skylord commited on
Commit
7272e79
1 Parent(s): 3ce1698

Add model files- updated model

Browse files
all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 5.0,
3
+ "init_mem_cpu_alloc_delta": 64150,
4
+ "init_mem_cpu_peaked_delta": 18306,
5
+ "init_mem_gpu_alloc_delta": 1261972480,
6
+ "init_mem_gpu_peaked_delta": 0,
7
+ "total_flos": 3.7562509460378655e+18,
8
+ "train_mem_cpu_alloc_delta": 0,
9
+ "train_mem_cpu_peaked_delta": 0,
10
+ "train_mem_gpu_alloc_delta": 5077884928,
11
+ "train_mem_gpu_peaked_delta": 0,
12
+ "train_runtime": 11825.7565,
13
+ "train_samples": 3717,
14
+ "train_samples_per_second": 0.049
15
+ }
config.json CHANGED
@@ -1,6 +1,6 @@
1
  {
2
- "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
3
- "activation_dropout": 0.1,
4
  "apply_spec_augment": true,
5
  "architectures": [
6
  "Wav2Vec2ForCTC"
1
  {
2
+ "_name_or_path": "/workspace/data2/el/wav2vec2-large-xlsr-greek/checkpoint-580",
3
+ "activation_dropout": 0.0,
4
  "apply_spec_augment": true,
5
  "architectures": [
6
  "Wav2Vec2ForCTC"
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:42a9506ca95227e7297c23b7ce1eeb2aafe115676f6510d9aa027b7b51f9fb64
3
  size 1262151127
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1c87d1a701a371a56922907aee8f97385396ccc8fbd9a3b5040ea1b2ba79367
3
  size 1262151127
train_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 5.0,
3
+ "init_mem_cpu_alloc_delta": 64150,
4
+ "init_mem_cpu_peaked_delta": 18306,
5
+ "init_mem_gpu_alloc_delta": 1261972480,
6
+ "init_mem_gpu_peaked_delta": 0,
7
+ "total_flos": 3.7562509460378655e+18,
8
+ "train_mem_cpu_alloc_delta": 0,
9
+ "train_mem_cpu_peaked_delta": 0,
10
+ "train_mem_gpu_alloc_delta": 5077884928,
11
+ "train_mem_gpu_peaked_delta": 0,
12
+ "train_runtime": 11825.7565,
13
+ "train_samples": 3717,
14
+ "train_samples_per_second": 0.049
15
+ }
trainer_state.json ADDED
@@ -0,0 +1,411 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.687998384980317,
3
+ "best_model_checkpoint": "/workspace/data2/el/wav2vec2-large-xlsr-greek/checkpoint-580",
4
+ "epoch": 4.9957081545064375,
5
+ "global_step": 580,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.09,
12
+ "learning_rate": 0.00015,
13
+ "loss": 18.1246,
14
+ "step": 10
15
+ },
16
+ {
17
+ "epoch": 0.17,
18
+ "learning_rate": 0.0003,
19
+ "loss": 16.1703,
20
+ "step": 20
21
+ },
22
+ {
23
+ "epoch": 0.26,
24
+ "learning_rate": 0.0002946428571428571,
25
+ "loss": 6.2477,
26
+ "step": 30
27
+ },
28
+ {
29
+ "epoch": 0.34,
30
+ "learning_rate": 0.00028928571428571425,
31
+ "loss": 3.5827,
32
+ "step": 40
33
+ },
34
+ {
35
+ "epoch": 0.43,
36
+ "learning_rate": 0.00028392857142857137,
37
+ "loss": 3.2333,
38
+ "step": 50
39
+ },
40
+ {
41
+ "epoch": 0.52,
42
+ "learning_rate": 0.00027857142857142854,
43
+ "loss": 3.1704,
44
+ "step": 60
45
+ },
46
+ {
47
+ "epoch": 0.6,
48
+ "learning_rate": 0.0002732142857142857,
49
+ "loss": 3.1395,
50
+ "step": 70
51
+ },
52
+ {
53
+ "epoch": 0.69,
54
+ "learning_rate": 0.00026785714285714287,
55
+ "loss": 3.1397,
56
+ "step": 80
57
+ },
58
+ {
59
+ "epoch": 0.77,
60
+ "learning_rate": 0.0002625,
61
+ "loss": 3.1298,
62
+ "step": 90
63
+ },
64
+ {
65
+ "epoch": 0.86,
66
+ "learning_rate": 0.0002571428571428571,
67
+ "loss": 3.1082,
68
+ "step": 100
69
+ },
70
+ {
71
+ "epoch": 0.94,
72
+ "learning_rate": 0.00025178571428571426,
73
+ "loss": 3.0917,
74
+ "step": 110
75
+ },
76
+ {
77
+ "epoch": 1.0,
78
+ "eval_loss": 3.068535089492798,
79
+ "eval_runtime": 645.4744,
80
+ "eval_samples_per_second": 2.358,
81
+ "eval_wer": 1.0,
82
+ "step": 116
83
+ },
84
+ {
85
+ "epoch": 1.03,
86
+ "learning_rate": 0.0002464285714285714,
87
+ "loss": 3.2085,
88
+ "step": 120
89
+ },
90
+ {
91
+ "epoch": 1.12,
92
+ "learning_rate": 0.00024107142857142857,
93
+ "loss": 3.0118,
94
+ "step": 130
95
+ },
96
+ {
97
+ "epoch": 1.21,
98
+ "learning_rate": 0.00023571428571428569,
99
+ "loss": 3.0139,
100
+ "step": 140
101
+ },
102
+ {
103
+ "epoch": 1.29,
104
+ "learning_rate": 0.00023035714285714285,
105
+ "loss": 3.0176,
106
+ "step": 150
107
+ },
108
+ {
109
+ "epoch": 1.38,
110
+ "learning_rate": 0.000225,
111
+ "loss": 2.9979,
112
+ "step": 160
113
+ },
114
+ {
115
+ "epoch": 1.46,
116
+ "learning_rate": 0.0002196428571428571,
117
+ "loss": 2.9721,
118
+ "step": 170
119
+ },
120
+ {
121
+ "epoch": 1.55,
122
+ "learning_rate": 0.00021428571428571427,
123
+ "loss": 2.9585,
124
+ "step": 180
125
+ },
126
+ {
127
+ "epoch": 1.64,
128
+ "learning_rate": 0.0002089285714285714,
129
+ "loss": 2.9578,
130
+ "step": 190
131
+ },
132
+ {
133
+ "epoch": 1.72,
134
+ "learning_rate": 0.00020357142857142856,
135
+ "loss": 2.9347,
136
+ "step": 200
137
+ },
138
+ {
139
+ "epoch": 1.81,
140
+ "learning_rate": 0.0001982142857142857,
141
+ "loss": 2.9348,
142
+ "step": 210
143
+ },
144
+ {
145
+ "epoch": 1.89,
146
+ "learning_rate": 0.00019285714285714286,
147
+ "loss": 2.9156,
148
+ "step": 220
149
+ },
150
+ {
151
+ "epoch": 1.98,
152
+ "learning_rate": 0.00018749999999999998,
153
+ "loss": 2.8911,
154
+ "step": 230
155
+ },
156
+ {
157
+ "epoch": 2.0,
158
+ "eval_loss": 2.9123377799987793,
159
+ "eval_runtime": 651.1025,
160
+ "eval_samples_per_second": 2.338,
161
+ "eval_wer": 1.0,
162
+ "step": 232
163
+ },
164
+ {
165
+ "epoch": 2.07,
166
+ "learning_rate": 0.00018214285714285712,
167
+ "loss": 3.002,
168
+ "step": 240
169
+ },
170
+ {
171
+ "epoch": 2.15,
172
+ "learning_rate": 0.00017678571428571428,
173
+ "loss": 2.7696,
174
+ "step": 250
175
+ },
176
+ {
177
+ "epoch": 2.24,
178
+ "learning_rate": 0.0001714285714285714,
179
+ "loss": 2.6308,
180
+ "step": 260
181
+ },
182
+ {
183
+ "epoch": 2.33,
184
+ "learning_rate": 0.00016607142857142857,
185
+ "loss": 2.3949,
186
+ "step": 270
187
+ },
188
+ {
189
+ "epoch": 2.41,
190
+ "learning_rate": 0.0001607142857142857,
191
+ "loss": 2.0719,
192
+ "step": 280
193
+ },
194
+ {
195
+ "epoch": 2.5,
196
+ "learning_rate": 0.00015535714285714285,
197
+ "loss": 1.7915,
198
+ "step": 290
199
+ },
200
+ {
201
+ "epoch": 2.58,
202
+ "learning_rate": 0.00015,
203
+ "loss": 1.5479,
204
+ "step": 300
205
+ },
206
+ {
207
+ "epoch": 2.67,
208
+ "learning_rate": 0.00014464285714285713,
209
+ "loss": 1.4167,
210
+ "step": 310
211
+ },
212
+ {
213
+ "epoch": 2.76,
214
+ "learning_rate": 0.00013928571428571427,
215
+ "loss": 1.2183,
216
+ "step": 320
217
+ },
218
+ {
219
+ "epoch": 2.84,
220
+ "learning_rate": 0.00013392857142857144,
221
+ "loss": 1.1713,
222
+ "step": 330
223
+ },
224
+ {
225
+ "epoch": 2.93,
226
+ "learning_rate": 0.00012857142857142855,
227
+ "loss": 1.0325,
228
+ "step": 340
229
+ },
230
+ {
231
+ "epoch": 3.0,
232
+ "eval_loss": 0.8776374459266663,
233
+ "eval_runtime": 651.452,
234
+ "eval_samples_per_second": 2.336,
235
+ "eval_wer": 0.8105380034319168,
236
+ "step": 348
237
+ },
238
+ {
239
+ "epoch": 3.02,
240
+ "learning_rate": 0.0001232142857142857,
241
+ "loss": 0.9978,
242
+ "step": 350
243
+ },
244
+ {
245
+ "epoch": 3.1,
246
+ "learning_rate": 0.00011785714285714284,
247
+ "loss": 0.8941,
248
+ "step": 360
249
+ },
250
+ {
251
+ "epoch": 3.19,
252
+ "learning_rate": 0.0001125,
253
+ "loss": 0.875,
254
+ "step": 370
255
+ },
256
+ {
257
+ "epoch": 3.27,
258
+ "learning_rate": 0.00010714285714285714,
259
+ "loss": 0.801,
260
+ "step": 380
261
+ },
262
+ {
263
+ "epoch": 3.36,
264
+ "learning_rate": 0.00010178571428571428,
265
+ "loss": 0.7847,
266
+ "step": 390
267
+ },
268
+ {
269
+ "epoch": 3.45,
270
+ "learning_rate": 9.642857142857143e-05,
271
+ "loss": 0.6949,
272
+ "step": 400
273
+ },
274
+ {
275
+ "epoch": 3.53,
276
+ "learning_rate": 9.107142857142856e-05,
277
+ "loss": 0.7307,
278
+ "step": 410
279
+ },
280
+ {
281
+ "epoch": 3.62,
282
+ "learning_rate": 8.57142857142857e-05,
283
+ "loss": 0.6729,
284
+ "step": 420
285
+ },
286
+ {
287
+ "epoch": 3.7,
288
+ "learning_rate": 8.035714285714285e-05,
289
+ "loss": 0.6699,
290
+ "step": 430
291
+ },
292
+ {
293
+ "epoch": 3.79,
294
+ "learning_rate": 7.5e-05,
295
+ "loss": 0.6768,
296
+ "step": 440
297
+ },
298
+ {
299
+ "epoch": 3.88,
300
+ "learning_rate": 6.964285714285713e-05,
301
+ "loss": 0.6459,
302
+ "step": 450
303
+ },
304
+ {
305
+ "epoch": 3.96,
306
+ "learning_rate": 6.428571428571427e-05,
307
+ "loss": 0.6589,
308
+ "step": 460
309
+ },
310
+ {
311
+ "epoch": 4.0,
312
+ "eval_loss": 0.6442593336105347,
313
+ "eval_runtime": 653.4327,
314
+ "eval_samples_per_second": 2.329,
315
+ "eval_wer": 0.7133340062582013,
316
+ "step": 464
317
+ },
318
+ {
319
+ "epoch": 4.05,
320
+ "learning_rate": 5.892857142857142e-05,
321
+ "loss": 0.6647,
322
+ "step": 470
323
+ },
324
+ {
325
+ "epoch": 4.14,
326
+ "learning_rate": 5.357142857142857e-05,
327
+ "loss": 0.6859,
328
+ "step": 480
329
+ },
330
+ {
331
+ "epoch": 4.22,
332
+ "learning_rate": 4.8214285714285716e-05,
333
+ "loss": 0.6298,
334
+ "step": 490
335
+ },
336
+ {
337
+ "epoch": 4.31,
338
+ "learning_rate": 4.285714285714285e-05,
339
+ "loss": 0.6076,
340
+ "step": 500
341
+ },
342
+ {
343
+ "epoch": 4.39,
344
+ "learning_rate": 3.75e-05,
345
+ "loss": 0.5485,
346
+ "step": 510
347
+ },
348
+ {
349
+ "epoch": 4.48,
350
+ "learning_rate": 3.214285714285714e-05,
351
+ "loss": 0.5852,
352
+ "step": 520
353
+ },
354
+ {
355
+ "epoch": 4.57,
356
+ "learning_rate": 2.6785714285714284e-05,
357
+ "loss": 0.5452,
358
+ "step": 530
359
+ },
360
+ {
361
+ "epoch": 4.65,
362
+ "learning_rate": 2.1428571428571425e-05,
363
+ "loss": 0.5593,
364
+ "step": 540
365
+ },
366
+ {
367
+ "epoch": 4.74,
368
+ "learning_rate": 1.607142857142857e-05,
369
+ "loss": 0.5316,
370
+ "step": 550
371
+ },
372
+ {
373
+ "epoch": 4.82,
374
+ "learning_rate": 1.0714285714285712e-05,
375
+ "loss": 0.5742,
376
+ "step": 560
377
+ },
378
+ {
379
+ "epoch": 4.91,
380
+ "learning_rate": 5.357142857142856e-06,
381
+ "loss": 0.5978,
382
+ "step": 570
383
+ },
384
+ {
385
+ "epoch": 5.0,
386
+ "learning_rate": 0.0,
387
+ "loss": 0.5398,
388
+ "step": 580
389
+ },
390
+ {
391
+ "epoch": 5.0,
392
+ "eval_loss": 0.6011346578598022,
393
+ "eval_runtime": 660.5778,
394
+ "eval_samples_per_second": 2.304,
395
+ "eval_wer": 0.687998384980317,
396
+ "step": 580
397
+ },
398
+ {
399
+ "epoch": 5.0,
400
+ "step": 580,
401
+ "total_flos": 3.7562509460378655e+18,
402
+ "train_runtime": 11825.7565,
403
+ "train_samples_per_second": 0.049
404
+ }
405
+ ],
406
+ "max_steps": 580,
407
+ "num_train_epochs": 5,
408
+ "total_flos": 3.7562509460378655e+18,
409
+ "trial_name": null,
410
+ "trial_params": null
411
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c64942f25d6a43c19ab5eca168eac25bb89dcdc54eca615e76bbf76a4b47494
3
+ size 2351
vocab.json CHANGED
@@ -1 +1 @@
1
- {"β": 0, "ψ": 1, "": 2, "κ": 3, "δ": 4, "σ": 5, "ό": 6, "λ": 7, "η": 8, "χ": 9, "υ": 10, "n": 11, "έ": 12, "φ": 13, "ί": 14, "ξ": 15, "ν": 16, "ϊ": 17, "τ": 18, "v": 19, "": 20, "α": 21, "g": 22, "m": 23, "ρ": 24, "π": 25, "ι": 26, "ζ": 27, "e": 28, "o": 29, "a": 30, "́": 31, "´": 32, "ΐ": 33, "h": 34, "»": 35, "ώ": 36, "θ": 37, "«": 38, "ε": 39, "μ": 40, "ή": 41, "ο": 42, "r": 44, "ά": 45, "ω": 46, "ύ": 47, "ς": 48, "γ": 49, "ϋ": 50, "|": 43, "[UNK]": 51, "[PAD]": 52}
1
+ {"σ": 0, "δ": 1, "γ": 2, "«": 3, "ώ": 4, "θ": 5, "r": 6, "τ": 7, "ή": 8, "»": 9, "ι": 10, "ΐ": 11, "φ": 12, "m": 13, "h": 14, "ί": 15, "π": 16, "n": 17, "´": 18, "λ": 19, "η": 20, "ψ": 21, "e": 22, "ζ": 23, "ό": 25, "ρ": 26, "β": 27, "α": 28, "ν": 29, "ϋ": 30, "v": 31, "ε": 32, "ά": 33, "g": 34, "ύ": 35, "κ": 36, "μ": 37, "έ": 38, "ς": 39, "ξ": 40, "ϊ": 41, "χ": 42, "o": 43, "ο": 44, "́": 45, "": 46, "'": 47, "a": 48, "ω": 49, "υ": 50, "|": 24, "[UNK]": 51, "[PAD]": 52}