lucio commited on
Commit
20d6d7c
1 Parent(s): 9636944

Training in progress, step 500

Browse files
.ipynb_checkpoints/config-checkpoint.json ADDED
@@ -0,0 +1,107 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-xls-r-300m",
3
+ "activation_dropout": 0.1,
4
+ "adapter_kernel_size": 3,
5
+ "adapter_stride": 2,
6
+ "add_adapter": false,
7
+ "apply_spec_augment": true,
8
+ "architectures": [
9
+ "Wav2Vec2ForCTC"
10
+ ],
11
+ "attention_dropout": 0.0,
12
+ "bos_token_id": 1,
13
+ "classifier_proj_size": 256,
14
+ "codevector_dim": 768,
15
+ "contrastive_logits_temperature": 0.1,
16
+ "conv_bias": true,
17
+ "conv_dim": [
18
+ 512,
19
+ 512,
20
+ 512,
21
+ 512,
22
+ 512,
23
+ 512,
24
+ 512
25
+ ],
26
+ "conv_kernel": [
27
+ 10,
28
+ 3,
29
+ 3,
30
+ 3,
31
+ 3,
32
+ 2,
33
+ 2
34
+ ],
35
+ "conv_stride": [
36
+ 5,
37
+ 2,
38
+ 2,
39
+ 2,
40
+ 2,
41
+ 2,
42
+ 2
43
+ ],
44
+ "ctc_loss_reduction": "mean",
45
+ "ctc_zero_infinity": false,
46
+ "diversity_loss_weight": 0.1,
47
+ "do_stable_layer_norm": true,
48
+ "eos_token_id": 2,
49
+ "feat_extract_activation": "gelu",
50
+ "feat_extract_dropout": 0.0,
51
+ "feat_extract_norm": "layer",
52
+ "feat_proj_dropout": 0.0,
53
+ "feat_quantizer_dropout": 0.0,
54
+ "final_dropout": 0.0,
55
+ "hidden_act": "gelu",
56
+ "hidden_dropout": 0.0,
57
+ "hidden_size": 1024,
58
+ "initializer_range": 0.02,
59
+ "intermediate_size": 4096,
60
+ "layer_norm_eps": 1e-05,
61
+ "layerdrop": 0.0,
62
+ "mask_feature_length": 64,
63
+ "mask_feature_min_masks": 0,
64
+ "mask_feature_prob": 0.25,
65
+ "mask_time_length": 10,
66
+ "mask_time_min_masks": 2,
67
+ "mask_time_prob": 0.75,
68
+ "model_type": "wav2vec2",
69
+ "num_adapter_layers": 3,
70
+ "num_attention_heads": 16,
71
+ "num_codevector_groups": 2,
72
+ "num_codevectors_per_group": 320,
73
+ "num_conv_pos_embedding_groups": 16,
74
+ "num_conv_pos_embeddings": 128,
75
+ "num_feat_extract_layers": 7,
76
+ "num_hidden_layers": 24,
77
+ "num_negatives": 100,
78
+ "output_hidden_size": 1024,
79
+ "pad_token_id": 43,
80
+ "proj_codevector_dim": 768,
81
+ "tdnn_dilation": [
82
+ 1,
83
+ 2,
84
+ 3,
85
+ 1,
86
+ 1
87
+ ],
88
+ "tdnn_dim": [
89
+ 512,
90
+ 512,
91
+ 512,
92
+ 512,
93
+ 1500
94
+ ],
95
+ "tdnn_kernel": [
96
+ 5,
97
+ 3,
98
+ 3,
99
+ 1,
100
+ 1
101
+ ],
102
+ "torch_dtype": "float32",
103
+ "transformers_version": "4.16.0.dev0",
104
+ "use_weighted_layer_sum": false,
105
+ "vocab_size": 46,
106
+ "xvector_output_dim": 512
107
+ }
.ipynb_checkpoints/eval_results-checkpoint.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 50.0,
3
+ "eval_loss": 0.226610004901886,
4
+ "eval_runtime": 133.034,
5
+ "eval_samples": 2620,
6
+ "eval_samples_per_second": 19.694,
7
+ "eval_steps_per_second": 2.466,
8
+ "eval_wer": 0.3654982085794519
9
+ }
.ipynb_checkpoints/preprocessor_config-checkpoint.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_normalize": true,
3
+ "feature_extractor_type": "Wav2Vec2FeatureExtractor",
4
+ "feature_size": 1,
5
+ "padding_side": "right",
6
+ "padding_value": 0,
7
+ "return_attention_mask": true,
8
+ "sampling_rate": 16000
9
+ }
.ipynb_checkpoints/tokenizer_config-checkpoint.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"unk_token": "[UNK]", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "[PAD]", "do_lower_case": false, "word_delimiter_token": "|", "special_tokens_map_file": null, "tokenizer_file": null, "name_or_path": "./xls-r-uyghur-cv8", "tokenizer_class": "Wav2Vec2CTCTokenizer"}
.ipynb_checkpoints/trainer_state-checkpoint.json ADDED
@@ -0,0 +1,733 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 50.0,
5
+ "global_step": 9150,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.55,
12
+ "learning_rate": 3.7125e-06,
13
+ "loss": 14.9109,
14
+ "step": 100
15
+ },
16
+ {
17
+ "epoch": 1.09,
18
+ "learning_rate": 7.4625e-06,
19
+ "loss": 7.5771,
20
+ "step": 200
21
+ },
22
+ {
23
+ "epoch": 1.64,
24
+ "learning_rate": 1.1212499999999998e-05,
25
+ "loss": 4.939,
26
+ "step": 300
27
+ },
28
+ {
29
+ "epoch": 2.19,
30
+ "learning_rate": 1.49625e-05,
31
+ "loss": 4.1873,
32
+ "step": 400
33
+ },
34
+ {
35
+ "epoch": 2.73,
36
+ "learning_rate": 1.8712499999999997e-05,
37
+ "loss": 3.6863,
38
+ "step": 500
39
+ },
40
+ {
41
+ "epoch": 2.73,
42
+ "eval_loss": 3.5362327098846436,
43
+ "eval_runtime": 132.4771,
44
+ "eval_samples_per_second": 19.777,
45
+ "eval_steps_per_second": 2.476,
46
+ "eval_wer": 1.0,
47
+ "step": 500
48
+ },
49
+ {
50
+ "epoch": 3.28,
51
+ "learning_rate": 2.2462499999999997e-05,
52
+ "loss": 3.3718,
53
+ "step": 600
54
+ },
55
+ {
56
+ "epoch": 3.83,
57
+ "learning_rate": 2.6212499999999997e-05,
58
+ "loss": 3.2608,
59
+ "step": 700
60
+ },
61
+ {
62
+ "epoch": 4.37,
63
+ "learning_rate": 2.99625e-05,
64
+ "loss": 3.2251,
65
+ "step": 800
66
+ },
67
+ {
68
+ "epoch": 4.92,
69
+ "learning_rate": 3.37125e-05,
70
+ "loss": 3.1741,
71
+ "step": 900
72
+ },
73
+ {
74
+ "epoch": 5.46,
75
+ "learning_rate": 3.7462499999999996e-05,
76
+ "loss": 3.1409,
77
+ "step": 1000
78
+ },
79
+ {
80
+ "epoch": 5.46,
81
+ "eval_loss": 3.132805347442627,
82
+ "eval_runtime": 132.2993,
83
+ "eval_samples_per_second": 19.804,
84
+ "eval_steps_per_second": 2.479,
85
+ "eval_wer": 1.0,
86
+ "step": 1000
87
+ },
88
+ {
89
+ "epoch": 6.01,
90
+ "learning_rate": 4.12125e-05,
91
+ "loss": 3.1047,
92
+ "step": 1100
93
+ },
94
+ {
95
+ "epoch": 6.56,
96
+ "learning_rate": 4.4962499999999995e-05,
97
+ "loss": 2.8994,
98
+ "step": 1200
99
+ },
100
+ {
101
+ "epoch": 7.1,
102
+ "learning_rate": 4.871249999999999e-05,
103
+ "loss": 2.5085,
104
+ "step": 1300
105
+ },
106
+ {
107
+ "epoch": 7.65,
108
+ "learning_rate": 5.2462499999999994e-05,
109
+ "loss": 2.1499,
110
+ "step": 1400
111
+ },
112
+ {
113
+ "epoch": 8.2,
114
+ "learning_rate": 5.6175e-05,
115
+ "loss": 1.8979,
116
+ "step": 1500
117
+ },
118
+ {
119
+ "epoch": 8.2,
120
+ "eval_loss": 0.9715010523796082,
121
+ "eval_runtime": 131.829,
122
+ "eval_samples_per_second": 19.874,
123
+ "eval_steps_per_second": 2.488,
124
+ "eval_wer": 0.8863658371259805,
125
+ "step": 1500
126
+ },
127
+ {
128
+ "epoch": 8.74,
129
+ "learning_rate": 5.9925e-05,
130
+ "loss": 1.7702,
131
+ "step": 1600
132
+ },
133
+ {
134
+ "epoch": 9.29,
135
+ "learning_rate": 6.367499999999999e-05,
136
+ "loss": 1.6541,
137
+ "step": 1700
138
+ },
139
+ {
140
+ "epoch": 9.84,
141
+ "learning_rate": 6.738749999999999e-05,
142
+ "loss": 1.6072,
143
+ "step": 1800
144
+ },
145
+ {
146
+ "epoch": 10.38,
147
+ "learning_rate": 7.11375e-05,
148
+ "loss": 1.5188,
149
+ "step": 1900
150
+ },
151
+ {
152
+ "epoch": 10.93,
153
+ "learning_rate": 7.48875e-05,
154
+ "loss": 1.4859,
155
+ "step": 2000
156
+ },
157
+ {
158
+ "epoch": 10.93,
159
+ "eval_loss": 0.5233765244483948,
160
+ "eval_runtime": 130.5844,
161
+ "eval_samples_per_second": 20.064,
162
+ "eval_steps_per_second": 2.512,
163
+ "eval_wer": 0.7062554468868016,
164
+ "step": 2000
165
+ },
166
+ {
167
+ "epoch": 11.48,
168
+ "learning_rate": 7.398251748251747e-05,
169
+ "loss": 1.4386,
170
+ "step": 2100
171
+ },
172
+ {
173
+ "epoch": 12.02,
174
+ "learning_rate": 7.293356643356642e-05,
175
+ "loss": 1.4333,
176
+ "step": 2200
177
+ },
178
+ {
179
+ "epoch": 12.57,
180
+ "learning_rate": 7.188461538461538e-05,
181
+ "loss": 1.3923,
182
+ "step": 2300
183
+ },
184
+ {
185
+ "epoch": 13.11,
186
+ "learning_rate": 7.083566433566433e-05,
187
+ "loss": 1.3623,
188
+ "step": 2400
189
+ },
190
+ {
191
+ "epoch": 13.66,
192
+ "learning_rate": 6.978671328671328e-05,
193
+ "loss": 1.3388,
194
+ "step": 2500
195
+ },
196
+ {
197
+ "epoch": 13.66,
198
+ "eval_loss": 0.4093627333641052,
199
+ "eval_runtime": 131.4899,
200
+ "eval_samples_per_second": 19.925,
201
+ "eval_steps_per_second": 2.494,
202
+ "eval_wer": 0.6203156773506343,
203
+ "step": 2500
204
+ },
205
+ {
206
+ "epoch": 14.21,
207
+ "learning_rate": 6.873776223776224e-05,
208
+ "loss": 1.3338,
209
+ "step": 2600
210
+ },
211
+ {
212
+ "epoch": 14.75,
213
+ "learning_rate": 6.768881118881118e-05,
214
+ "loss": 1.3039,
215
+ "step": 2700
216
+ },
217
+ {
218
+ "epoch": 15.3,
219
+ "learning_rate": 6.663986013986013e-05,
220
+ "loss": 1.2935,
221
+ "step": 2800
222
+ },
223
+ {
224
+ "epoch": 15.85,
225
+ "learning_rate": 6.559090909090908e-05,
226
+ "loss": 1.2818,
227
+ "step": 2900
228
+ },
229
+ {
230
+ "epoch": 16.39,
231
+ "learning_rate": 6.454195804195804e-05,
232
+ "loss": 1.2531,
233
+ "step": 3000
234
+ },
235
+ {
236
+ "epoch": 16.39,
237
+ "eval_loss": 0.3596232533454895,
238
+ "eval_runtime": 131.7209,
239
+ "eval_samples_per_second": 19.891,
240
+ "eval_steps_per_second": 2.49,
241
+ "eval_wer": 0.5184952067396146,
242
+ "step": 3000
243
+ },
244
+ {
245
+ "epoch": 16.94,
246
+ "learning_rate": 6.349300699300699e-05,
247
+ "loss": 1.2531,
248
+ "step": 3100
249
+ },
250
+ {
251
+ "epoch": 17.49,
252
+ "learning_rate": 6.244405594405593e-05,
253
+ "loss": 1.2282,
254
+ "step": 3200
255
+ },
256
+ {
257
+ "epoch": 18.03,
258
+ "learning_rate": 6.139510489510488e-05,
259
+ "loss": 1.2088,
260
+ "step": 3300
261
+ },
262
+ {
263
+ "epoch": 18.58,
264
+ "learning_rate": 6.034615384615384e-05,
265
+ "loss": 1.2041,
266
+ "step": 3400
267
+ },
268
+ {
269
+ "epoch": 19.13,
270
+ "learning_rate": 5.9297202797202796e-05,
271
+ "loss": 1.1992,
272
+ "step": 3500
273
+ },
274
+ {
275
+ "epoch": 19.13,
276
+ "eval_loss": 0.32211676239967346,
277
+ "eval_runtime": 130.9541,
278
+ "eval_samples_per_second": 20.007,
279
+ "eval_steps_per_second": 2.505,
280
+ "eval_wer": 0.4854265517575288,
281
+ "step": 3500
282
+ },
283
+ {
284
+ "epoch": 19.67,
285
+ "learning_rate": 5.8248251748251737e-05,
286
+ "loss": 1.1776,
287
+ "step": 3600
288
+ },
289
+ {
290
+ "epoch": 20.22,
291
+ "learning_rate": 5.719930069930069e-05,
292
+ "loss": 1.1815,
293
+ "step": 3700
294
+ },
295
+ {
296
+ "epoch": 20.77,
297
+ "learning_rate": 5.6160839160839156e-05,
298
+ "loss": 1.1624,
299
+ "step": 3800
300
+ },
301
+ {
302
+ "epoch": 21.31,
303
+ "learning_rate": 5.511188811188811e-05,
304
+ "loss": 1.1576,
305
+ "step": 3900
306
+ },
307
+ {
308
+ "epoch": 21.86,
309
+ "learning_rate": 5.4062937062937064e-05,
310
+ "loss": 1.1589,
311
+ "step": 4000
312
+ },
313
+ {
314
+ "epoch": 21.86,
315
+ "eval_loss": 0.30402758717536926,
316
+ "eval_runtime": 130.3078,
317
+ "eval_samples_per_second": 20.106,
318
+ "eval_steps_per_second": 2.517,
319
+ "eval_wer": 0.4610244988864143,
320
+ "step": 4000
321
+ },
322
+ {
323
+ "epoch": 22.4,
324
+ "learning_rate": 5.3013986013986004e-05,
325
+ "loss": 1.1705,
326
+ "step": 4100
327
+ },
328
+ {
329
+ "epoch": 22.95,
330
+ "learning_rate": 5.196503496503496e-05,
331
+ "loss": 1.1486,
332
+ "step": 4200
333
+ },
334
+ {
335
+ "epoch": 23.5,
336
+ "learning_rate": 5.091608391608391e-05,
337
+ "loss": 1.1266,
338
+ "step": 4300
339
+ },
340
+ {
341
+ "epoch": 24.04,
342
+ "learning_rate": 4.9867132867132866e-05,
343
+ "loss": 1.1348,
344
+ "step": 4400
345
+ },
346
+ {
347
+ "epoch": 24.59,
348
+ "learning_rate": 4.8818181818181813e-05,
349
+ "loss": 1.1345,
350
+ "step": 4500
351
+ },
352
+ {
353
+ "epoch": 24.59,
354
+ "eval_loss": 0.2906973361968994,
355
+ "eval_runtime": 132.6128,
356
+ "eval_samples_per_second": 19.757,
357
+ "eval_steps_per_second": 2.473,
358
+ "eval_wer": 0.4449501307252832,
359
+ "step": 4500
360
+ },
361
+ {
362
+ "epoch": 25.14,
363
+ "learning_rate": 4.776923076923077e-05,
364
+ "loss": 1.108,
365
+ "step": 4600
366
+ },
367
+ {
368
+ "epoch": 25.68,
369
+ "learning_rate": 4.672027972027972e-05,
370
+ "loss": 1.1114,
371
+ "step": 4700
372
+ },
373
+ {
374
+ "epoch": 26.23,
375
+ "learning_rate": 4.567132867132866e-05,
376
+ "loss": 1.1071,
377
+ "step": 4800
378
+ },
379
+ {
380
+ "epoch": 26.78,
381
+ "learning_rate": 4.4622377622377616e-05,
382
+ "loss": 1.1009,
383
+ "step": 4900
384
+ },
385
+ {
386
+ "epoch": 27.32,
387
+ "learning_rate": 4.357342657342657e-05,
388
+ "loss": 1.086,
389
+ "step": 5000
390
+ },
391
+ {
392
+ "epoch": 27.32,
393
+ "eval_loss": 0.27438458800315857,
394
+ "eval_runtime": 131.0041,
395
+ "eval_samples_per_second": 19.999,
396
+ "eval_steps_per_second": 2.504,
397
+ "eval_wer": 0.4298925147671153,
398
+ "step": 5000
399
+ },
400
+ {
401
+ "epoch": 27.87,
402
+ "learning_rate": 4.2524475524475524e-05,
403
+ "loss": 1.0843,
404
+ "step": 5100
405
+ },
406
+ {
407
+ "epoch": 28.42,
408
+ "learning_rate": 4.147552447552447e-05,
409
+ "loss": 1.0761,
410
+ "step": 5200
411
+ },
412
+ {
413
+ "epoch": 28.96,
414
+ "learning_rate": 4.0426573426573425e-05,
415
+ "loss": 1.0779,
416
+ "step": 5300
417
+ },
418
+ {
419
+ "epoch": 29.51,
420
+ "learning_rate": 3.937762237762237e-05,
421
+ "loss": 1.074,
422
+ "step": 5400
423
+ },
424
+ {
425
+ "epoch": 30.05,
426
+ "learning_rate": 3.832867132867133e-05,
427
+ "loss": 1.0697,
428
+ "step": 5500
429
+ },
430
+ {
431
+ "epoch": 30.05,
432
+ "eval_loss": 0.2617148756980896,
433
+ "eval_runtime": 131.2229,
434
+ "eval_samples_per_second": 19.966,
435
+ "eval_steps_per_second": 2.5,
436
+ "eval_wer": 0.4147864820373777,
437
+ "step": 5500
438
+ },
439
+ {
440
+ "epoch": 30.6,
441
+ "learning_rate": 3.727972027972028e-05,
442
+ "loss": 1.0599,
443
+ "step": 5600
444
+ },
445
+ {
446
+ "epoch": 31.15,
447
+ "learning_rate": 3.623076923076923e-05,
448
+ "loss": 1.0666,
449
+ "step": 5700
450
+ },
451
+ {
452
+ "epoch": 31.69,
453
+ "learning_rate": 3.518181818181818e-05,
454
+ "loss": 1.0574,
455
+ "step": 5800
456
+ },
457
+ {
458
+ "epoch": 32.24,
459
+ "learning_rate": 3.413286713286713e-05,
460
+ "loss": 1.0534,
461
+ "step": 5900
462
+ },
463
+ {
464
+ "epoch": 32.79,
465
+ "learning_rate": 3.308391608391608e-05,
466
+ "loss": 1.0518,
467
+ "step": 6000
468
+ },
469
+ {
470
+ "epoch": 32.79,
471
+ "eval_loss": 0.25627899169921875,
472
+ "eval_runtime": 130.5236,
473
+ "eval_samples_per_second": 20.073,
474
+ "eval_steps_per_second": 2.513,
475
+ "eval_wer": 0.4033117071753655,
476
+ "step": 6000
477
+ },
478
+ {
479
+ "epoch": 33.33,
480
+ "learning_rate": 3.204545454545454e-05,
481
+ "loss": 1.0304,
482
+ "step": 6100
483
+ },
484
+ {
485
+ "epoch": 33.88,
486
+ "learning_rate": 3.0996503496503495e-05,
487
+ "loss": 1.0336,
488
+ "step": 6200
489
+ },
490
+ {
491
+ "epoch": 34.43,
492
+ "learning_rate": 2.9947552447552443e-05,
493
+ "loss": 1.0289,
494
+ "step": 6300
495
+ },
496
+ {
497
+ "epoch": 34.97,
498
+ "learning_rate": 2.8898601398601397e-05,
499
+ "loss": 1.027,
500
+ "step": 6400
501
+ },
502
+ {
503
+ "epoch": 35.52,
504
+ "learning_rate": 2.7849650349650347e-05,
505
+ "loss": 1.0101,
506
+ "step": 6500
507
+ },
508
+ {
509
+ "epoch": 35.52,
510
+ "eval_loss": 0.24797162413597107,
511
+ "eval_runtime": 130.8457,
512
+ "eval_samples_per_second": 20.024,
513
+ "eval_steps_per_second": 2.507,
514
+ "eval_wer": 0.3934346857751525,
515
+ "step": 6500
516
+ },
517
+ {
518
+ "epoch": 36.07,
519
+ "learning_rate": 2.6800699300699298e-05,
520
+ "loss": 1.0278,
521
+ "step": 6600
522
+ },
523
+ {
524
+ "epoch": 36.61,
525
+ "learning_rate": 2.575174825174825e-05,
526
+ "loss": 0.9947,
527
+ "step": 6700
528
+ },
529
+ {
530
+ "epoch": 37.16,
531
+ "learning_rate": 2.4702797202797203e-05,
532
+ "loss": 1.0011,
533
+ "step": 6800
534
+ },
535
+ {
536
+ "epoch": 37.7,
537
+ "learning_rate": 2.365384615384615e-05,
538
+ "loss": 1.0041,
539
+ "step": 6900
540
+ },
541
+ {
542
+ "epoch": 38.25,
543
+ "learning_rate": 2.2604895104895104e-05,
544
+ "loss": 1.0013,
545
+ "step": 7000
546
+ },
547
+ {
548
+ "epoch": 38.25,
549
+ "eval_loss": 0.2412029653787613,
550
+ "eval_runtime": 130.9316,
551
+ "eval_samples_per_second": 20.01,
552
+ "eval_steps_per_second": 2.505,
553
+ "eval_wer": 0.38549433523772636,
554
+ "step": 7000
555
+ },
556
+ {
557
+ "epoch": 38.8,
558
+ "learning_rate": 2.1555944055944055e-05,
559
+ "loss": 0.9881,
560
+ "step": 7100
561
+ },
562
+ {
563
+ "epoch": 39.34,
564
+ "learning_rate": 2.0506993006993005e-05,
565
+ "loss": 0.9897,
566
+ "step": 7200
567
+ },
568
+ {
569
+ "epoch": 39.89,
570
+ "learning_rate": 1.9458041958041956e-05,
571
+ "loss": 0.9853,
572
+ "step": 7300
573
+ },
574
+ {
575
+ "epoch": 40.44,
576
+ "learning_rate": 1.8409090909090907e-05,
577
+ "loss": 0.9873,
578
+ "step": 7400
579
+ },
580
+ {
581
+ "epoch": 40.98,
582
+ "learning_rate": 1.7360139860139857e-05,
583
+ "loss": 0.9845,
584
+ "step": 7500
585
+ },
586
+ {
587
+ "epoch": 40.98,
588
+ "eval_loss": 0.23970751464366913,
589
+ "eval_runtime": 131.9142,
590
+ "eval_samples_per_second": 19.861,
591
+ "eval_steps_per_second": 2.486,
592
+ "eval_wer": 0.37711823375617315,
593
+ "step": 7500
594
+ },
595
+ {
596
+ "epoch": 41.53,
597
+ "learning_rate": 1.631118881118881e-05,
598
+ "loss": 0.9802,
599
+ "step": 7600
600
+ },
601
+ {
602
+ "epoch": 42.08,
603
+ "learning_rate": 1.5262237762237762e-05,
604
+ "loss": 0.9856,
605
+ "step": 7700
606
+ },
607
+ {
608
+ "epoch": 42.62,
609
+ "learning_rate": 1.4213286713286713e-05,
610
+ "loss": 0.9738,
611
+ "step": 7800
612
+ },
613
+ {
614
+ "epoch": 43.17,
615
+ "learning_rate": 1.3164335664335663e-05,
616
+ "loss": 0.9703,
617
+ "step": 7900
618
+ },
619
+ {
620
+ "epoch": 43.72,
621
+ "learning_rate": 1.2115384615384615e-05,
622
+ "loss": 0.9739,
623
+ "step": 8000
624
+ },
625
+ {
626
+ "epoch": 43.72,
627
+ "eval_loss": 0.2302592694759369,
628
+ "eval_runtime": 132.2355,
629
+ "eval_samples_per_second": 19.813,
630
+ "eval_steps_per_second": 2.48,
631
+ "eval_wer": 0.3726154740001937,
632
+ "step": 8000
633
+ },
634
+ {
635
+ "epoch": 44.26,
636
+ "learning_rate": 1.1066433566433566e-05,
637
+ "loss": 0.9687,
638
+ "step": 8100
639
+ },
640
+ {
641
+ "epoch": 44.81,
642
+ "learning_rate": 1.0038461538461537e-05,
643
+ "loss": 0.9697,
644
+ "step": 8200
645
+ },
646
+ {
647
+ "epoch": 45.36,
648
+ "learning_rate": 8.98951048951049e-06,
649
+ "loss": 0.9671,
650
+ "step": 8300
651
+ },
652
+ {
653
+ "epoch": 45.9,
654
+ "learning_rate": 7.94055944055944e-06,
655
+ "loss": 0.9578,
656
+ "step": 8400
657
+ },
658
+ {
659
+ "epoch": 46.45,
660
+ "learning_rate": 6.891608391608391e-06,
661
+ "loss": 0.9636,
662
+ "step": 8500
663
+ },
664
+ {
665
+ "epoch": 46.45,
666
+ "eval_loss": 0.22845281660556793,
667
+ "eval_runtime": 132.3979,
668
+ "eval_samples_per_second": 19.789,
669
+ "eval_steps_per_second": 2.477,
670
+ "eval_wer": 0.36869371550305025,
671
+ "step": 8500
672
+ },
673
+ {
674
+ "epoch": 46.99,
675
+ "learning_rate": 5.842657342657342e-06,
676
+ "loss": 0.955,
677
+ "step": 8600
678
+ },
679
+ {
680
+ "epoch": 47.54,
681
+ "learning_rate": 4.7937062937062936e-06,
682
+ "loss": 0.9495,
683
+ "step": 8700
684
+ },
685
+ {
686
+ "epoch": 48.09,
687
+ "learning_rate": 3.7447552447552446e-06,
688
+ "loss": 0.9612,
689
+ "step": 8800
690
+ },
691
+ {
692
+ "epoch": 48.63,
693
+ "learning_rate": 2.6958041958041957e-06,
694
+ "loss": 0.9415,
695
+ "step": 8900
696
+ },
697
+ {
698
+ "epoch": 49.18,
699
+ "learning_rate": 1.6468531468531468e-06,
700
+ "loss": 0.9466,
701
+ "step": 9000
702
+ },
703
+ {
704
+ "epoch": 49.18,
705
+ "eval_loss": 0.2261359691619873,
706
+ "eval_runtime": 131.8096,
707
+ "eval_samples_per_second": 19.877,
708
+ "eval_steps_per_second": 2.488,
709
+ "eval_wer": 0.36477195700590687,
710
+ "step": 9000
711
+ },
712
+ {
713
+ "epoch": 49.73,
714
+ "learning_rate": 5.979020979020978e-07,
715
+ "loss": 0.9533,
716
+ "step": 9100
717
+ },
718
+ {
719
+ "epoch": 50.0,
720
+ "step": 9150,
721
+ "total_flos": 5.14420411350113e+19,
722
+ "train_loss": 1.6428977303426773,
723
+ "train_runtime": 23684.6778,
724
+ "train_samples_per_second": 12.36,
725
+ "train_steps_per_second": 0.386
726
+ }
727
+ ],
728
+ "max_steps": 9150,
729
+ "num_train_epochs": 50,
730
+ "total_flos": 5.14420411350113e+19,
731
+ "trial_name": null,
732
+ "trial_params": null
733
+ }
.ipynb_checkpoints/vocab-checkpoint.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"\u0626": 1, "\u0627": 2, "\u0628": 3, "\u062a": 4, "\u062c": 5, "\u062e": 6, "\u062f": 7, "\u0631": 8, "\u0632": 9, "\u0633": 10, "\u0634": 11, "\u063a": 12, "\u0641": 13, "\u0642": 14, "\u0643": 15, "\u0644": 16, "\u0645": 17, "\u0646": 18, "\u0648": 19, "\u0649": 20, "\u064a": 21, "\u067e": 22, "\u0686": 23, "\u0698": 24, "\u06ad": 25, "\u06af": 26, "\u06be": 27, "\u06c6": 28, "\u06c7": 29, "\u06c8": 30, "\u06cb": 31, "\u06d0": 32, "\u06d5": 33, "|": 0, "[UNK]": 34, "[PAD]": 35}
added_tokens.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"<s>": 36, "</s>": 37}
config.json CHANGED
@@ -49,7 +49,7 @@
49
  "feat_extract_activation": "gelu",
50
  "feat_extract_dropout": 0.0,
51
  "feat_extract_norm": "layer",
52
- "feat_proj_dropout": 0.0,
53
  "feat_quantizer_dropout": 0.0,
54
  "final_dropout": 0.0,
55
  "hidden_act": "gelu",
@@ -58,7 +58,7 @@
58
  "initializer_range": 0.02,
59
  "intermediate_size": 4096,
60
  "layer_norm_eps": 1e-05,
61
- "layerdrop": 0.0,
62
  "mask_feature_length": 64,
63
  "mask_feature_min_masks": 0,
64
  "mask_feature_prob": 0.25,
@@ -76,7 +76,7 @@
76
  "num_hidden_layers": 24,
77
  "num_negatives": 100,
78
  "output_hidden_size": 1024,
79
- "pad_token_id": 43,
80
  "proj_codevector_dim": 768,
81
  "tdnn_dilation": [
82
  1,
@@ -102,6 +102,6 @@
102
  "torch_dtype": "float32",
103
  "transformers_version": "4.16.0.dev0",
104
  "use_weighted_layer_sum": false,
105
- "vocab_size": 46,
106
  "xvector_output_dim": 512
107
  }
 
49
  "feat_extract_activation": "gelu",
50
  "feat_extract_dropout": 0.0,
51
  "feat_extract_norm": "layer",
52
+ "feat_proj_dropout": 0.01,
53
  "feat_quantizer_dropout": 0.0,
54
  "final_dropout": 0.0,
55
  "hidden_act": "gelu",
 
58
  "initializer_range": 0.02,
59
  "intermediate_size": 4096,
60
  "layer_norm_eps": 1e-05,
61
+ "layerdrop": 0.01,
62
  "mask_feature_length": 64,
63
  "mask_feature_min_masks": 0,
64
  "mask_feature_prob": 0.25,
 
76
  "num_hidden_layers": 24,
77
  "num_negatives": 100,
78
  "output_hidden_size": 1024,
79
+ "pad_token_id": 35,
80
  "proj_codevector_dim": 768,
81
  "tdnn_dilation": [
82
  1,
 
102
  "torch_dtype": "float32",
103
  "transformers_version": "4.16.0.dev0",
104
  "use_weighted_layer_sum": false,
105
+ "vocab_size": 38,
106
  "xvector_output_dim": 512
107
  }
eval.py CHANGED
@@ -49,7 +49,7 @@ def log_results(result: Dataset, args: Dict[str, str]):
49
  def normalize_text(text: str) -> str:
50
  """DO ADAPT FOR YOUR USE CASE. this function normalizes the target text."""
51
 
52
- chars_to_ignore_regex = '[,?.!\-\;\:"“%‘”�—’…–]' # noqa: W605 IMPORTANT: this should correspond to the chars that were ignored during training
53
 
54
  text = re.sub(chars_to_ignore_regex, "", text.lower())
55
 
 
49
  def normalize_text(text: str) -> str:
50
  """DO ADAPT FOR YOUR USE CASE. this function normalizes the target text."""
51
 
52
+ chars_to_ignore_regex = '[!"%,.:;?\\_|©«¬»،؛؟‒–—’“”„…‹›−☺♂�\\\\-]' # noqa: W605 IMPORTANT: this should correspond to the chars that were ignored during training
53
 
54
  text = re.sub(chars_to_ignore_regex, "", text.lower())
55
 
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3d0766bfc29cd39a379c7d68fbb74740dd0e49ae35d358b18714f5d81a4a265e
3
- size 1262112241
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58b49521325b8f7f03fadd8108710e86cf7648a27e4d95bca7a100d6f0bcafe7
3
+ size 1262079473
runs/Jan30_06-27-48_job-0074bb36-c67f-4775-b1b6-176eb09b0ba4/1643524110.4381003/events.out.tfevents.1643524110.job-0074bb36-c67f-4775-b1b6-176eb09b0ba4.1718926.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06a18bc0c8a722b001f276930e89460c1e003c727968acc2809c3f49c07b923e
3
+ size 4802
runs/Jan30_06-27-48_job-0074bb36-c67f-4775-b1b6-176eb09b0ba4/events.out.tfevents.1643524110.job-0074bb36-c67f-4775-b1b6-176eb09b0ba4.1718926.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8346aada8c8b5a627d85ab3bdffb1288a6c44fef4fa75797d72f46c75f9b976d
3
+ size 5833
special_tokens_map.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]", "additional_special_tokens": [{"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}]}
tokenizer_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"unk_token": "[UNK]", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "[PAD]", "do_lower_case": false, "word_delimiter_token": "|", "special_tokens_map_file": null, "tokenizer_file": null, "name_or_path": "./xls-r-uyghur-cv7", "tokenizer_class": "Wav2Vec2CTCTokenizer"}
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ce6ca494ec7cd57f84af014b087832b0acfbae962f708eeaaad1fab88853da2f
3
- size 2991
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca9ce329cfce769b1c7b1253f8967dc41f08fdc7b753833175ccc88d99a88024
3
+ size 3055
vocab.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"ئ": 1, "ا": 2, "ب": 3, "ت": 4, "ج": 5, "خ": 6, "د": 7, "ر": 8, "ز": 9, "س": 10, "ش": 11, "غ": 12, "ف": 13, "ق": 14, "ك": 15, "ل": 16, "م": 17, "ن": 18, "و": 19, "ى": 20, "ي": 21, "پ": 22, "چ": 23, "ژ": 24, "ڭ": 25, "گ": 26, "ھ": 27, "ۆ": 28, "ۇ": 29, "ۈ": 30, "ۋ": 31, "ې": 32, "ە": 33, "|": 0, "[UNK]": 34, "[PAD]": 35}