EYEDOL commited on
Commit
32e5646
·
verified ·
1 Parent(s): 1337ff5

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. checkpoint-1000/config.json +47 -0
  2. checkpoint-1000/generation_config.json +9 -0
  3. checkpoint-1000/model.safetensors +3 -0
  4. checkpoint-1000/optimizer.pt +3 -0
  5. checkpoint-1000/preprocessor_config.json +15 -0
  6. checkpoint-1000/rng_state.pth +3 -0
  7. checkpoint-1000/scaler.pt +3 -0
  8. checkpoint-1000/scheduler.pt +3 -0
  9. checkpoint-1000/trainer_state.json +402 -0
  10. checkpoint-1000/training_args.bin +3 -0
  11. checkpoint-1500/config.json +47 -0
  12. checkpoint-1500/generation_config.json +9 -0
  13. checkpoint-1500/model.safetensors +3 -0
  14. checkpoint-1500/optimizer.pt +3 -0
  15. checkpoint-1500/preprocessor_config.json +15 -0
  16. checkpoint-1500/rng_state.pth +3 -0
  17. checkpoint-1500/scaler.pt +3 -0
  18. checkpoint-1500/scheduler.pt +3 -0
  19. checkpoint-1500/trainer_state.json +586 -0
  20. checkpoint-1500/training_args.bin +3 -0
  21. checkpoint-2000/config.json +47 -0
  22. checkpoint-2000/generation_config.json +9 -0
  23. checkpoint-2000/model.safetensors +3 -0
  24. checkpoint-2000/optimizer.pt +3 -0
  25. checkpoint-2000/preprocessor_config.json +15 -0
  26. checkpoint-2000/rng_state.pth +3 -0
  27. checkpoint-2000/scaler.pt +3 -0
  28. checkpoint-2000/scheduler.pt +3 -0
  29. checkpoint-2000/trainer_state.json +770 -0
  30. checkpoint-2000/training_args.bin +3 -0
  31. checkpoint-2500/config.json +47 -0
  32. checkpoint-2500/generation_config.json +9 -0
  33. checkpoint-2500/model.safetensors +3 -0
  34. checkpoint-2500/optimizer.pt +3 -0
  35. checkpoint-2500/preprocessor_config.json +15 -0
  36. checkpoint-2500/rng_state.pth +3 -0
  37. checkpoint-2500/scaler.pt +3 -0
  38. checkpoint-2500/scheduler.pt +3 -0
  39. checkpoint-2500/trainer_state.json +954 -0
  40. checkpoint-2500/training_args.bin +3 -0
  41. checkpoint-3000/config.json +47 -0
  42. checkpoint-3000/generation_config.json +9 -0
  43. checkpoint-3000/model.safetensors +3 -0
  44. checkpoint-3000/optimizer.pt +3 -0
  45. checkpoint-3000/preprocessor_config.json +15 -0
  46. checkpoint-3000/rng_state.pth +3 -0
  47. checkpoint-3000/scaler.pt +3 -0
  48. checkpoint-3000/scheduler.pt +3 -0
  49. checkpoint-3000/trainer_state.json +1138 -0
  50. checkpoint-3000/training_args.bin +3 -0
checkpoint-1000/config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.0,
3
+ "activation_function": "gelu",
4
+ "apply_spec_augment": false,
5
+ "architectures": [
6
+ "WhisperForConditionalGeneration"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "begin_suppress_tokens": null,
10
+ "bos_token_id": 50257,
11
+ "classifier_proj_size": 256,
12
+ "d_model": 768,
13
+ "decoder_attention_heads": 12,
14
+ "decoder_ffn_dim": 3072,
15
+ "decoder_layerdrop": 0.0,
16
+ "decoder_layers": 12,
17
+ "decoder_start_token_id": 50258,
18
+ "dropout": 0.0,
19
+ "encoder_attention_heads": 12,
20
+ "encoder_ffn_dim": 3072,
21
+ "encoder_layerdrop": 0.0,
22
+ "encoder_layers": 12,
23
+ "eos_token_id": 50257,
24
+ "forced_decoder_ids": null,
25
+ "init_std": 0.02,
26
+ "is_encoder_decoder": true,
27
+ "mask_feature_length": 10,
28
+ "mask_feature_min_masks": 0,
29
+ "mask_feature_prob": 0.0,
30
+ "mask_time_length": 10,
31
+ "mask_time_min_masks": 2,
32
+ "mask_time_prob": 0.05,
33
+ "max_length": null,
34
+ "max_source_positions": 1500,
35
+ "max_target_positions": 448,
36
+ "median_filter_width": 7,
37
+ "model_type": "whisper",
38
+ "num_hidden_layers": 12,
39
+ "num_mel_bins": 80,
40
+ "pad_token_id": 50257,
41
+ "scale_embedding": false,
42
+ "torch_dtype": "float32",
43
+ "transformers_version": "4.51.3",
44
+ "use_cache": false,
45
+ "use_weighted_layer_sum": false,
46
+ "vocab_size": 51865
47
+ }
checkpoint-1000/generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 50257,
3
+ "decoder_start_token_id": 50258,
4
+ "eos_token_id": 50257,
5
+ "pad_token_id": 50257,
6
+ "suppress_tokens": [],
7
+ "transformers_version": "4.51.3",
8
+ "use_cache": false
9
+ }
checkpoint-1000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd44d89d0147eedf670e3f3dba96f1821950c516fd7f57b976543e0256c67f1c
3
+ size 966995080
checkpoint-1000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e446455b26d8c3490f570cceb19fd1d68ec10cd8adda69fb8c9716facc3d454c
3
+ size 1925064044
checkpoint-1000/preprocessor_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "chunk_length": 30,
3
+ "dither": 0.0,
4
+ "feature_extractor_type": "WhisperFeatureExtractor",
5
+ "feature_size": 80,
6
+ "hop_length": 160,
7
+ "n_fft": 400,
8
+ "n_samples": 480000,
9
+ "nb_max_frames": 3000,
10
+ "padding_side": "right",
11
+ "padding_value": 0.0,
12
+ "processor_class": "WhisperProcessor",
13
+ "return_attention_mask": false,
14
+ "sampling_rate": 16000
15
+ }
checkpoint-1000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d4f95eca64235b7f040ceb7c0bbe77a9a52bb931b9122581a1cf284ad8efef39
3
+ size 14244
checkpoint-1000/scaler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b20682d68cd70b8b0b208c3bd5f62edfdf7322c345d5f8fb0fc9912581d14875
3
+ size 988
checkpoint-1000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:746d3afa0d117ef9f30a06dd95b2e34c08a35c2d0208de96d411b73e96ff1923
3
+ size 1064
checkpoint-1000/trainer_state.json ADDED
@@ -0,0 +1,402 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 1000,
3
+ "best_metric": 14.071299035983584,
4
+ "best_model_checkpoint": "./SALAMA_C3/checkpoint-1000",
5
+ "epoch": 1.675041876046901,
6
+ "eval_steps": 500,
7
+ "global_step": 1000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.03350083752093802,
14
+ "grad_norm": 11.757207870483398,
15
+ "learning_rate": 6.333333333333334e-07,
16
+ "loss": 0.5112,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.06700167504187604,
21
+ "grad_norm": 10.197953224182129,
22
+ "learning_rate": 1.3e-06,
23
+ "loss": 0.5283,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.10050251256281408,
28
+ "grad_norm": 12.811074256896973,
29
+ "learning_rate": 1.9666666666666668e-06,
30
+ "loss": 0.474,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.13400335008375208,
35
+ "grad_norm": 8.559342384338379,
36
+ "learning_rate": 2.6e-06,
37
+ "loss": 0.424,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.16750418760469013,
42
+ "grad_norm": 6.9629225730896,
43
+ "learning_rate": 3.266666666666667e-06,
44
+ "loss": 0.4465,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.20100502512562815,
49
+ "grad_norm": 7.833058834075928,
50
+ "learning_rate": 3.9333333333333335e-06,
51
+ "loss": 0.3764,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.23450586264656617,
56
+ "grad_norm": 6.882424831390381,
57
+ "learning_rate": 4.600000000000001e-06,
58
+ "loss": 0.3705,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.26800670016750416,
63
+ "grad_norm": 5.4183244705200195,
64
+ "learning_rate": 5.2666666666666665e-06,
65
+ "loss": 0.4116,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.3015075376884422,
70
+ "grad_norm": 6.103787899017334,
71
+ "learning_rate": 5.933333333333335e-06,
72
+ "loss": 0.3903,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.33500837520938026,
77
+ "grad_norm": 6.084160804748535,
78
+ "learning_rate": 6.600000000000001e-06,
79
+ "loss": 0.3794,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.3685092127303183,
84
+ "grad_norm": 4.531406879425049,
85
+ "learning_rate": 7.266666666666668e-06,
86
+ "loss": 0.3904,
87
+ "step": 220
88
+ },
89
+ {
90
+ "epoch": 0.4020100502512563,
91
+ "grad_norm": 6.702854156494141,
92
+ "learning_rate": 7.933333333333334e-06,
93
+ "loss": 0.3672,
94
+ "step": 240
95
+ },
96
+ {
97
+ "epoch": 0.4355108877721943,
98
+ "grad_norm": 6.386377811431885,
99
+ "learning_rate": 8.6e-06,
100
+ "loss": 0.3379,
101
+ "step": 260
102
+ },
103
+ {
104
+ "epoch": 0.46901172529313234,
105
+ "grad_norm": 5.33281135559082,
106
+ "learning_rate": 9.266666666666667e-06,
107
+ "loss": 0.4102,
108
+ "step": 280
109
+ },
110
+ {
111
+ "epoch": 0.5025125628140703,
112
+ "grad_norm": 6.031435012817383,
113
+ "learning_rate": 9.933333333333334e-06,
114
+ "loss": 0.3815,
115
+ "step": 300
116
+ },
117
+ {
118
+ "epoch": 0.5360134003350083,
119
+ "grad_norm": 4.8638176918029785,
120
+ "learning_rate": 9.968253968253969e-06,
121
+ "loss": 0.358,
122
+ "step": 320
123
+ },
124
+ {
125
+ "epoch": 0.5695142378559463,
126
+ "grad_norm": 5.352113723754883,
127
+ "learning_rate": 9.932980599647268e-06,
128
+ "loss": 0.3578,
129
+ "step": 340
130
+ },
131
+ {
132
+ "epoch": 0.6030150753768844,
133
+ "grad_norm": 5.560739040374756,
134
+ "learning_rate": 9.897707231040565e-06,
135
+ "loss": 0.3795,
136
+ "step": 360
137
+ },
138
+ {
139
+ "epoch": 0.6365159128978225,
140
+ "grad_norm": 6.186940670013428,
141
+ "learning_rate": 9.862433862433864e-06,
142
+ "loss": 0.3852,
143
+ "step": 380
144
+ },
145
+ {
146
+ "epoch": 0.6700167504187605,
147
+ "grad_norm": 5.37507438659668,
148
+ "learning_rate": 9.827160493827161e-06,
149
+ "loss": 0.3761,
150
+ "step": 400
151
+ },
152
+ {
153
+ "epoch": 0.7035175879396985,
154
+ "grad_norm": 6.373449802398682,
155
+ "learning_rate": 9.79188712522046e-06,
156
+ "loss": 0.3674,
157
+ "step": 420
158
+ },
159
+ {
160
+ "epoch": 0.7370184254606366,
161
+ "grad_norm": 5.702625274658203,
162
+ "learning_rate": 9.756613756613757e-06,
163
+ "loss": 0.4013,
164
+ "step": 440
165
+ },
166
+ {
167
+ "epoch": 0.7705192629815746,
168
+ "grad_norm": 5.5675153732299805,
169
+ "learning_rate": 9.721340388007056e-06,
170
+ "loss": 0.3495,
171
+ "step": 460
172
+ },
173
+ {
174
+ "epoch": 0.8040201005025126,
175
+ "grad_norm": 6.296374320983887,
176
+ "learning_rate": 9.686067019400353e-06,
177
+ "loss": 0.3704,
178
+ "step": 480
179
+ },
180
+ {
181
+ "epoch": 0.8375209380234506,
182
+ "grad_norm": 4.830463886260986,
183
+ "learning_rate": 9.650793650793652e-06,
184
+ "loss": 0.36,
185
+ "step": 500
186
+ },
187
+ {
188
+ "epoch": 0.8375209380234506,
189
+ "eval_loss": 0.27764827013015747,
190
+ "eval_runtime": 1752.3826,
191
+ "eval_samples_per_second": 2.725,
192
+ "eval_steps_per_second": 0.341,
193
+ "eval_wer": 19.59291781998664,
194
+ "step": 500
195
+ },
196
+ {
197
+ "epoch": 0.8710217755443886,
198
+ "grad_norm": 4.863058090209961,
199
+ "learning_rate": 9.61552028218695e-06,
200
+ "loss": 0.3874,
201
+ "step": 520
202
+ },
203
+ {
204
+ "epoch": 0.9045226130653267,
205
+ "grad_norm": 6.121027946472168,
206
+ "learning_rate": 9.580246913580248e-06,
207
+ "loss": 0.382,
208
+ "step": 540
209
+ },
210
+ {
211
+ "epoch": 0.9380234505862647,
212
+ "grad_norm": 4.538082599639893,
213
+ "learning_rate": 9.544973544973546e-06,
214
+ "loss": 0.3611,
215
+ "step": 560
216
+ },
217
+ {
218
+ "epoch": 0.9715242881072027,
219
+ "grad_norm": 4.8772382736206055,
220
+ "learning_rate": 9.509700176366844e-06,
221
+ "loss": 0.3584,
222
+ "step": 580
223
+ },
224
+ {
225
+ "epoch": 1.0050251256281406,
226
+ "grad_norm": 4.039212703704834,
227
+ "learning_rate": 9.474426807760142e-06,
228
+ "loss": 0.3701,
229
+ "step": 600
230
+ },
231
+ {
232
+ "epoch": 1.0385259631490786,
233
+ "grad_norm": 4.787687301635742,
234
+ "learning_rate": 9.43915343915344e-06,
235
+ "loss": 0.2058,
236
+ "step": 620
237
+ },
238
+ {
239
+ "epoch": 1.0720268006700167,
240
+ "grad_norm": 4.503021717071533,
241
+ "learning_rate": 9.403880070546738e-06,
242
+ "loss": 0.2455,
243
+ "step": 640
244
+ },
245
+ {
246
+ "epoch": 1.1055276381909547,
247
+ "grad_norm": 6.663857936859131,
248
+ "learning_rate": 9.368606701940036e-06,
249
+ "loss": 0.2179,
250
+ "step": 660
251
+ },
252
+ {
253
+ "epoch": 1.1390284757118927,
254
+ "grad_norm": 4.486196041107178,
255
+ "learning_rate": 9.333333333333334e-06,
256
+ "loss": 0.2196,
257
+ "step": 680
258
+ },
259
+ {
260
+ "epoch": 1.1725293132328307,
261
+ "grad_norm": 4.667060852050781,
262
+ "learning_rate": 9.298059964726633e-06,
263
+ "loss": 0.2183,
264
+ "step": 700
265
+ },
266
+ {
267
+ "epoch": 1.2060301507537687,
268
+ "grad_norm": 5.608316898345947,
269
+ "learning_rate": 9.26278659611993e-06,
270
+ "loss": 0.2161,
271
+ "step": 720
272
+ },
273
+ {
274
+ "epoch": 1.2395309882747068,
275
+ "grad_norm": 4.2184271812438965,
276
+ "learning_rate": 9.227513227513229e-06,
277
+ "loss": 0.2382,
278
+ "step": 740
279
+ },
280
+ {
281
+ "epoch": 1.2730318257956448,
282
+ "grad_norm": 3.871945381164551,
283
+ "learning_rate": 9.192239858906526e-06,
284
+ "loss": 0.2214,
285
+ "step": 760
286
+ },
287
+ {
288
+ "epoch": 1.3065326633165828,
289
+ "grad_norm": 3.730222225189209,
290
+ "learning_rate": 9.156966490299825e-06,
291
+ "loss": 0.2213,
292
+ "step": 780
293
+ },
294
+ {
295
+ "epoch": 1.3400335008375208,
296
+ "grad_norm": 4.740777015686035,
297
+ "learning_rate": 9.121693121693122e-06,
298
+ "loss": 0.2503,
299
+ "step": 800
300
+ },
301
+ {
302
+ "epoch": 1.3735343383584588,
303
+ "grad_norm": 4.123469352722168,
304
+ "learning_rate": 9.086419753086421e-06,
305
+ "loss": 0.2443,
306
+ "step": 820
307
+ },
308
+ {
309
+ "epoch": 1.4070351758793969,
310
+ "grad_norm": 3.988917827606201,
311
+ "learning_rate": 9.051146384479718e-06,
312
+ "loss": 0.241,
313
+ "step": 840
314
+ },
315
+ {
316
+ "epoch": 1.4405360134003349,
317
+ "grad_norm": 5.4730305671691895,
318
+ "learning_rate": 9.015873015873017e-06,
319
+ "loss": 0.2163,
320
+ "step": 860
321
+ },
322
+ {
323
+ "epoch": 1.474036850921273,
324
+ "grad_norm": 5.020652770996094,
325
+ "learning_rate": 8.980599647266314e-06,
326
+ "loss": 0.2387,
327
+ "step": 880
328
+ },
329
+ {
330
+ "epoch": 1.507537688442211,
331
+ "grad_norm": 4.231486797332764,
332
+ "learning_rate": 8.945326278659613e-06,
333
+ "loss": 0.2523,
334
+ "step": 900
335
+ },
336
+ {
337
+ "epoch": 1.541038525963149,
338
+ "grad_norm": 6.197975158691406,
339
+ "learning_rate": 8.910052910052912e-06,
340
+ "loss": 0.2255,
341
+ "step": 920
342
+ },
343
+ {
344
+ "epoch": 1.574539363484087,
345
+ "grad_norm": 5.489210605621338,
346
+ "learning_rate": 8.874779541446209e-06,
347
+ "loss": 0.2397,
348
+ "step": 940
349
+ },
350
+ {
351
+ "epoch": 1.608040201005025,
352
+ "grad_norm": 4.32737398147583,
353
+ "learning_rate": 8.839506172839508e-06,
354
+ "loss": 0.2298,
355
+ "step": 960
356
+ },
357
+ {
358
+ "epoch": 1.641541038525963,
359
+ "grad_norm": 4.504214763641357,
360
+ "learning_rate": 8.804232804232805e-06,
361
+ "loss": 0.2301,
362
+ "step": 980
363
+ },
364
+ {
365
+ "epoch": 1.675041876046901,
366
+ "grad_norm": 3.9694759845733643,
367
+ "learning_rate": 8.768959435626104e-06,
368
+ "loss": 0.2338,
369
+ "step": 1000
370
+ },
371
+ {
372
+ "epoch": 1.675041876046901,
373
+ "eval_loss": 0.18838582932949066,
374
+ "eval_runtime": 1770.3449,
375
+ "eval_samples_per_second": 2.698,
376
+ "eval_steps_per_second": 0.337,
377
+ "eval_wer": 14.071299035983584,
378
+ "step": 1000
379
+ }
380
+ ],
381
+ "logging_steps": 20,
382
+ "max_steps": 5970,
383
+ "num_input_tokens_seen": 0,
384
+ "num_train_epochs": 10,
385
+ "save_steps": 500,
386
+ "stateful_callbacks": {
387
+ "TrainerControl": {
388
+ "args": {
389
+ "should_epoch_stop": false,
390
+ "should_evaluate": false,
391
+ "should_log": false,
392
+ "should_save": true,
393
+ "should_training_stop": false
394
+ },
395
+ "attributes": {}
396
+ }
397
+ },
398
+ "total_flos": 9.23386705403904e+18,
399
+ "train_batch_size": 16,
400
+ "trial_name": null,
401
+ "trial_params": null
402
+ }
checkpoint-1000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41a3c645624f7463029286b7d5e1ad6924cfc44a0d153880431b77c40e5deb73
3
+ size 5432
checkpoint-1500/config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.0,
3
+ "activation_function": "gelu",
4
+ "apply_spec_augment": false,
5
+ "architectures": [
6
+ "WhisperForConditionalGeneration"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "begin_suppress_tokens": null,
10
+ "bos_token_id": 50257,
11
+ "classifier_proj_size": 256,
12
+ "d_model": 768,
13
+ "decoder_attention_heads": 12,
14
+ "decoder_ffn_dim": 3072,
15
+ "decoder_layerdrop": 0.0,
16
+ "decoder_layers": 12,
17
+ "decoder_start_token_id": 50258,
18
+ "dropout": 0.0,
19
+ "encoder_attention_heads": 12,
20
+ "encoder_ffn_dim": 3072,
21
+ "encoder_layerdrop": 0.0,
22
+ "encoder_layers": 12,
23
+ "eos_token_id": 50257,
24
+ "forced_decoder_ids": null,
25
+ "init_std": 0.02,
26
+ "is_encoder_decoder": true,
27
+ "mask_feature_length": 10,
28
+ "mask_feature_min_masks": 0,
29
+ "mask_feature_prob": 0.0,
30
+ "mask_time_length": 10,
31
+ "mask_time_min_masks": 2,
32
+ "mask_time_prob": 0.05,
33
+ "max_length": null,
34
+ "max_source_positions": 1500,
35
+ "max_target_positions": 448,
36
+ "median_filter_width": 7,
37
+ "model_type": "whisper",
38
+ "num_hidden_layers": 12,
39
+ "num_mel_bins": 80,
40
+ "pad_token_id": 50257,
41
+ "scale_embedding": false,
42
+ "torch_dtype": "float32",
43
+ "transformers_version": "4.51.3",
44
+ "use_cache": false,
45
+ "use_weighted_layer_sum": false,
46
+ "vocab_size": 51865
47
+ }
checkpoint-1500/generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 50257,
3
+ "decoder_start_token_id": 50258,
4
+ "eos_token_id": 50257,
5
+ "pad_token_id": 50257,
6
+ "suppress_tokens": [],
7
+ "transformers_version": "4.51.3",
8
+ "use_cache": false
9
+ }
checkpoint-1500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1383039729a87311da55da8166aab4f70de2077cf353120cbe4019867d195421
3
+ size 966995080
checkpoint-1500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59853d5c8ad3a215d900e6fe789b56aa9301c83f2f20ecda92d4504fea9a52e7
3
+ size 1925064044
checkpoint-1500/preprocessor_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "chunk_length": 30,
3
+ "dither": 0.0,
4
+ "feature_extractor_type": "WhisperFeatureExtractor",
5
+ "feature_size": 80,
6
+ "hop_length": 160,
7
+ "n_fft": 400,
8
+ "n_samples": 480000,
9
+ "nb_max_frames": 3000,
10
+ "padding_side": "right",
11
+ "padding_value": 0.0,
12
+ "processor_class": "WhisperProcessor",
13
+ "return_attention_mask": false,
14
+ "sampling_rate": 16000
15
+ }
checkpoint-1500/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:641ea6afeeb91ac119db140ccc4b56a307b61b309a6e3509792e0c955cb04c57
3
+ size 14244
checkpoint-1500/scaler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4a0fb91661a47fbe0e37f999cf89e4b80985f1c000b17f3f08e9e22356182db
3
+ size 988
checkpoint-1500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9fa73076154129d09f32d7071b6e9a88bc719841a440c20606e35875e46afcd
3
+ size 1064
checkpoint-1500/trainer_state.json ADDED
@@ -0,0 +1,586 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 1500,
3
+ "best_metric": 11.713753937195762,
4
+ "best_model_checkpoint": "./SALAMA_C3/checkpoint-1500",
5
+ "epoch": 2.5125628140703515,
6
+ "eval_steps": 500,
7
+ "global_step": 1500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.03350083752093802,
14
+ "grad_norm": 11.757207870483398,
15
+ "learning_rate": 6.333333333333334e-07,
16
+ "loss": 0.5112,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.06700167504187604,
21
+ "grad_norm": 10.197953224182129,
22
+ "learning_rate": 1.3e-06,
23
+ "loss": 0.5283,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.10050251256281408,
28
+ "grad_norm": 12.811074256896973,
29
+ "learning_rate": 1.9666666666666668e-06,
30
+ "loss": 0.474,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.13400335008375208,
35
+ "grad_norm": 8.559342384338379,
36
+ "learning_rate": 2.6e-06,
37
+ "loss": 0.424,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.16750418760469013,
42
+ "grad_norm": 6.9629225730896,
43
+ "learning_rate": 3.266666666666667e-06,
44
+ "loss": 0.4465,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.20100502512562815,
49
+ "grad_norm": 7.833058834075928,
50
+ "learning_rate": 3.9333333333333335e-06,
51
+ "loss": 0.3764,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.23450586264656617,
56
+ "grad_norm": 6.882424831390381,
57
+ "learning_rate": 4.600000000000001e-06,
58
+ "loss": 0.3705,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.26800670016750416,
63
+ "grad_norm": 5.4183244705200195,
64
+ "learning_rate": 5.2666666666666665e-06,
65
+ "loss": 0.4116,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.3015075376884422,
70
+ "grad_norm": 6.103787899017334,
71
+ "learning_rate": 5.933333333333335e-06,
72
+ "loss": 0.3903,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.33500837520938026,
77
+ "grad_norm": 6.084160804748535,
78
+ "learning_rate": 6.600000000000001e-06,
79
+ "loss": 0.3794,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.3685092127303183,
84
+ "grad_norm": 4.531406879425049,
85
+ "learning_rate": 7.266666666666668e-06,
86
+ "loss": 0.3904,
87
+ "step": 220
88
+ },
89
+ {
90
+ "epoch": 0.4020100502512563,
91
+ "grad_norm": 6.702854156494141,
92
+ "learning_rate": 7.933333333333334e-06,
93
+ "loss": 0.3672,
94
+ "step": 240
95
+ },
96
+ {
97
+ "epoch": 0.4355108877721943,
98
+ "grad_norm": 6.386377811431885,
99
+ "learning_rate": 8.6e-06,
100
+ "loss": 0.3379,
101
+ "step": 260
102
+ },
103
+ {
104
+ "epoch": 0.46901172529313234,
105
+ "grad_norm": 5.33281135559082,
106
+ "learning_rate": 9.266666666666667e-06,
107
+ "loss": 0.4102,
108
+ "step": 280
109
+ },
110
+ {
111
+ "epoch": 0.5025125628140703,
112
+ "grad_norm": 6.031435012817383,
113
+ "learning_rate": 9.933333333333334e-06,
114
+ "loss": 0.3815,
115
+ "step": 300
116
+ },
117
+ {
118
+ "epoch": 0.5360134003350083,
119
+ "grad_norm": 4.8638176918029785,
120
+ "learning_rate": 9.968253968253969e-06,
121
+ "loss": 0.358,
122
+ "step": 320
123
+ },
124
+ {
125
+ "epoch": 0.5695142378559463,
126
+ "grad_norm": 5.352113723754883,
127
+ "learning_rate": 9.932980599647268e-06,
128
+ "loss": 0.3578,
129
+ "step": 340
130
+ },
131
+ {
132
+ "epoch": 0.6030150753768844,
133
+ "grad_norm": 5.560739040374756,
134
+ "learning_rate": 9.897707231040565e-06,
135
+ "loss": 0.3795,
136
+ "step": 360
137
+ },
138
+ {
139
+ "epoch": 0.6365159128978225,
140
+ "grad_norm": 6.186940670013428,
141
+ "learning_rate": 9.862433862433864e-06,
142
+ "loss": 0.3852,
143
+ "step": 380
144
+ },
145
+ {
146
+ "epoch": 0.6700167504187605,
147
+ "grad_norm": 5.37507438659668,
148
+ "learning_rate": 9.827160493827161e-06,
149
+ "loss": 0.3761,
150
+ "step": 400
151
+ },
152
+ {
153
+ "epoch": 0.7035175879396985,
154
+ "grad_norm": 6.373449802398682,
155
+ "learning_rate": 9.79188712522046e-06,
156
+ "loss": 0.3674,
157
+ "step": 420
158
+ },
159
+ {
160
+ "epoch": 0.7370184254606366,
161
+ "grad_norm": 5.702625274658203,
162
+ "learning_rate": 9.756613756613757e-06,
163
+ "loss": 0.4013,
164
+ "step": 440
165
+ },
166
+ {
167
+ "epoch": 0.7705192629815746,
168
+ "grad_norm": 5.5675153732299805,
169
+ "learning_rate": 9.721340388007056e-06,
170
+ "loss": 0.3495,
171
+ "step": 460
172
+ },
173
+ {
174
+ "epoch": 0.8040201005025126,
175
+ "grad_norm": 6.296374320983887,
176
+ "learning_rate": 9.686067019400353e-06,
177
+ "loss": 0.3704,
178
+ "step": 480
179
+ },
180
+ {
181
+ "epoch": 0.8375209380234506,
182
+ "grad_norm": 4.830463886260986,
183
+ "learning_rate": 9.650793650793652e-06,
184
+ "loss": 0.36,
185
+ "step": 500
186
+ },
187
+ {
188
+ "epoch": 0.8375209380234506,
189
+ "eval_loss": 0.27764827013015747,
190
+ "eval_runtime": 1752.3826,
191
+ "eval_samples_per_second": 2.725,
192
+ "eval_steps_per_second": 0.341,
193
+ "eval_wer": 19.59291781998664,
194
+ "step": 500
195
+ },
196
+ {
197
+ "epoch": 0.8710217755443886,
198
+ "grad_norm": 4.863058090209961,
199
+ "learning_rate": 9.61552028218695e-06,
200
+ "loss": 0.3874,
201
+ "step": 520
202
+ },
203
+ {
204
+ "epoch": 0.9045226130653267,
205
+ "grad_norm": 6.121027946472168,
206
+ "learning_rate": 9.580246913580248e-06,
207
+ "loss": 0.382,
208
+ "step": 540
209
+ },
210
+ {
211
+ "epoch": 0.9380234505862647,
212
+ "grad_norm": 4.538082599639893,
213
+ "learning_rate": 9.544973544973546e-06,
214
+ "loss": 0.3611,
215
+ "step": 560
216
+ },
217
+ {
218
+ "epoch": 0.9715242881072027,
219
+ "grad_norm": 4.8772382736206055,
220
+ "learning_rate": 9.509700176366844e-06,
221
+ "loss": 0.3584,
222
+ "step": 580
223
+ },
224
+ {
225
+ "epoch": 1.0050251256281406,
226
+ "grad_norm": 4.039212703704834,
227
+ "learning_rate": 9.474426807760142e-06,
228
+ "loss": 0.3701,
229
+ "step": 600
230
+ },
231
+ {
232
+ "epoch": 1.0385259631490786,
233
+ "grad_norm": 4.787687301635742,
234
+ "learning_rate": 9.43915343915344e-06,
235
+ "loss": 0.2058,
236
+ "step": 620
237
+ },
238
+ {
239
+ "epoch": 1.0720268006700167,
240
+ "grad_norm": 4.503021717071533,
241
+ "learning_rate": 9.403880070546738e-06,
242
+ "loss": 0.2455,
243
+ "step": 640
244
+ },
245
+ {
246
+ "epoch": 1.1055276381909547,
247
+ "grad_norm": 6.663857936859131,
248
+ "learning_rate": 9.368606701940036e-06,
249
+ "loss": 0.2179,
250
+ "step": 660
251
+ },
252
+ {
253
+ "epoch": 1.1390284757118927,
254
+ "grad_norm": 4.486196041107178,
255
+ "learning_rate": 9.333333333333334e-06,
256
+ "loss": 0.2196,
257
+ "step": 680
258
+ },
259
+ {
260
+ "epoch": 1.1725293132328307,
261
+ "grad_norm": 4.667060852050781,
262
+ "learning_rate": 9.298059964726633e-06,
263
+ "loss": 0.2183,
264
+ "step": 700
265
+ },
266
+ {
267
+ "epoch": 1.2060301507537687,
268
+ "grad_norm": 5.608316898345947,
269
+ "learning_rate": 9.26278659611993e-06,
270
+ "loss": 0.2161,
271
+ "step": 720
272
+ },
273
+ {
274
+ "epoch": 1.2395309882747068,
275
+ "grad_norm": 4.2184271812438965,
276
+ "learning_rate": 9.227513227513229e-06,
277
+ "loss": 0.2382,
278
+ "step": 740
279
+ },
280
+ {
281
+ "epoch": 1.2730318257956448,
282
+ "grad_norm": 3.871945381164551,
283
+ "learning_rate": 9.192239858906526e-06,
284
+ "loss": 0.2214,
285
+ "step": 760
286
+ },
287
+ {
288
+ "epoch": 1.3065326633165828,
289
+ "grad_norm": 3.730222225189209,
290
+ "learning_rate": 9.156966490299825e-06,
291
+ "loss": 0.2213,
292
+ "step": 780
293
+ },
294
+ {
295
+ "epoch": 1.3400335008375208,
296
+ "grad_norm": 4.740777015686035,
297
+ "learning_rate": 9.121693121693122e-06,
298
+ "loss": 0.2503,
299
+ "step": 800
300
+ },
301
+ {
302
+ "epoch": 1.3735343383584588,
303
+ "grad_norm": 4.123469352722168,
304
+ "learning_rate": 9.086419753086421e-06,
305
+ "loss": 0.2443,
306
+ "step": 820
307
+ },
308
+ {
309
+ "epoch": 1.4070351758793969,
310
+ "grad_norm": 3.988917827606201,
311
+ "learning_rate": 9.051146384479718e-06,
312
+ "loss": 0.241,
313
+ "step": 840
314
+ },
315
+ {
316
+ "epoch": 1.4405360134003349,
317
+ "grad_norm": 5.4730305671691895,
318
+ "learning_rate": 9.015873015873017e-06,
319
+ "loss": 0.2163,
320
+ "step": 860
321
+ },
322
+ {
323
+ "epoch": 1.474036850921273,
324
+ "grad_norm": 5.020652770996094,
325
+ "learning_rate": 8.980599647266314e-06,
326
+ "loss": 0.2387,
327
+ "step": 880
328
+ },
329
+ {
330
+ "epoch": 1.507537688442211,
331
+ "grad_norm": 4.231486797332764,
332
+ "learning_rate": 8.945326278659613e-06,
333
+ "loss": 0.2523,
334
+ "step": 900
335
+ },
336
+ {
337
+ "epoch": 1.541038525963149,
338
+ "grad_norm": 6.197975158691406,
339
+ "learning_rate": 8.910052910052912e-06,
340
+ "loss": 0.2255,
341
+ "step": 920
342
+ },
343
+ {
344
+ "epoch": 1.574539363484087,
345
+ "grad_norm": 5.489210605621338,
346
+ "learning_rate": 8.874779541446209e-06,
347
+ "loss": 0.2397,
348
+ "step": 940
349
+ },
350
+ {
351
+ "epoch": 1.608040201005025,
352
+ "grad_norm": 4.32737398147583,
353
+ "learning_rate": 8.839506172839508e-06,
354
+ "loss": 0.2298,
355
+ "step": 960
356
+ },
357
+ {
358
+ "epoch": 1.641541038525963,
359
+ "grad_norm": 4.504214763641357,
360
+ "learning_rate": 8.804232804232805e-06,
361
+ "loss": 0.2301,
362
+ "step": 980
363
+ },
364
+ {
365
+ "epoch": 1.675041876046901,
366
+ "grad_norm": 3.9694759845733643,
367
+ "learning_rate": 8.768959435626104e-06,
368
+ "loss": 0.2338,
369
+ "step": 1000
370
+ },
371
+ {
372
+ "epoch": 1.675041876046901,
373
+ "eval_loss": 0.18838582932949066,
374
+ "eval_runtime": 1770.3449,
375
+ "eval_samples_per_second": 2.698,
376
+ "eval_steps_per_second": 0.337,
377
+ "eval_wer": 14.071299035983584,
378
+ "step": 1000
379
+ },
380
+ {
381
+ "epoch": 1.708542713567839,
382
+ "grad_norm": 4.972934722900391,
383
+ "learning_rate": 8.7336860670194e-06,
384
+ "loss": 0.2388,
385
+ "step": 1020
386
+ },
387
+ {
388
+ "epoch": 1.742043551088777,
389
+ "grad_norm": 5.178994655609131,
390
+ "learning_rate": 8.6984126984127e-06,
391
+ "loss": 0.2531,
392
+ "step": 1040
393
+ },
394
+ {
395
+ "epoch": 1.775544388609715,
396
+ "grad_norm": 3.560372829437256,
397
+ "learning_rate": 8.663139329805997e-06,
398
+ "loss": 0.2187,
399
+ "step": 1060
400
+ },
401
+ {
402
+ "epoch": 1.809045226130653,
403
+ "grad_norm": 4.647324562072754,
404
+ "learning_rate": 8.627865961199296e-06,
405
+ "loss": 0.2413,
406
+ "step": 1080
407
+ },
408
+ {
409
+ "epoch": 1.8425460636515911,
410
+ "grad_norm": 3.6784422397613525,
411
+ "learning_rate": 8.592592592592593e-06,
412
+ "loss": 0.2242,
413
+ "step": 1100
414
+ },
415
+ {
416
+ "epoch": 1.8760469011725294,
417
+ "grad_norm": 5.928018093109131,
418
+ "learning_rate": 8.557319223985891e-06,
419
+ "loss": 0.2406,
420
+ "step": 1120
421
+ },
422
+ {
423
+ "epoch": 1.9095477386934674,
424
+ "grad_norm": 6.358487606048584,
425
+ "learning_rate": 8.52204585537919e-06,
426
+ "loss": 0.2482,
427
+ "step": 1140
428
+ },
429
+ {
430
+ "epoch": 1.9430485762144054,
431
+ "grad_norm": 4.409506797790527,
432
+ "learning_rate": 8.486772486772487e-06,
433
+ "loss": 0.2415,
434
+ "step": 1160
435
+ },
436
+ {
437
+ "epoch": 1.9765494137353434,
438
+ "grad_norm": 5.169639587402344,
439
+ "learning_rate": 8.451499118165786e-06,
440
+ "loss": 0.222,
441
+ "step": 1180
442
+ },
443
+ {
444
+ "epoch": 2.0100502512562812,
445
+ "grad_norm": 2.8841328620910645,
446
+ "learning_rate": 8.416225749559083e-06,
447
+ "loss": 0.2109,
448
+ "step": 1200
449
+ },
450
+ {
451
+ "epoch": 2.0435510887772192,
452
+ "grad_norm": 3.835498332977295,
453
+ "learning_rate": 8.380952380952382e-06,
454
+ "loss": 0.1284,
455
+ "step": 1220
456
+ },
457
+ {
458
+ "epoch": 2.0770519262981573,
459
+ "grad_norm": 2.584859848022461,
460
+ "learning_rate": 8.34567901234568e-06,
461
+ "loss": 0.1059,
462
+ "step": 1240
463
+ },
464
+ {
465
+ "epoch": 2.1105527638190953,
466
+ "grad_norm": 2.797682285308838,
467
+ "learning_rate": 8.310405643738978e-06,
468
+ "loss": 0.0967,
469
+ "step": 1260
470
+ },
471
+ {
472
+ "epoch": 2.1440536013400333,
473
+ "grad_norm": 3.2937309741973877,
474
+ "learning_rate": 8.275132275132275e-06,
475
+ "loss": 0.1267,
476
+ "step": 1280
477
+ },
478
+ {
479
+ "epoch": 2.1775544388609713,
480
+ "grad_norm": 3.344325065612793,
481
+ "learning_rate": 8.239858906525574e-06,
482
+ "loss": 0.1354,
483
+ "step": 1300
484
+ },
485
+ {
486
+ "epoch": 2.2110552763819094,
487
+ "grad_norm": 2.955522060394287,
488
+ "learning_rate": 8.204585537918873e-06,
489
+ "loss": 0.104,
490
+ "step": 1320
491
+ },
492
+ {
493
+ "epoch": 2.2445561139028474,
494
+ "grad_norm": 5.180908203125,
495
+ "learning_rate": 8.16931216931217e-06,
496
+ "loss": 0.1229,
497
+ "step": 1340
498
+ },
499
+ {
500
+ "epoch": 2.2780569514237854,
501
+ "grad_norm": 3.639417886734009,
502
+ "learning_rate": 8.134038800705469e-06,
503
+ "loss": 0.104,
504
+ "step": 1360
505
+ },
506
+ {
507
+ "epoch": 2.3115577889447234,
508
+ "grad_norm": 3.7380902767181396,
509
+ "learning_rate": 8.098765432098766e-06,
510
+ "loss": 0.1141,
511
+ "step": 1380
512
+ },
513
+ {
514
+ "epoch": 2.3450586264656614,
515
+ "grad_norm": 4.100318908691406,
516
+ "learning_rate": 8.063492063492065e-06,
517
+ "loss": 0.1123,
518
+ "step": 1400
519
+ },
520
+ {
521
+ "epoch": 2.3785594639865995,
522
+ "grad_norm": 2.8147048950195312,
523
+ "learning_rate": 8.028218694885362e-06,
524
+ "loss": 0.1039,
525
+ "step": 1420
526
+ },
527
+ {
528
+ "epoch": 2.4120603015075375,
529
+ "grad_norm": 3.7376410961151123,
530
+ "learning_rate": 7.99294532627866e-06,
531
+ "loss": 0.1179,
532
+ "step": 1440
533
+ },
534
+ {
535
+ "epoch": 2.4455611390284755,
536
+ "grad_norm": 4.320065975189209,
537
+ "learning_rate": 7.957671957671958e-06,
538
+ "loss": 0.1167,
539
+ "step": 1460
540
+ },
541
+ {
542
+ "epoch": 2.4790619765494135,
543
+ "grad_norm": 4.415127277374268,
544
+ "learning_rate": 7.922398589065257e-06,
545
+ "loss": 0.125,
546
+ "step": 1480
547
+ },
548
+ {
549
+ "epoch": 2.5125628140703515,
550
+ "grad_norm": 2.631763219833374,
551
+ "learning_rate": 7.887125220458554e-06,
552
+ "loss": 0.1188,
553
+ "step": 1500
554
+ },
555
+ {
556
+ "epoch": 2.5125628140703515,
557
+ "eval_loss": 0.13635103404521942,
558
+ "eval_runtime": 1806.1099,
559
+ "eval_samples_per_second": 2.644,
560
+ "eval_steps_per_second": 0.331,
561
+ "eval_wer": 11.713753937195762,
562
+ "step": 1500
563
+ }
564
+ ],
565
+ "logging_steps": 20,
566
+ "max_steps": 5970,
567
+ "num_input_tokens_seen": 0,
568
+ "num_train_epochs": 10,
569
+ "save_steps": 500,
570
+ "stateful_callbacks": {
571
+ "TrainerControl": {
572
+ "args": {
573
+ "should_epoch_stop": false,
574
+ "should_evaluate": false,
575
+ "should_log": false,
576
+ "should_save": true,
577
+ "should_training_stop": false
578
+ },
579
+ "attributes": {}
580
+ }
581
+ },
582
+ "total_flos": 1.385036770295808e+19,
583
+ "train_batch_size": 16,
584
+ "trial_name": null,
585
+ "trial_params": null
586
+ }
checkpoint-1500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41a3c645624f7463029286b7d5e1ad6924cfc44a0d153880431b77c40e5deb73
3
+ size 5432
checkpoint-2000/config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.0,
3
+ "activation_function": "gelu",
4
+ "apply_spec_augment": false,
5
+ "architectures": [
6
+ "WhisperForConditionalGeneration"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "begin_suppress_tokens": null,
10
+ "bos_token_id": 50257,
11
+ "classifier_proj_size": 256,
12
+ "d_model": 768,
13
+ "decoder_attention_heads": 12,
14
+ "decoder_ffn_dim": 3072,
15
+ "decoder_layerdrop": 0.0,
16
+ "decoder_layers": 12,
17
+ "decoder_start_token_id": 50258,
18
+ "dropout": 0.0,
19
+ "encoder_attention_heads": 12,
20
+ "encoder_ffn_dim": 3072,
21
+ "encoder_layerdrop": 0.0,
22
+ "encoder_layers": 12,
23
+ "eos_token_id": 50257,
24
+ "forced_decoder_ids": null,
25
+ "init_std": 0.02,
26
+ "is_encoder_decoder": true,
27
+ "mask_feature_length": 10,
28
+ "mask_feature_min_masks": 0,
29
+ "mask_feature_prob": 0.0,
30
+ "mask_time_length": 10,
31
+ "mask_time_min_masks": 2,
32
+ "mask_time_prob": 0.05,
33
+ "max_length": null,
34
+ "max_source_positions": 1500,
35
+ "max_target_positions": 448,
36
+ "median_filter_width": 7,
37
+ "model_type": "whisper",
38
+ "num_hidden_layers": 12,
39
+ "num_mel_bins": 80,
40
+ "pad_token_id": 50257,
41
+ "scale_embedding": false,
42
+ "torch_dtype": "float32",
43
+ "transformers_version": "4.51.3",
44
+ "use_cache": false,
45
+ "use_weighted_layer_sum": false,
46
+ "vocab_size": 51865
47
+ }
checkpoint-2000/generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 50257,
3
+ "decoder_start_token_id": 50258,
4
+ "eos_token_id": 50257,
5
+ "pad_token_id": 50257,
6
+ "suppress_tokens": [],
7
+ "transformers_version": "4.51.3",
8
+ "use_cache": false
9
+ }
checkpoint-2000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c498ee290ccc95138b06ba4e9df6e40f71bbcd1469a11d559d1fda78500860d6
3
+ size 966995080
checkpoint-2000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d55d01888dd6a30defbf104a69ca6ce62571bbebec904e0246886e33bea442ef
3
+ size 1925064044
checkpoint-2000/preprocessor_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "chunk_length": 30,
3
+ "dither": 0.0,
4
+ "feature_extractor_type": "WhisperFeatureExtractor",
5
+ "feature_size": 80,
6
+ "hop_length": 160,
7
+ "n_fft": 400,
8
+ "n_samples": 480000,
9
+ "nb_max_frames": 3000,
10
+ "padding_side": "right",
11
+ "padding_value": 0.0,
12
+ "processor_class": "WhisperProcessor",
13
+ "return_attention_mask": false,
14
+ "sampling_rate": 16000
15
+ }
checkpoint-2000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91d9a982d0a102bb658e38ae6685c0a7ee27a2849f974142ce6a350968204948
3
+ size 14244
checkpoint-2000/scaler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd89000c206de186c97dd611bd50d2dd7f764488fa801cf3f469d639f7c1b66f
3
+ size 988
checkpoint-2000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c19a31a6a8b6605fe25dc22b645addbe28ac65d7d2baa70210b9ac1d12e57193
3
+ size 1064
checkpoint-2000/trainer_state.json ADDED
@@ -0,0 +1,770 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 2000,
3
+ "best_metric": 7.172854824854443,
4
+ "best_model_checkpoint": "./SALAMA_C3/checkpoint-2000",
5
+ "epoch": 3.3500837520938025,
6
+ "eval_steps": 500,
7
+ "global_step": 2000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.03350083752093802,
14
+ "grad_norm": 11.757207870483398,
15
+ "learning_rate": 6.333333333333334e-07,
16
+ "loss": 0.5112,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.06700167504187604,
21
+ "grad_norm": 10.197953224182129,
22
+ "learning_rate": 1.3e-06,
23
+ "loss": 0.5283,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.10050251256281408,
28
+ "grad_norm": 12.811074256896973,
29
+ "learning_rate": 1.9666666666666668e-06,
30
+ "loss": 0.474,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.13400335008375208,
35
+ "grad_norm": 8.559342384338379,
36
+ "learning_rate": 2.6e-06,
37
+ "loss": 0.424,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.16750418760469013,
42
+ "grad_norm": 6.9629225730896,
43
+ "learning_rate": 3.266666666666667e-06,
44
+ "loss": 0.4465,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.20100502512562815,
49
+ "grad_norm": 7.833058834075928,
50
+ "learning_rate": 3.9333333333333335e-06,
51
+ "loss": 0.3764,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.23450586264656617,
56
+ "grad_norm": 6.882424831390381,
57
+ "learning_rate": 4.600000000000001e-06,
58
+ "loss": 0.3705,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.26800670016750416,
63
+ "grad_norm": 5.4183244705200195,
64
+ "learning_rate": 5.2666666666666665e-06,
65
+ "loss": 0.4116,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.3015075376884422,
70
+ "grad_norm": 6.103787899017334,
71
+ "learning_rate": 5.933333333333335e-06,
72
+ "loss": 0.3903,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.33500837520938026,
77
+ "grad_norm": 6.084160804748535,
78
+ "learning_rate": 6.600000000000001e-06,
79
+ "loss": 0.3794,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.3685092127303183,
84
+ "grad_norm": 4.531406879425049,
85
+ "learning_rate": 7.266666666666668e-06,
86
+ "loss": 0.3904,
87
+ "step": 220
88
+ },
89
+ {
90
+ "epoch": 0.4020100502512563,
91
+ "grad_norm": 6.702854156494141,
92
+ "learning_rate": 7.933333333333334e-06,
93
+ "loss": 0.3672,
94
+ "step": 240
95
+ },
96
+ {
97
+ "epoch": 0.4355108877721943,
98
+ "grad_norm": 6.386377811431885,
99
+ "learning_rate": 8.6e-06,
100
+ "loss": 0.3379,
101
+ "step": 260
102
+ },
103
+ {
104
+ "epoch": 0.46901172529313234,
105
+ "grad_norm": 5.33281135559082,
106
+ "learning_rate": 9.266666666666667e-06,
107
+ "loss": 0.4102,
108
+ "step": 280
109
+ },
110
+ {
111
+ "epoch": 0.5025125628140703,
112
+ "grad_norm": 6.031435012817383,
113
+ "learning_rate": 9.933333333333334e-06,
114
+ "loss": 0.3815,
115
+ "step": 300
116
+ },
117
+ {
118
+ "epoch": 0.5360134003350083,
119
+ "grad_norm": 4.8638176918029785,
120
+ "learning_rate": 9.968253968253969e-06,
121
+ "loss": 0.358,
122
+ "step": 320
123
+ },
124
+ {
125
+ "epoch": 0.5695142378559463,
126
+ "grad_norm": 5.352113723754883,
127
+ "learning_rate": 9.932980599647268e-06,
128
+ "loss": 0.3578,
129
+ "step": 340
130
+ },
131
+ {
132
+ "epoch": 0.6030150753768844,
133
+ "grad_norm": 5.560739040374756,
134
+ "learning_rate": 9.897707231040565e-06,
135
+ "loss": 0.3795,
136
+ "step": 360
137
+ },
138
+ {
139
+ "epoch": 0.6365159128978225,
140
+ "grad_norm": 6.186940670013428,
141
+ "learning_rate": 9.862433862433864e-06,
142
+ "loss": 0.3852,
143
+ "step": 380
144
+ },
145
+ {
146
+ "epoch": 0.6700167504187605,
147
+ "grad_norm": 5.37507438659668,
148
+ "learning_rate": 9.827160493827161e-06,
149
+ "loss": 0.3761,
150
+ "step": 400
151
+ },
152
+ {
153
+ "epoch": 0.7035175879396985,
154
+ "grad_norm": 6.373449802398682,
155
+ "learning_rate": 9.79188712522046e-06,
156
+ "loss": 0.3674,
157
+ "step": 420
158
+ },
159
+ {
160
+ "epoch": 0.7370184254606366,
161
+ "grad_norm": 5.702625274658203,
162
+ "learning_rate": 9.756613756613757e-06,
163
+ "loss": 0.4013,
164
+ "step": 440
165
+ },
166
+ {
167
+ "epoch": 0.7705192629815746,
168
+ "grad_norm": 5.5675153732299805,
169
+ "learning_rate": 9.721340388007056e-06,
170
+ "loss": 0.3495,
171
+ "step": 460
172
+ },
173
+ {
174
+ "epoch": 0.8040201005025126,
175
+ "grad_norm": 6.296374320983887,
176
+ "learning_rate": 9.686067019400353e-06,
177
+ "loss": 0.3704,
178
+ "step": 480
179
+ },
180
+ {
181
+ "epoch": 0.8375209380234506,
182
+ "grad_norm": 4.830463886260986,
183
+ "learning_rate": 9.650793650793652e-06,
184
+ "loss": 0.36,
185
+ "step": 500
186
+ },
187
+ {
188
+ "epoch": 0.8375209380234506,
189
+ "eval_loss": 0.27764827013015747,
190
+ "eval_runtime": 1752.3826,
191
+ "eval_samples_per_second": 2.725,
192
+ "eval_steps_per_second": 0.341,
193
+ "eval_wer": 19.59291781998664,
194
+ "step": 500
195
+ },
196
+ {
197
+ "epoch": 0.8710217755443886,
198
+ "grad_norm": 4.863058090209961,
199
+ "learning_rate": 9.61552028218695e-06,
200
+ "loss": 0.3874,
201
+ "step": 520
202
+ },
203
+ {
204
+ "epoch": 0.9045226130653267,
205
+ "grad_norm": 6.121027946472168,
206
+ "learning_rate": 9.580246913580248e-06,
207
+ "loss": 0.382,
208
+ "step": 540
209
+ },
210
+ {
211
+ "epoch": 0.9380234505862647,
212
+ "grad_norm": 4.538082599639893,
213
+ "learning_rate": 9.544973544973546e-06,
214
+ "loss": 0.3611,
215
+ "step": 560
216
+ },
217
+ {
218
+ "epoch": 0.9715242881072027,
219
+ "grad_norm": 4.8772382736206055,
220
+ "learning_rate": 9.509700176366844e-06,
221
+ "loss": 0.3584,
222
+ "step": 580
223
+ },
224
+ {
225
+ "epoch": 1.0050251256281406,
226
+ "grad_norm": 4.039212703704834,
227
+ "learning_rate": 9.474426807760142e-06,
228
+ "loss": 0.3701,
229
+ "step": 600
230
+ },
231
+ {
232
+ "epoch": 1.0385259631490786,
233
+ "grad_norm": 4.787687301635742,
234
+ "learning_rate": 9.43915343915344e-06,
235
+ "loss": 0.2058,
236
+ "step": 620
237
+ },
238
+ {
239
+ "epoch": 1.0720268006700167,
240
+ "grad_norm": 4.503021717071533,
241
+ "learning_rate": 9.403880070546738e-06,
242
+ "loss": 0.2455,
243
+ "step": 640
244
+ },
245
+ {
246
+ "epoch": 1.1055276381909547,
247
+ "grad_norm": 6.663857936859131,
248
+ "learning_rate": 9.368606701940036e-06,
249
+ "loss": 0.2179,
250
+ "step": 660
251
+ },
252
+ {
253
+ "epoch": 1.1390284757118927,
254
+ "grad_norm": 4.486196041107178,
255
+ "learning_rate": 9.333333333333334e-06,
256
+ "loss": 0.2196,
257
+ "step": 680
258
+ },
259
+ {
260
+ "epoch": 1.1725293132328307,
261
+ "grad_norm": 4.667060852050781,
262
+ "learning_rate": 9.298059964726633e-06,
263
+ "loss": 0.2183,
264
+ "step": 700
265
+ },
266
+ {
267
+ "epoch": 1.2060301507537687,
268
+ "grad_norm": 5.608316898345947,
269
+ "learning_rate": 9.26278659611993e-06,
270
+ "loss": 0.2161,
271
+ "step": 720
272
+ },
273
+ {
274
+ "epoch": 1.2395309882747068,
275
+ "grad_norm": 4.2184271812438965,
276
+ "learning_rate": 9.227513227513229e-06,
277
+ "loss": 0.2382,
278
+ "step": 740
279
+ },
280
+ {
281
+ "epoch": 1.2730318257956448,
282
+ "grad_norm": 3.871945381164551,
283
+ "learning_rate": 9.192239858906526e-06,
284
+ "loss": 0.2214,
285
+ "step": 760
286
+ },
287
+ {
288
+ "epoch": 1.3065326633165828,
289
+ "grad_norm": 3.730222225189209,
290
+ "learning_rate": 9.156966490299825e-06,
291
+ "loss": 0.2213,
292
+ "step": 780
293
+ },
294
+ {
295
+ "epoch": 1.3400335008375208,
296
+ "grad_norm": 4.740777015686035,
297
+ "learning_rate": 9.121693121693122e-06,
298
+ "loss": 0.2503,
299
+ "step": 800
300
+ },
301
+ {
302
+ "epoch": 1.3735343383584588,
303
+ "grad_norm": 4.123469352722168,
304
+ "learning_rate": 9.086419753086421e-06,
305
+ "loss": 0.2443,
306
+ "step": 820
307
+ },
308
+ {
309
+ "epoch": 1.4070351758793969,
310
+ "grad_norm": 3.988917827606201,
311
+ "learning_rate": 9.051146384479718e-06,
312
+ "loss": 0.241,
313
+ "step": 840
314
+ },
315
+ {
316
+ "epoch": 1.4405360134003349,
317
+ "grad_norm": 5.4730305671691895,
318
+ "learning_rate": 9.015873015873017e-06,
319
+ "loss": 0.2163,
320
+ "step": 860
321
+ },
322
+ {
323
+ "epoch": 1.474036850921273,
324
+ "grad_norm": 5.020652770996094,
325
+ "learning_rate": 8.980599647266314e-06,
326
+ "loss": 0.2387,
327
+ "step": 880
328
+ },
329
+ {
330
+ "epoch": 1.507537688442211,
331
+ "grad_norm": 4.231486797332764,
332
+ "learning_rate": 8.945326278659613e-06,
333
+ "loss": 0.2523,
334
+ "step": 900
335
+ },
336
+ {
337
+ "epoch": 1.541038525963149,
338
+ "grad_norm": 6.197975158691406,
339
+ "learning_rate": 8.910052910052912e-06,
340
+ "loss": 0.2255,
341
+ "step": 920
342
+ },
343
+ {
344
+ "epoch": 1.574539363484087,
345
+ "grad_norm": 5.489210605621338,
346
+ "learning_rate": 8.874779541446209e-06,
347
+ "loss": 0.2397,
348
+ "step": 940
349
+ },
350
+ {
351
+ "epoch": 1.608040201005025,
352
+ "grad_norm": 4.32737398147583,
353
+ "learning_rate": 8.839506172839508e-06,
354
+ "loss": 0.2298,
355
+ "step": 960
356
+ },
357
+ {
358
+ "epoch": 1.641541038525963,
359
+ "grad_norm": 4.504214763641357,
360
+ "learning_rate": 8.804232804232805e-06,
361
+ "loss": 0.2301,
362
+ "step": 980
363
+ },
364
+ {
365
+ "epoch": 1.675041876046901,
366
+ "grad_norm": 3.9694759845733643,
367
+ "learning_rate": 8.768959435626104e-06,
368
+ "loss": 0.2338,
369
+ "step": 1000
370
+ },
371
+ {
372
+ "epoch": 1.675041876046901,
373
+ "eval_loss": 0.18838582932949066,
374
+ "eval_runtime": 1770.3449,
375
+ "eval_samples_per_second": 2.698,
376
+ "eval_steps_per_second": 0.337,
377
+ "eval_wer": 14.071299035983584,
378
+ "step": 1000
379
+ },
380
+ {
381
+ "epoch": 1.708542713567839,
382
+ "grad_norm": 4.972934722900391,
383
+ "learning_rate": 8.7336860670194e-06,
384
+ "loss": 0.2388,
385
+ "step": 1020
386
+ },
387
+ {
388
+ "epoch": 1.742043551088777,
389
+ "grad_norm": 5.178994655609131,
390
+ "learning_rate": 8.6984126984127e-06,
391
+ "loss": 0.2531,
392
+ "step": 1040
393
+ },
394
+ {
395
+ "epoch": 1.775544388609715,
396
+ "grad_norm": 3.560372829437256,
397
+ "learning_rate": 8.663139329805997e-06,
398
+ "loss": 0.2187,
399
+ "step": 1060
400
+ },
401
+ {
402
+ "epoch": 1.809045226130653,
403
+ "grad_norm": 4.647324562072754,
404
+ "learning_rate": 8.627865961199296e-06,
405
+ "loss": 0.2413,
406
+ "step": 1080
407
+ },
408
+ {
409
+ "epoch": 1.8425460636515911,
410
+ "grad_norm": 3.6784422397613525,
411
+ "learning_rate": 8.592592592592593e-06,
412
+ "loss": 0.2242,
413
+ "step": 1100
414
+ },
415
+ {
416
+ "epoch": 1.8760469011725294,
417
+ "grad_norm": 5.928018093109131,
418
+ "learning_rate": 8.557319223985891e-06,
419
+ "loss": 0.2406,
420
+ "step": 1120
421
+ },
422
+ {
423
+ "epoch": 1.9095477386934674,
424
+ "grad_norm": 6.358487606048584,
425
+ "learning_rate": 8.52204585537919e-06,
426
+ "loss": 0.2482,
427
+ "step": 1140
428
+ },
429
+ {
430
+ "epoch": 1.9430485762144054,
431
+ "grad_norm": 4.409506797790527,
432
+ "learning_rate": 8.486772486772487e-06,
433
+ "loss": 0.2415,
434
+ "step": 1160
435
+ },
436
+ {
437
+ "epoch": 1.9765494137353434,
438
+ "grad_norm": 5.169639587402344,
439
+ "learning_rate": 8.451499118165786e-06,
440
+ "loss": 0.222,
441
+ "step": 1180
442
+ },
443
+ {
444
+ "epoch": 2.0100502512562812,
445
+ "grad_norm": 2.8841328620910645,
446
+ "learning_rate": 8.416225749559083e-06,
447
+ "loss": 0.2109,
448
+ "step": 1200
449
+ },
450
+ {
451
+ "epoch": 2.0435510887772192,
452
+ "grad_norm": 3.835498332977295,
453
+ "learning_rate": 8.380952380952382e-06,
454
+ "loss": 0.1284,
455
+ "step": 1220
456
+ },
457
+ {
458
+ "epoch": 2.0770519262981573,
459
+ "grad_norm": 2.584859848022461,
460
+ "learning_rate": 8.34567901234568e-06,
461
+ "loss": 0.1059,
462
+ "step": 1240
463
+ },
464
+ {
465
+ "epoch": 2.1105527638190953,
466
+ "grad_norm": 2.797682285308838,
467
+ "learning_rate": 8.310405643738978e-06,
468
+ "loss": 0.0967,
469
+ "step": 1260
470
+ },
471
+ {
472
+ "epoch": 2.1440536013400333,
473
+ "grad_norm": 3.2937309741973877,
474
+ "learning_rate": 8.275132275132275e-06,
475
+ "loss": 0.1267,
476
+ "step": 1280
477
+ },
478
+ {
479
+ "epoch": 2.1775544388609713,
480
+ "grad_norm": 3.344325065612793,
481
+ "learning_rate": 8.239858906525574e-06,
482
+ "loss": 0.1354,
483
+ "step": 1300
484
+ },
485
+ {
486
+ "epoch": 2.2110552763819094,
487
+ "grad_norm": 2.955522060394287,
488
+ "learning_rate": 8.204585537918873e-06,
489
+ "loss": 0.104,
490
+ "step": 1320
491
+ },
492
+ {
493
+ "epoch": 2.2445561139028474,
494
+ "grad_norm": 5.180908203125,
495
+ "learning_rate": 8.16931216931217e-06,
496
+ "loss": 0.1229,
497
+ "step": 1340
498
+ },
499
+ {
500
+ "epoch": 2.2780569514237854,
501
+ "grad_norm": 3.639417886734009,
502
+ "learning_rate": 8.134038800705469e-06,
503
+ "loss": 0.104,
504
+ "step": 1360
505
+ },
506
+ {
507
+ "epoch": 2.3115577889447234,
508
+ "grad_norm": 3.7380902767181396,
509
+ "learning_rate": 8.098765432098766e-06,
510
+ "loss": 0.1141,
511
+ "step": 1380
512
+ },
513
+ {
514
+ "epoch": 2.3450586264656614,
515
+ "grad_norm": 4.100318908691406,
516
+ "learning_rate": 8.063492063492065e-06,
517
+ "loss": 0.1123,
518
+ "step": 1400
519
+ },
520
+ {
521
+ "epoch": 2.3785594639865995,
522
+ "grad_norm": 2.8147048950195312,
523
+ "learning_rate": 8.028218694885362e-06,
524
+ "loss": 0.1039,
525
+ "step": 1420
526
+ },
527
+ {
528
+ "epoch": 2.4120603015075375,
529
+ "grad_norm": 3.7376410961151123,
530
+ "learning_rate": 7.99294532627866e-06,
531
+ "loss": 0.1179,
532
+ "step": 1440
533
+ },
534
+ {
535
+ "epoch": 2.4455611390284755,
536
+ "grad_norm": 4.320065975189209,
537
+ "learning_rate": 7.957671957671958e-06,
538
+ "loss": 0.1167,
539
+ "step": 1460
540
+ },
541
+ {
542
+ "epoch": 2.4790619765494135,
543
+ "grad_norm": 4.415127277374268,
544
+ "learning_rate": 7.922398589065257e-06,
545
+ "loss": 0.125,
546
+ "step": 1480
547
+ },
548
+ {
549
+ "epoch": 2.5125628140703515,
550
+ "grad_norm": 2.631763219833374,
551
+ "learning_rate": 7.887125220458554e-06,
552
+ "loss": 0.1188,
553
+ "step": 1500
554
+ },
555
+ {
556
+ "epoch": 2.5125628140703515,
557
+ "eval_loss": 0.13635103404521942,
558
+ "eval_runtime": 1806.1099,
559
+ "eval_samples_per_second": 2.644,
560
+ "eval_steps_per_second": 0.331,
561
+ "eval_wer": 11.713753937195762,
562
+ "step": 1500
563
+ },
564
+ {
565
+ "epoch": 2.5460636515912896,
566
+ "grad_norm": 3.608773946762085,
567
+ "learning_rate": 7.851851851851853e-06,
568
+ "loss": 0.1222,
569
+ "step": 1520
570
+ },
571
+ {
572
+ "epoch": 2.5795644891122276,
573
+ "grad_norm": 3.4494569301605225,
574
+ "learning_rate": 7.816578483245151e-06,
575
+ "loss": 0.1087,
576
+ "step": 1540
577
+ },
578
+ {
579
+ "epoch": 2.6130653266331656,
580
+ "grad_norm": 3.6112048625946045,
581
+ "learning_rate": 7.781305114638449e-06,
582
+ "loss": 0.1097,
583
+ "step": 1560
584
+ },
585
+ {
586
+ "epoch": 2.6465661641541036,
587
+ "grad_norm": 3.1978707313537598,
588
+ "learning_rate": 7.746031746031747e-06,
589
+ "loss": 0.1101,
590
+ "step": 1580
591
+ },
592
+ {
593
+ "epoch": 2.6800670016750416,
594
+ "grad_norm": 3.6376733779907227,
595
+ "learning_rate": 7.710758377425045e-06,
596
+ "loss": 0.1056,
597
+ "step": 1600
598
+ },
599
+ {
600
+ "epoch": 2.7135678391959797,
601
+ "grad_norm": 3.754915237426758,
602
+ "learning_rate": 7.675485008818343e-06,
603
+ "loss": 0.1103,
604
+ "step": 1620
605
+ },
606
+ {
607
+ "epoch": 2.7470686767169177,
608
+ "grad_norm": 3.329240322113037,
609
+ "learning_rate": 7.64021164021164e-06,
610
+ "loss": 0.1193,
611
+ "step": 1640
612
+ },
613
+ {
614
+ "epoch": 2.7805695142378557,
615
+ "grad_norm": 3.9300169944763184,
616
+ "learning_rate": 7.604938271604939e-06,
617
+ "loss": 0.1232,
618
+ "step": 1660
619
+ },
620
+ {
621
+ "epoch": 2.8140703517587937,
622
+ "grad_norm": 4.290626049041748,
623
+ "learning_rate": 7.569664902998237e-06,
624
+ "loss": 0.1374,
625
+ "step": 1680
626
+ },
627
+ {
628
+ "epoch": 2.8475711892797317,
629
+ "grad_norm": 3.8987998962402344,
630
+ "learning_rate": 7.534391534391535e-06,
631
+ "loss": 0.1062,
632
+ "step": 1700
633
+ },
634
+ {
635
+ "epoch": 2.8810720268006698,
636
+ "grad_norm": 4.234960079193115,
637
+ "learning_rate": 7.499118165784833e-06,
638
+ "loss": 0.111,
639
+ "step": 1720
640
+ },
641
+ {
642
+ "epoch": 2.914572864321608,
643
+ "grad_norm": 3.103458881378174,
644
+ "learning_rate": 7.463844797178131e-06,
645
+ "loss": 0.1076,
646
+ "step": 1740
647
+ },
648
+ {
649
+ "epoch": 2.948073701842546,
650
+ "grad_norm": 3.270204782485962,
651
+ "learning_rate": 7.428571428571429e-06,
652
+ "loss": 0.1224,
653
+ "step": 1760
654
+ },
655
+ {
656
+ "epoch": 2.981574539363484,
657
+ "grad_norm": 4.261337757110596,
658
+ "learning_rate": 7.393298059964727e-06,
659
+ "loss": 0.1224,
660
+ "step": 1780
661
+ },
662
+ {
663
+ "epoch": 3.0150753768844223,
664
+ "grad_norm": 1.9531301259994507,
665
+ "learning_rate": 7.358024691358025e-06,
666
+ "loss": 0.0913,
667
+ "step": 1800
668
+ },
669
+ {
670
+ "epoch": 3.0485762144053603,
671
+ "grad_norm": 1.866215467453003,
672
+ "learning_rate": 7.322751322751324e-06,
673
+ "loss": 0.0466,
674
+ "step": 1820
675
+ },
676
+ {
677
+ "epoch": 3.0820770519262983,
678
+ "grad_norm": 3.9456610679626465,
679
+ "learning_rate": 7.287477954144622e-06,
680
+ "loss": 0.0451,
681
+ "step": 1840
682
+ },
683
+ {
684
+ "epoch": 3.1155778894472363,
685
+ "grad_norm": 1.9518849849700928,
686
+ "learning_rate": 7.25220458553792e-06,
687
+ "loss": 0.0414,
688
+ "step": 1860
689
+ },
690
+ {
691
+ "epoch": 3.1490787269681744,
692
+ "grad_norm": 2.232792854309082,
693
+ "learning_rate": 7.216931216931218e-06,
694
+ "loss": 0.0502,
695
+ "step": 1880
696
+ },
697
+ {
698
+ "epoch": 3.1825795644891124,
699
+ "grad_norm": 2.074127674102783,
700
+ "learning_rate": 7.181657848324516e-06,
701
+ "loss": 0.057,
702
+ "step": 1900
703
+ },
704
+ {
705
+ "epoch": 3.2160804020100504,
706
+ "grad_norm": 3.5036733150482178,
707
+ "learning_rate": 7.146384479717814e-06,
708
+ "loss": 0.0515,
709
+ "step": 1920
710
+ },
711
+ {
712
+ "epoch": 3.2495812395309884,
713
+ "grad_norm": 2.324014186859131,
714
+ "learning_rate": 7.111111111111112e-06,
715
+ "loss": 0.0454,
716
+ "step": 1940
717
+ },
718
+ {
719
+ "epoch": 3.2830820770519265,
720
+ "grad_norm": 2.710326671600342,
721
+ "learning_rate": 7.07583774250441e-06,
722
+ "loss": 0.0473,
723
+ "step": 1960
724
+ },
725
+ {
726
+ "epoch": 3.3165829145728645,
727
+ "grad_norm": 2.8943896293640137,
728
+ "learning_rate": 7.040564373897708e-06,
729
+ "loss": 0.046,
730
+ "step": 1980
731
+ },
732
+ {
733
+ "epoch": 3.3500837520938025,
734
+ "grad_norm": 2.319986581802368,
735
+ "learning_rate": 7.005291005291006e-06,
736
+ "loss": 0.0449,
737
+ "step": 2000
738
+ },
739
+ {
740
+ "epoch": 3.3500837520938025,
741
+ "eval_loss": 0.10934468358755112,
742
+ "eval_runtime": 1762.4506,
743
+ "eval_samples_per_second": 2.71,
744
+ "eval_steps_per_second": 0.339,
745
+ "eval_wer": 7.172854824854443,
746
+ "step": 2000
747
+ }
748
+ ],
749
+ "logging_steps": 20,
750
+ "max_steps": 5970,
751
+ "num_input_tokens_seen": 0,
752
+ "num_train_epochs": 10,
753
+ "save_steps": 500,
754
+ "stateful_callbacks": {
755
+ "TrainerControl": {
756
+ "args": {
757
+ "should_epoch_stop": false,
758
+ "should_evaluate": false,
759
+ "should_log": false,
760
+ "should_save": true,
761
+ "should_training_stop": false
762
+ },
763
+ "attributes": {}
764
+ }
765
+ },
766
+ "total_flos": 1.846686835187712e+19,
767
+ "train_batch_size": 16,
768
+ "trial_name": null,
769
+ "trial_params": null
770
+ }
checkpoint-2000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41a3c645624f7463029286b7d5e1ad6924cfc44a0d153880431b77c40e5deb73
3
+ size 5432
checkpoint-2500/config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.0,
3
+ "activation_function": "gelu",
4
+ "apply_spec_augment": false,
5
+ "architectures": [
6
+ "WhisperForConditionalGeneration"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "begin_suppress_tokens": null,
10
+ "bos_token_id": 50257,
11
+ "classifier_proj_size": 256,
12
+ "d_model": 768,
13
+ "decoder_attention_heads": 12,
14
+ "decoder_ffn_dim": 3072,
15
+ "decoder_layerdrop": 0.0,
16
+ "decoder_layers": 12,
17
+ "decoder_start_token_id": 50258,
18
+ "dropout": 0.0,
19
+ "encoder_attention_heads": 12,
20
+ "encoder_ffn_dim": 3072,
21
+ "encoder_layerdrop": 0.0,
22
+ "encoder_layers": 12,
23
+ "eos_token_id": 50257,
24
+ "forced_decoder_ids": null,
25
+ "init_std": 0.02,
26
+ "is_encoder_decoder": true,
27
+ "mask_feature_length": 10,
28
+ "mask_feature_min_masks": 0,
29
+ "mask_feature_prob": 0.0,
30
+ "mask_time_length": 10,
31
+ "mask_time_min_masks": 2,
32
+ "mask_time_prob": 0.05,
33
+ "max_length": null,
34
+ "max_source_positions": 1500,
35
+ "max_target_positions": 448,
36
+ "median_filter_width": 7,
37
+ "model_type": "whisper",
38
+ "num_hidden_layers": 12,
39
+ "num_mel_bins": 80,
40
+ "pad_token_id": 50257,
41
+ "scale_embedding": false,
42
+ "torch_dtype": "float32",
43
+ "transformers_version": "4.51.3",
44
+ "use_cache": false,
45
+ "use_weighted_layer_sum": false,
46
+ "vocab_size": 51865
47
+ }
checkpoint-2500/generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 50257,
3
+ "decoder_start_token_id": 50258,
4
+ "eos_token_id": 50257,
5
+ "pad_token_id": 50257,
6
+ "suppress_tokens": [],
7
+ "transformers_version": "4.51.3",
8
+ "use_cache": false
9
+ }
checkpoint-2500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:983ad47253d47f33288d03a6dc02f69916cf46fab753a6c1fc58916fd035349a
3
+ size 966995080
checkpoint-2500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30d507a0591cd28124a7b5e2ceb3f1b6db701163d043cdf64538d377ff5de8c1
3
+ size 1925064044
checkpoint-2500/preprocessor_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "chunk_length": 30,
3
+ "dither": 0.0,
4
+ "feature_extractor_type": "WhisperFeatureExtractor",
5
+ "feature_size": 80,
6
+ "hop_length": 160,
7
+ "n_fft": 400,
8
+ "n_samples": 480000,
9
+ "nb_max_frames": 3000,
10
+ "padding_side": "right",
11
+ "padding_value": 0.0,
12
+ "processor_class": "WhisperProcessor",
13
+ "return_attention_mask": false,
14
+ "sampling_rate": 16000
15
+ }
checkpoint-2500/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:846458c7462562ffbdb4648906c3b912a1d942419c6cf6acd73bf7bd018aae2c
3
+ size 14244
checkpoint-2500/scaler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f518493be2847dc24c07ed29ea8a55ff05bdfa05590734e42ae275f087cf2300
3
+ size 988
checkpoint-2500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d4dcde2bb7ed571bf37cb3c2010a2bd3c91906a89169b61a4be9525501656fe
3
+ size 1064
checkpoint-2500/trainer_state.json ADDED
@@ -0,0 +1,954 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 2500,
3
+ "best_metric": 6.707549871146321,
4
+ "best_model_checkpoint": "./SALAMA_C3/checkpoint-2500",
5
+ "epoch": 4.187604690117253,
6
+ "eval_steps": 500,
7
+ "global_step": 2500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.03350083752093802,
14
+ "grad_norm": 11.757207870483398,
15
+ "learning_rate": 6.333333333333334e-07,
16
+ "loss": 0.5112,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.06700167504187604,
21
+ "grad_norm": 10.197953224182129,
22
+ "learning_rate": 1.3e-06,
23
+ "loss": 0.5283,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.10050251256281408,
28
+ "grad_norm": 12.811074256896973,
29
+ "learning_rate": 1.9666666666666668e-06,
30
+ "loss": 0.474,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.13400335008375208,
35
+ "grad_norm": 8.559342384338379,
36
+ "learning_rate": 2.6e-06,
37
+ "loss": 0.424,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.16750418760469013,
42
+ "grad_norm": 6.9629225730896,
43
+ "learning_rate": 3.266666666666667e-06,
44
+ "loss": 0.4465,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.20100502512562815,
49
+ "grad_norm": 7.833058834075928,
50
+ "learning_rate": 3.9333333333333335e-06,
51
+ "loss": 0.3764,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.23450586264656617,
56
+ "grad_norm": 6.882424831390381,
57
+ "learning_rate": 4.600000000000001e-06,
58
+ "loss": 0.3705,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.26800670016750416,
63
+ "grad_norm": 5.4183244705200195,
64
+ "learning_rate": 5.2666666666666665e-06,
65
+ "loss": 0.4116,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.3015075376884422,
70
+ "grad_norm": 6.103787899017334,
71
+ "learning_rate": 5.933333333333335e-06,
72
+ "loss": 0.3903,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.33500837520938026,
77
+ "grad_norm": 6.084160804748535,
78
+ "learning_rate": 6.600000000000001e-06,
79
+ "loss": 0.3794,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.3685092127303183,
84
+ "grad_norm": 4.531406879425049,
85
+ "learning_rate": 7.266666666666668e-06,
86
+ "loss": 0.3904,
87
+ "step": 220
88
+ },
89
+ {
90
+ "epoch": 0.4020100502512563,
91
+ "grad_norm": 6.702854156494141,
92
+ "learning_rate": 7.933333333333334e-06,
93
+ "loss": 0.3672,
94
+ "step": 240
95
+ },
96
+ {
97
+ "epoch": 0.4355108877721943,
98
+ "grad_norm": 6.386377811431885,
99
+ "learning_rate": 8.6e-06,
100
+ "loss": 0.3379,
101
+ "step": 260
102
+ },
103
+ {
104
+ "epoch": 0.46901172529313234,
105
+ "grad_norm": 5.33281135559082,
106
+ "learning_rate": 9.266666666666667e-06,
107
+ "loss": 0.4102,
108
+ "step": 280
109
+ },
110
+ {
111
+ "epoch": 0.5025125628140703,
112
+ "grad_norm": 6.031435012817383,
113
+ "learning_rate": 9.933333333333334e-06,
114
+ "loss": 0.3815,
115
+ "step": 300
116
+ },
117
+ {
118
+ "epoch": 0.5360134003350083,
119
+ "grad_norm": 4.8638176918029785,
120
+ "learning_rate": 9.968253968253969e-06,
121
+ "loss": 0.358,
122
+ "step": 320
123
+ },
124
+ {
125
+ "epoch": 0.5695142378559463,
126
+ "grad_norm": 5.352113723754883,
127
+ "learning_rate": 9.932980599647268e-06,
128
+ "loss": 0.3578,
129
+ "step": 340
130
+ },
131
+ {
132
+ "epoch": 0.6030150753768844,
133
+ "grad_norm": 5.560739040374756,
134
+ "learning_rate": 9.897707231040565e-06,
135
+ "loss": 0.3795,
136
+ "step": 360
137
+ },
138
+ {
139
+ "epoch": 0.6365159128978225,
140
+ "grad_norm": 6.186940670013428,
141
+ "learning_rate": 9.862433862433864e-06,
142
+ "loss": 0.3852,
143
+ "step": 380
144
+ },
145
+ {
146
+ "epoch": 0.6700167504187605,
147
+ "grad_norm": 5.37507438659668,
148
+ "learning_rate": 9.827160493827161e-06,
149
+ "loss": 0.3761,
150
+ "step": 400
151
+ },
152
+ {
153
+ "epoch": 0.7035175879396985,
154
+ "grad_norm": 6.373449802398682,
155
+ "learning_rate": 9.79188712522046e-06,
156
+ "loss": 0.3674,
157
+ "step": 420
158
+ },
159
+ {
160
+ "epoch": 0.7370184254606366,
161
+ "grad_norm": 5.702625274658203,
162
+ "learning_rate": 9.756613756613757e-06,
163
+ "loss": 0.4013,
164
+ "step": 440
165
+ },
166
+ {
167
+ "epoch": 0.7705192629815746,
168
+ "grad_norm": 5.5675153732299805,
169
+ "learning_rate": 9.721340388007056e-06,
170
+ "loss": 0.3495,
171
+ "step": 460
172
+ },
173
+ {
174
+ "epoch": 0.8040201005025126,
175
+ "grad_norm": 6.296374320983887,
176
+ "learning_rate": 9.686067019400353e-06,
177
+ "loss": 0.3704,
178
+ "step": 480
179
+ },
180
+ {
181
+ "epoch": 0.8375209380234506,
182
+ "grad_norm": 4.830463886260986,
183
+ "learning_rate": 9.650793650793652e-06,
184
+ "loss": 0.36,
185
+ "step": 500
186
+ },
187
+ {
188
+ "epoch": 0.8375209380234506,
189
+ "eval_loss": 0.27764827013015747,
190
+ "eval_runtime": 1752.3826,
191
+ "eval_samples_per_second": 2.725,
192
+ "eval_steps_per_second": 0.341,
193
+ "eval_wer": 19.59291781998664,
194
+ "step": 500
195
+ },
196
+ {
197
+ "epoch": 0.8710217755443886,
198
+ "grad_norm": 4.863058090209961,
199
+ "learning_rate": 9.61552028218695e-06,
200
+ "loss": 0.3874,
201
+ "step": 520
202
+ },
203
+ {
204
+ "epoch": 0.9045226130653267,
205
+ "grad_norm": 6.121027946472168,
206
+ "learning_rate": 9.580246913580248e-06,
207
+ "loss": 0.382,
208
+ "step": 540
209
+ },
210
+ {
211
+ "epoch": 0.9380234505862647,
212
+ "grad_norm": 4.538082599639893,
213
+ "learning_rate": 9.544973544973546e-06,
214
+ "loss": 0.3611,
215
+ "step": 560
216
+ },
217
+ {
218
+ "epoch": 0.9715242881072027,
219
+ "grad_norm": 4.8772382736206055,
220
+ "learning_rate": 9.509700176366844e-06,
221
+ "loss": 0.3584,
222
+ "step": 580
223
+ },
224
+ {
225
+ "epoch": 1.0050251256281406,
226
+ "grad_norm": 4.039212703704834,
227
+ "learning_rate": 9.474426807760142e-06,
228
+ "loss": 0.3701,
229
+ "step": 600
230
+ },
231
+ {
232
+ "epoch": 1.0385259631490786,
233
+ "grad_norm": 4.787687301635742,
234
+ "learning_rate": 9.43915343915344e-06,
235
+ "loss": 0.2058,
236
+ "step": 620
237
+ },
238
+ {
239
+ "epoch": 1.0720268006700167,
240
+ "grad_norm": 4.503021717071533,
241
+ "learning_rate": 9.403880070546738e-06,
242
+ "loss": 0.2455,
243
+ "step": 640
244
+ },
245
+ {
246
+ "epoch": 1.1055276381909547,
247
+ "grad_norm": 6.663857936859131,
248
+ "learning_rate": 9.368606701940036e-06,
249
+ "loss": 0.2179,
250
+ "step": 660
251
+ },
252
+ {
253
+ "epoch": 1.1390284757118927,
254
+ "grad_norm": 4.486196041107178,
255
+ "learning_rate": 9.333333333333334e-06,
256
+ "loss": 0.2196,
257
+ "step": 680
258
+ },
259
+ {
260
+ "epoch": 1.1725293132328307,
261
+ "grad_norm": 4.667060852050781,
262
+ "learning_rate": 9.298059964726633e-06,
263
+ "loss": 0.2183,
264
+ "step": 700
265
+ },
266
+ {
267
+ "epoch": 1.2060301507537687,
268
+ "grad_norm": 5.608316898345947,
269
+ "learning_rate": 9.26278659611993e-06,
270
+ "loss": 0.2161,
271
+ "step": 720
272
+ },
273
+ {
274
+ "epoch": 1.2395309882747068,
275
+ "grad_norm": 4.2184271812438965,
276
+ "learning_rate": 9.227513227513229e-06,
277
+ "loss": 0.2382,
278
+ "step": 740
279
+ },
280
+ {
281
+ "epoch": 1.2730318257956448,
282
+ "grad_norm": 3.871945381164551,
283
+ "learning_rate": 9.192239858906526e-06,
284
+ "loss": 0.2214,
285
+ "step": 760
286
+ },
287
+ {
288
+ "epoch": 1.3065326633165828,
289
+ "grad_norm": 3.730222225189209,
290
+ "learning_rate": 9.156966490299825e-06,
291
+ "loss": 0.2213,
292
+ "step": 780
293
+ },
294
+ {
295
+ "epoch": 1.3400335008375208,
296
+ "grad_norm": 4.740777015686035,
297
+ "learning_rate": 9.121693121693122e-06,
298
+ "loss": 0.2503,
299
+ "step": 800
300
+ },
301
+ {
302
+ "epoch": 1.3735343383584588,
303
+ "grad_norm": 4.123469352722168,
304
+ "learning_rate": 9.086419753086421e-06,
305
+ "loss": 0.2443,
306
+ "step": 820
307
+ },
308
+ {
309
+ "epoch": 1.4070351758793969,
310
+ "grad_norm": 3.988917827606201,
311
+ "learning_rate": 9.051146384479718e-06,
312
+ "loss": 0.241,
313
+ "step": 840
314
+ },
315
+ {
316
+ "epoch": 1.4405360134003349,
317
+ "grad_norm": 5.4730305671691895,
318
+ "learning_rate": 9.015873015873017e-06,
319
+ "loss": 0.2163,
320
+ "step": 860
321
+ },
322
+ {
323
+ "epoch": 1.474036850921273,
324
+ "grad_norm": 5.020652770996094,
325
+ "learning_rate": 8.980599647266314e-06,
326
+ "loss": 0.2387,
327
+ "step": 880
328
+ },
329
+ {
330
+ "epoch": 1.507537688442211,
331
+ "grad_norm": 4.231486797332764,
332
+ "learning_rate": 8.945326278659613e-06,
333
+ "loss": 0.2523,
334
+ "step": 900
335
+ },
336
+ {
337
+ "epoch": 1.541038525963149,
338
+ "grad_norm": 6.197975158691406,
339
+ "learning_rate": 8.910052910052912e-06,
340
+ "loss": 0.2255,
341
+ "step": 920
342
+ },
343
+ {
344
+ "epoch": 1.574539363484087,
345
+ "grad_norm": 5.489210605621338,
346
+ "learning_rate": 8.874779541446209e-06,
347
+ "loss": 0.2397,
348
+ "step": 940
349
+ },
350
+ {
351
+ "epoch": 1.608040201005025,
352
+ "grad_norm": 4.32737398147583,
353
+ "learning_rate": 8.839506172839508e-06,
354
+ "loss": 0.2298,
355
+ "step": 960
356
+ },
357
+ {
358
+ "epoch": 1.641541038525963,
359
+ "grad_norm": 4.504214763641357,
360
+ "learning_rate": 8.804232804232805e-06,
361
+ "loss": 0.2301,
362
+ "step": 980
363
+ },
364
+ {
365
+ "epoch": 1.675041876046901,
366
+ "grad_norm": 3.9694759845733643,
367
+ "learning_rate": 8.768959435626104e-06,
368
+ "loss": 0.2338,
369
+ "step": 1000
370
+ },
371
+ {
372
+ "epoch": 1.675041876046901,
373
+ "eval_loss": 0.18838582932949066,
374
+ "eval_runtime": 1770.3449,
375
+ "eval_samples_per_second": 2.698,
376
+ "eval_steps_per_second": 0.337,
377
+ "eval_wer": 14.071299035983584,
378
+ "step": 1000
379
+ },
380
+ {
381
+ "epoch": 1.708542713567839,
382
+ "grad_norm": 4.972934722900391,
383
+ "learning_rate": 8.7336860670194e-06,
384
+ "loss": 0.2388,
385
+ "step": 1020
386
+ },
387
+ {
388
+ "epoch": 1.742043551088777,
389
+ "grad_norm": 5.178994655609131,
390
+ "learning_rate": 8.6984126984127e-06,
391
+ "loss": 0.2531,
392
+ "step": 1040
393
+ },
394
+ {
395
+ "epoch": 1.775544388609715,
396
+ "grad_norm": 3.560372829437256,
397
+ "learning_rate": 8.663139329805997e-06,
398
+ "loss": 0.2187,
399
+ "step": 1060
400
+ },
401
+ {
402
+ "epoch": 1.809045226130653,
403
+ "grad_norm": 4.647324562072754,
404
+ "learning_rate": 8.627865961199296e-06,
405
+ "loss": 0.2413,
406
+ "step": 1080
407
+ },
408
+ {
409
+ "epoch": 1.8425460636515911,
410
+ "grad_norm": 3.6784422397613525,
411
+ "learning_rate": 8.592592592592593e-06,
412
+ "loss": 0.2242,
413
+ "step": 1100
414
+ },
415
+ {
416
+ "epoch": 1.8760469011725294,
417
+ "grad_norm": 5.928018093109131,
418
+ "learning_rate": 8.557319223985891e-06,
419
+ "loss": 0.2406,
420
+ "step": 1120
421
+ },
422
+ {
423
+ "epoch": 1.9095477386934674,
424
+ "grad_norm": 6.358487606048584,
425
+ "learning_rate": 8.52204585537919e-06,
426
+ "loss": 0.2482,
427
+ "step": 1140
428
+ },
429
+ {
430
+ "epoch": 1.9430485762144054,
431
+ "grad_norm": 4.409506797790527,
432
+ "learning_rate": 8.486772486772487e-06,
433
+ "loss": 0.2415,
434
+ "step": 1160
435
+ },
436
+ {
437
+ "epoch": 1.9765494137353434,
438
+ "grad_norm": 5.169639587402344,
439
+ "learning_rate": 8.451499118165786e-06,
440
+ "loss": 0.222,
441
+ "step": 1180
442
+ },
443
+ {
444
+ "epoch": 2.0100502512562812,
445
+ "grad_norm": 2.8841328620910645,
446
+ "learning_rate": 8.416225749559083e-06,
447
+ "loss": 0.2109,
448
+ "step": 1200
449
+ },
450
+ {
451
+ "epoch": 2.0435510887772192,
452
+ "grad_norm": 3.835498332977295,
453
+ "learning_rate": 8.380952380952382e-06,
454
+ "loss": 0.1284,
455
+ "step": 1220
456
+ },
457
+ {
458
+ "epoch": 2.0770519262981573,
459
+ "grad_norm": 2.584859848022461,
460
+ "learning_rate": 8.34567901234568e-06,
461
+ "loss": 0.1059,
462
+ "step": 1240
463
+ },
464
+ {
465
+ "epoch": 2.1105527638190953,
466
+ "grad_norm": 2.797682285308838,
467
+ "learning_rate": 8.310405643738978e-06,
468
+ "loss": 0.0967,
469
+ "step": 1260
470
+ },
471
+ {
472
+ "epoch": 2.1440536013400333,
473
+ "grad_norm": 3.2937309741973877,
474
+ "learning_rate": 8.275132275132275e-06,
475
+ "loss": 0.1267,
476
+ "step": 1280
477
+ },
478
+ {
479
+ "epoch": 2.1775544388609713,
480
+ "grad_norm": 3.344325065612793,
481
+ "learning_rate": 8.239858906525574e-06,
482
+ "loss": 0.1354,
483
+ "step": 1300
484
+ },
485
+ {
486
+ "epoch": 2.2110552763819094,
487
+ "grad_norm": 2.955522060394287,
488
+ "learning_rate": 8.204585537918873e-06,
489
+ "loss": 0.104,
490
+ "step": 1320
491
+ },
492
+ {
493
+ "epoch": 2.2445561139028474,
494
+ "grad_norm": 5.180908203125,
495
+ "learning_rate": 8.16931216931217e-06,
496
+ "loss": 0.1229,
497
+ "step": 1340
498
+ },
499
+ {
500
+ "epoch": 2.2780569514237854,
501
+ "grad_norm": 3.639417886734009,
502
+ "learning_rate": 8.134038800705469e-06,
503
+ "loss": 0.104,
504
+ "step": 1360
505
+ },
506
+ {
507
+ "epoch": 2.3115577889447234,
508
+ "grad_norm": 3.7380902767181396,
509
+ "learning_rate": 8.098765432098766e-06,
510
+ "loss": 0.1141,
511
+ "step": 1380
512
+ },
513
+ {
514
+ "epoch": 2.3450586264656614,
515
+ "grad_norm": 4.100318908691406,
516
+ "learning_rate": 8.063492063492065e-06,
517
+ "loss": 0.1123,
518
+ "step": 1400
519
+ },
520
+ {
521
+ "epoch": 2.3785594639865995,
522
+ "grad_norm": 2.8147048950195312,
523
+ "learning_rate": 8.028218694885362e-06,
524
+ "loss": 0.1039,
525
+ "step": 1420
526
+ },
527
+ {
528
+ "epoch": 2.4120603015075375,
529
+ "grad_norm": 3.7376410961151123,
530
+ "learning_rate": 7.99294532627866e-06,
531
+ "loss": 0.1179,
532
+ "step": 1440
533
+ },
534
+ {
535
+ "epoch": 2.4455611390284755,
536
+ "grad_norm": 4.320065975189209,
537
+ "learning_rate": 7.957671957671958e-06,
538
+ "loss": 0.1167,
539
+ "step": 1460
540
+ },
541
+ {
542
+ "epoch": 2.4790619765494135,
543
+ "grad_norm": 4.415127277374268,
544
+ "learning_rate": 7.922398589065257e-06,
545
+ "loss": 0.125,
546
+ "step": 1480
547
+ },
548
+ {
549
+ "epoch": 2.5125628140703515,
550
+ "grad_norm": 2.631763219833374,
551
+ "learning_rate": 7.887125220458554e-06,
552
+ "loss": 0.1188,
553
+ "step": 1500
554
+ },
555
+ {
556
+ "epoch": 2.5125628140703515,
557
+ "eval_loss": 0.13635103404521942,
558
+ "eval_runtime": 1806.1099,
559
+ "eval_samples_per_second": 2.644,
560
+ "eval_steps_per_second": 0.331,
561
+ "eval_wer": 11.713753937195762,
562
+ "step": 1500
563
+ },
564
+ {
565
+ "epoch": 2.5460636515912896,
566
+ "grad_norm": 3.608773946762085,
567
+ "learning_rate": 7.851851851851853e-06,
568
+ "loss": 0.1222,
569
+ "step": 1520
570
+ },
571
+ {
572
+ "epoch": 2.5795644891122276,
573
+ "grad_norm": 3.4494569301605225,
574
+ "learning_rate": 7.816578483245151e-06,
575
+ "loss": 0.1087,
576
+ "step": 1540
577
+ },
578
+ {
579
+ "epoch": 2.6130653266331656,
580
+ "grad_norm": 3.6112048625946045,
581
+ "learning_rate": 7.781305114638449e-06,
582
+ "loss": 0.1097,
583
+ "step": 1560
584
+ },
585
+ {
586
+ "epoch": 2.6465661641541036,
587
+ "grad_norm": 3.1978707313537598,
588
+ "learning_rate": 7.746031746031747e-06,
589
+ "loss": 0.1101,
590
+ "step": 1580
591
+ },
592
+ {
593
+ "epoch": 2.6800670016750416,
594
+ "grad_norm": 3.6376733779907227,
595
+ "learning_rate": 7.710758377425045e-06,
596
+ "loss": 0.1056,
597
+ "step": 1600
598
+ },
599
+ {
600
+ "epoch": 2.7135678391959797,
601
+ "grad_norm": 3.754915237426758,
602
+ "learning_rate": 7.675485008818343e-06,
603
+ "loss": 0.1103,
604
+ "step": 1620
605
+ },
606
+ {
607
+ "epoch": 2.7470686767169177,
608
+ "grad_norm": 3.329240322113037,
609
+ "learning_rate": 7.64021164021164e-06,
610
+ "loss": 0.1193,
611
+ "step": 1640
612
+ },
613
+ {
614
+ "epoch": 2.7805695142378557,
615
+ "grad_norm": 3.9300169944763184,
616
+ "learning_rate": 7.604938271604939e-06,
617
+ "loss": 0.1232,
618
+ "step": 1660
619
+ },
620
+ {
621
+ "epoch": 2.8140703517587937,
622
+ "grad_norm": 4.290626049041748,
623
+ "learning_rate": 7.569664902998237e-06,
624
+ "loss": 0.1374,
625
+ "step": 1680
626
+ },
627
+ {
628
+ "epoch": 2.8475711892797317,
629
+ "grad_norm": 3.8987998962402344,
630
+ "learning_rate": 7.534391534391535e-06,
631
+ "loss": 0.1062,
632
+ "step": 1700
633
+ },
634
+ {
635
+ "epoch": 2.8810720268006698,
636
+ "grad_norm": 4.234960079193115,
637
+ "learning_rate": 7.499118165784833e-06,
638
+ "loss": 0.111,
639
+ "step": 1720
640
+ },
641
+ {
642
+ "epoch": 2.914572864321608,
643
+ "grad_norm": 3.103458881378174,
644
+ "learning_rate": 7.463844797178131e-06,
645
+ "loss": 0.1076,
646
+ "step": 1740
647
+ },
648
+ {
649
+ "epoch": 2.948073701842546,
650
+ "grad_norm": 3.270204782485962,
651
+ "learning_rate": 7.428571428571429e-06,
652
+ "loss": 0.1224,
653
+ "step": 1760
654
+ },
655
+ {
656
+ "epoch": 2.981574539363484,
657
+ "grad_norm": 4.261337757110596,
658
+ "learning_rate": 7.393298059964727e-06,
659
+ "loss": 0.1224,
660
+ "step": 1780
661
+ },
662
+ {
663
+ "epoch": 3.0150753768844223,
664
+ "grad_norm": 1.9531301259994507,
665
+ "learning_rate": 7.358024691358025e-06,
666
+ "loss": 0.0913,
667
+ "step": 1800
668
+ },
669
+ {
670
+ "epoch": 3.0485762144053603,
671
+ "grad_norm": 1.866215467453003,
672
+ "learning_rate": 7.322751322751324e-06,
673
+ "loss": 0.0466,
674
+ "step": 1820
675
+ },
676
+ {
677
+ "epoch": 3.0820770519262983,
678
+ "grad_norm": 3.9456610679626465,
679
+ "learning_rate": 7.287477954144622e-06,
680
+ "loss": 0.0451,
681
+ "step": 1840
682
+ },
683
+ {
684
+ "epoch": 3.1155778894472363,
685
+ "grad_norm": 1.9518849849700928,
686
+ "learning_rate": 7.25220458553792e-06,
687
+ "loss": 0.0414,
688
+ "step": 1860
689
+ },
690
+ {
691
+ "epoch": 3.1490787269681744,
692
+ "grad_norm": 2.232792854309082,
693
+ "learning_rate": 7.216931216931218e-06,
694
+ "loss": 0.0502,
695
+ "step": 1880
696
+ },
697
+ {
698
+ "epoch": 3.1825795644891124,
699
+ "grad_norm": 2.074127674102783,
700
+ "learning_rate": 7.181657848324516e-06,
701
+ "loss": 0.057,
702
+ "step": 1900
703
+ },
704
+ {
705
+ "epoch": 3.2160804020100504,
706
+ "grad_norm": 3.5036733150482178,
707
+ "learning_rate": 7.146384479717814e-06,
708
+ "loss": 0.0515,
709
+ "step": 1920
710
+ },
711
+ {
712
+ "epoch": 3.2495812395309884,
713
+ "grad_norm": 2.324014186859131,
714
+ "learning_rate": 7.111111111111112e-06,
715
+ "loss": 0.0454,
716
+ "step": 1940
717
+ },
718
+ {
719
+ "epoch": 3.2830820770519265,
720
+ "grad_norm": 2.710326671600342,
721
+ "learning_rate": 7.07583774250441e-06,
722
+ "loss": 0.0473,
723
+ "step": 1960
724
+ },
725
+ {
726
+ "epoch": 3.3165829145728645,
727
+ "grad_norm": 2.8943896293640137,
728
+ "learning_rate": 7.040564373897708e-06,
729
+ "loss": 0.046,
730
+ "step": 1980
731
+ },
732
+ {
733
+ "epoch": 3.3500837520938025,
734
+ "grad_norm": 2.319986581802368,
735
+ "learning_rate": 7.005291005291006e-06,
736
+ "loss": 0.0449,
737
+ "step": 2000
738
+ },
739
+ {
740
+ "epoch": 3.3500837520938025,
741
+ "eval_loss": 0.10934468358755112,
742
+ "eval_runtime": 1762.4506,
743
+ "eval_samples_per_second": 2.71,
744
+ "eval_steps_per_second": 0.339,
745
+ "eval_wer": 7.172854824854443,
746
+ "step": 2000
747
+ },
748
+ {
749
+ "epoch": 3.3835845896147405,
750
+ "grad_norm": 3.317129135131836,
751
+ "learning_rate": 6.9700176366843046e-06,
752
+ "loss": 0.0645,
753
+ "step": 2020
754
+ },
755
+ {
756
+ "epoch": 3.4170854271356785,
757
+ "grad_norm": 1.9533768892288208,
758
+ "learning_rate": 6.9347442680776025e-06,
759
+ "loss": 0.0432,
760
+ "step": 2040
761
+ },
762
+ {
763
+ "epoch": 3.4505862646566166,
764
+ "grad_norm": 2.0508453845977783,
765
+ "learning_rate": 6.8994708994709005e-06,
766
+ "loss": 0.0521,
767
+ "step": 2060
768
+ },
769
+ {
770
+ "epoch": 3.4840871021775546,
771
+ "grad_norm": 2.163236141204834,
772
+ "learning_rate": 6.8641975308641985e-06,
773
+ "loss": 0.0529,
774
+ "step": 2080
775
+ },
776
+ {
777
+ "epoch": 3.5175879396984926,
778
+ "grad_norm": 2.7154581546783447,
779
+ "learning_rate": 6.8289241622574965e-06,
780
+ "loss": 0.0452,
781
+ "step": 2100
782
+ },
783
+ {
784
+ "epoch": 3.5510887772194306,
785
+ "grad_norm": 3.0822432041168213,
786
+ "learning_rate": 6.7936507936507944e-06,
787
+ "loss": 0.0546,
788
+ "step": 2120
789
+ },
790
+ {
791
+ "epoch": 3.5845896147403686,
792
+ "grad_norm": 4.19010591506958,
793
+ "learning_rate": 6.758377425044092e-06,
794
+ "loss": 0.0529,
795
+ "step": 2140
796
+ },
797
+ {
798
+ "epoch": 3.6180904522613067,
799
+ "grad_norm": 2.9883594512939453,
800
+ "learning_rate": 6.72310405643739e-06,
801
+ "loss": 0.0503,
802
+ "step": 2160
803
+ },
804
+ {
805
+ "epoch": 3.6515912897822447,
806
+ "grad_norm": 2.3664371967315674,
807
+ "learning_rate": 6.687830687830688e-06,
808
+ "loss": 0.0498,
809
+ "step": 2180
810
+ },
811
+ {
812
+ "epoch": 3.6850921273031827,
813
+ "grad_norm": 2.0549991130828857,
814
+ "learning_rate": 6.652557319223986e-06,
815
+ "loss": 0.051,
816
+ "step": 2200
817
+ },
818
+ {
819
+ "epoch": 3.7185929648241207,
820
+ "grad_norm": 2.5339038372039795,
821
+ "learning_rate": 6.617283950617285e-06,
822
+ "loss": 0.0568,
823
+ "step": 2220
824
+ },
825
+ {
826
+ "epoch": 3.7520938023450587,
827
+ "grad_norm": 1.9988099336624146,
828
+ "learning_rate": 6.582010582010583e-06,
829
+ "loss": 0.051,
830
+ "step": 2240
831
+ },
832
+ {
833
+ "epoch": 3.7855946398659968,
834
+ "grad_norm": 2.5243782997131348,
835
+ "learning_rate": 6.546737213403881e-06,
836
+ "loss": 0.056,
837
+ "step": 2260
838
+ },
839
+ {
840
+ "epoch": 3.819095477386935,
841
+ "grad_norm": 3.157158136367798,
842
+ "learning_rate": 6.511463844797179e-06,
843
+ "loss": 0.0497,
844
+ "step": 2280
845
+ },
846
+ {
847
+ "epoch": 3.852596314907873,
848
+ "grad_norm": 1.9286202192306519,
849
+ "learning_rate": 6.476190476190477e-06,
850
+ "loss": 0.0426,
851
+ "step": 2300
852
+ },
853
+ {
854
+ "epoch": 3.886097152428811,
855
+ "grad_norm": 3.808802604675293,
856
+ "learning_rate": 6.440917107583775e-06,
857
+ "loss": 0.0499,
858
+ "step": 2320
859
+ },
860
+ {
861
+ "epoch": 3.919597989949749,
862
+ "grad_norm": 2.506671667098999,
863
+ "learning_rate": 6.405643738977073e-06,
864
+ "loss": 0.052,
865
+ "step": 2340
866
+ },
867
+ {
868
+ "epoch": 3.953098827470687,
869
+ "grad_norm": 2.9451920986175537,
870
+ "learning_rate": 6.370370370370371e-06,
871
+ "loss": 0.0552,
872
+ "step": 2360
873
+ },
874
+ {
875
+ "epoch": 3.986599664991625,
876
+ "grad_norm": 2.592744827270508,
877
+ "learning_rate": 6.335097001763669e-06,
878
+ "loss": 0.0527,
879
+ "step": 2380
880
+ },
881
+ {
882
+ "epoch": 4.0201005025125625,
883
+ "grad_norm": 1.8891575336456299,
884
+ "learning_rate": 6.299823633156967e-06,
885
+ "loss": 0.0289,
886
+ "step": 2400
887
+ },
888
+ {
889
+ "epoch": 4.0536013400335005,
890
+ "grad_norm": 1.8053243160247803,
891
+ "learning_rate": 6.264550264550266e-06,
892
+ "loss": 0.0192,
893
+ "step": 2420
894
+ },
895
+ {
896
+ "epoch": 4.0871021775544385,
897
+ "grad_norm": 2.0084407329559326,
898
+ "learning_rate": 6.229276895943564e-06,
899
+ "loss": 0.0242,
900
+ "step": 2440
901
+ },
902
+ {
903
+ "epoch": 4.1206030150753765,
904
+ "grad_norm": 1.5919119119644165,
905
+ "learning_rate": 6.194003527336862e-06,
906
+ "loss": 0.0211,
907
+ "step": 2460
908
+ },
909
+ {
910
+ "epoch": 4.1541038525963145,
911
+ "grad_norm": 1.9214613437652588,
912
+ "learning_rate": 6.15873015873016e-06,
913
+ "loss": 0.0233,
914
+ "step": 2480
915
+ },
916
+ {
917
+ "epoch": 4.187604690117253,
918
+ "grad_norm": 1.2652311325073242,
919
+ "learning_rate": 6.123456790123458e-06,
920
+ "loss": 0.0199,
921
+ "step": 2500
922
+ },
923
+ {
924
+ "epoch": 4.187604690117253,
925
+ "eval_loss": 0.0981329157948494,
926
+ "eval_runtime": 1779.0213,
927
+ "eval_samples_per_second": 2.685,
928
+ "eval_steps_per_second": 0.336,
929
+ "eval_wer": 6.707549871146321,
930
+ "step": 2500
931
+ }
932
+ ],
933
+ "logging_steps": 20,
934
+ "max_steps": 5970,
935
+ "num_input_tokens_seen": 0,
936
+ "num_train_epochs": 10,
937
+ "save_steps": 500,
938
+ "stateful_callbacks": {
939
+ "TrainerControl": {
940
+ "args": {
941
+ "should_epoch_stop": false,
942
+ "should_evaluate": false,
943
+ "should_log": false,
944
+ "should_save": true,
945
+ "should_training_stop": false
946
+ },
947
+ "attributes": {}
948
+ }
949
+ },
950
+ "total_flos": 2.308336900079616e+19,
951
+ "train_batch_size": 16,
952
+ "trial_name": null,
953
+ "trial_params": null
954
+ }
checkpoint-2500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41a3c645624f7463029286b7d5e1ad6924cfc44a0d153880431b77c40e5deb73
3
+ size 5432
checkpoint-3000/config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.0,
3
+ "activation_function": "gelu",
4
+ "apply_spec_augment": false,
5
+ "architectures": [
6
+ "WhisperForConditionalGeneration"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "begin_suppress_tokens": null,
10
+ "bos_token_id": 50257,
11
+ "classifier_proj_size": 256,
12
+ "d_model": 768,
13
+ "decoder_attention_heads": 12,
14
+ "decoder_ffn_dim": 3072,
15
+ "decoder_layerdrop": 0.0,
16
+ "decoder_layers": 12,
17
+ "decoder_start_token_id": 50258,
18
+ "dropout": 0.0,
19
+ "encoder_attention_heads": 12,
20
+ "encoder_ffn_dim": 3072,
21
+ "encoder_layerdrop": 0.0,
22
+ "encoder_layers": 12,
23
+ "eos_token_id": 50257,
24
+ "forced_decoder_ids": null,
25
+ "init_std": 0.02,
26
+ "is_encoder_decoder": true,
27
+ "mask_feature_length": 10,
28
+ "mask_feature_min_masks": 0,
29
+ "mask_feature_prob": 0.0,
30
+ "mask_time_length": 10,
31
+ "mask_time_min_masks": 2,
32
+ "mask_time_prob": 0.05,
33
+ "max_length": null,
34
+ "max_source_positions": 1500,
35
+ "max_target_positions": 448,
36
+ "median_filter_width": 7,
37
+ "model_type": "whisper",
38
+ "num_hidden_layers": 12,
39
+ "num_mel_bins": 80,
40
+ "pad_token_id": 50257,
41
+ "scale_embedding": false,
42
+ "torch_dtype": "float32",
43
+ "transformers_version": "4.51.3",
44
+ "use_cache": false,
45
+ "use_weighted_layer_sum": false,
46
+ "vocab_size": 51865
47
+ }
checkpoint-3000/generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 50257,
3
+ "decoder_start_token_id": 50258,
4
+ "eos_token_id": 50257,
5
+ "pad_token_id": 50257,
6
+ "suppress_tokens": [],
7
+ "transformers_version": "4.51.3",
8
+ "use_cache": false
9
+ }
checkpoint-3000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0abcf9469cdebe6dab6eca1ce89a6fe70fb85e3e8364817748f1b9707ee82afd
3
+ size 966995080
checkpoint-3000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:833b2c174e0af52976ea986fa4f02df78ec80fedb5801e9aaf6ac3a26d0689c5
3
+ size 1925064044
checkpoint-3000/preprocessor_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "chunk_length": 30,
3
+ "dither": 0.0,
4
+ "feature_extractor_type": "WhisperFeatureExtractor",
5
+ "feature_size": 80,
6
+ "hop_length": 160,
7
+ "n_fft": 400,
8
+ "n_samples": 480000,
9
+ "nb_max_frames": 3000,
10
+ "padding_side": "right",
11
+ "padding_value": 0.0,
12
+ "processor_class": "WhisperProcessor",
13
+ "return_attention_mask": false,
14
+ "sampling_rate": 16000
15
+ }
checkpoint-3000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27c93067937ab1be348de455c016b3ce31c80733910a246e9c9ccaeebb34c39a
3
+ size 14244
checkpoint-3000/scaler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67811e0c62d0a1a53aaafe03633e26905661076c1c0120aa29a847c54a088eed
3
+ size 988
checkpoint-3000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b8d74175fdadc4a37aa16ac28556cb53a4295de37309a86bc03e28b0ad0b32e
3
+ size 1064
checkpoint-3000/trainer_state.json ADDED
@@ -0,0 +1,1138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 3000,
3
+ "best_metric": 5.500143170754987,
4
+ "best_model_checkpoint": "./SALAMA_C3/checkpoint-3000",
5
+ "epoch": 5.025125628140704,
6
+ "eval_steps": 500,
7
+ "global_step": 3000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.03350083752093802,
14
+ "grad_norm": 11.757207870483398,
15
+ "learning_rate": 6.333333333333334e-07,
16
+ "loss": 0.5112,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.06700167504187604,
21
+ "grad_norm": 10.197953224182129,
22
+ "learning_rate": 1.3e-06,
23
+ "loss": 0.5283,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.10050251256281408,
28
+ "grad_norm": 12.811074256896973,
29
+ "learning_rate": 1.9666666666666668e-06,
30
+ "loss": 0.474,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.13400335008375208,
35
+ "grad_norm": 8.559342384338379,
36
+ "learning_rate": 2.6e-06,
37
+ "loss": 0.424,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.16750418760469013,
42
+ "grad_norm": 6.9629225730896,
43
+ "learning_rate": 3.266666666666667e-06,
44
+ "loss": 0.4465,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.20100502512562815,
49
+ "grad_norm": 7.833058834075928,
50
+ "learning_rate": 3.9333333333333335e-06,
51
+ "loss": 0.3764,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.23450586264656617,
56
+ "grad_norm": 6.882424831390381,
57
+ "learning_rate": 4.600000000000001e-06,
58
+ "loss": 0.3705,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.26800670016750416,
63
+ "grad_norm": 5.4183244705200195,
64
+ "learning_rate": 5.2666666666666665e-06,
65
+ "loss": 0.4116,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.3015075376884422,
70
+ "grad_norm": 6.103787899017334,
71
+ "learning_rate": 5.933333333333335e-06,
72
+ "loss": 0.3903,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.33500837520938026,
77
+ "grad_norm": 6.084160804748535,
78
+ "learning_rate": 6.600000000000001e-06,
79
+ "loss": 0.3794,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.3685092127303183,
84
+ "grad_norm": 4.531406879425049,
85
+ "learning_rate": 7.266666666666668e-06,
86
+ "loss": 0.3904,
87
+ "step": 220
88
+ },
89
+ {
90
+ "epoch": 0.4020100502512563,
91
+ "grad_norm": 6.702854156494141,
92
+ "learning_rate": 7.933333333333334e-06,
93
+ "loss": 0.3672,
94
+ "step": 240
95
+ },
96
+ {
97
+ "epoch": 0.4355108877721943,
98
+ "grad_norm": 6.386377811431885,
99
+ "learning_rate": 8.6e-06,
100
+ "loss": 0.3379,
101
+ "step": 260
102
+ },
103
+ {
104
+ "epoch": 0.46901172529313234,
105
+ "grad_norm": 5.33281135559082,
106
+ "learning_rate": 9.266666666666667e-06,
107
+ "loss": 0.4102,
108
+ "step": 280
109
+ },
110
+ {
111
+ "epoch": 0.5025125628140703,
112
+ "grad_norm": 6.031435012817383,
113
+ "learning_rate": 9.933333333333334e-06,
114
+ "loss": 0.3815,
115
+ "step": 300
116
+ },
117
+ {
118
+ "epoch": 0.5360134003350083,
119
+ "grad_norm": 4.8638176918029785,
120
+ "learning_rate": 9.968253968253969e-06,
121
+ "loss": 0.358,
122
+ "step": 320
123
+ },
124
+ {
125
+ "epoch": 0.5695142378559463,
126
+ "grad_norm": 5.352113723754883,
127
+ "learning_rate": 9.932980599647268e-06,
128
+ "loss": 0.3578,
129
+ "step": 340
130
+ },
131
+ {
132
+ "epoch": 0.6030150753768844,
133
+ "grad_norm": 5.560739040374756,
134
+ "learning_rate": 9.897707231040565e-06,
135
+ "loss": 0.3795,
136
+ "step": 360
137
+ },
138
+ {
139
+ "epoch": 0.6365159128978225,
140
+ "grad_norm": 6.186940670013428,
141
+ "learning_rate": 9.862433862433864e-06,
142
+ "loss": 0.3852,
143
+ "step": 380
144
+ },
145
+ {
146
+ "epoch": 0.6700167504187605,
147
+ "grad_norm": 5.37507438659668,
148
+ "learning_rate": 9.827160493827161e-06,
149
+ "loss": 0.3761,
150
+ "step": 400
151
+ },
152
+ {
153
+ "epoch": 0.7035175879396985,
154
+ "grad_norm": 6.373449802398682,
155
+ "learning_rate": 9.79188712522046e-06,
156
+ "loss": 0.3674,
157
+ "step": 420
158
+ },
159
+ {
160
+ "epoch": 0.7370184254606366,
161
+ "grad_norm": 5.702625274658203,
162
+ "learning_rate": 9.756613756613757e-06,
163
+ "loss": 0.4013,
164
+ "step": 440
165
+ },
166
+ {
167
+ "epoch": 0.7705192629815746,
168
+ "grad_norm": 5.5675153732299805,
169
+ "learning_rate": 9.721340388007056e-06,
170
+ "loss": 0.3495,
171
+ "step": 460
172
+ },
173
+ {
174
+ "epoch": 0.8040201005025126,
175
+ "grad_norm": 6.296374320983887,
176
+ "learning_rate": 9.686067019400353e-06,
177
+ "loss": 0.3704,
178
+ "step": 480
179
+ },
180
+ {
181
+ "epoch": 0.8375209380234506,
182
+ "grad_norm": 4.830463886260986,
183
+ "learning_rate": 9.650793650793652e-06,
184
+ "loss": 0.36,
185
+ "step": 500
186
+ },
187
+ {
188
+ "epoch": 0.8375209380234506,
189
+ "eval_loss": 0.27764827013015747,
190
+ "eval_runtime": 1752.3826,
191
+ "eval_samples_per_second": 2.725,
192
+ "eval_steps_per_second": 0.341,
193
+ "eval_wer": 19.59291781998664,
194
+ "step": 500
195
+ },
196
+ {
197
+ "epoch": 0.8710217755443886,
198
+ "grad_norm": 4.863058090209961,
199
+ "learning_rate": 9.61552028218695e-06,
200
+ "loss": 0.3874,
201
+ "step": 520
202
+ },
203
+ {
204
+ "epoch": 0.9045226130653267,
205
+ "grad_norm": 6.121027946472168,
206
+ "learning_rate": 9.580246913580248e-06,
207
+ "loss": 0.382,
208
+ "step": 540
209
+ },
210
+ {
211
+ "epoch": 0.9380234505862647,
212
+ "grad_norm": 4.538082599639893,
213
+ "learning_rate": 9.544973544973546e-06,
214
+ "loss": 0.3611,
215
+ "step": 560
216
+ },
217
+ {
218
+ "epoch": 0.9715242881072027,
219
+ "grad_norm": 4.8772382736206055,
220
+ "learning_rate": 9.509700176366844e-06,
221
+ "loss": 0.3584,
222
+ "step": 580
223
+ },
224
+ {
225
+ "epoch": 1.0050251256281406,
226
+ "grad_norm": 4.039212703704834,
227
+ "learning_rate": 9.474426807760142e-06,
228
+ "loss": 0.3701,
229
+ "step": 600
230
+ },
231
+ {
232
+ "epoch": 1.0385259631490786,
233
+ "grad_norm": 4.787687301635742,
234
+ "learning_rate": 9.43915343915344e-06,
235
+ "loss": 0.2058,
236
+ "step": 620
237
+ },
238
+ {
239
+ "epoch": 1.0720268006700167,
240
+ "grad_norm": 4.503021717071533,
241
+ "learning_rate": 9.403880070546738e-06,
242
+ "loss": 0.2455,
243
+ "step": 640
244
+ },
245
+ {
246
+ "epoch": 1.1055276381909547,
247
+ "grad_norm": 6.663857936859131,
248
+ "learning_rate": 9.368606701940036e-06,
249
+ "loss": 0.2179,
250
+ "step": 660
251
+ },
252
+ {
253
+ "epoch": 1.1390284757118927,
254
+ "grad_norm": 4.486196041107178,
255
+ "learning_rate": 9.333333333333334e-06,
256
+ "loss": 0.2196,
257
+ "step": 680
258
+ },
259
+ {
260
+ "epoch": 1.1725293132328307,
261
+ "grad_norm": 4.667060852050781,
262
+ "learning_rate": 9.298059964726633e-06,
263
+ "loss": 0.2183,
264
+ "step": 700
265
+ },
266
+ {
267
+ "epoch": 1.2060301507537687,
268
+ "grad_norm": 5.608316898345947,
269
+ "learning_rate": 9.26278659611993e-06,
270
+ "loss": 0.2161,
271
+ "step": 720
272
+ },
273
+ {
274
+ "epoch": 1.2395309882747068,
275
+ "grad_norm": 4.2184271812438965,
276
+ "learning_rate": 9.227513227513229e-06,
277
+ "loss": 0.2382,
278
+ "step": 740
279
+ },
280
+ {
281
+ "epoch": 1.2730318257956448,
282
+ "grad_norm": 3.871945381164551,
283
+ "learning_rate": 9.192239858906526e-06,
284
+ "loss": 0.2214,
285
+ "step": 760
286
+ },
287
+ {
288
+ "epoch": 1.3065326633165828,
289
+ "grad_norm": 3.730222225189209,
290
+ "learning_rate": 9.156966490299825e-06,
291
+ "loss": 0.2213,
292
+ "step": 780
293
+ },
294
+ {
295
+ "epoch": 1.3400335008375208,
296
+ "grad_norm": 4.740777015686035,
297
+ "learning_rate": 9.121693121693122e-06,
298
+ "loss": 0.2503,
299
+ "step": 800
300
+ },
301
+ {
302
+ "epoch": 1.3735343383584588,
303
+ "grad_norm": 4.123469352722168,
304
+ "learning_rate": 9.086419753086421e-06,
305
+ "loss": 0.2443,
306
+ "step": 820
307
+ },
308
+ {
309
+ "epoch": 1.4070351758793969,
310
+ "grad_norm": 3.988917827606201,
311
+ "learning_rate": 9.051146384479718e-06,
312
+ "loss": 0.241,
313
+ "step": 840
314
+ },
315
+ {
316
+ "epoch": 1.4405360134003349,
317
+ "grad_norm": 5.4730305671691895,
318
+ "learning_rate": 9.015873015873017e-06,
319
+ "loss": 0.2163,
320
+ "step": 860
321
+ },
322
+ {
323
+ "epoch": 1.474036850921273,
324
+ "grad_norm": 5.020652770996094,
325
+ "learning_rate": 8.980599647266314e-06,
326
+ "loss": 0.2387,
327
+ "step": 880
328
+ },
329
+ {
330
+ "epoch": 1.507537688442211,
331
+ "grad_norm": 4.231486797332764,
332
+ "learning_rate": 8.945326278659613e-06,
333
+ "loss": 0.2523,
334
+ "step": 900
335
+ },
336
+ {
337
+ "epoch": 1.541038525963149,
338
+ "grad_norm": 6.197975158691406,
339
+ "learning_rate": 8.910052910052912e-06,
340
+ "loss": 0.2255,
341
+ "step": 920
342
+ },
343
+ {
344
+ "epoch": 1.574539363484087,
345
+ "grad_norm": 5.489210605621338,
346
+ "learning_rate": 8.874779541446209e-06,
347
+ "loss": 0.2397,
348
+ "step": 940
349
+ },
350
+ {
351
+ "epoch": 1.608040201005025,
352
+ "grad_norm": 4.32737398147583,
353
+ "learning_rate": 8.839506172839508e-06,
354
+ "loss": 0.2298,
355
+ "step": 960
356
+ },
357
+ {
358
+ "epoch": 1.641541038525963,
359
+ "grad_norm": 4.504214763641357,
360
+ "learning_rate": 8.804232804232805e-06,
361
+ "loss": 0.2301,
362
+ "step": 980
363
+ },
364
+ {
365
+ "epoch": 1.675041876046901,
366
+ "grad_norm": 3.9694759845733643,
367
+ "learning_rate": 8.768959435626104e-06,
368
+ "loss": 0.2338,
369
+ "step": 1000
370
+ },
371
+ {
372
+ "epoch": 1.675041876046901,
373
+ "eval_loss": 0.18838582932949066,
374
+ "eval_runtime": 1770.3449,
375
+ "eval_samples_per_second": 2.698,
376
+ "eval_steps_per_second": 0.337,
377
+ "eval_wer": 14.071299035983584,
378
+ "step": 1000
379
+ },
380
+ {
381
+ "epoch": 1.708542713567839,
382
+ "grad_norm": 4.972934722900391,
383
+ "learning_rate": 8.7336860670194e-06,
384
+ "loss": 0.2388,
385
+ "step": 1020
386
+ },
387
+ {
388
+ "epoch": 1.742043551088777,
389
+ "grad_norm": 5.178994655609131,
390
+ "learning_rate": 8.6984126984127e-06,
391
+ "loss": 0.2531,
392
+ "step": 1040
393
+ },
394
+ {
395
+ "epoch": 1.775544388609715,
396
+ "grad_norm": 3.560372829437256,
397
+ "learning_rate": 8.663139329805997e-06,
398
+ "loss": 0.2187,
399
+ "step": 1060
400
+ },
401
+ {
402
+ "epoch": 1.809045226130653,
403
+ "grad_norm": 4.647324562072754,
404
+ "learning_rate": 8.627865961199296e-06,
405
+ "loss": 0.2413,
406
+ "step": 1080
407
+ },
408
+ {
409
+ "epoch": 1.8425460636515911,
410
+ "grad_norm": 3.6784422397613525,
411
+ "learning_rate": 8.592592592592593e-06,
412
+ "loss": 0.2242,
413
+ "step": 1100
414
+ },
415
+ {
416
+ "epoch": 1.8760469011725294,
417
+ "grad_norm": 5.928018093109131,
418
+ "learning_rate": 8.557319223985891e-06,
419
+ "loss": 0.2406,
420
+ "step": 1120
421
+ },
422
+ {
423
+ "epoch": 1.9095477386934674,
424
+ "grad_norm": 6.358487606048584,
425
+ "learning_rate": 8.52204585537919e-06,
426
+ "loss": 0.2482,
427
+ "step": 1140
428
+ },
429
+ {
430
+ "epoch": 1.9430485762144054,
431
+ "grad_norm": 4.409506797790527,
432
+ "learning_rate": 8.486772486772487e-06,
433
+ "loss": 0.2415,
434
+ "step": 1160
435
+ },
436
+ {
437
+ "epoch": 1.9765494137353434,
438
+ "grad_norm": 5.169639587402344,
439
+ "learning_rate": 8.451499118165786e-06,
440
+ "loss": 0.222,
441
+ "step": 1180
442
+ },
443
+ {
444
+ "epoch": 2.0100502512562812,
445
+ "grad_norm": 2.8841328620910645,
446
+ "learning_rate": 8.416225749559083e-06,
447
+ "loss": 0.2109,
448
+ "step": 1200
449
+ },
450
+ {
451
+ "epoch": 2.0435510887772192,
452
+ "grad_norm": 3.835498332977295,
453
+ "learning_rate": 8.380952380952382e-06,
454
+ "loss": 0.1284,
455
+ "step": 1220
456
+ },
457
+ {
458
+ "epoch": 2.0770519262981573,
459
+ "grad_norm": 2.584859848022461,
460
+ "learning_rate": 8.34567901234568e-06,
461
+ "loss": 0.1059,
462
+ "step": 1240
463
+ },
464
+ {
465
+ "epoch": 2.1105527638190953,
466
+ "grad_norm": 2.797682285308838,
467
+ "learning_rate": 8.310405643738978e-06,
468
+ "loss": 0.0967,
469
+ "step": 1260
470
+ },
471
+ {
472
+ "epoch": 2.1440536013400333,
473
+ "grad_norm": 3.2937309741973877,
474
+ "learning_rate": 8.275132275132275e-06,
475
+ "loss": 0.1267,
476
+ "step": 1280
477
+ },
478
+ {
479
+ "epoch": 2.1775544388609713,
480
+ "grad_norm": 3.344325065612793,
481
+ "learning_rate": 8.239858906525574e-06,
482
+ "loss": 0.1354,
483
+ "step": 1300
484
+ },
485
+ {
486
+ "epoch": 2.2110552763819094,
487
+ "grad_norm": 2.955522060394287,
488
+ "learning_rate": 8.204585537918873e-06,
489
+ "loss": 0.104,
490
+ "step": 1320
491
+ },
492
+ {
493
+ "epoch": 2.2445561139028474,
494
+ "grad_norm": 5.180908203125,
495
+ "learning_rate": 8.16931216931217e-06,
496
+ "loss": 0.1229,
497
+ "step": 1340
498
+ },
499
+ {
500
+ "epoch": 2.2780569514237854,
501
+ "grad_norm": 3.639417886734009,
502
+ "learning_rate": 8.134038800705469e-06,
503
+ "loss": 0.104,
504
+ "step": 1360
505
+ },
506
+ {
507
+ "epoch": 2.3115577889447234,
508
+ "grad_norm": 3.7380902767181396,
509
+ "learning_rate": 8.098765432098766e-06,
510
+ "loss": 0.1141,
511
+ "step": 1380
512
+ },
513
+ {
514
+ "epoch": 2.3450586264656614,
515
+ "grad_norm": 4.100318908691406,
516
+ "learning_rate": 8.063492063492065e-06,
517
+ "loss": 0.1123,
518
+ "step": 1400
519
+ },
520
+ {
521
+ "epoch": 2.3785594639865995,
522
+ "grad_norm": 2.8147048950195312,
523
+ "learning_rate": 8.028218694885362e-06,
524
+ "loss": 0.1039,
525
+ "step": 1420
526
+ },
527
+ {
528
+ "epoch": 2.4120603015075375,
529
+ "grad_norm": 3.7376410961151123,
530
+ "learning_rate": 7.99294532627866e-06,
531
+ "loss": 0.1179,
532
+ "step": 1440
533
+ },
534
+ {
535
+ "epoch": 2.4455611390284755,
536
+ "grad_norm": 4.320065975189209,
537
+ "learning_rate": 7.957671957671958e-06,
538
+ "loss": 0.1167,
539
+ "step": 1460
540
+ },
541
+ {
542
+ "epoch": 2.4790619765494135,
543
+ "grad_norm": 4.415127277374268,
544
+ "learning_rate": 7.922398589065257e-06,
545
+ "loss": 0.125,
546
+ "step": 1480
547
+ },
548
+ {
549
+ "epoch": 2.5125628140703515,
550
+ "grad_norm": 2.631763219833374,
551
+ "learning_rate": 7.887125220458554e-06,
552
+ "loss": 0.1188,
553
+ "step": 1500
554
+ },
555
+ {
556
+ "epoch": 2.5125628140703515,
557
+ "eval_loss": 0.13635103404521942,
558
+ "eval_runtime": 1806.1099,
559
+ "eval_samples_per_second": 2.644,
560
+ "eval_steps_per_second": 0.331,
561
+ "eval_wer": 11.713753937195762,
562
+ "step": 1500
563
+ },
564
+ {
565
+ "epoch": 2.5460636515912896,
566
+ "grad_norm": 3.608773946762085,
567
+ "learning_rate": 7.851851851851853e-06,
568
+ "loss": 0.1222,
569
+ "step": 1520
570
+ },
571
+ {
572
+ "epoch": 2.5795644891122276,
573
+ "grad_norm": 3.4494569301605225,
574
+ "learning_rate": 7.816578483245151e-06,
575
+ "loss": 0.1087,
576
+ "step": 1540
577
+ },
578
+ {
579
+ "epoch": 2.6130653266331656,
580
+ "grad_norm": 3.6112048625946045,
581
+ "learning_rate": 7.781305114638449e-06,
582
+ "loss": 0.1097,
583
+ "step": 1560
584
+ },
585
+ {
586
+ "epoch": 2.6465661641541036,
587
+ "grad_norm": 3.1978707313537598,
588
+ "learning_rate": 7.746031746031747e-06,
589
+ "loss": 0.1101,
590
+ "step": 1580
591
+ },
592
+ {
593
+ "epoch": 2.6800670016750416,
594
+ "grad_norm": 3.6376733779907227,
595
+ "learning_rate": 7.710758377425045e-06,
596
+ "loss": 0.1056,
597
+ "step": 1600
598
+ },
599
+ {
600
+ "epoch": 2.7135678391959797,
601
+ "grad_norm": 3.754915237426758,
602
+ "learning_rate": 7.675485008818343e-06,
603
+ "loss": 0.1103,
604
+ "step": 1620
605
+ },
606
+ {
607
+ "epoch": 2.7470686767169177,
608
+ "grad_norm": 3.329240322113037,
609
+ "learning_rate": 7.64021164021164e-06,
610
+ "loss": 0.1193,
611
+ "step": 1640
612
+ },
613
+ {
614
+ "epoch": 2.7805695142378557,
615
+ "grad_norm": 3.9300169944763184,
616
+ "learning_rate": 7.604938271604939e-06,
617
+ "loss": 0.1232,
618
+ "step": 1660
619
+ },
620
+ {
621
+ "epoch": 2.8140703517587937,
622
+ "grad_norm": 4.290626049041748,
623
+ "learning_rate": 7.569664902998237e-06,
624
+ "loss": 0.1374,
625
+ "step": 1680
626
+ },
627
+ {
628
+ "epoch": 2.8475711892797317,
629
+ "grad_norm": 3.8987998962402344,
630
+ "learning_rate": 7.534391534391535e-06,
631
+ "loss": 0.1062,
632
+ "step": 1700
633
+ },
634
+ {
635
+ "epoch": 2.8810720268006698,
636
+ "grad_norm": 4.234960079193115,
637
+ "learning_rate": 7.499118165784833e-06,
638
+ "loss": 0.111,
639
+ "step": 1720
640
+ },
641
+ {
642
+ "epoch": 2.914572864321608,
643
+ "grad_norm": 3.103458881378174,
644
+ "learning_rate": 7.463844797178131e-06,
645
+ "loss": 0.1076,
646
+ "step": 1740
647
+ },
648
+ {
649
+ "epoch": 2.948073701842546,
650
+ "grad_norm": 3.270204782485962,
651
+ "learning_rate": 7.428571428571429e-06,
652
+ "loss": 0.1224,
653
+ "step": 1760
654
+ },
655
+ {
656
+ "epoch": 2.981574539363484,
657
+ "grad_norm": 4.261337757110596,
658
+ "learning_rate": 7.393298059964727e-06,
659
+ "loss": 0.1224,
660
+ "step": 1780
661
+ },
662
+ {
663
+ "epoch": 3.0150753768844223,
664
+ "grad_norm": 1.9531301259994507,
665
+ "learning_rate": 7.358024691358025e-06,
666
+ "loss": 0.0913,
667
+ "step": 1800
668
+ },
669
+ {
670
+ "epoch": 3.0485762144053603,
671
+ "grad_norm": 1.866215467453003,
672
+ "learning_rate": 7.322751322751324e-06,
673
+ "loss": 0.0466,
674
+ "step": 1820
675
+ },
676
+ {
677
+ "epoch": 3.0820770519262983,
678
+ "grad_norm": 3.9456610679626465,
679
+ "learning_rate": 7.287477954144622e-06,
680
+ "loss": 0.0451,
681
+ "step": 1840
682
+ },
683
+ {
684
+ "epoch": 3.1155778894472363,
685
+ "grad_norm": 1.9518849849700928,
686
+ "learning_rate": 7.25220458553792e-06,
687
+ "loss": 0.0414,
688
+ "step": 1860
689
+ },
690
+ {
691
+ "epoch": 3.1490787269681744,
692
+ "grad_norm": 2.232792854309082,
693
+ "learning_rate": 7.216931216931218e-06,
694
+ "loss": 0.0502,
695
+ "step": 1880
696
+ },
697
+ {
698
+ "epoch": 3.1825795644891124,
699
+ "grad_norm": 2.074127674102783,
700
+ "learning_rate": 7.181657848324516e-06,
701
+ "loss": 0.057,
702
+ "step": 1900
703
+ },
704
+ {
705
+ "epoch": 3.2160804020100504,
706
+ "grad_norm": 3.5036733150482178,
707
+ "learning_rate": 7.146384479717814e-06,
708
+ "loss": 0.0515,
709
+ "step": 1920
710
+ },
711
+ {
712
+ "epoch": 3.2495812395309884,
713
+ "grad_norm": 2.324014186859131,
714
+ "learning_rate": 7.111111111111112e-06,
715
+ "loss": 0.0454,
716
+ "step": 1940
717
+ },
718
+ {
719
+ "epoch": 3.2830820770519265,
720
+ "grad_norm": 2.710326671600342,
721
+ "learning_rate": 7.07583774250441e-06,
722
+ "loss": 0.0473,
723
+ "step": 1960
724
+ },
725
+ {
726
+ "epoch": 3.3165829145728645,
727
+ "grad_norm": 2.8943896293640137,
728
+ "learning_rate": 7.040564373897708e-06,
729
+ "loss": 0.046,
730
+ "step": 1980
731
+ },
732
+ {
733
+ "epoch": 3.3500837520938025,
734
+ "grad_norm": 2.319986581802368,
735
+ "learning_rate": 7.005291005291006e-06,
736
+ "loss": 0.0449,
737
+ "step": 2000
738
+ },
739
+ {
740
+ "epoch": 3.3500837520938025,
741
+ "eval_loss": 0.10934468358755112,
742
+ "eval_runtime": 1762.4506,
743
+ "eval_samples_per_second": 2.71,
744
+ "eval_steps_per_second": 0.339,
745
+ "eval_wer": 7.172854824854443,
746
+ "step": 2000
747
+ },
748
+ {
749
+ "epoch": 3.3835845896147405,
750
+ "grad_norm": 3.317129135131836,
751
+ "learning_rate": 6.9700176366843046e-06,
752
+ "loss": 0.0645,
753
+ "step": 2020
754
+ },
755
+ {
756
+ "epoch": 3.4170854271356785,
757
+ "grad_norm": 1.9533768892288208,
758
+ "learning_rate": 6.9347442680776025e-06,
759
+ "loss": 0.0432,
760
+ "step": 2040
761
+ },
762
+ {
763
+ "epoch": 3.4505862646566166,
764
+ "grad_norm": 2.0508453845977783,
765
+ "learning_rate": 6.8994708994709005e-06,
766
+ "loss": 0.0521,
767
+ "step": 2060
768
+ },
769
+ {
770
+ "epoch": 3.4840871021775546,
771
+ "grad_norm": 2.163236141204834,
772
+ "learning_rate": 6.8641975308641985e-06,
773
+ "loss": 0.0529,
774
+ "step": 2080
775
+ },
776
+ {
777
+ "epoch": 3.5175879396984926,
778
+ "grad_norm": 2.7154581546783447,
779
+ "learning_rate": 6.8289241622574965e-06,
780
+ "loss": 0.0452,
781
+ "step": 2100
782
+ },
783
+ {
784
+ "epoch": 3.5510887772194306,
785
+ "grad_norm": 3.0822432041168213,
786
+ "learning_rate": 6.7936507936507944e-06,
787
+ "loss": 0.0546,
788
+ "step": 2120
789
+ },
790
+ {
791
+ "epoch": 3.5845896147403686,
792
+ "grad_norm": 4.19010591506958,
793
+ "learning_rate": 6.758377425044092e-06,
794
+ "loss": 0.0529,
795
+ "step": 2140
796
+ },
797
+ {
798
+ "epoch": 3.6180904522613067,
799
+ "grad_norm": 2.9883594512939453,
800
+ "learning_rate": 6.72310405643739e-06,
801
+ "loss": 0.0503,
802
+ "step": 2160
803
+ },
804
+ {
805
+ "epoch": 3.6515912897822447,
806
+ "grad_norm": 2.3664371967315674,
807
+ "learning_rate": 6.687830687830688e-06,
808
+ "loss": 0.0498,
809
+ "step": 2180
810
+ },
811
+ {
812
+ "epoch": 3.6850921273031827,
813
+ "grad_norm": 2.0549991130828857,
814
+ "learning_rate": 6.652557319223986e-06,
815
+ "loss": 0.051,
816
+ "step": 2200
817
+ },
818
+ {
819
+ "epoch": 3.7185929648241207,
820
+ "grad_norm": 2.5339038372039795,
821
+ "learning_rate": 6.617283950617285e-06,
822
+ "loss": 0.0568,
823
+ "step": 2220
824
+ },
825
+ {
826
+ "epoch": 3.7520938023450587,
827
+ "grad_norm": 1.9988099336624146,
828
+ "learning_rate": 6.582010582010583e-06,
829
+ "loss": 0.051,
830
+ "step": 2240
831
+ },
832
+ {
833
+ "epoch": 3.7855946398659968,
834
+ "grad_norm": 2.5243782997131348,
835
+ "learning_rate": 6.546737213403881e-06,
836
+ "loss": 0.056,
837
+ "step": 2260
838
+ },
839
+ {
840
+ "epoch": 3.819095477386935,
841
+ "grad_norm": 3.157158136367798,
842
+ "learning_rate": 6.511463844797179e-06,
843
+ "loss": 0.0497,
844
+ "step": 2280
845
+ },
846
+ {
847
+ "epoch": 3.852596314907873,
848
+ "grad_norm": 1.9286202192306519,
849
+ "learning_rate": 6.476190476190477e-06,
850
+ "loss": 0.0426,
851
+ "step": 2300
852
+ },
853
+ {
854
+ "epoch": 3.886097152428811,
855
+ "grad_norm": 3.808802604675293,
856
+ "learning_rate": 6.440917107583775e-06,
857
+ "loss": 0.0499,
858
+ "step": 2320
859
+ },
860
+ {
861
+ "epoch": 3.919597989949749,
862
+ "grad_norm": 2.506671667098999,
863
+ "learning_rate": 6.405643738977073e-06,
864
+ "loss": 0.052,
865
+ "step": 2340
866
+ },
867
+ {
868
+ "epoch": 3.953098827470687,
869
+ "grad_norm": 2.9451920986175537,
870
+ "learning_rate": 6.370370370370371e-06,
871
+ "loss": 0.0552,
872
+ "step": 2360
873
+ },
874
+ {
875
+ "epoch": 3.986599664991625,
876
+ "grad_norm": 2.592744827270508,
877
+ "learning_rate": 6.335097001763669e-06,
878
+ "loss": 0.0527,
879
+ "step": 2380
880
+ },
881
+ {
882
+ "epoch": 4.0201005025125625,
883
+ "grad_norm": 1.8891575336456299,
884
+ "learning_rate": 6.299823633156967e-06,
885
+ "loss": 0.0289,
886
+ "step": 2400
887
+ },
888
+ {
889
+ "epoch": 4.0536013400335005,
890
+ "grad_norm": 1.8053243160247803,
891
+ "learning_rate": 6.264550264550266e-06,
892
+ "loss": 0.0192,
893
+ "step": 2420
894
+ },
895
+ {
896
+ "epoch": 4.0871021775544385,
897
+ "grad_norm": 2.0084407329559326,
898
+ "learning_rate": 6.229276895943564e-06,
899
+ "loss": 0.0242,
900
+ "step": 2440
901
+ },
902
+ {
903
+ "epoch": 4.1206030150753765,
904
+ "grad_norm": 1.5919119119644165,
905
+ "learning_rate": 6.194003527336862e-06,
906
+ "loss": 0.0211,
907
+ "step": 2460
908
+ },
909
+ {
910
+ "epoch": 4.1541038525963145,
911
+ "grad_norm": 1.9214613437652588,
912
+ "learning_rate": 6.15873015873016e-06,
913
+ "loss": 0.0233,
914
+ "step": 2480
915
+ },
916
+ {
917
+ "epoch": 4.187604690117253,
918
+ "grad_norm": 1.2652311325073242,
919
+ "learning_rate": 6.123456790123458e-06,
920
+ "loss": 0.0199,
921
+ "step": 2500
922
+ },
923
+ {
924
+ "epoch": 4.187604690117253,
925
+ "eval_loss": 0.0981329157948494,
926
+ "eval_runtime": 1779.0213,
927
+ "eval_samples_per_second": 2.685,
928
+ "eval_steps_per_second": 0.336,
929
+ "eval_wer": 6.707549871146321,
930
+ "step": 2500
931
+ },
932
+ {
933
+ "epoch": 4.221105527638191,
934
+ "grad_norm": 2.026528835296631,
935
+ "learning_rate": 6.088183421516756e-06,
936
+ "loss": 0.0217,
937
+ "step": 2520
938
+ },
939
+ {
940
+ "epoch": 4.254606365159129,
941
+ "grad_norm": 1.596919059753418,
942
+ "learning_rate": 6.052910052910054e-06,
943
+ "loss": 0.0167,
944
+ "step": 2540
945
+ },
946
+ {
947
+ "epoch": 4.288107202680067,
948
+ "grad_norm": 2.9445090293884277,
949
+ "learning_rate": 6.017636684303352e-06,
950
+ "loss": 0.0225,
951
+ "step": 2560
952
+ },
953
+ {
954
+ "epoch": 4.321608040201005,
955
+ "grad_norm": 2.4160282611846924,
956
+ "learning_rate": 5.9823633156966496e-06,
957
+ "loss": 0.0253,
958
+ "step": 2580
959
+ },
960
+ {
961
+ "epoch": 4.355108877721943,
962
+ "grad_norm": 1.461127758026123,
963
+ "learning_rate": 5.9470899470899475e-06,
964
+ "loss": 0.0197,
965
+ "step": 2600
966
+ },
967
+ {
968
+ "epoch": 4.388609715242881,
969
+ "grad_norm": 2.7892863750457764,
970
+ "learning_rate": 5.911816578483246e-06,
971
+ "loss": 0.022,
972
+ "step": 2620
973
+ },
974
+ {
975
+ "epoch": 4.422110552763819,
976
+ "grad_norm": 1.651208758354187,
977
+ "learning_rate": 5.876543209876544e-06,
978
+ "loss": 0.0215,
979
+ "step": 2640
980
+ },
981
+ {
982
+ "epoch": 4.455611390284757,
983
+ "grad_norm": 2.2500391006469727,
984
+ "learning_rate": 5.841269841269842e-06,
985
+ "loss": 0.0247,
986
+ "step": 2660
987
+ },
988
+ {
989
+ "epoch": 4.489112227805695,
990
+ "grad_norm": 4.447635173797607,
991
+ "learning_rate": 5.80599647266314e-06,
992
+ "loss": 0.0263,
993
+ "step": 2680
994
+ },
995
+ {
996
+ "epoch": 4.522613065326633,
997
+ "grad_norm": 0.8300407528877258,
998
+ "learning_rate": 5.770723104056438e-06,
999
+ "loss": 0.0209,
1000
+ "step": 2700
1001
+ },
1002
+ {
1003
+ "epoch": 4.556113902847571,
1004
+ "grad_norm": 1.6874111890792847,
1005
+ "learning_rate": 5.735449735449736e-06,
1006
+ "loss": 0.0195,
1007
+ "step": 2720
1008
+ },
1009
+ {
1010
+ "epoch": 4.589614740368509,
1011
+ "grad_norm": 2.4045815467834473,
1012
+ "learning_rate": 5.700176366843034e-06,
1013
+ "loss": 0.0224,
1014
+ "step": 2740
1015
+ },
1016
+ {
1017
+ "epoch": 4.623115577889447,
1018
+ "grad_norm": 2.3160908222198486,
1019
+ "learning_rate": 5.664902998236332e-06,
1020
+ "loss": 0.0179,
1021
+ "step": 2760
1022
+ },
1023
+ {
1024
+ "epoch": 4.656616415410385,
1025
+ "grad_norm": 1.6684287786483765,
1026
+ "learning_rate": 5.62962962962963e-06,
1027
+ "loss": 0.0238,
1028
+ "step": 2780
1029
+ },
1030
+ {
1031
+ "epoch": 4.690117252931323,
1032
+ "grad_norm": 1.973906397819519,
1033
+ "learning_rate": 5.594356261022928e-06,
1034
+ "loss": 0.0226,
1035
+ "step": 2800
1036
+ },
1037
+ {
1038
+ "epoch": 4.723618090452261,
1039
+ "grad_norm": 2.270906686782837,
1040
+ "learning_rate": 5.559082892416227e-06,
1041
+ "loss": 0.0232,
1042
+ "step": 2820
1043
+ },
1044
+ {
1045
+ "epoch": 4.757118927973199,
1046
+ "grad_norm": 1.8875011205673218,
1047
+ "learning_rate": 5.523809523809525e-06,
1048
+ "loss": 0.0218,
1049
+ "step": 2840
1050
+ },
1051
+ {
1052
+ "epoch": 4.790619765494137,
1053
+ "grad_norm": 1.1312583684921265,
1054
+ "learning_rate": 5.488536155202823e-06,
1055
+ "loss": 0.0212,
1056
+ "step": 2860
1057
+ },
1058
+ {
1059
+ "epoch": 4.824120603015075,
1060
+ "grad_norm": 0.864783525466919,
1061
+ "learning_rate": 5.453262786596121e-06,
1062
+ "loss": 0.023,
1063
+ "step": 2880
1064
+ },
1065
+ {
1066
+ "epoch": 4.857621440536013,
1067
+ "grad_norm": 1.2935965061187744,
1068
+ "learning_rate": 5.417989417989419e-06,
1069
+ "loss": 0.019,
1070
+ "step": 2900
1071
+ },
1072
+ {
1073
+ "epoch": 4.891122278056951,
1074
+ "grad_norm": 2.4576382637023926,
1075
+ "learning_rate": 5.382716049382717e-06,
1076
+ "loss": 0.0193,
1077
+ "step": 2920
1078
+ },
1079
+ {
1080
+ "epoch": 4.924623115577889,
1081
+ "grad_norm": 2.71472430229187,
1082
+ "learning_rate": 5.347442680776015e-06,
1083
+ "loss": 0.0253,
1084
+ "step": 2940
1085
+ },
1086
+ {
1087
+ "epoch": 4.958123953098827,
1088
+ "grad_norm": 2.84940505027771,
1089
+ "learning_rate": 5.312169312169313e-06,
1090
+ "loss": 0.0218,
1091
+ "step": 2960
1092
+ },
1093
+ {
1094
+ "epoch": 4.991624790619765,
1095
+ "grad_norm": 1.8483999967575073,
1096
+ "learning_rate": 5.276895943562611e-06,
1097
+ "loss": 0.0226,
1098
+ "step": 2980
1099
+ },
1100
+ {
1101
+ "epoch": 5.025125628140704,
1102
+ "grad_norm": 0.6126876473426819,
1103
+ "learning_rate": 5.241622574955909e-06,
1104
+ "loss": 0.0101,
1105
+ "step": 3000
1106
+ },
1107
+ {
1108
+ "epoch": 5.025125628140704,
1109
+ "eval_loss": 0.09390027821063995,
1110
+ "eval_runtime": 1767.4512,
1111
+ "eval_samples_per_second": 2.702,
1112
+ "eval_steps_per_second": 0.338,
1113
+ "eval_wer": 5.500143170754987,
1114
+ "step": 3000
1115
+ }
1116
+ ],
1117
+ "logging_steps": 20,
1118
+ "max_steps": 5970,
1119
+ "num_input_tokens_seen": 0,
1120
+ "num_train_epochs": 10,
1121
+ "save_steps": 500,
1122
+ "stateful_callbacks": {
1123
+ "TrainerControl": {
1124
+ "args": {
1125
+ "should_epoch_stop": false,
1126
+ "should_evaluate": false,
1127
+ "should_log": false,
1128
+ "should_save": true,
1129
+ "should_training_stop": false
1130
+ },
1131
+ "attributes": {}
1132
+ }
1133
+ },
1134
+ "total_flos": 2.76998696497152e+19,
1135
+ "train_batch_size": 16,
1136
+ "trial_name": null,
1137
+ "trial_params": null
1138
+ }
checkpoint-3000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41a3c645624f7463029286b7d5e1ad6924cfc44a0d153880431b77c40e5deb73
3
+ size 5432