gigant commited on
Commit
1981104
1 Parent(s): a924022

End of training

Browse files
all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 50.0,
3
+ "eval_cer": 0.029360946316590746,
4
+ "eval_loss": 0.15534016489982605,
5
+ "eval_runtime": 138.9476,
6
+ "eval_samples": 3843,
7
+ "eval_samples_per_second": 27.658,
8
+ "eval_steps_per_second": 3.462,
9
+ "eval_wer": 0.11739302551552884,
10
+ "train_loss": 0.17004985039278586,
11
+ "train_runtime": 77839.5185,
12
+ "train_samples": 30943,
13
+ "train_samples_per_second": 19.876,
14
+ "train_steps_per_second": 0.414
15
+ }
eval_results.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 50.0,
3
+ "eval_cer": 0.029360946316590746,
4
+ "eval_loss": 0.15534016489982605,
5
+ "eval_runtime": 138.9476,
6
+ "eval_samples": 3843,
7
+ "eval_samples_per_second": 27.658,
8
+ "eval_steps_per_second": 3.462,
9
+ "eval_wer": 0.11739302551552884
10
+ }
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:da47c33e2dcad1c1db5e0e7fce10a1c809005414f139c82d18016d368275913d
3
  size 1262071281
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0b801c7a4fa42854484db39077806d1d0997f8bb7dab06721d2d7f5a6e90ccc
3
  size 1262071281
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 50.0,
3
+ "train_loss": 0.17004985039278586,
4
+ "train_runtime": 77839.5185,
5
+ "train_samples": 30943,
6
+ "train_samples_per_second": 19.876,
7
+ "train_steps_per_second": 0.414
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,1049 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 49.99896587383661,
5
+ "global_step": 32200,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.78,
12
+ "learning_rate": 0.000998,
13
+ "loss": 2.9272,
14
+ "step": 500
15
+ },
16
+ {
17
+ "epoch": 0.78,
18
+ "eval_cer": 0.23553399465944927,
19
+ "eval_loss": 0.7603499889373779,
20
+ "eval_runtime": 139.9494,
21
+ "eval_samples_per_second": 27.46,
22
+ "eval_steps_per_second": 3.437,
23
+ "eval_wer": 0.7733832367653939,
24
+ "step": 500
25
+ },
26
+ {
27
+ "epoch": 1.55,
28
+ "learning_rate": 0.0009842586750788643,
29
+ "loss": 0.6157,
30
+ "step": 1000
31
+ },
32
+ {
33
+ "epoch": 1.55,
34
+ "eval_cer": 0.12468736029385112,
35
+ "eval_loss": 0.40030696988105774,
36
+ "eval_runtime": 143.7328,
37
+ "eval_samples_per_second": 26.737,
38
+ "eval_steps_per_second": 3.346,
39
+ "eval_wer": 0.48655846259041996,
40
+ "step": 1000
41
+ },
42
+ {
43
+ "epoch": 2.33,
44
+ "learning_rate": 0.0009684858044164038,
45
+ "loss": 0.4452,
46
+ "step": 1500
47
+ },
48
+ {
49
+ "epoch": 2.33,
50
+ "eval_cer": 0.0910430989693462,
51
+ "eval_loss": 0.2960306406021118,
52
+ "eval_runtime": 141.1265,
53
+ "eval_samples_per_second": 27.231,
54
+ "eval_steps_per_second": 3.408,
55
+ "eval_wer": 0.3688775326591572,
56
+ "step": 1500
57
+ },
58
+ {
59
+ "epoch": 3.11,
60
+ "learning_rate": 0.0009527129337539432,
61
+ "loss": 0.3631,
62
+ "step": 2000
63
+ },
64
+ {
65
+ "epoch": 3.11,
66
+ "eval_cer": 0.07961287048560348,
67
+ "eval_loss": 0.25798073410987854,
68
+ "eval_runtime": 138.8783,
69
+ "eval_samples_per_second": 27.672,
70
+ "eval_steps_per_second": 3.463,
71
+ "eval_wer": 0.3205455788678159,
72
+ "step": 2000
73
+ },
74
+ {
75
+ "epoch": 3.88,
76
+ "learning_rate": 0.0009369400630914827,
77
+ "loss": 0.3153,
78
+ "step": 2500
79
+ },
80
+ {
81
+ "epoch": 3.88,
82
+ "eval_cer": 0.07467709000398728,
83
+ "eval_loss": 0.24647507071495056,
84
+ "eval_runtime": 136.7566,
85
+ "eval_samples_per_second": 28.101,
86
+ "eval_steps_per_second": 3.517,
87
+ "eval_wer": 0.29772915392089827,
88
+ "step": 2500
89
+ },
90
+ {
91
+ "epoch": 4.66,
92
+ "learning_rate": 0.0009211671924290221,
93
+ "loss": 0.2795,
94
+ "step": 3000
95
+ },
96
+ {
97
+ "epoch": 4.66,
98
+ "eval_cer": 0.06940299409156266,
99
+ "eval_loss": 0.2273630052804947,
100
+ "eval_runtime": 144.7547,
101
+ "eval_samples_per_second": 26.548,
102
+ "eval_steps_per_second": 3.323,
103
+ "eval_wer": 0.27890740274228953,
104
+ "step": 3000
105
+ },
106
+ {
107
+ "epoch": 5.43,
108
+ "learning_rate": 0.0009053943217665615,
109
+ "loss": 0.2615,
110
+ "step": 3500
111
+ },
112
+ {
113
+ "epoch": 5.43,
114
+ "eval_cer": 0.06749392844628638,
115
+ "eval_loss": 0.22766220569610596,
116
+ "eval_runtime": 139.1146,
117
+ "eval_samples_per_second": 27.625,
118
+ "eval_steps_per_second": 3.458,
119
+ "eval_wer": 0.26850685572389965,
120
+ "step": 3500
121
+ },
122
+ {
123
+ "epoch": 6.21,
124
+ "learning_rate": 0.000889621451104101,
125
+ "loss": 0.2389,
126
+ "step": 4000
127
+ },
128
+ {
129
+ "epoch": 6.21,
130
+ "eval_cer": 0.06267897490424465,
131
+ "eval_loss": 0.21347470581531525,
132
+ "eval_runtime": 143.4039,
133
+ "eval_samples_per_second": 26.798,
134
+ "eval_steps_per_second": 3.354,
135
+ "eval_wer": 0.2517724115593623,
136
+ "step": 4000
137
+ },
138
+ {
139
+ "epoch": 6.99,
140
+ "learning_rate": 0.0008738485804416404,
141
+ "loss": 0.2229,
142
+ "step": 4500
143
+ },
144
+ {
145
+ "epoch": 6.99,
146
+ "eval_cer": 0.06139216799777678,
147
+ "eval_loss": 0.20536719262599945,
148
+ "eval_runtime": 140.3579,
149
+ "eval_samples_per_second": 27.38,
150
+ "eval_steps_per_second": 3.427,
151
+ "eval_wer": 0.24493468168568036,
152
+ "step": 4500
153
+ },
154
+ {
155
+ "epoch": 7.76,
156
+ "learning_rate": 0.0008580757097791798,
157
+ "loss": 0.2067,
158
+ "step": 5000
159
+ },
160
+ {
161
+ "epoch": 7.76,
162
+ "eval_cer": 0.05967038410884091,
163
+ "eval_loss": 0.20964160561561584,
164
+ "eval_runtime": 144.9834,
165
+ "eval_samples_per_second": 26.506,
166
+ "eval_steps_per_second": 3.318,
167
+ "eval_wer": 0.23780904739626443,
168
+ "step": 5000
169
+ },
170
+ {
171
+ "epoch": 8.54,
172
+ "learning_rate": 0.0008423028391167192,
173
+ "loss": 0.1977,
174
+ "step": 5500
175
+ },
176
+ {
177
+ "epoch": 8.54,
178
+ "eval_cer": 0.06000265819267064,
179
+ "eval_loss": 0.20422270894050598,
180
+ "eval_runtime": 139.8836,
181
+ "eval_samples_per_second": 27.473,
182
+ "eval_steps_per_second": 3.439,
183
+ "eval_wer": 0.23874473674739985,
184
+ "step": 5500
185
+ },
186
+ {
187
+ "epoch": 9.32,
188
+ "learning_rate": 0.0008265299684542586,
189
+ "loss": 0.1896,
190
+ "step": 6000
191
+ },
192
+ {
193
+ "epoch": 9.32,
194
+ "eval_cer": 0.059458936964585624,
195
+ "eval_loss": 0.2110028713941574,
196
+ "eval_runtime": 139.3458,
197
+ "eval_samples_per_second": 27.579,
198
+ "eval_steps_per_second": 3.452,
199
+ "eval_wer": 0.23827689207183214,
200
+ "step": 6000
201
+ },
202
+ {
203
+ "epoch": 10.09,
204
+ "learning_rate": 0.000810757097791798,
205
+ "loss": 0.1801,
206
+ "step": 6500
207
+ },
208
+ {
209
+ "epoch": 10.09,
210
+ "eval_cer": 0.05477689305607578,
211
+ "eval_loss": 0.19087238609790802,
212
+ "eval_runtime": 138.0429,
213
+ "eval_samples_per_second": 27.839,
214
+ "eval_steps_per_second": 3.484,
215
+ "eval_wer": 0.21650412063195018,
216
+ "step": 6500
217
+ },
218
+ {
219
+ "epoch": 10.87,
220
+ "learning_rate": 0.0007949842271293376,
221
+ "loss": 0.174,
222
+ "step": 7000
223
+ },
224
+ {
225
+ "epoch": 10.87,
226
+ "eval_cer": 0.05590662494109687,
227
+ "eval_loss": 0.18831488490104675,
228
+ "eval_runtime": 140.5023,
229
+ "eval_samples_per_second": 27.352,
230
+ "eval_steps_per_second": 3.423,
231
+ "eval_wer": 0.2206427466081261,
232
+ "step": 7000
233
+ },
234
+ {
235
+ "epoch": 11.65,
236
+ "learning_rate": 0.000779211356466877,
237
+ "loss": 0.1685,
238
+ "step": 7500
239
+ },
240
+ {
241
+ "epoch": 11.65,
242
+ "eval_cer": 0.052807413941012286,
243
+ "eval_loss": 0.1847681999206543,
244
+ "eval_runtime": 140.5872,
245
+ "eval_samples_per_second": 27.335,
246
+ "eval_steps_per_second": 3.421,
247
+ "eval_wer": 0.209702378810235,
248
+ "step": 7500
249
+ },
250
+ {
251
+ "epoch": 12.42,
252
+ "learning_rate": 0.0007634384858044164,
253
+ "loss": 0.1591,
254
+ "step": 8000
255
+ },
256
+ {
257
+ "epoch": 12.42,
258
+ "eval_cer": 0.05136353201309764,
259
+ "eval_loss": 0.18508584797382355,
260
+ "eval_runtime": 138.1047,
261
+ "eval_samples_per_second": 27.827,
262
+ "eval_steps_per_second": 3.483,
263
+ "eval_wer": 0.20394429049555549,
264
+ "step": 8000
265
+ },
266
+ {
267
+ "epoch": 13.2,
268
+ "learning_rate": 0.0007476971608832807,
269
+ "loss": 0.1537,
270
+ "step": 8500
271
+ },
272
+ {
273
+ "epoch": 13.2,
274
+ "eval_cer": 0.05175017821973587,
275
+ "eval_loss": 0.18811114132404327,
276
+ "eval_runtime": 143.7666,
277
+ "eval_samples_per_second": 26.731,
278
+ "eval_steps_per_second": 3.346,
279
+ "eval_wer": 0.20653543023716125,
280
+ "step": 8500
281
+ },
282
+ {
283
+ "epoch": 13.97,
284
+ "learning_rate": 0.0007319242902208201,
285
+ "loss": 0.1504,
286
+ "step": 9000
287
+ },
288
+ {
289
+ "epoch": 13.97,
290
+ "eval_cer": 0.04994381547309788,
291
+ "eval_loss": 0.18396763503551483,
292
+ "eval_runtime": 144.751,
293
+ "eval_samples_per_second": 26.549,
294
+ "eval_steps_per_second": 3.323,
295
+ "eval_wer": 0.1972145247777738,
296
+ "step": 9000
297
+ },
298
+ {
299
+ "epoch": 14.75,
300
+ "learning_rate": 0.0007161514195583596,
301
+ "loss": 0.145,
302
+ "step": 9500
303
+ },
304
+ {
305
+ "epoch": 14.75,
306
+ "eval_cer": 0.05168976474994865,
307
+ "eval_loss": 0.1845300793647766,
308
+ "eval_runtime": 142.1949,
309
+ "eval_samples_per_second": 27.026,
310
+ "eval_steps_per_second": 3.383,
311
+ "eval_wer": 0.20293662504048657,
312
+ "step": 9500
313
+ },
314
+ {
315
+ "epoch": 15.53,
316
+ "learning_rate": 0.000700378548895899,
317
+ "loss": 0.1417,
318
+ "step": 10000
319
+ },
320
+ {
321
+ "epoch": 15.53,
322
+ "eval_cer": 0.050717107886374346,
323
+ "eval_loss": 0.18841156363487244,
324
+ "eval_runtime": 137.4986,
325
+ "eval_samples_per_second": 27.949,
326
+ "eval_steps_per_second": 3.498,
327
+ "eval_wer": 0.20027350919494727,
328
+ "step": 10000
329
+ },
330
+ {
331
+ "epoch": 16.3,
332
+ "learning_rate": 0.0006846056782334384,
333
+ "loss": 0.1364,
334
+ "step": 10500
335
+ },
336
+ {
337
+ "epoch": 16.3,
338
+ "eval_cer": 0.05172601283182098,
339
+ "eval_loss": 0.20098073780536652,
340
+ "eval_runtime": 138.6706,
341
+ "eval_samples_per_second": 27.713,
342
+ "eval_steps_per_second": 3.469,
343
+ "eval_wer": 0.2036563860798215,
344
+ "step": 10500
345
+ },
346
+ {
347
+ "epoch": 17.08,
348
+ "learning_rate": 0.0006688643533123028,
349
+ "loss": 0.1331,
350
+ "step": 11000
351
+ },
352
+ {
353
+ "epoch": 17.08,
354
+ "eval_cer": 0.04831265178884284,
355
+ "eval_loss": 0.183824360370636,
356
+ "eval_runtime": 137.3455,
357
+ "eval_samples_per_second": 27.981,
358
+ "eval_steps_per_second": 3.502,
359
+ "eval_wer": 0.19232014971029618,
360
+ "step": 11000
361
+ },
362
+ {
363
+ "epoch": 17.86,
364
+ "learning_rate": 0.0006530914826498424,
365
+ "loss": 0.129,
366
+ "step": 11500
367
+ },
368
+ {
369
+ "epoch": 17.86,
370
+ "eval_cer": 0.048886579751821466,
371
+ "eval_loss": 0.18181012570858002,
372
+ "eval_runtime": 140.6277,
373
+ "eval_samples_per_second": 27.327,
374
+ "eval_steps_per_second": 3.42,
375
+ "eval_wer": 0.19221218555439595,
376
+ "step": 11500
377
+ },
378
+ {
379
+ "epoch": 18.63,
380
+ "learning_rate": 0.0006373186119873818,
381
+ "loss": 0.1198,
382
+ "step": 12000
383
+ },
384
+ {
385
+ "epoch": 18.63,
386
+ "eval_cer": 0.04653649577709846,
387
+ "eval_loss": 0.17603321373462677,
388
+ "eval_runtime": 144.3916,
389
+ "eval_samples_per_second": 26.615,
390
+ "eval_steps_per_second": 3.331,
391
+ "eval_wer": 0.1861302047720157,
392
+ "step": 12000
393
+ },
394
+ {
395
+ "epoch": 19.41,
396
+ "learning_rate": 0.0006215457413249212,
397
+ "loss": 0.1203,
398
+ "step": 12500
399
+ },
400
+ {
401
+ "epoch": 19.41,
402
+ "eval_cer": 0.04645795826637507,
403
+ "eval_loss": 0.168550506234169,
404
+ "eval_runtime": 142.0056,
405
+ "eval_samples_per_second": 27.062,
406
+ "eval_steps_per_second": 3.387,
407
+ "eval_wer": 0.18393493360204413,
408
+ "step": 12500
409
+ },
410
+ {
411
+ "epoch": 20.19,
412
+ "learning_rate": 0.0006058044164037855,
413
+ "loss": 0.1225,
414
+ "step": 13000
415
+ },
416
+ {
417
+ "epoch": 20.19,
418
+ "eval_cer": 0.047938088276162055,
419
+ "eval_loss": 0.18283818662166595,
420
+ "eval_runtime": 142.8894,
421
+ "eval_samples_per_second": 26.895,
422
+ "eval_steps_per_second": 3.366,
423
+ "eval_wer": 0.19199625724259545,
424
+ "step": 13000
425
+ },
426
+ {
427
+ "epoch": 20.96,
428
+ "learning_rate": 0.0005900315457413249,
429
+ "loss": 0.1145,
430
+ "step": 13500
431
+ },
432
+ {
433
+ "epoch": 20.96,
434
+ "eval_cer": 0.04463347147880092,
435
+ "eval_loss": 0.16731047630310059,
436
+ "eval_runtime": 141.4047,
437
+ "eval_samples_per_second": 27.177,
438
+ "eval_steps_per_second": 3.402,
439
+ "eval_wer": 0.17839277359916508,
440
+ "step": 13500
441
+ },
442
+ {
443
+ "epoch": 21.74,
444
+ "learning_rate": 0.0005742586750788644,
445
+ "loss": 0.1053,
446
+ "step": 14000
447
+ },
448
+ {
449
+ "epoch": 21.74,
450
+ "eval_cer": 0.04560008699539649,
451
+ "eval_loss": 0.1801901012659073,
452
+ "eval_runtime": 142.8084,
453
+ "eval_samples_per_second": 26.91,
454
+ "eval_steps_per_second": 3.368,
455
+ "eval_wer": 0.18098391334077085,
456
+ "step": 14000
457
+ },
458
+ {
459
+ "epoch": 22.51,
460
+ "learning_rate": 0.0005584858044164039,
461
+ "loss": 0.1071,
462
+ "step": 14500
463
+ },
464
+ {
465
+ "epoch": 22.51,
466
+ "eval_cer": 0.044385776252673294,
467
+ "eval_loss": 0.17691390216350555,
468
+ "eval_runtime": 140.9354,
469
+ "eval_samples_per_second": 27.268,
470
+ "eval_steps_per_second": 3.413,
471
+ "eval_wer": 0.17752906035196314,
472
+ "step": 14500
473
+ },
474
+ {
475
+ "epoch": 23.29,
476
+ "learning_rate": 0.0005427444794952682,
477
+ "loss": 0.1053,
478
+ "step": 15000
479
+ },
480
+ {
481
+ "epoch": 23.29,
482
+ "eval_cer": 0.04566654181216244,
483
+ "eval_loss": 0.19200138747692108,
484
+ "eval_runtime": 142.1681,
485
+ "eval_samples_per_second": 27.031,
486
+ "eval_steps_per_second": 3.383,
487
+ "eval_wer": 0.17832079749523158,
488
+ "step": 15000
489
+ },
490
+ {
491
+ "epoch": 24.07,
492
+ "learning_rate": 0.0005270031545741325,
493
+ "loss": 0.1024,
494
+ "step": 15500
495
+ },
496
+ {
497
+ "epoch": 24.07,
498
+ "eval_cer": 0.04464555417275836,
499
+ "eval_loss": 0.19041383266448975,
500
+ "eval_runtime": 138.4185,
501
+ "eval_samples_per_second": 27.764,
502
+ "eval_steps_per_second": 3.475,
503
+ "eval_wer": 0.17752906035196314,
504
+ "step": 15500
505
+ },
506
+ {
507
+ "epoch": 24.84,
508
+ "learning_rate": 0.000511230283911672,
509
+ "loss": 0.0987,
510
+ "step": 16000
511
+ },
512
+ {
513
+ "epoch": 24.84,
514
+ "eval_cer": 0.044615347437864744,
515
+ "eval_loss": 0.17931906878948212,
516
+ "eval_runtime": 138.8631,
517
+ "eval_samples_per_second": 27.675,
518
+ "eval_steps_per_second": 3.464,
519
+ "eval_wer": 0.17616151437722677,
520
+ "step": 16000
521
+ },
522
+ {
523
+ "epoch": 25.62,
524
+ "learning_rate": 0.0004954574132492114,
525
+ "loss": 0.0949,
526
+ "step": 16500
527
+ },
528
+ {
529
+ "epoch": 25.62,
530
+ "eval_cer": 0.04433744547684352,
531
+ "eval_loss": 0.1801224648952484,
532
+ "eval_runtime": 140.9756,
533
+ "eval_samples_per_second": 27.26,
534
+ "eval_steps_per_second": 3.412,
535
+ "eval_wer": 0.17655738294886097,
536
+ "step": 16500
537
+ },
538
+ {
539
+ "epoch": 26.4,
540
+ "learning_rate": 0.00047968454258675084,
541
+ "loss": 0.0942,
542
+ "step": 17000
543
+ },
544
+ {
545
+ "epoch": 26.4,
546
+ "eval_cer": 0.04232567693292896,
547
+ "eval_loss": 0.17306886613368988,
548
+ "eval_runtime": 140.9004,
549
+ "eval_samples_per_second": 27.275,
550
+ "eval_steps_per_second": 3.414,
551
+ "eval_wer": 0.1659409076186706,
552
+ "step": 17000
553
+ },
554
+ {
555
+ "epoch": 27.17,
556
+ "learning_rate": 0.00046391167192429025,
557
+ "loss": 0.0906,
558
+ "step": 17500
559
+ },
560
+ {
561
+ "epoch": 27.17,
562
+ "eval_cer": 0.042374007708758744,
563
+ "eval_loss": 0.17762278020381927,
564
+ "eval_runtime": 142.288,
565
+ "eval_samples_per_second": 27.009,
566
+ "eval_steps_per_second": 3.38,
567
+ "eval_wer": 0.16975564112714578,
568
+ "step": 17500
569
+ },
570
+ {
571
+ "epoch": 27.95,
572
+ "learning_rate": 0.00044813880126182967,
573
+ "loss": 0.0861,
574
+ "step": 18000
575
+ },
576
+ {
577
+ "epoch": 27.95,
578
+ "eval_cer": 0.040622017084929254,
579
+ "eval_loss": 0.17157311737537384,
580
+ "eval_runtime": 139.3451,
581
+ "eval_samples_per_second": 27.579,
582
+ "eval_steps_per_second": 3.452,
583
+ "eval_wer": 0.16000287904415733,
584
+ "step": 18000
585
+ },
586
+ {
587
+ "epoch": 28.73,
588
+ "learning_rate": 0.00043239747634069405,
589
+ "loss": 0.0851,
590
+ "step": 18500
591
+ },
592
+ {
593
+ "epoch": 28.73,
594
+ "eval_cer": 0.041008663291567486,
595
+ "eval_loss": 0.16616342961788177,
596
+ "eval_runtime": 138.87,
597
+ "eval_samples_per_second": 27.673,
598
+ "eval_steps_per_second": 3.464,
599
+ "eval_wer": 0.1629538993054306,
600
+ "step": 18500
601
+ },
602
+ {
603
+ "epoch": 29.5,
604
+ "learning_rate": 0.00041662460567823347,
605
+ "loss": 0.0844,
606
+ "step": 19000
607
+ },
608
+ {
609
+ "epoch": 29.5,
610
+ "eval_cer": 0.039298962096589056,
611
+ "eval_loss": 0.16713382303714752,
612
+ "eval_runtime": 139.8577,
613
+ "eval_samples_per_second": 27.478,
614
+ "eval_steps_per_second": 3.439,
615
+ "eval_wer": 0.1572317990427178,
616
+ "step": 19000
617
+ },
618
+ {
619
+ "epoch": 30.28,
620
+ "learning_rate": 0.0004008517350157729,
621
+ "loss": 0.0792,
622
+ "step": 19500
623
+ },
624
+ {
625
+ "epoch": 30.28,
626
+ "eval_cer": 0.04074888537148243,
627
+ "eval_loss": 0.17681923508644104,
628
+ "eval_runtime": 140.8425,
629
+ "eval_samples_per_second": 27.286,
630
+ "eval_steps_per_second": 3.415,
631
+ "eval_wer": 0.15993090294022386,
632
+ "step": 19500
633
+ },
634
+ {
635
+ "epoch": 31.06,
636
+ "learning_rate": 0.0003850788643533123,
637
+ "loss": 0.0798,
638
+ "step": 20000
639
+ },
640
+ {
641
+ "epoch": 31.06,
642
+ "eval_cer": 0.039365416913355,
643
+ "eval_loss": 0.17318624258041382,
644
+ "eval_runtime": 138.8235,
645
+ "eval_samples_per_second": 27.683,
646
+ "eval_steps_per_second": 3.465,
647
+ "eval_wer": 0.15579227696404793,
648
+ "step": 20000
649
+ },
650
+ {
651
+ "epoch": 31.83,
652
+ "learning_rate": 0.0003693059936908517,
653
+ "loss": 0.0779,
654
+ "step": 20500
655
+ },
656
+ {
657
+ "epoch": 31.83,
658
+ "eval_cer": 0.03882773703224871,
659
+ "eval_loss": 0.1694292426109314,
660
+ "eval_runtime": 142.3851,
661
+ "eval_samples_per_second": 26.99,
662
+ "eval_steps_per_second": 3.378,
663
+ "eval_wer": 0.15442473098931156,
664
+ "step": 20500
665
+ },
666
+ {
667
+ "epoch": 32.61,
668
+ "learning_rate": 0.0003535331230283912,
669
+ "loss": 0.0718,
670
+ "step": 21000
671
+ },
672
+ {
673
+ "epoch": 32.61,
674
+ "eval_cer": 0.03985476601863151,
675
+ "eval_loss": 0.17093372344970703,
676
+ "eval_runtime": 144.4958,
677
+ "eval_samples_per_second": 26.596,
678
+ "eval_steps_per_second": 3.329,
679
+ "eval_wer": 0.15784359592615252,
680
+ "step": 21000
681
+ },
682
+ {
683
+ "epoch": 33.38,
684
+ "learning_rate": 0.0003377917981072555,
685
+ "loss": 0.0732,
686
+ "step": 21500
687
+ },
688
+ {
689
+ "epoch": 33.38,
690
+ "eval_cer": 0.03907543225837633,
691
+ "eval_loss": 0.1696743220090866,
692
+ "eval_runtime": 141.4742,
693
+ "eval_samples_per_second": 27.164,
694
+ "eval_steps_per_second": 3.4,
695
+ "eval_wer": 0.15226544787130672,
696
+ "step": 21500
697
+ },
698
+ {
699
+ "epoch": 34.16,
700
+ "learning_rate": 0.00032201892744479493,
701
+ "loss": 0.0708,
702
+ "step": 22000
703
+ },
704
+ {
705
+ "epoch": 34.16,
706
+ "eval_cer": 0.037462392615057453,
707
+ "eval_loss": 0.16164687275886536,
708
+ "eval_runtime": 142.4307,
709
+ "eval_samples_per_second": 26.982,
710
+ "eval_steps_per_second": 3.377,
711
+ "eval_wer": 0.14744304890776264,
712
+ "step": 22000
713
+ },
714
+ {
715
+ "epoch": 34.94,
716
+ "learning_rate": 0.0003062460567823344,
717
+ "loss": 0.0678,
718
+ "step": 22500
719
+ },
720
+ {
721
+ "epoch": 34.94,
722
+ "eval_cer": 0.037498640696929784,
723
+ "eval_loss": 0.16977904736995697,
724
+ "eval_runtime": 140.9926,
725
+ "eval_samples_per_second": 27.257,
726
+ "eval_steps_per_second": 3.412,
727
+ "eval_wer": 0.14737107280382913,
728
+ "step": 22500
729
+ },
730
+ {
731
+ "epoch": 35.71,
732
+ "learning_rate": 0.0002904731861198738,
733
+ "loss": 0.0642,
734
+ "step": 23000
735
+ },
736
+ {
737
+ "epoch": 35.71,
738
+ "eval_cer": 0.036900547346036275,
739
+ "eval_loss": 0.16813962161540985,
740
+ "eval_runtime": 139.7128,
741
+ "eval_samples_per_second": 27.506,
742
+ "eval_steps_per_second": 3.443,
743
+ "eval_wer": 0.14593155072515926,
744
+ "step": 23000
745
+ },
746
+ {
747
+ "epoch": 36.49,
748
+ "learning_rate": 0.00027470031545741324,
749
+ "loss": 0.0661,
750
+ "step": 23500
751
+ },
752
+ {
753
+ "epoch": 36.49,
754
+ "eval_cer": 0.03566207121539818,
755
+ "eval_loss": 0.1611974984407425,
756
+ "eval_runtime": 141.2001,
757
+ "eval_samples_per_second": 27.217,
758
+ "eval_steps_per_second": 3.407,
759
+ "eval_wer": 0.14110915176161515,
760
+ "step": 23500
761
+ },
762
+ {
763
+ "epoch": 37.27,
764
+ "learning_rate": 0.00025892744479495266,
765
+ "loss": 0.0629,
766
+ "step": 24000
767
+ },
768
+ {
769
+ "epoch": 37.27,
770
+ "eval_cer": 0.035529161581866295,
771
+ "eval_loss": 0.1662248969078064,
772
+ "eval_runtime": 141.4579,
773
+ "eval_samples_per_second": 27.167,
774
+ "eval_steps_per_second": 3.4,
775
+ "eval_wer": 0.14143304422931588,
776
+ "step": 24000
777
+ },
778
+ {
779
+ "epoch": 38.04,
780
+ "learning_rate": 0.00024315457413249213,
781
+ "loss": 0.0587,
782
+ "step": 24500
783
+ },
784
+ {
785
+ "epoch": 38.04,
786
+ "eval_cer": 0.03514251537522806,
787
+ "eval_loss": 0.16591240465641022,
788
+ "eval_runtime": 138.7975,
789
+ "eval_samples_per_second": 27.688,
790
+ "eval_steps_per_second": 3.465,
791
+ "eval_wer": 0.14078525929391442,
792
+ "step": 24500
793
+ },
794
+ {
795
+ "epoch": 38.82,
796
+ "learning_rate": 0.00022738170347003155,
797
+ "loss": 0.0581,
798
+ "step": 25000
799
+ },
800
+ {
801
+ "epoch": 38.82,
802
+ "eval_cer": 0.03515459806918551,
803
+ "eval_loss": 0.1612222045660019,
804
+ "eval_runtime": 140.4043,
805
+ "eval_samples_per_second": 27.371,
806
+ "eval_steps_per_second": 3.426,
807
+ "eval_wer": 0.13823010760427537,
808
+ "step": 25000
809
+ },
810
+ {
811
+ "epoch": 39.6,
812
+ "learning_rate": 0.000211608832807571,
813
+ "loss": 0.0556,
814
+ "step": 25500
815
+ },
816
+ {
817
+ "epoch": 39.6,
818
+ "eval_cer": 0.03452629798339838,
819
+ "eval_loss": 0.16471649706363678,
820
+ "eval_runtime": 139.8247,
821
+ "eval_samples_per_second": 27.484,
822
+ "eval_steps_per_second": 3.44,
823
+ "eval_wer": 0.1376183107208407,
824
+ "step": 25500
825
+ },
826
+ {
827
+ "epoch": 40.37,
828
+ "learning_rate": 0.00019586750788643534,
829
+ "loss": 0.0543,
830
+ "step": 26000
831
+ },
832
+ {
833
+ "epoch": 40.37,
834
+ "eval_cer": 0.03374092287616447,
835
+ "eval_loss": 0.16579236090183258,
836
+ "eval_runtime": 140.5138,
837
+ "eval_samples_per_second": 27.35,
838
+ "eval_steps_per_second": 3.423,
839
+ "eval_wer": 0.13347968474466476,
840
+ "step": 26000
841
+ },
842
+ {
843
+ "epoch": 41.15,
844
+ "learning_rate": 0.00018009463722397476,
845
+ "loss": 0.052,
846
+ "step": 26500
847
+ },
848
+ {
849
+ "epoch": 41.15,
850
+ "eval_cer": 0.03429068545122821,
851
+ "eval_loss": 0.17156867682933807,
852
+ "eval_runtime": 144.8911,
853
+ "eval_samples_per_second": 26.523,
854
+ "eval_steps_per_second": 3.32,
855
+ "eval_wer": 0.1369345377334725,
856
+ "step": 26500
857
+ },
858
+ {
859
+ "epoch": 41.92,
860
+ "learning_rate": 0.0001643217665615142,
861
+ "loss": 0.0513,
862
+ "step": 27000
863
+ },
864
+ {
865
+ "epoch": 41.92,
866
+ "eval_cer": 0.03304012662663267,
867
+ "eval_loss": 0.15995080769062042,
868
+ "eval_runtime": 141.0194,
869
+ "eval_samples_per_second": 27.252,
870
+ "eval_steps_per_second": 3.411,
871
+ "eval_wer": 0.13168028214632743,
872
+ "step": 27000
873
+ },
874
+ {
875
+ "epoch": 42.7,
876
+ "learning_rate": 0.00014854889589905362,
877
+ "loss": 0.0491,
878
+ "step": 27500
879
+ },
880
+ {
881
+ "epoch": 42.7,
882
+ "eval_cer": 0.03279243140050506,
883
+ "eval_loss": 0.16710308194160461,
884
+ "eval_runtime": 143.6022,
885
+ "eval_samples_per_second": 26.761,
886
+ "eval_steps_per_second": 3.35,
887
+ "eval_wer": 0.13110447331485947,
888
+ "step": 27500
889
+ },
890
+ {
891
+ "epoch": 43.48,
892
+ "learning_rate": 0.0001328391167192429,
893
+ "loss": 0.0463,
894
+ "step": 28000
895
+ },
896
+ {
897
+ "epoch": 43.48,
898
+ "eval_cer": 0.032375578458973216,
899
+ "eval_loss": 0.16127389669418335,
900
+ "eval_runtime": 141.8046,
901
+ "eval_samples_per_second": 27.101,
902
+ "eval_steps_per_second": 3.392,
903
+ "eval_wer": 0.1289092021448879,
904
+ "step": 28000
905
+ },
906
+ {
907
+ "epoch": 44.25,
908
+ "learning_rate": 0.00011706624605678232,
909
+ "loss": 0.0468,
910
+ "step": 28500
911
+ },
912
+ {
913
+ "epoch": 44.25,
914
+ "eval_cer": 0.031523748534973355,
915
+ "eval_loss": 0.15992802381515503,
916
+ "eval_runtime": 142.9591,
917
+ "eval_samples_per_second": 26.882,
918
+ "eval_steps_per_second": 3.365,
919
+ "eval_wer": 0.12599416993558138,
920
+ "step": 28500
921
+ },
922
+ {
923
+ "epoch": 45.03,
924
+ "learning_rate": 0.00010129337539432177,
925
+ "loss": 0.0435,
926
+ "step": 29000
927
+ },
928
+ {
929
+ "epoch": 45.03,
930
+ "eval_cer": 0.030816910938462838,
931
+ "eval_loss": 0.15560874342918396,
932
+ "eval_runtime": 140.0872,
933
+ "eval_samples_per_second": 27.433,
934
+ "eval_steps_per_second": 3.434,
935
+ "eval_wer": 0.12318710188217512,
936
+ "step": 29000
937
+ },
938
+ {
939
+ "epoch": 45.81,
940
+ "learning_rate": 8.555205047318612e-05,
941
+ "loss": 0.043,
942
+ "step": 29500
943
+ },
944
+ {
945
+ "epoch": 45.81,
946
+ "eval_cer": 0.03093773787803729,
947
+ "eval_loss": 0.1587720513343811,
948
+ "eval_runtime": 140.3288,
949
+ "eval_samples_per_second": 27.386,
950
+ "eval_steps_per_second": 3.428,
951
+ "eval_wer": 0.12397883902544356,
952
+ "step": 29500
953
+ },
954
+ {
955
+ "epoch": 46.58,
956
+ "learning_rate": 6.977917981072555e-05,
957
+ "loss": 0.0421,
958
+ "step": 30000
959
+ },
960
+ {
961
+ "epoch": 46.58,
962
+ "eval_cer": 0.03079274555054795,
963
+ "eval_loss": 0.1567341834306717,
964
+ "eval_runtime": 141.7614,
965
+ "eval_samples_per_second": 27.109,
966
+ "eval_steps_per_second": 3.393,
967
+ "eval_wer": 0.12167560369957174,
968
+ "step": 30000
969
+ },
970
+ {
971
+ "epoch": 47.36,
972
+ "learning_rate": 5.400630914826499e-05,
973
+ "loss": 0.04,
974
+ "step": 30500
975
+ },
976
+ {
977
+ "epoch": 47.36,
978
+ "eval_cer": 0.0301584041177821,
979
+ "eval_loss": 0.1532924771308899,
980
+ "eval_runtime": 143.1401,
981
+ "eval_samples_per_second": 26.848,
982
+ "eval_steps_per_second": 3.36,
983
+ "eval_wer": 0.11984021304926765,
984
+ "step": 30500
985
+ },
986
+ {
987
+ "epoch": 48.14,
988
+ "learning_rate": 3.823343848580442e-05,
989
+ "loss": 0.0389,
990
+ "step": 31000
991
+ },
992
+ {
993
+ "epoch": 48.14,
994
+ "eval_cer": 0.0296992617473992,
995
+ "eval_loss": 0.15820840001106262,
996
+ "eval_runtime": 144.2309,
997
+ "eval_samples_per_second": 26.645,
998
+ "eval_steps_per_second": 3.335,
999
+ "eval_wer": 0.11847266707453126,
1000
+ "step": 31000
1001
+ },
1002
+ {
1003
+ "epoch": 48.91,
1004
+ "learning_rate": 2.2460567823343848e-05,
1005
+ "loss": 0.0387,
1006
+ "step": 31500
1007
+ },
1008
+ {
1009
+ "epoch": 48.91,
1010
+ "eval_cer": 0.029675096359484312,
1011
+ "eval_loss": 0.1575988531112671,
1012
+ "eval_runtime": 140.1473,
1013
+ "eval_samples_per_second": 27.421,
1014
+ "eval_steps_per_second": 3.432,
1015
+ "eval_wer": 0.11872458343829849,
1016
+ "step": 31500
1017
+ },
1018
+ {
1019
+ "epoch": 49.69,
1020
+ "learning_rate": 6.687697160883281e-06,
1021
+ "loss": 0.0376,
1022
+ "step": 32000
1023
+ },
1024
+ {
1025
+ "epoch": 49.69,
1026
+ "eval_cer": 0.02951802133803753,
1027
+ "eval_loss": 0.15600277483463287,
1028
+ "eval_runtime": 140.2646,
1029
+ "eval_samples_per_second": 27.398,
1030
+ "eval_steps_per_second": 3.429,
1031
+ "eval_wer": 0.11822075071076403,
1032
+ "step": 32000
1033
+ },
1034
+ {
1035
+ "epoch": 50.0,
1036
+ "step": 32200,
1037
+ "total_flos": 1.8729227608976848e+20,
1038
+ "train_loss": 0.17004985039278586,
1039
+ "train_runtime": 77839.5185,
1040
+ "train_samples_per_second": 19.876,
1041
+ "train_steps_per_second": 0.414
1042
+ }
1043
+ ],
1044
+ "max_steps": 32200,
1045
+ "num_train_epochs": 50,
1046
+ "total_flos": 1.8729227608976848e+20,
1047
+ "trial_name": null,
1048
+ "trial_params": null
1049
+ }