1326
+ "step": 17000
1327
+ },
1328
+ {
1329
+ "epoch": 160.38,
1330
+ "eval_loss": 0.5331774353981018,
1331
+ "eval_runtime": 1.7792,
1332
+ "eval_samples_per_second": 402.997,
1333
+ "eval_steps_per_second": 3.372,
1334
+ "eval_wer": 0.17751084417751084,
1335
+ "step": 17000
1336
+ },
1337
+ {
1338
+ "epoch": 161.32,
1339
+ "learning_rate": 2.234761372819577e-06,
1340
+ "loss": 0.0511,
1341
+ "step": 17100
1342
+ },
1343
+ {
1344
+ "epoch": 162.26,
1345
+ "learning_rate": 2.132072350836164e-06,
1346
+ "loss": 0.0416,
1347
+ "step": 17200
1348
+ },
1349
+ {
1350
+ "epoch": 163.21,
1351
+ "learning_rate": 2.031516342089438e-06,
1352
+ "loss": 0.0544,
1353
+ "step": 17300
1354
+ },
1355
+ {
1356
+ "epoch": 164.15,
1357
+ "learning_rate": 1.93312060754407e-06,
1358
+ "loss": 0.0577,
1359
+ "step": 17400
1360
+ },
1361
+ {
1362
+ "epoch": 165.09,
1363
+ "learning_rate": 1.8369118225094419e-06,
1364
+ "loss": 0.0481,
1365
+ "step": 17500
1366
+ },
1367
+ {
1368
+ "epoch": 165.09,
1369
+ "eval_loss": 0.5373131036758423,
1370
+ "eval_runtime": 1.687,
1371
+ "eval_samples_per_second": 425.014,
1372
+ "eval_steps_per_second": 3.557,
1373
+ "eval_wer": 0.17842842842842843,
1374
+ "step": 17500
1375
+ },
1376
+ {
1377
+ "epoch": 166.04,
1378
+ "learning_rate": 1.7429160694078983e-06,
1379
+ "loss": 0.0421,
1380
+ "step": 17600
1381
+ },
1382
+ {
1383
+ "epoch": 166.98,
1384
+ "learning_rate": 1.6511588307037596e-06,
1385
+ "loss": 0.046,
1386
+ "step": 17700
1387
+ },
1388
+ {
1389
+ "epoch": 167.92,
1390
+ "learning_rate": 1.5616649819949492e-06,
1391
+ "loss": 0.0488,
1392
+ "step": 17800
1393
+ },
1394
+ {
1395
+ "epoch": 168.87,
1396
+ "learning_rate": 1.474458785269176e-06,
1397
+ "loss": 0.047,
1398
+ "step": 17900
1399
+ },
1400
+ {
1401
+ "epoch": 169.81,
1402
+ "learning_rate": 1.3895638823264447e-06,
1403
+ "loss": 0.0459,
1404
+ "step": 18000
1405
+ },
1406
+ {
1407
+ "epoch": 169.81,
1408
+ "eval_loss": 0.5335074067115784,
1409
+ "eval_runtime": 1.7158,
1410
+ "eval_samples_per_second": 417.873,
1411
+ "eval_steps_per_second": 3.497,
1412
+ "eval_wer": 0.17559225892559227,
1413
+ "step": 18000
1414
+ },
1415
+ {
1416
+ "epoch": 170.75,
1417
+ "learning_rate": 1.3070032883697149e-06,
1418
+ "loss": 0.0511,
1419
+ "step": 18100
1420
+ },
1421
+ {
1422
+ "epoch": 171.7,
1423
+ "learning_rate": 1.2267993857654182e-06,
1424
+ "loss": 0.0454,
1425
+ "step": 18200
1426
+ },
1427
+ {
1428
+ "epoch": 172.64,
1429
+ "learning_rate": 1.1489739179755622e-06,
1430
+ "loss": 0.0542,
1431
+ "step": 18300
1432
+ },
1433
+ {
1434
+ "epoch": 173.58,
1435
+ "learning_rate": 1.0735479836630136e-06,
1436
+ "loss": 0.0423,
1437
+ "step": 18400
1438
+ },
1439
+ {
1440
+ "epoch": 174.53,
1441
+ "learning_rate": 1.0005420309716128e-06,
1442
+ "loss": 0.0473,
1443
+ "step": 18500
1444
+ },
1445
+ {
1446
+ "epoch": 174.53,
1447
+ "eval_loss": 0.5360230207443237,
1448
+ "eval_runtime": 1.7847,
1449
+ "eval_samples_per_second": 401.751,
1450
+ "eval_steps_per_second": 3.362,
1451
+ "eval_wer": 0.18084751418084752,
1452
+ "step": 18500
1453
+ },
1454
+ {
1455
+ "epoch": 175.47,
1456
+ "learning_rate": 9.299758519826274e-07,
1457
+ "loss": 0.0488,
1458
+ "step": 18600
1459
+ },
1460
+ {
1461
+ "epoch": 176.42,
1462
+ "learning_rate": 8.625374173369939e-07,
1463
+ "loss": 0.0481,
1464
+ "step": 18700
1465
+ },
1466
+ {
1467
+ "epoch": 177.36,
1468
+ "learning_rate": 7.968826482175651e-07,
1469
+ "loss": 0.0558,
1470
+ "step": 18800
1471
+ },
1472
+ {
1473
+ "epoch": 178.3,
1474
+ "learning_rate": 7.337228653264616e-07,
1475
+ "loss": 0.0482,
1476
+ "step": 18900
1477
+ },
1478
+ {
1479
+ "epoch": 179.25,
1480
+ "learning_rate": 6.730751914257305e-07,
1481
+ "loss": 0.0512,
1482
+ "step": 19000
1483
+ },
1484
+ {
1485
+ "epoch": 179.25,
1486
+ "eval_loss": 0.5346848368644714,
1487
+ "eval_runtime": 1.7297,
1488
+ "eval_samples_per_second": 414.521,
1489
+ "eval_steps_per_second": 3.469,
1490
+ "eval_wer": 0.17909576242909575,
1491
+ "step": 19000
1492
+ },
1493
+ {
1494
+ "epoch": 180.19,
1495
+ "learning_rate": 6.149560682389066e-07,
1496
+ "loss": 0.0428,
1497
+ "step": 19100
1498
+ },
1499
+ {
1500
+ "epoch": 181.13,
1501
+ "learning_rate": 5.59381251993627e-07,
1502
+ "loss": 0.0549,
1503
+ "step": 19200
1504
+ },
1505
+ {
1506
+ "epoch": 182.08,
1507
+ "learning_rate": 5.063658091500756e-07,
1508
+ "loss": 0.0477,
1509
+ "step": 19300
1510
+ },
1511
+ {
1512
+ "epoch": 183.02,
1513
+ "learning_rate": 4.559241123164404e-07,
1514
+ "loss": 0.0389,
1515
+ "step": 19400
1516
+ },
1517
+ {
1518
+ "epoch": 183.96,
1519
+ "learning_rate": 4.080698363524582e-07,
1520
+ "loss": 0.046,
1521
+ "step": 19500
1522
+ },
1523
+ {
1524
+ "epoch": 183.96,
1525
+ "eval_loss": 0.536734402179718,
1526
+ "eval_runtime": 1.8385,
1527
+ "eval_samples_per_second": 390.002,
1528
+ "eval_steps_per_second": 3.264,
1529
+ "eval_wer": 0.17776109442776108,
1530
+ "step": 19500
1531
+ },
1532
+ {
1533
+ "epoch": 184.91,
1534
+ "learning_rate": 3.6281595466213127e-07,
1535
+ "loss": 0.0475,
1536
+ "step": 19600
1537
+ },
1538
+ {
1539
+ "epoch": 185.85,
1540
+ "learning_rate": 3.201747356766105e-07,
1541
+ "loss": 0.0451,
1542
+ "step": 19700
1543
+ },
1544
+ {
1545
+ "epoch": 186.79,
1546
+ "learning_rate": 2.8054488337360244e-07,
1547
+ "loss": 0.0458,
1548
+ "step": 19800
1549
+ },
1550
+ {
1551
+ "epoch": 187.74,
1552
+ "learning_rate": 2.43136556688629e-07,
1553
+ "loss": 0.0468,
1554
+ "step": 19900
1555
+ },
1556
+ {
1557
+ "epoch": 188.68,
1558
+ "learning_rate": 2.0837333806782877e-07,
1559
+ "loss": 0.048,
1560
+ "step": 20000
1561
+ },
1562
+ {
1563
+ "epoch": 188.68,
1564
+ "eval_loss": 0.5353638529777527,
1565
+ "eval_runtime": 2.0173,
1566
+ "eval_samples_per_second": 355.431,
1567
+ "eval_steps_per_second": 2.974,
1568
+ "eval_wer": 0.17834501167834502,
1569
+ "step": 20000
1570
+ },
1571
+ {
1572
+ "epoch": 189.62,
1573
+ "learning_rate": 1.7626465189952146e-07,
1574
+ "loss": 0.0442,
1575
+ "step": 20100
1576
+ },
1577
+ {
1578
+ "epoch": 190.57,
1579
+ "learning_rate": 1.4681920292219598e-07,
1580
+ "loss": 0.0453,
1581
+ "step": 20200
1582
+ },
1583
+ {
1584
+ "epoch": 191.51,
1585
+ "learning_rate": 1.2004497386462476e-07,
1586
+ "loss": 0.0508,
1587
+ "step": 20300
1588
+ },
1589
+ {
1590
+ "epoch": 192.45,
1591
+ "learning_rate": 9.594922328173051e-08,
1592
+ "loss": 0.0483,
1593
+ "step": 20400
1594
+ },
1595
+ {
1596
+ "epoch": 193.4,
1597
+ "learning_rate": 7.453848358678018e-08,
1598
+ "loss": 0.0471,
1599
+ "step": 20500
1600
+ },
1601
+ {
1602
+ "epoch": 193.4,
1603
+ "eval_loss": 0.5366040468215942,
1604
+ "eval_runtime": 1.9042,
1605
+ "eval_samples_per_second": 376.544,
1606
+ "eval_steps_per_second": 3.151,
1607
+ "eval_wer": 0.18143143143143142,
1608
+ "step": 20500
1609
+ },
1610
+ {
1611
+ "epoch": 194.34,
1612
+ "learning_rate": 5.5818559280425014e-08,
1613
+ "loss": 0.0496,
1614
+ "step": 20600
1615
+ },
1616
+ {
1617
+ "epoch": 195.28,
1618
+ "learning_rate": 3.979452537709261e-08,
1619
+ "loss": 0.0584,
1620
+ "step": 20700
1621
+ },
1622
+ {
1623
+ "epoch": 196.23,
1624
+ "learning_rate": 2.647072602913747e-08,
1625
+ "loss": 0.0657,
1626
+ "step": 20800
1627
+ },
1628
+ {
1629
+ "epoch": 197.17,
1630
+ "learning_rate": 1.58507733491331e-08,
1631
+ "loss": 0.0578,
1632
+ "step": 20900
1633
+ },
1634
+ {
1635
+ "epoch": 198.11,
1636
+ "learning_rate": 7.937546430613063e-09,
1637
+ "loss": 0.0419,
1638
+ "step": 21000
1639
+ },
1640
+ {
1641
+ "epoch": 198.11,
1642
+ "eval_loss": 0.5363670587539673,
1643
+ "eval_runtime": 1.7593,
1644
+ "eval_samples_per_second": 407.539,
1645
+ "eval_steps_per_second": 3.41,
1646
+ "eval_wer": 0.18084751418084752,
1647
+ "step": 21000
1648
+ },
1649
+ {
1650
+ "epoch": 199.06,
1651
+ "learning_rate": 2.7331905675509473e-09,
1652
+ "loss": 0.0467,
1653
+ "step": 21100
1654
+ },
1655
+ {
1656
+ "epoch": 200.0,
1657
+ "learning_rate": 2.3911667276110826e-10,
1658
+ "loss": 0.0503,
1659
+ "step": 21200
1660
+ },
1661
+ {
1662
+ "epoch": 200.0,
1663
+ "step": 21200,
1664
+ "total_flos": 5.684431605865225e+19,
1665
+ "train_loss": 1.4449969133341087,
1666
+ "train_runtime": 6910.1807,
1667
+ "train_samples_per_second": 97.942,
1668
+ "train_steps_per_second": 3.068
1669
+ }
1670
+ ],
1671
+ "logging_steps": 100,
1672
+ "max_steps": 21200,
1673
+ "num_train_epochs": 200,
1674
+ "save_steps": 500,
1675
+ "total_flos": 5.684431605865225e+19,
1676
+ "trial_name": null,
1677
+ "trial_params": null
1678
+ }