tanmaylaud commited on
Commit
8629923
1 Parent(s): aafdd3b

Updated weights WER:24.94

Browse files
Files changed (7) hide show
  1. config.json +8 -8
  2. optimizer.pt +3 -0
  3. pytorch_model.bin +2 -2
  4. scheduler.pt +3 -0
  5. trainer_state.json +1016 -0
  6. training_args.bin +1 -1
  7. vocab.json +1 -1
config.json CHANGED
@@ -1,11 +1,11 @@
1
  {
2
- "_name_or_path": "/workspace/output_models/hi-mr/wav2vec2-large-xlsr-hindi-marathi-2/checkpoint-2000",
3
- "activation_dropout": 0.06,
4
  "apply_spec_augment": true,
5
  "architectures": [
6
  "Wav2Vec2ForCTC"
7
  ],
8
- "attention_dropout": 0.1,
9
  "bos_token_id": 1,
10
  "conv_bias": true,
11
  "conv_dim": [
@@ -42,8 +42,8 @@
42
  "feat_extract_activation": "gelu",
43
  "feat_extract_dropout": 0.0,
44
  "feat_extract_norm": "layer",
45
- "feat_proj_dropout": 0.04,
46
- "final_dropout": 0.0,
47
  "gradient_checkpointing": true,
48
  "hidden_act": "gelu",
49
  "hidden_dropout": 0.05,
@@ -62,7 +62,7 @@
62
  "mask_time_length": 10,
63
  "mask_time_min_space": 1,
64
  "mask_time_other": 0.0,
65
- "mask_time_prob": 0.082,
66
  "mask_time_selection": "static",
67
  "model_type": "wav2vec2",
68
  "num_attention_heads": 16,
@@ -70,7 +70,7 @@
70
  "num_conv_pos_embeddings": 128,
71
  "num_feat_extract_layers": 7,
72
  "num_hidden_layers": 24,
73
- "pad_token_id": 72,
74
  "transformers_version": "4.5.0.dev0",
75
- "vocab_size": 73
76
  }
1
  {
2
+ "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
3
+ "activation_dropout": 0.1,
4
  "apply_spec_augment": true,
5
  "architectures": [
6
  "Wav2Vec2ForCTC"
7
  ],
8
+ "attention_dropout": 0.2,
9
  "bos_token_id": 1,
10
  "conv_bias": true,
11
  "conv_dim": [
42
  "feat_extract_activation": "gelu",
43
  "feat_extract_dropout": 0.0,
44
  "feat_extract_norm": "layer",
45
+ "feat_proj_dropout": 0.1,
46
+ "final_dropout": 0.1,
47
  "gradient_checkpointing": true,
48
  "hidden_act": "gelu",
49
  "hidden_dropout": 0.05,
62
  "mask_time_length": 10,
63
  "mask_time_min_space": 1,
64
  "mask_time_other": 0.0,
65
+ "mask_time_prob": 0.05,
66
  "mask_time_selection": "static",
67
  "model_type": "wav2vec2",
68
  "num_attention_heads": 16,
70
  "num_conv_pos_embeddings": 128,
71
  "num_feat_extract_layers": 7,
72
  "num_hidden_layers": 24,
73
+ "pad_token_id": 73,
74
  "transformers_version": "4.5.0.dev0",
75
+ "vocab_size": 74
76
  }
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b538512208981b86bad40c84dca35d6debe0480643b51f75675349c77e210c3
3
+ size 2490683911
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6d9f134a8368c7c58d6495a4345d607845c7e9118a6346b63f6ec40b9b14472d
3
- size 1262233111
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d2ba642a7c83d88a9b82b15757e40514fecf6e95a473f3915b831030070f15d
3
+ size 1262237207
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f676aeb2802a2b9a559dffe2909c5347f7a37bf0aa25486203d2b34b9322c69a
3
+ size 623
trainer_state.json ADDED
@@ -0,0 +1,1016 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.24944954621126686,
3
+ "best_model_checkpoint": "/workspace/output_models/hi-mr/wav2vec2-large-xlsr-hindi-marathi-3/checkpoint-10000",
4
+ "epoch": 16.0,
5
+ "global_step": 10000,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.16,
12
+ "learning_rate": 6.666666666666666e-05,
13
+ "loss": 16.623,
14
+ "step": 100
15
+ },
16
+ {
17
+ "epoch": 0.32,
18
+ "learning_rate": 0.0001333333333333333,
19
+ "loss": 3.8395,
20
+ "step": 200
21
+ },
22
+ {
23
+ "epoch": 0.32,
24
+ "eval_loss": 3.426492214202881,
25
+ "eval_runtime": 80.6897,
26
+ "eval_samples_per_second": 24.786,
27
+ "eval_wer": 1.0,
28
+ "step": 200
29
+ },
30
+ {
31
+ "epoch": 0.48,
32
+ "learning_rate": 0.00019999999999999998,
33
+ "loss": 3.4548,
34
+ "step": 300
35
+ },
36
+ {
37
+ "epoch": 0.64,
38
+ "learning_rate": 0.0002666666666666666,
39
+ "loss": 3.3471,
40
+ "step": 400
41
+ },
42
+ {
43
+ "epoch": 0.64,
44
+ "eval_loss": 2.9866673946380615,
45
+ "eval_runtime": 81.4064,
46
+ "eval_samples_per_second": 24.568,
47
+ "eval_wer": 1.0,
48
+ "step": 400
49
+ },
50
+ {
51
+ "epoch": 0.8,
52
+ "learning_rate": 0.0002991803278688524,
53
+ "loss": 1.9911,
54
+ "step": 500
55
+ },
56
+ {
57
+ "epoch": 0.96,
58
+ "learning_rate": 0.00029754098360655737,
59
+ "loss": 0.9874,
60
+ "step": 600
61
+ },
62
+ {
63
+ "epoch": 0.96,
64
+ "eval_loss": 0.6858933568000793,
65
+ "eval_runtime": 82.4721,
66
+ "eval_samples_per_second": 24.251,
67
+ "eval_wer": 0.5436335320337253,
68
+ "step": 600
69
+ },
70
+ {
71
+ "epoch": 1.12,
72
+ "learning_rate": 0.00029590163934426226,
73
+ "loss": 0.7698,
74
+ "step": 700
75
+ },
76
+ {
77
+ "epoch": 1.28,
78
+ "learning_rate": 0.00029426229508196716,
79
+ "loss": 0.6471,
80
+ "step": 800
81
+ },
82
+ {
83
+ "epoch": 1.28,
84
+ "eval_loss": 0.5174924731254578,
85
+ "eval_runtime": 82.661,
86
+ "eval_samples_per_second": 24.195,
87
+ "eval_wer": 0.43928897481338275,
88
+ "step": 800
89
+ },
90
+ {
91
+ "epoch": 1.44,
92
+ "learning_rate": 0.0002926229508196721,
93
+ "loss": 0.568,
94
+ "step": 900
95
+ },
96
+ {
97
+ "epoch": 1.6,
98
+ "learning_rate": 0.000290983606557377,
99
+ "loss": 0.5393,
100
+ "step": 1000
101
+ },
102
+ {
103
+ "epoch": 1.6,
104
+ "eval_loss": 0.4453641474246979,
105
+ "eval_runtime": 83.0998,
106
+ "eval_samples_per_second": 24.067,
107
+ "eval_wer": 0.39015090489232584,
108
+ "step": 1000
109
+ },
110
+ {
111
+ "epoch": 1.76,
112
+ "learning_rate": 0.00028934426229508195,
113
+ "loss": 0.5009,
114
+ "step": 1100
115
+ },
116
+ {
117
+ "epoch": 1.92,
118
+ "learning_rate": 0.00028770491803278684,
119
+ "loss": 0.4932,
120
+ "step": 1200
121
+ },
122
+ {
123
+ "epoch": 1.92,
124
+ "eval_loss": 0.41159647703170776,
125
+ "eval_runtime": 83.9345,
126
+ "eval_samples_per_second": 23.828,
127
+ "eval_wer": 0.37189194994898234,
128
+ "step": 1200
129
+ },
130
+ {
131
+ "epoch": 2.08,
132
+ "learning_rate": 0.00028606557377049174,
133
+ "loss": 0.4407,
134
+ "step": 1300
135
+ },
136
+ {
137
+ "epoch": 2.24,
138
+ "learning_rate": 0.00028442622950819674,
139
+ "loss": 0.3891,
140
+ "step": 1400
141
+ },
142
+ {
143
+ "epoch": 2.24,
144
+ "eval_loss": 0.4060937166213989,
145
+ "eval_runtime": 82.5906,
146
+ "eval_samples_per_second": 24.216,
147
+ "eval_wer": 0.3500886096342839,
148
+ "step": 1400
149
+ },
150
+ {
151
+ "epoch": 2.4,
152
+ "learning_rate": 0.00028278688524590163,
153
+ "loss": 0.393,
154
+ "step": 1500
155
+ },
156
+ {
157
+ "epoch": 2.56,
158
+ "learning_rate": 0.00028114754098360653,
159
+ "loss": 0.3723,
160
+ "step": 1600
161
+ },
162
+ {
163
+ "epoch": 2.56,
164
+ "eval_loss": 0.37875330448150635,
165
+ "eval_runtime": 83.1067,
166
+ "eval_samples_per_second": 24.065,
167
+ "eval_wer": 0.3370388271306589,
168
+ "step": 1600
169
+ },
170
+ {
171
+ "epoch": 2.72,
172
+ "learning_rate": 0.0002795081967213115,
173
+ "loss": 0.3732,
174
+ "step": 1700
175
+ },
176
+ {
177
+ "epoch": 2.88,
178
+ "learning_rate": 0.00027786885245901637,
179
+ "loss": 0.3536,
180
+ "step": 1800
181
+ },
182
+ {
183
+ "epoch": 2.88,
184
+ "eval_loss": 0.3747707009315491,
185
+ "eval_runtime": 83.4226,
186
+ "eval_samples_per_second": 23.974,
187
+ "eval_wer": 0.3268889962945062,
188
+ "step": 1800
189
+ },
190
+ {
191
+ "epoch": 3.04,
192
+ "learning_rate": 0.0002762295081967213,
193
+ "loss": 0.3394,
194
+ "step": 1900
195
+ },
196
+ {
197
+ "epoch": 3.2,
198
+ "learning_rate": 0.0002745901639344262,
199
+ "loss": 0.2952,
200
+ "step": 2000
201
+ },
202
+ {
203
+ "epoch": 3.2,
204
+ "eval_loss": 0.3621469736099243,
205
+ "eval_runtime": 82.21,
206
+ "eval_samples_per_second": 24.328,
207
+ "eval_wer": 0.31018742280221256,
208
+ "step": 2000
209
+ },
210
+ {
211
+ "epoch": 3.36,
212
+ "learning_rate": 0.0002729508196721311,
213
+ "loss": 0.3014,
214
+ "step": 2100
215
+ },
216
+ {
217
+ "epoch": 3.52,
218
+ "learning_rate": 0.00027131147540983606,
219
+ "loss": 0.2984,
220
+ "step": 2200
221
+ },
222
+ {
223
+ "epoch": 3.52,
224
+ "eval_loss": 0.35865798592567444,
225
+ "eval_runtime": 83.2434,
226
+ "eval_samples_per_second": 24.026,
227
+ "eval_wer": 0.2970302346812738,
228
+ "step": 2200
229
+ },
230
+ {
231
+ "epoch": 3.68,
232
+ "learning_rate": 0.00026967213114754095,
233
+ "loss": 0.2826,
234
+ "step": 2300
235
+ },
236
+ {
237
+ "epoch": 3.84,
238
+ "learning_rate": 0.0002680327868852459,
239
+ "loss": 0.2821,
240
+ "step": 2400
241
+ },
242
+ {
243
+ "epoch": 3.84,
244
+ "eval_loss": 0.3647235035896301,
245
+ "eval_runtime": 83.2284,
246
+ "eval_samples_per_second": 24.03,
247
+ "eval_wer": 0.30664303743085763,
248
+ "step": 2400
249
+ },
250
+ {
251
+ "epoch": 4.0,
252
+ "learning_rate": 0.0002663934426229508,
253
+ "loss": 0.27,
254
+ "step": 2500
255
+ },
256
+ {
257
+ "epoch": 4.16,
258
+ "learning_rate": 0.0002647540983606557,
259
+ "loss": 0.2433,
260
+ "step": 2600
261
+ },
262
+ {
263
+ "epoch": 4.16,
264
+ "eval_loss": 0.3836052119731903,
265
+ "eval_runtime": 82.8815,
266
+ "eval_samples_per_second": 24.131,
267
+ "eval_wer": 0.3012190537565115,
268
+ "step": 2600
269
+ },
270
+ {
271
+ "epoch": 4.32,
272
+ "learning_rate": 0.00026311475409836063,
273
+ "loss": 0.2373,
274
+ "step": 2700
275
+ },
276
+ {
277
+ "epoch": 4.48,
278
+ "learning_rate": 0.00026147540983606553,
279
+ "loss": 0.2304,
280
+ "step": 2800
281
+ },
282
+ {
283
+ "epoch": 4.48,
284
+ "eval_loss": 0.3565887212753296,
285
+ "eval_runtime": 83.2398,
286
+ "eval_samples_per_second": 24.027,
287
+ "eval_wer": 0.2935395521185758,
288
+ "step": 2800
289
+ },
290
+ {
291
+ "epoch": 4.64,
292
+ "learning_rate": 0.0002598360655737705,
293
+ "loss": 0.2262,
294
+ "step": 2900
295
+ },
296
+ {
297
+ "epoch": 4.8,
298
+ "learning_rate": 0.00025819672131147537,
299
+ "loss": 0.2249,
300
+ "step": 3000
301
+ },
302
+ {
303
+ "epoch": 4.8,
304
+ "eval_loss": 0.371224969625473,
305
+ "eval_runtime": 83.7191,
306
+ "eval_samples_per_second": 23.889,
307
+ "eval_wer": 0.285913753289297,
308
+ "step": 3000
309
+ },
310
+ {
311
+ "epoch": 4.96,
312
+ "learning_rate": 0.00025655737704918027,
313
+ "loss": 0.2232,
314
+ "step": 3100
315
+ },
316
+ {
317
+ "epoch": 5.12,
318
+ "learning_rate": 0.0002549180327868852,
319
+ "loss": 0.2055,
320
+ "step": 3200
321
+ },
322
+ {
323
+ "epoch": 5.12,
324
+ "eval_loss": 0.36438417434692383,
325
+ "eval_runtime": 82.4039,
326
+ "eval_samples_per_second": 24.271,
327
+ "eval_wer": 0.28048976961495087,
328
+ "step": 3200
329
+ },
330
+ {
331
+ "epoch": 5.28,
332
+ "learning_rate": 0.00025327868852459016,
333
+ "loss": 0.1999,
334
+ "step": 3300
335
+ },
336
+ {
337
+ "epoch": 5.44,
338
+ "learning_rate": 0.00025163934426229506,
339
+ "loss": 0.1969,
340
+ "step": 3400
341
+ },
342
+ {
343
+ "epoch": 5.44,
344
+ "eval_loss": 0.371999055147171,
345
+ "eval_runtime": 83.0961,
346
+ "eval_samples_per_second": 24.069,
347
+ "eval_wer": 0.28204715106600076,
348
+ "step": 3400
349
+ },
350
+ {
351
+ "epoch": 5.6,
352
+ "learning_rate": 0.00025,
353
+ "loss": 0.1949,
354
+ "step": 3500
355
+ },
356
+ {
357
+ "epoch": 5.76,
358
+ "learning_rate": 0.0002483606557377049,
359
+ "loss": 0.1945,
360
+ "step": 3600
361
+ },
362
+ {
363
+ "epoch": 5.76,
364
+ "eval_loss": 0.3662562370300293,
365
+ "eval_runtime": 82.7035,
366
+ "eval_samples_per_second": 24.183,
367
+ "eval_wer": 0.28435637183824714,
368
+ "step": 3600
369
+ },
370
+ {
371
+ "epoch": 5.92,
372
+ "learning_rate": 0.00024672131147540985,
373
+ "loss": 0.191,
374
+ "step": 3700
375
+ },
376
+ {
377
+ "epoch": 6.08,
378
+ "learning_rate": 0.00024508196721311474,
379
+ "loss": 0.1689,
380
+ "step": 3800
381
+ },
382
+ {
383
+ "epoch": 6.08,
384
+ "eval_loss": 0.3731316328048706,
385
+ "eval_runtime": 83.1862,
386
+ "eval_samples_per_second": 24.042,
387
+ "eval_wer": 0.2787712797379303,
388
+ "step": 3800
389
+ },
390
+ {
391
+ "epoch": 6.24,
392
+ "learning_rate": 0.00024344262295081966,
393
+ "loss": 0.1631,
394
+ "step": 3900
395
+ },
396
+ {
397
+ "epoch": 6.4,
398
+ "learning_rate": 0.00024180327868852458,
399
+ "loss": 0.1568,
400
+ "step": 4000
401
+ },
402
+ {
403
+ "epoch": 6.4,
404
+ "eval_loss": 0.37625598907470703,
405
+ "eval_runtime": 83.6104,
406
+ "eval_samples_per_second": 23.92,
407
+ "eval_wer": 0.2781805488427045,
408
+ "step": 4000
409
+ },
410
+ {
411
+ "epoch": 6.56,
412
+ "learning_rate": 0.00024016393442622948,
413
+ "loss": 0.1665,
414
+ "step": 4100
415
+ },
416
+ {
417
+ "epoch": 6.72,
418
+ "learning_rate": 0.0002385245901639344,
419
+ "loss": 0.1581,
420
+ "step": 4200
421
+ },
422
+ {
423
+ "epoch": 6.72,
424
+ "eval_loss": 0.3642520606517792,
425
+ "eval_runtime": 83.3706,
426
+ "eval_samples_per_second": 23.989,
427
+ "eval_wer": 0.2725954567423876,
428
+ "step": 4200
429
+ },
430
+ {
431
+ "epoch": 6.88,
432
+ "learning_rate": 0.00023688524590163932,
433
+ "loss": 0.1517,
434
+ "step": 4300
435
+ },
436
+ {
437
+ "epoch": 7.04,
438
+ "learning_rate": 0.00023524590163934424,
439
+ "loss": 0.1529,
440
+ "step": 4400
441
+ },
442
+ {
443
+ "epoch": 7.04,
444
+ "eval_loss": 0.3819567859172821,
445
+ "eval_runtime": 83.7194,
446
+ "eval_samples_per_second": 23.889,
447
+ "eval_wer": 0.27238064550776003,
448
+ "step": 4400
449
+ },
450
+ {
451
+ "epoch": 7.2,
452
+ "learning_rate": 0.00023360655737704916,
453
+ "loss": 0.1344,
454
+ "step": 4500
455
+ },
456
+ {
457
+ "epoch": 7.36,
458
+ "learning_rate": 0.00023196721311475406,
459
+ "loss": 0.1363,
460
+ "step": 4600
461
+ },
462
+ {
463
+ "epoch": 7.36,
464
+ "eval_loss": 0.38544762134552,
465
+ "eval_runtime": 83.3319,
466
+ "eval_samples_per_second": 24.0,
467
+ "eval_wer": 0.2733472960635841,
468
+ "step": 4600
469
+ },
470
+ {
471
+ "epoch": 7.52,
472
+ "learning_rate": 0.00023032786885245898,
473
+ "loss": 0.1276,
474
+ "step": 4700
475
+ },
476
+ {
477
+ "epoch": 7.68,
478
+ "learning_rate": 0.0002286885245901639,
479
+ "loss": 0.1252,
480
+ "step": 4800
481
+ },
482
+ {
483
+ "epoch": 7.68,
484
+ "eval_loss": 0.37892764806747437,
485
+ "eval_runtime": 83.6054,
486
+ "eval_samples_per_second": 23.922,
487
+ "eval_wer": 0.2722732398904463,
488
+ "step": 4800
489
+ },
490
+ {
491
+ "epoch": 7.84,
492
+ "learning_rate": 0.00022704918032786882,
493
+ "loss": 0.1328,
494
+ "step": 4900
495
+ },
496
+ {
497
+ "epoch": 8.0,
498
+ "learning_rate": 0.00022540983606557374,
499
+ "loss": 0.1327,
500
+ "step": 5000
501
+ },
502
+ {
503
+ "epoch": 8.0,
504
+ "eval_loss": 0.3974834084510803,
505
+ "eval_runtime": 83.8203,
506
+ "eval_samples_per_second": 23.861,
507
+ "eval_wer": 0.2739380269588099,
508
+ "step": 5000
509
+ },
510
+ {
511
+ "epoch": 8.16,
512
+ "learning_rate": 0.0002237704918032787,
513
+ "loss": 0.1176,
514
+ "step": 5100
515
+ },
516
+ {
517
+ "epoch": 8.32,
518
+ "learning_rate": 0.0002221311475409836,
519
+ "loss": 0.1157,
520
+ "step": 5200
521
+ },
522
+ {
523
+ "epoch": 8.32,
524
+ "eval_loss": 0.3960728347301483,
525
+ "eval_runtime": 84.4842,
526
+ "eval_samples_per_second": 23.673,
527
+ "eval_wer": 0.2711991837173084,
528
+ "step": 5200
529
+ },
530
+ {
531
+ "epoch": 8.48,
532
+ "learning_rate": 0.0002204918032786885,
533
+ "loss": 0.1151,
534
+ "step": 5300
535
+ },
536
+ {
537
+ "epoch": 8.64,
538
+ "learning_rate": 0.00021885245901639343,
539
+ "loss": 0.1173,
540
+ "step": 5400
541
+ },
542
+ {
543
+ "epoch": 8.64,
544
+ "eval_loss": 0.40709516406059265,
545
+ "eval_runtime": 82.9277,
546
+ "eval_samples_per_second": 24.117,
547
+ "eval_wer": 0.2727565651683583,
548
+ "step": 5400
549
+ },
550
+ {
551
+ "epoch": 8.8,
552
+ "learning_rate": 0.00021721311475409835,
553
+ "loss": 0.1143,
554
+ "step": 5500
555
+ },
556
+ {
557
+ "epoch": 8.96,
558
+ "learning_rate": 0.00021557377049180327,
559
+ "loss": 0.1166,
560
+ "step": 5600
561
+ },
562
+ {
563
+ "epoch": 8.96,
564
+ "eval_loss": 0.41475972533226013,
565
+ "eval_runtime": 84.4573,
566
+ "eval_samples_per_second": 23.681,
567
+ "eval_wer": 0.2736695129155255,
568
+ "step": 5600
569
+ },
570
+ {
571
+ "epoch": 9.12,
572
+ "learning_rate": 0.0002139344262295082,
573
+ "loss": 0.1028,
574
+ "step": 5700
575
+ },
576
+ {
577
+ "epoch": 9.28,
578
+ "learning_rate": 0.00021229508196721309,
579
+ "loss": 0.1026,
580
+ "step": 5800
581
+ },
582
+ {
583
+ "epoch": 9.28,
584
+ "eval_loss": 0.4114053547382355,
585
+ "eval_runtime": 83.5505,
586
+ "eval_samples_per_second": 23.938,
587
+ "eval_wer": 0.2681381236238655,
588
+ "step": 5800
589
+ },
590
+ {
591
+ "epoch": 9.44,
592
+ "learning_rate": 0.000210655737704918,
593
+ "loss": 0.1004,
594
+ "step": 5900
595
+ },
596
+ {
597
+ "epoch": 9.6,
598
+ "learning_rate": 0.00020901639344262293,
599
+ "loss": 0.1121,
600
+ "step": 6000
601
+ },
602
+ {
603
+ "epoch": 9.6,
604
+ "eval_loss": 0.4229777753353119,
605
+ "eval_runtime": 83.5437,
606
+ "eval_samples_per_second": 23.94,
607
+ "eval_wer": 0.27060845282208257,
608
+ "step": 6000
609
+ },
610
+ {
611
+ "epoch": 9.76,
612
+ "learning_rate": 0.00020737704918032785,
613
+ "loss": 0.1026,
614
+ "step": 6100
615
+ },
616
+ {
617
+ "epoch": 9.92,
618
+ "learning_rate": 0.00020573770491803277,
619
+ "loss": 0.1041,
620
+ "step": 6200
621
+ },
622
+ {
623
+ "epoch": 9.92,
624
+ "eval_loss": 0.40711304545402527,
625
+ "eval_runtime": 83.5856,
626
+ "eval_samples_per_second": 23.928,
627
+ "eval_wer": 0.27017883035282747,
628
+ "step": 6200
629
+ },
630
+ {
631
+ "epoch": 10.08,
632
+ "learning_rate": 0.0002040983606557377,
633
+ "loss": 0.0877,
634
+ "step": 6300
635
+ },
636
+ {
637
+ "epoch": 10.24,
638
+ "learning_rate": 0.0002024590163934426,
639
+ "loss": 0.088,
640
+ "step": 6400
641
+ },
642
+ {
643
+ "epoch": 10.24,
644
+ "eval_loss": 0.437641978263855,
645
+ "eval_runtime": 83.2287,
646
+ "eval_samples_per_second": 24.03,
647
+ "eval_wer": 0.26797701519789485,
648
+ "step": 6400
649
+ },
650
+ {
651
+ "epoch": 10.4,
652
+ "learning_rate": 0.0002008196721311475,
653
+ "loss": 0.0941,
654
+ "step": 6500
655
+ },
656
+ {
657
+ "epoch": 10.56,
658
+ "learning_rate": 0.00019918032786885243,
659
+ "loss": 0.0884,
660
+ "step": 6600
661
+ },
662
+ {
663
+ "epoch": 10.56,
664
+ "eval_loss": 0.4451253414154053,
665
+ "eval_runtime": 84.0696,
666
+ "eval_samples_per_second": 23.79,
667
+ "eval_wer": 0.26529187476505023,
668
+ "step": 6600
669
+ },
670
+ {
671
+ "epoch": 10.72,
672
+ "learning_rate": 0.00019754098360655735,
673
+ "loss": 0.0867,
674
+ "step": 6700
675
+ },
676
+ {
677
+ "epoch": 10.88,
678
+ "learning_rate": 0.00019590163934426227,
679
+ "loss": 0.0978,
680
+ "step": 6800
681
+ },
682
+ {
683
+ "epoch": 10.88,
684
+ "eval_loss": 0.4322107136249542,
685
+ "eval_runtime": 83.5762,
686
+ "eval_samples_per_second": 23.93,
687
+ "eval_wer": 0.26475484667848126,
688
+ "step": 6800
689
+ },
690
+ {
691
+ "epoch": 11.04,
692
+ "learning_rate": 0.00019426229508196722,
693
+ "loss": 0.0834,
694
+ "step": 6900
695
+ },
696
+ {
697
+ "epoch": 11.2,
698
+ "learning_rate": 0.00019262295081967214,
699
+ "loss": 0.0785,
700
+ "step": 7000
701
+ },
702
+ {
703
+ "epoch": 11.2,
704
+ "eval_loss": 0.4244215190410614,
705
+ "eval_runtime": 83.288,
706
+ "eval_samples_per_second": 24.013,
707
+ "eval_wer": 0.25970678266473335,
708
+ "step": 7000
709
+ },
710
+ {
711
+ "epoch": 11.36,
712
+ "learning_rate": 0.00019098360655737704,
713
+ "loss": 0.0723,
714
+ "step": 7100
715
+ },
716
+ {
717
+ "epoch": 11.52,
718
+ "learning_rate": 0.00018934426229508196,
719
+ "loss": 0.0771,
720
+ "step": 7200
721
+ },
722
+ {
723
+ "epoch": 11.52,
724
+ "eval_loss": 0.43486374616622925,
725
+ "eval_runtime": 83.8141,
726
+ "eval_samples_per_second": 23.862,
727
+ "eval_wer": 0.25755867031845764,
728
+ "step": 7200
729
+ },
730
+ {
731
+ "epoch": 11.68,
732
+ "learning_rate": 0.00018770491803278688,
733
+ "loss": 0.0787,
734
+ "step": 7300
735
+ },
736
+ {
737
+ "epoch": 11.84,
738
+ "learning_rate": 0.0001860655737704918,
739
+ "loss": 0.0809,
740
+ "step": 7400
741
+ },
742
+ {
743
+ "epoch": 11.84,
744
+ "eval_loss": 0.44311779737472534,
745
+ "eval_runtime": 85.6363,
746
+ "eval_samples_per_second": 23.355,
747
+ "eval_wer": 0.2628215455668331,
748
+ "step": 7400
749
+ },
750
+ {
751
+ "epoch": 12.0,
752
+ "learning_rate": 0.00018442622950819672,
753
+ "loss": 0.0778,
754
+ "step": 7500
755
+ },
756
+ {
757
+ "epoch": 12.16,
758
+ "learning_rate": 0.00018278688524590162,
759
+ "loss": 0.071,
760
+ "step": 7600
761
+ },
762
+ {
763
+ "epoch": 12.16,
764
+ "eval_loss": 0.47199195623397827,
765
+ "eval_runtime": 82.9761,
766
+ "eval_samples_per_second": 24.103,
767
+ "eval_wer": 0.2588475377262231,
768
+ "step": 7600
769
+ },
770
+ {
771
+ "epoch": 12.32,
772
+ "learning_rate": 0.00018114754098360654,
773
+ "loss": 0.074,
774
+ "step": 7700
775
+ },
776
+ {
777
+ "epoch": 12.48,
778
+ "learning_rate": 0.00017950819672131146,
779
+ "loss": 0.0677,
780
+ "step": 7800
781
+ },
782
+ {
783
+ "epoch": 12.48,
784
+ "eval_loss": 0.4387381672859192,
785
+ "eval_runtime": 83.0897,
786
+ "eval_samples_per_second": 24.07,
787
+ "eval_wer": 0.259867891090704,
788
+ "step": 7800
789
+ },
790
+ {
791
+ "epoch": 12.64,
792
+ "learning_rate": 0.00017786885245901638,
793
+ "loss": 0.0703,
794
+ "step": 7900
795
+ },
796
+ {
797
+ "epoch": 12.8,
798
+ "learning_rate": 0.0001762295081967213,
799
+ "loss": 0.0765,
800
+ "step": 8000
801
+ },
802
+ {
803
+ "epoch": 12.8,
804
+ "eval_loss": 0.4592679738998413,
805
+ "eval_runtime": 84.0398,
806
+ "eval_samples_per_second": 23.798,
807
+ "eval_wer": 0.2572364534665163,
808
+ "step": 8000
809
+ },
810
+ {
811
+ "epoch": 12.96,
812
+ "learning_rate": 0.0001745901639344262,
813
+ "loss": 0.0683,
814
+ "step": 8100
815
+ },
816
+ {
817
+ "epoch": 13.12,
818
+ "learning_rate": 0.00017295081967213112,
819
+ "loss": 0.0621,
820
+ "step": 8200
821
+ },
822
+ {
823
+ "epoch": 13.12,
824
+ "eval_loss": 0.4588267505168915,
825
+ "eval_runtime": 82.8224,
826
+ "eval_samples_per_second": 24.148,
827
+ "eval_wer": 0.2633585736534021,
828
+ "step": 8200
829
+ },
830
+ {
831
+ "epoch": 13.28,
832
+ "learning_rate": 0.00017131147540983604,
833
+ "loss": 0.065,
834
+ "step": 8300
835
+ },
836
+ {
837
+ "epoch": 13.44,
838
+ "learning_rate": 0.00016967213114754096,
839
+ "loss": 0.067,
840
+ "step": 8400
841
+ },
842
+ {
843
+ "epoch": 13.44,
844
+ "eval_loss": 0.4433707296848297,
845
+ "eval_runtime": 84.8017,
846
+ "eval_samples_per_second": 23.584,
847
+ "eval_wer": 0.25922345738682134,
848
+ "step": 8400
849
+ },
850
+ {
851
+ "epoch": 13.6,
852
+ "learning_rate": 0.00016803278688524588,
853
+ "loss": 0.067,
854
+ "step": 8500
855
+ },
856
+ {
857
+ "epoch": 13.76,
858
+ "learning_rate": 0.0001663934426229508,
859
+ "loss": 0.0633,
860
+ "step": 8600
861
+ },
862
+ {
863
+ "epoch": 13.76,
864
+ "eval_loss": 0.4525003433227539,
865
+ "eval_runtime": 82.3869,
866
+ "eval_samples_per_second": 24.276,
867
+ "eval_wer": 0.2647011438698244,
868
+ "step": 8600
869
+ },
870
+ {
871
+ "epoch": 13.92,
872
+ "learning_rate": 0.00016475409836065575,
873
+ "loss": 0.0704,
874
+ "step": 8700
875
+ },
876
+ {
877
+ "epoch": 14.08,
878
+ "learning_rate": 0.00016311475409836064,
879
+ "loss": 0.0635,
880
+ "step": 8800
881
+ },
882
+ {
883
+ "epoch": 14.08,
884
+ "eval_loss": 0.46746838092803955,
885
+ "eval_runtime": 82.2147,
886
+ "eval_samples_per_second": 24.327,
887
+ "eval_wer": 0.2580419955963697,
888
+ "step": 8800
889
+ },
890
+ {
891
+ "epoch": 14.24,
892
+ "learning_rate": 0.00016147540983606556,
893
+ "loss": 0.0568,
894
+ "step": 8900
895
+ },
896
+ {
897
+ "epoch": 14.4,
898
+ "learning_rate": 0.00015983606557377049,
899
+ "loss": 0.06,
900
+ "step": 9000
901
+ },
902
+ {
903
+ "epoch": 14.4,
904
+ "eval_loss": 0.46421578526496887,
905
+ "eval_runtime": 84.6457,
906
+ "eval_samples_per_second": 23.628,
907
+ "eval_wer": 0.25841791525696794,
908
+ "step": 9000
909
+ },
910
+ {
911
+ "epoch": 14.56,
912
+ "learning_rate": 0.0001581967213114754,
913
+ "loss": 0.0588,
914
+ "step": 9100
915
+ },
916
+ {
917
+ "epoch": 14.72,
918
+ "learning_rate": 0.00015655737704918033,
919
+ "loss": 0.0618,
920
+ "step": 9200
921
+ },
922
+ {
923
+ "epoch": 14.72,
924
+ "eval_loss": 0.465658038854599,
925
+ "eval_runtime": 83.6497,
926
+ "eval_samples_per_second": 23.909,
927
+ "eval_wer": 0.2606734332205574,
928
+ "step": 9200
929
+ },
930
+ {
931
+ "epoch": 14.88,
932
+ "learning_rate": 0.00015491803278688525,
933
+ "loss": 0.0578,
934
+ "step": 9300
935
+ },
936
+ {
937
+ "epoch": 15.04,
938
+ "learning_rate": 0.00015327868852459014,
939
+ "loss": 0.0562,
940
+ "step": 9400
941
+ },
942
+ {
943
+ "epoch": 15.04,
944
+ "eval_loss": 0.4753575325012207,
945
+ "eval_runtime": 82.0241,
946
+ "eval_samples_per_second": 24.383,
947
+ "eval_wer": 0.25831050963965413,
948
+ "step": 9400
949
+ },
950
+ {
951
+ "epoch": 15.2,
952
+ "learning_rate": 0.00015163934426229507,
953
+ "loss": 0.0489,
954
+ "step": 9500
955
+ },
956
+ {
957
+ "epoch": 15.36,
958
+ "learning_rate": 0.00015,
959
+ "loss": 0.0528,
960
+ "step": 9600
961
+ },
962
+ {
963
+ "epoch": 15.36,
964
+ "eval_loss": 0.48737889528274536,
965
+ "eval_runtime": 85.0733,
966
+ "eval_samples_per_second": 23.509,
967
+ "eval_wer": 0.2623382202889211,
968
+ "step": 9600
969
+ },
970
+ {
971
+ "epoch": 15.52,
972
+ "learning_rate": 0.0001483606557377049,
973
+ "loss": 0.0522,
974
+ "step": 9700
975
+ },
976
+ {
977
+ "epoch": 15.68,
978
+ "learning_rate": 0.00014672131147540983,
979
+ "loss": 0.0528,
980
+ "step": 9800
981
+ },
982
+ {
983
+ "epoch": 15.68,
984
+ "eval_loss": 0.47928667068481445,
985
+ "eval_runtime": 84.0501,
986
+ "eval_samples_per_second": 23.795,
987
+ "eval_wer": 0.26206970624563664,
988
+ "step": 9800
989
+ },
990
+ {
991
+ "epoch": 15.84,
992
+ "learning_rate": 0.00014508196721311472,
993
+ "loss": 0.0503,
994
+ "step": 9900
995
+ },
996
+ {
997
+ "epoch": 16.0,
998
+ "learning_rate": 0.00014344262295081964,
999
+ "loss": 0.0507,
1000
+ "step": 10000
1001
+ },
1002
+ {
1003
+ "epoch": 16.0,
1004
+ "eval_loss": 0.46980950236320496,
1005
+ "eval_runtime": 82.8709,
1006
+ "eval_samples_per_second": 24.134,
1007
+ "eval_wer": 0.24944954621126686,
1008
+ "step": 10000
1009
+ }
1010
+ ],
1011
+ "max_steps": 18750,
1012
+ "num_train_epochs": 30,
1013
+ "total_flos": 7.756303740936756e+19,
1014
+ "trial_name": null,
1015
+ "trial_params": null
1016
+ }
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cfe564fad0e81969d7ce7124ee6b55472766e7da3d46a2de73dc20a62151ed22
3
  size 2351
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9f1cf07032e95b729864528346e3e38cb097599f2b03403cc89e7a553263ec2
3
  size 2351
vocab.json CHANGED
@@ -1 +1 @@
1
- {"ँ": 1, "ं": 2, "ः": 3, "अ": 4, "आ": 5, "इ": 6, "ई": 7, "उ": 8, "ऊ": 9, "ऋ": 10, "ऍ": 11, "ए": 12, "ऐ": 13, "ऑ": 14, "ओ": 15, "औ": 16, "क": 17, "ख": 18, "ग": 19, "घ": 20, "च": 21, "छ": 22, "ज": 23, "झ": 24, "ञ": 25, "ट": 26, "ठ": 27, "ड": 28, "ढ": 29, "ण": 30, "त": 31, "थ": 32, "द": 33, "ध": 34, "न": 35, "प": 36, "फ": 37, "ब": 38, "भ": 39, "म": 40, "य": 41, "र": 42, "ल": 43, "ळ": 44, "व": 45, "श": 46, "ष": 47, "स": 48, "ह": 49, "़": 50, "ा": 51, "ि": 52, "ी": 53, "ु": 54, "ू": 55, "ृ": 56, "ॅ": 57, "े": 58, "ै": 59, "ॉ": 60, "ो": 61, "ौ": 62, "्": 63, "": 64, "": 65, "": 66, "": 67, "": 68, "": 69, "": 70, "|": 0, "[UNK]": 71, "[PAD]": 72}
1
+ {"ँ": 1, "ं": 2, "ः": 3, "अ": 4, "आ": 5, "इ": 6, "ई": 7, "उ": 8, "ऊ": 9, "ऋ": 10, "ऍ": 11, "ए": 12, "ऐ": 13, "ऑ": 14, "ओ": 15, "औ": 16, "क": 17, "ख": 18, "ग": 19, "घ": 20, "च": 21, "छ": 22, "ज": 23, "झ": 24, "ञ": 25, "ट": 26, "ठ": 27, "ड": 28, "ढ": 29, "ण": 30, "त": 31, "थ": 32, "द": 33, "ध": 34, "न": 35, "प": 36, "फ": 37, "ब": 38, "भ": 39, "म": 40, "य": 41, "र": 42, "ल": 43, "ळ": 44, "व": 45, "श": 46, "ष": 47, "स": 48, "ह": 49, "़": 50, "ा": 51, "ि": 52, "ी": 53, "ु": 54, "ू": 55, "ृ": 56, "ॅ": 57, "े": 58, "ै": 59, "ॉ": 60, "ो": 61, "ौ": 62, "्": 63, "": 64, "": 65, "": 66, "": 67, "": 68, "": 69, "": 70, "ॠ": 71, "|": 0, "[UNK]": 72, "[PAD]": 73}