afaqalinagra commited on
Commit
1160f76
·
verified ·
1 Parent(s): 7e9b56d

Delete trainer_state (2).json

Browse files
Files changed (1) hide show
  1. trainer_state (2).json +0 -386
trainer_state (2).json DELETED
@@ -1,386 +0,0 @@
1
- {
2
- "best_global_step": 400,
3
- "best_metric": 39.02045209903122,
4
- "best_model_checkpoint": "./whisper-small-ps/checkpoint-400",
5
- "epoch": 7.017817371937639,
6
- "eval_steps": 50,
7
- "global_step": 400,
8
- "is_hyper_param_search": false,
9
- "is_local_process_zero": true,
10
- "is_world_process_zero": true,
11
- "log_history": [
12
- {
13
- "epoch": 0.17817371937639198,
14
- "grad_norm": 25.405925750732422,
15
- "learning_rate": 9e-06,
16
- "loss": 3.7318,
17
- "step": 10
18
- },
19
- {
20
- "epoch": 0.35634743875278396,
21
- "grad_norm": 15.668482780456543,
22
- "learning_rate": 1.9e-05,
23
- "loss": 1.9857,
24
- "step": 20
25
- },
26
- {
27
- "epoch": 0.534521158129176,
28
- "grad_norm": 28.658483505249023,
29
- "learning_rate": 1.9625e-05,
30
- "loss": 1.4661,
31
- "step": 30
32
- },
33
- {
34
- "epoch": 0.7126948775055679,
35
- "grad_norm": 18.580156326293945,
36
- "learning_rate": 1.9208333333333337e-05,
37
- "loss": 1.3166,
38
- "step": 40
39
- },
40
- {
41
- "epoch": 0.89086859688196,
42
- "grad_norm": 14.863431930541992,
43
- "learning_rate": 1.8791666666666668e-05,
44
- "loss": 1.1666,
45
- "step": 50
46
- },
47
- {
48
- "epoch": 0.89086859688196,
49
- "eval_loss": 1.0706024169921875,
50
- "eval_runtime": 172.0309,
51
- "eval_samples_per_second": 0.657,
52
- "eval_steps_per_second": 0.657,
53
- "eval_wer": 58.50376749192681,
54
- "step": 50
55
- },
56
- {
57
- "epoch": 1.0534521158129175,
58
- "grad_norm": 11.499765396118164,
59
- "learning_rate": 1.8375e-05,
60
- "loss": 1.0377,
61
- "step": 60
62
- },
63
- {
64
- "epoch": 1.2316258351893095,
65
- "grad_norm": 11.639135360717773,
66
- "learning_rate": 1.7958333333333334e-05,
67
- "loss": 0.786,
68
- "step": 70
69
- },
70
- {
71
- "epoch": 1.4097995545657016,
72
- "grad_norm": 11.476306915283203,
73
- "learning_rate": 1.754166666666667e-05,
74
- "loss": 0.7532,
75
- "step": 80
76
- },
77
- {
78
- "epoch": 1.5879732739420935,
79
- "grad_norm": 9.792712211608887,
80
- "learning_rate": 1.7125e-05,
81
- "loss": 0.6477,
82
- "step": 90
83
- },
84
- {
85
- "epoch": 1.7661469933184857,
86
- "grad_norm": 9.391416549682617,
87
- "learning_rate": 1.6708333333333334e-05,
88
- "loss": 0.6258,
89
- "step": 100
90
- },
91
- {
92
- "epoch": 1.7661469933184857,
93
- "eval_loss": 0.7860051393508911,
94
- "eval_runtime": 144.957,
95
- "eval_samples_per_second": 0.78,
96
- "eval_steps_per_second": 0.78,
97
- "eval_wer": 46.716899892357375,
98
- "step": 100
99
- },
100
- {
101
- "epoch": 1.9443207126948776,
102
- "grad_norm": 8.822958946228027,
103
- "learning_rate": 1.629166666666667e-05,
104
- "loss": 0.5398,
105
- "step": 110
106
- },
107
- {
108
- "epoch": 2.106904231625835,
109
- "grad_norm": 6.613277435302734,
110
- "learning_rate": 1.5875e-05,
111
- "loss": 0.4014,
112
- "step": 120
113
- },
114
- {
115
- "epoch": 2.285077951002227,
116
- "grad_norm": 5.975327491760254,
117
- "learning_rate": 1.5458333333333334e-05,
118
- "loss": 0.3024,
119
- "step": 130
120
- },
121
- {
122
- "epoch": 2.463251670378619,
123
- "grad_norm": 6.35911226272583,
124
- "learning_rate": 1.5041666666666667e-05,
125
- "loss": 0.2958,
126
- "step": 140
127
- },
128
- {
129
- "epoch": 2.6414253897550113,
130
- "grad_norm": 7.164131164550781,
131
- "learning_rate": 1.4625e-05,
132
- "loss": 0.3249,
133
- "step": 150
134
- },
135
- {
136
- "epoch": 2.6414253897550113,
137
- "eval_loss": 0.7021090984344482,
138
- "eval_runtime": 142.4441,
139
- "eval_samples_per_second": 0.793,
140
- "eval_steps_per_second": 0.793,
141
- "eval_wer": 44.40258342303552,
142
- "step": 150
143
- },
144
- {
145
- "epoch": 2.819599109131403,
146
- "grad_norm": 7.799057483673096,
147
- "learning_rate": 1.4208333333333336e-05,
148
- "loss": 0.2982,
149
- "step": 160
150
- },
151
- {
152
- "epoch": 2.997772828507795,
153
- "grad_norm": 8.640966415405273,
154
- "learning_rate": 1.3791666666666667e-05,
155
- "loss": 0.2577,
156
- "step": 170
157
- },
158
- {
159
- "epoch": 3.1603563474387526,
160
- "grad_norm": 4.808525562286377,
161
- "learning_rate": 1.3375e-05,
162
- "loss": 0.1443,
163
- "step": 180
164
- },
165
- {
166
- "epoch": 3.338530066815145,
167
- "grad_norm": 6.588418960571289,
168
- "learning_rate": 1.2958333333333334e-05,
169
- "loss": 0.1435,
170
- "step": 190
171
- },
172
- {
173
- "epoch": 3.516703786191537,
174
- "grad_norm": 4.835176944732666,
175
- "learning_rate": 1.2541666666666667e-05,
176
- "loss": 0.1268,
177
- "step": 200
178
- },
179
- {
180
- "epoch": 3.516703786191537,
181
- "eval_loss": 0.7087017893791199,
182
- "eval_runtime": 140.8539,
183
- "eval_samples_per_second": 0.802,
184
- "eval_steps_per_second": 0.802,
185
- "eval_wer": 41.980624327233585,
186
- "step": 200
187
- },
188
- {
189
- "epoch": 3.694877505567929,
190
- "grad_norm": 7.006318092346191,
191
- "learning_rate": 1.2125e-05,
192
- "loss": 0.1368,
193
- "step": 210
194
- },
195
- {
196
- "epoch": 3.8730512249443207,
197
- "grad_norm": 5.319632530212402,
198
- "learning_rate": 1.1708333333333334e-05,
199
- "loss": 0.1536,
200
- "step": 220
201
- },
202
- {
203
- "epoch": 4.035634743875279,
204
- "grad_norm": 3.353001594543457,
205
- "learning_rate": 1.1291666666666667e-05,
206
- "loss": 0.1396,
207
- "step": 230
208
- },
209
- {
210
- "epoch": 4.21380846325167,
211
- "grad_norm": 4.065522193908691,
212
- "learning_rate": 1.0875e-05,
213
- "loss": 0.0629,
214
- "step": 240
215
- },
216
- {
217
- "epoch": 4.3919821826280625,
218
- "grad_norm": 2.539537191390991,
219
- "learning_rate": 1.0458333333333335e-05,
220
- "loss": 0.0651,
221
- "step": 250
222
- },
223
- {
224
- "epoch": 4.3919821826280625,
225
- "eval_loss": 0.7520610094070435,
226
- "eval_runtime": 144.9719,
227
- "eval_samples_per_second": 0.779,
228
- "eval_steps_per_second": 0.779,
229
- "eval_wer": 40.74273412271259,
230
- "step": 250
231
- },
232
- {
233
- "epoch": 4.570155902004454,
234
- "grad_norm": 3.364603281021118,
235
- "learning_rate": 1.0041666666666667e-05,
236
- "loss": 0.0498,
237
- "step": 260
238
- },
239
- {
240
- "epoch": 4.748329621380846,
241
- "grad_norm": 4.829373359680176,
242
- "learning_rate": 9.625e-06,
243
- "loss": 0.076,
244
- "step": 270
245
- },
246
- {
247
- "epoch": 4.926503340757238,
248
- "grad_norm": 2.1371731758117676,
249
- "learning_rate": 9.208333333333333e-06,
250
- "loss": 0.0671,
251
- "step": 280
252
- },
253
- {
254
- "epoch": 5.089086859688196,
255
- "grad_norm": 2.170490264892578,
256
- "learning_rate": 8.791666666666667e-06,
257
- "loss": 0.0477,
258
- "step": 290
259
- },
260
- {
261
- "epoch": 5.267260579064588,
262
- "grad_norm": 3.078660249710083,
263
- "learning_rate": 8.375e-06,
264
- "loss": 0.0278,
265
- "step": 300
266
- },
267
- {
268
- "epoch": 5.267260579064588,
269
- "eval_loss": 0.766233503818512,
270
- "eval_runtime": 140.8607,
271
- "eval_samples_per_second": 0.802,
272
- "eval_steps_per_second": 0.802,
273
- "eval_wer": 40.635091496232505,
274
- "step": 300
275
- },
276
- {
277
- "epoch": 5.44543429844098,
278
- "grad_norm": 0.5123298764228821,
279
- "learning_rate": 7.958333333333333e-06,
280
- "loss": 0.0222,
281
- "step": 310
282
- },
283
- {
284
- "epoch": 5.6236080178173715,
285
- "grad_norm": 4.809408664703369,
286
- "learning_rate": 7.541666666666667e-06,
287
- "loss": 0.025,
288
- "step": 320
289
- },
290
- {
291
- "epoch": 5.801781737193764,
292
- "grad_norm": 1.5985106229782104,
293
- "learning_rate": 7.125e-06,
294
- "loss": 0.0359,
295
- "step": 330
296
- },
297
- {
298
- "epoch": 5.979955456570156,
299
- "grad_norm": 1.767372727394104,
300
- "learning_rate": 6.708333333333333e-06,
301
- "loss": 0.0352,
302
- "step": 340
303
- },
304
- {
305
- "epoch": 6.142538975501114,
306
- "grad_norm": 1.209718108177185,
307
- "learning_rate": 6.291666666666667e-06,
308
- "loss": 0.0194,
309
- "step": 350
310
- },
311
- {
312
- "epoch": 6.142538975501114,
313
- "eval_loss": 0.7994771003723145,
314
- "eval_runtime": 141.1898,
315
- "eval_samples_per_second": 0.8,
316
- "eval_steps_per_second": 0.8,
317
- "eval_wer": 39.82777179763186,
318
- "step": 350
319
- },
320
- {
321
- "epoch": 6.320712694877505,
322
- "grad_norm": 0.5679605007171631,
323
- "learning_rate": 5.8750000000000005e-06,
324
- "loss": 0.0173,
325
- "step": 360
326
- },
327
- {
328
- "epoch": 6.498886414253898,
329
- "grad_norm": 1.3990226984024048,
330
- "learning_rate": 5.458333333333333e-06,
331
- "loss": 0.0144,
332
- "step": 370
333
- },
334
- {
335
- "epoch": 6.67706013363029,
336
- "grad_norm": 0.8713251352310181,
337
- "learning_rate": 5.041666666666667e-06,
338
- "loss": 0.0088,
339
- "step": 380
340
- },
341
- {
342
- "epoch": 6.855233853006681,
343
- "grad_norm": 0.9836070537567139,
344
- "learning_rate": 4.625000000000001e-06,
345
- "loss": 0.01,
346
- "step": 390
347
- },
348
- {
349
- "epoch": 7.017817371937639,
350
- "grad_norm": 0.4024389982223511,
351
- "learning_rate": 4.208333333333333e-06,
352
- "loss": 0.0125,
353
- "step": 400
354
- },
355
- {
356
- "epoch": 7.017817371937639,
357
- "eval_loss": 0.805569052696228,
358
- "eval_runtime": 141.5404,
359
- "eval_samples_per_second": 0.798,
360
- "eval_steps_per_second": 0.798,
361
- "eval_wer": 39.02045209903122,
362
- "step": 400
363
- }
364
- ],
365
- "logging_steps": 10,
366
- "max_steps": 500,
367
- "num_input_tokens_seen": 0,
368
- "num_train_epochs": 9,
369
- "save_steps": 50,
370
- "stateful_callbacks": {
371
- "TrainerControl": {
372
- "args": {
373
- "should_epoch_stop": false,
374
- "should_evaluate": false,
375
- "should_log": false,
376
- "should_save": true,
377
- "should_training_stop": false
378
- },
379
- "attributes": {}
380
- }
381
- },
382
- "total_flos": 9.0933259640832e+17,
383
- "train_batch_size": 1,
384
- "trial_name": null,
385
- "trial_params": null
386
- }