Hugging Face's logo

speech31
/
XLS-R-300m-korean-ipa2

Automatic Speech Recognition
Transformers
PyTorch
wav2vec2
Inference Endpoints
Model card Files Community
speech31 commited on
Commit
50c424c
•
1 Parent(s): 984d6c7

upload model

Browse files
Files changed (12) hide show
  1. config.json +79 -0
  2. optimizer.pt +3 -0
  3. preprocessor_config.json +9 -0
  4. pytorch_model.bin +3 -0
  5. rng_state.pth +3 -0
  6. scaler.pt +3 -0
  7. scheduler.pt +3 -0
  8. special_tokens_map.json +1 -0
  9. tokenizer_config.json +1 -0
  10. trainer_state.json +640 -0
  11. training_args.bin +3 -0
  12. vocab.json +1 -0
config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-xls-r-300m",
3
+ "activation_dropout": 0.0,
4
+ "apply_spec_augment": true,
5
+ "architectures": [
6
+ "Wav2Vec2ForCTC"
7
+ ],
8
+ "attention_dropout": 0.0,
9
+ "bos_token_id": 1,
10
+ "classifier_proj_size": 256,
11
+ "codevector_dim": 768,
12
+ "contrastive_logits_temperature": 0.1,
13
+ "conv_bias": true,
14
+ "conv_dim": [
15
+ 512,
16
+ 512,
17
+ 512,
18
+ 512,
19
+ 512,
20
+ 512,
21
+ 512
22
+ ],
23
+ "conv_kernel": [
24
+ 10,
25
+ 3,
26
+ 3,
27
+ 3,
28
+ 3,
29
+ 2,
30
+ 2
31
+ ],
32
+ "conv_stride": [
33
+ 5,
34
+ 2,
35
+ 2,
36
+ 2,
37
+ 2,
38
+ 2,
39
+ 2
40
+ ],
41
+ "ctc_loss_reduction": "mean",
42
+ "ctc_zero_infinity": false,
43
+ "diversity_loss_weight": 0.1,
44
+ "do_stable_layer_norm": true,
45
+ "eos_token_id": 2,
46
+ "feat_extract_activation": "gelu",
47
+ "feat_extract_dropout": 0.0,
48
+ "feat_extract_norm": "layer",
49
+ "feat_proj_dropout": 0.0,
50
+ "feat_quantizer_dropout": 0.0,
51
+ "final_dropout": 0.0,
52
+ "gradient_checkpointing": false,
53
+ "hidden_act": "gelu",
54
+ "hidden_dropout": 0.0,
55
+ "hidden_size": 1024,
56
+ "initializer_range": 0.02,
57
+ "intermediate_size": 4096,
58
+ "layer_norm_eps": 1e-05,
59
+ "layerdrop": 0.0,
60
+ "mask_feature_length": 10,
61
+ "mask_feature_prob": 0.0,
62
+ "mask_time_length": 10,
63
+ "mask_time_prob": 0.05,
64
+ "model_type": "wav2vec2",
65
+ "num_attention_heads": 16,
66
+ "num_codevector_groups": 2,
67
+ "num_codevectors_per_group": 320,
68
+ "num_conv_pos_embedding_groups": 16,
69
+ "num_conv_pos_embeddings": 128,
70
+ "num_feat_extract_layers": 7,
71
+ "num_hidden_layers": 24,
72
+ "num_negatives": 100,
73
+ "pad_token_id": 46,
74
+ "proj_codevector_dim": 768,
75
+ "torch_dtype": "float32",
76
+ "transformers_version": "4.11.3",
77
+ "use_weighted_layer_sum": false,
78
+ "vocab_size": 47
79
+ }
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8ff0e526e18ad0c84e09193d837483bd143d07c92f345642e9abc797b951672
3
+ size 2490444305
preprocessor_config.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": 0.0,
7
+ "return_attention_mask": true,
8
+ "sampling_rate": 16000
9
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14bf7370fd97822b295c4f14dff71b07608ba06255bc54dd82eb68878f99607f
3
+ size 1262116337
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8605a5a54219f9ade287f38a8e6f5b27551b139b40122ff0a7049d9b4413f1ba
3
+ size 14567
scaler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:584a9d0b409b9b55bbeb95bba851d1cee0a3e95320365da04eb6e74d6482d859
3
+ size 559
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76334fd9a6ef2b5044231db8387512e2d806f167f7b7970220348cfe5c08de36
3
+ size 623
special_tokens_map.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]"}
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": "|", "tokenizer_class": "Wav2Vec2CTCTokenizer"}
trainer_state.json ADDED
@@ -0,0 +1,640 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.09535224735736847,
3
+ "best_model_checkpoint": "XLS-R-300m-korean-ipa_v2/checkpoint-8000",
4
+ "epoch": 12.778930566640064,
5
+ "global_step": 8000,
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": 2.9999999999999997e-05,
13
+ "loss": 5.6413,
14
+ "step": 100
15
+ },
16
+ {
17
+ "epoch": 0.32,
18
+ "learning_rate": 5.9999999999999995e-05,
19
+ "loss": 2.9849,
20
+ "step": 200
21
+ },
22
+ {
23
+ "epoch": 0.48,
24
+ "learning_rate": 8.999999999999999e-05,
25
+ "loss": 2.6488,
26
+ "step": 300
27
+ },
28
+ {
29
+ "epoch": 0.64,
30
+ "learning_rate": 0.00011999999999999999,
31
+ "loss": 2.1382,
32
+ "step": 400
33
+ },
34
+ {
35
+ "epoch": 0.8,
36
+ "learning_rate": 0.00015,
37
+ "loss": 0.8032,
38
+ "step": 500
39
+ },
40
+ {
41
+ "epoch": 0.8,
42
+ "eval_cer": 0.28992062691770265,
43
+ "eval_loss": 0.5041582584381104,
44
+ "eval_runtime": 249.7666,
45
+ "eval_samples_per_second": 8.916,
46
+ "eval_steps_per_second": 1.117,
47
+ "step": 500
48
+ },
49
+ {
50
+ "epoch": 0.96,
51
+ "learning_rate": 0.00017999999999999998,
52
+ "loss": 0.4962,
53
+ "step": 600
54
+ },
55
+ {
56
+ "epoch": 1.12,
57
+ "learning_rate": 0.00020999999999999998,
58
+ "loss": 0.4136,
59
+ "step": 700
60
+ },
61
+ {
62
+ "epoch": 1.28,
63
+ "learning_rate": 0.00023999999999999998,
64
+ "loss": 0.3741,
65
+ "step": 800
66
+ },
67
+ {
68
+ "epoch": 1.44,
69
+ "learning_rate": 0.00027,
70
+ "loss": 0.3477,
71
+ "step": 900
72
+ },
73
+ {
74
+ "epoch": 1.6,
75
+ "learning_rate": 0.0003,
76
+ "loss": 0.326,
77
+ "step": 1000
78
+ },
79
+ {
80
+ "epoch": 1.6,
81
+ "eval_cer": 0.17689681499301343,
82
+ "eval_loss": 0.26741623878479004,
83
+ "eval_runtime": 208.1626,
84
+ "eval_samples_per_second": 10.698,
85
+ "eval_steps_per_second": 1.34,
86
+ "step": 1000
87
+ },
88
+ {
89
+ "epoch": 1.76,
90
+ "learning_rate": 0.0002973958333333333,
91
+ "loss": 0.3141,
92
+ "step": 1100
93
+ },
94
+ {
95
+ "epoch": 1.92,
96
+ "learning_rate": 0.0002947916666666666,
97
+ "loss": 0.3045,
98
+ "step": 1200
99
+ },
100
+ {
101
+ "epoch": 2.08,
102
+ "learning_rate": 0.00029218749999999997,
103
+ "loss": 0.2645,
104
+ "step": 1300
105
+ },
106
+ {
107
+ "epoch": 2.24,
108
+ "learning_rate": 0.0002895833333333333,
109
+ "loss": 0.2399,
110
+ "step": 1400
111
+ },
112
+ {
113
+ "epoch": 2.4,
114
+ "learning_rate": 0.0002869791666666666,
115
+ "loss": 0.2343,
116
+ "step": 1500
117
+ },
118
+ {
119
+ "epoch": 2.4,
120
+ "eval_cer": 0.12242325588095478,
121
+ "eval_loss": 0.18896125257015228,
122
+ "eval_runtime": 208.4383,
123
+ "eval_samples_per_second": 10.684,
124
+ "eval_steps_per_second": 1.339,
125
+ "step": 1500
126
+ },
127
+ {
128
+ "epoch": 2.56,
129
+ "learning_rate": 0.00028437499999999996,
130
+ "loss": 0.2198,
131
+ "step": 1600
132
+ },
133
+ {
134
+ "epoch": 2.72,
135
+ "learning_rate": 0.0002817708333333333,
136
+ "loss": 0.2218,
137
+ "step": 1700
138
+ },
139
+ {
140
+ "epoch": 2.87,
141
+ "learning_rate": 0.00027916666666666666,
142
+ "loss": 0.2139,
143
+ "step": 1800
144
+ },
145
+ {
146
+ "epoch": 3.04,
147
+ "learning_rate": 0.00027656249999999995,
148
+ "loss": 0.2009,
149
+ "step": 1900
150
+ },
151
+ {
152
+ "epoch": 3.19,
153
+ "learning_rate": 0.0002739583333333333,
154
+ "loss": 0.1696,
155
+ "step": 2000
156
+ },
157
+ {
158
+ "epoch": 3.19,
159
+ "eval_cer": 0.10347311254843775,
160
+ "eval_loss": 0.16067682206630707,
161
+ "eval_runtime": 204.5024,
162
+ "eval_samples_per_second": 10.89,
163
+ "eval_steps_per_second": 1.364,
164
+ "step": 2000
165
+ },
166
+ {
167
+ "epoch": 3.35,
168
+ "learning_rate": 0.00027135416666666665,
169
+ "loss": 0.1707,
170
+ "step": 2100
171
+ },
172
+ {
173
+ "epoch": 3.51,
174
+ "learning_rate": 0.00026875,
175
+ "loss": 0.1663,
176
+ "step": 2200
177
+ },
178
+ {
179
+ "epoch": 3.67,
180
+ "learning_rate": 0.0002661458333333333,
181
+ "loss": 0.1577,
182
+ "step": 2300
183
+ },
184
+ {
185
+ "epoch": 3.83,
186
+ "learning_rate": 0.00026354166666666664,
187
+ "loss": 0.1598,
188
+ "step": 2400
189
+ },
190
+ {
191
+ "epoch": 3.99,
192
+ "learning_rate": 0.00026093749999999994,
193
+ "loss": 0.1608,
194
+ "step": 2500
195
+ },
196
+ {
197
+ "epoch": 3.99,
198
+ "eval_cer": 0.0870294876042582,
199
+ "eval_loss": 0.14269110560417175,
200
+ "eval_runtime": 200.8042,
201
+ "eval_samples_per_second": 11.09,
202
+ "eval_steps_per_second": 1.389,
203
+ "step": 2500
204
+ },
205
+ {
206
+ "epoch": 4.15,
207
+ "learning_rate": 0.00025833333333333334,
208
+ "loss": 0.1371,
209
+ "step": 2600
210
+ },
211
+ {
212
+ "epoch": 4.31,
213
+ "learning_rate": 0.00025572916666666664,
214
+ "loss": 0.1334,
215
+ "step": 2700
216
+ },
217
+ {
218
+ "epoch": 4.47,
219
+ "learning_rate": 0.000253125,
220
+ "loss": 0.1318,
221
+ "step": 2800
222
+ },
223
+ {
224
+ "epoch": 4.63,
225
+ "learning_rate": 0.0002505208333333333,
226
+ "loss": 0.1305,
227
+ "step": 2900
228
+ },
229
+ {
230
+ "epoch": 4.79,
231
+ "learning_rate": 0.00024791666666666663,
232
+ "loss": 0.1275,
233
+ "step": 3000
234
+ },
235
+ {
236
+ "epoch": 4.79,
237
+ "eval_cer": 0.07236491450467451,
238
+ "eval_loss": 0.12700112164020538,
239
+ "eval_runtime": 194.3447,
240
+ "eval_samples_per_second": 11.459,
241
+ "eval_steps_per_second": 1.436,
242
+ "step": 3000
243
+ },
244
+ {
245
+ "epoch": 4.95,
246
+ "learning_rate": 0.0002453125,
247
+ "loss": 0.1329,
248
+ "step": 3100
249
+ },
250
+ {
251
+ "epoch": 5.11,
252
+ "learning_rate": 0.00024270833333333333,
253
+ "loss": 0.1233,
254
+ "step": 3200
255
+ },
256
+ {
257
+ "epoch": 5.27,
258
+ "learning_rate": 0.00024010416666666665,
259
+ "loss": 0.1117,
260
+ "step": 3300
261
+ },
262
+ {
263
+ "epoch": 5.43,
264
+ "learning_rate": 0.00023749999999999997,
265
+ "loss": 0.1176,
266
+ "step": 3400
267
+ },
268
+ {
269
+ "epoch": 5.59,
270
+ "learning_rate": 0.00023489583333333332,
271
+ "loss": 0.1107,
272
+ "step": 3500
273
+ },
274
+ {
275
+ "epoch": 5.59,
276
+ "eval_cer": 0.07061467321626644,
277
+ "eval_loss": 0.12084876000881195,
278
+ "eval_runtime": 194.5834,
279
+ "eval_samples_per_second": 11.445,
280
+ "eval_steps_per_second": 1.434,
281
+ "step": 3500
282
+ },
283
+ {
284
+ "epoch": 5.75,
285
+ "learning_rate": 0.00023229166666666664,
286
+ "loss": 0.1075,
287
+ "step": 3600
288
+ },
289
+ {
290
+ "epoch": 5.91,
291
+ "learning_rate": 0.0002296875,
292
+ "loss": 0.1108,
293
+ "step": 3700
294
+ },
295
+ {
296
+ "epoch": 6.07,
297
+ "learning_rate": 0.00022708333333333331,
298
+ "loss": 0.1066,
299
+ "step": 3800
300
+ },
301
+ {
302
+ "epoch": 6.23,
303
+ "learning_rate": 0.00022447916666666664,
304
+ "loss": 0.0922,
305
+ "step": 3900
306
+ },
307
+ {
308
+ "epoch": 6.39,
309
+ "learning_rate": 0.00022187499999999999,
310
+ "loss": 0.0935,
311
+ "step": 4000
312
+ },
313
+ {
314
+ "epoch": 6.39,
315
+ "eval_cer": 0.066206658119535,
316
+ "eval_loss": 0.12421450763940811,
317
+ "eval_runtime": 193.406,
318
+ "eval_samples_per_second": 11.515,
319
+ "eval_steps_per_second": 1.443,
320
+ "step": 4000
321
+ },
322
+ {
323
+ "epoch": 6.55,
324
+ "learning_rate": 0.0002192708333333333,
325
+ "loss": 0.1012,
326
+ "step": 4100
327
+ },
328
+ {
329
+ "epoch": 6.71,
330
+ "learning_rate": 0.00021666666666666666,
331
+ "loss": 0.1015,
332
+ "step": 4200
333
+ },
334
+ {
335
+ "epoch": 6.87,
336
+ "learning_rate": 0.00021406249999999998,
337
+ "loss": 0.1004,
338
+ "step": 4300
339
+ },
340
+ {
341
+ "epoch": 7.03,
342
+ "learning_rate": 0.00021145833333333333,
343
+ "loss": 0.0979,
344
+ "step": 4400
345
+ },
346
+ {
347
+ "epoch": 7.19,
348
+ "learning_rate": 0.00020885416666666665,
349
+ "loss": 0.0827,
350
+ "step": 4500
351
+ },
352
+ {
353
+ "epoch": 7.19,
354
+ "eval_cer": 0.05754908598510494,
355
+ "eval_loss": 0.1160559132695198,
356
+ "eval_runtime": 196.9923,
357
+ "eval_samples_per_second": 11.305,
358
+ "eval_steps_per_second": 1.416,
359
+ "step": 4500
360
+ },
361
+ {
362
+ "epoch": 7.35,
363
+ "learning_rate": 0.00020624999999999997,
364
+ "loss": 0.0851,
365
+ "step": 4600
366
+ },
367
+ {
368
+ "epoch": 7.51,
369
+ "learning_rate": 0.00020364583333333332,
370
+ "loss": 0.0842,
371
+ "step": 4700
372
+ },
373
+ {
374
+ "epoch": 7.67,
375
+ "learning_rate": 0.00020104166666666664,
376
+ "loss": 0.0846,
377
+ "step": 4800
378
+ },
379
+ {
380
+ "epoch": 7.83,
381
+ "learning_rate": 0.0001984375,
382
+ "loss": 0.0869,
383
+ "step": 4900
384
+ },
385
+ {
386
+ "epoch": 7.99,
387
+ "learning_rate": 0.00019583333333333331,
388
+ "loss": 0.0885,
389
+ "step": 5000
390
+ },
391
+ {
392
+ "epoch": 7.99,
393
+ "eval_cer": 0.053594837148331145,
394
+ "eval_loss": 0.10113094747066498,
395
+ "eval_runtime": 209.0359,
396
+ "eval_samples_per_second": 10.654,
397
+ "eval_steps_per_second": 1.335,
398
+ "step": 5000
399
+ },
400
+ {
401
+ "epoch": 8.15,
402
+ "learning_rate": 0.00019322916666666664,
403
+ "loss": 0.0755,
404
+ "step": 5100
405
+ },
406
+ {
407
+ "epoch": 8.31,
408
+ "learning_rate": 0.00019062499999999998,
409
+ "loss": 0.0754,
410
+ "step": 5200
411
+ },
412
+ {
413
+ "epoch": 8.47,
414
+ "learning_rate": 0.00018802083333333333,
415
+ "loss": 0.0766,
416
+ "step": 5300
417
+ },
418
+ {
419
+ "epoch": 8.63,
420
+ "learning_rate": 0.00018541666666666666,
421
+ "loss": 0.0747,
422
+ "step": 5400
423
+ },
424
+ {
425
+ "epoch": 8.79,
426
+ "learning_rate": 0.00018281249999999998,
427
+ "loss": 0.0744,
428
+ "step": 5500
429
+ },
430
+ {
431
+ "epoch": 8.79,
432
+ "eval_cer": 0.052132701421800945,
433
+ "eval_loss": 0.10384104400873184,
434
+ "eval_runtime": 213.8935,
435
+ "eval_samples_per_second": 10.412,
436
+ "eval_steps_per_second": 1.304,
437
+ "step": 5500
438
+ },
439
+ {
440
+ "epoch": 8.94,
441
+ "learning_rate": 0.0001802083333333333,
442
+ "loss": 0.076,
443
+ "step": 5600
444
+ },
445
+ {
446
+ "epoch": 9.11,
447
+ "learning_rate": 0.00017760416666666665,
448
+ "loss": 0.0723,
449
+ "step": 5700
450
+ },
451
+ {
452
+ "epoch": 9.26,
453
+ "learning_rate": 0.000175,
454
+ "loss": 0.0705,
455
+ "step": 5800
456
+ },
457
+ {
458
+ "epoch": 9.42,
459
+ "learning_rate": 0.00017239583333333332,
460
+ "loss": 0.0707,
461
+ "step": 5900
462
+ },
463
+ {
464
+ "epoch": 9.58,
465
+ "learning_rate": 0.00016979166666666664,
466
+ "loss": 0.0721,
467
+ "step": 6000
468
+ },
469
+ {
470
+ "epoch": 9.58,
471
+ "eval_cer": 0.05004393609818637,
472
+ "eval_loss": 0.10492419451475143,
473
+ "eval_runtime": 214.0314,
474
+ "eval_samples_per_second": 10.405,
475
+ "eval_steps_per_second": 1.304,
476
+ "step": 6000
477
+ },
478
+ {
479
+ "epoch": 9.74,
480
+ "learning_rate": 0.00016718749999999996,
481
+ "loss": 0.0684,
482
+ "step": 6100
483
+ },
484
+ {
485
+ "epoch": 9.9,
486
+ "learning_rate": 0.00016458333333333334,
487
+ "loss": 0.0652,
488
+ "step": 6200
489
+ },
490
+ {
491
+ "epoch": 10.06,
492
+ "learning_rate": 0.00016197916666666666,
493
+ "loss": 0.0656,
494
+ "step": 6300
495
+ },
496
+ {
497
+ "epoch": 10.22,
498
+ "learning_rate": 0.00015937499999999998,
499
+ "loss": 0.0631,
500
+ "step": 6400
501
+ },
502
+ {
503
+ "epoch": 10.38,
504
+ "learning_rate": 0.0001567708333333333,
505
+ "loss": 0.0646,
506
+ "step": 6500
507
+ },
508
+ {
509
+ "epoch": 10.38,
510
+ "eval_cer": 0.048113628833604635,
511
+ "eval_loss": 0.1038593202829361,
512
+ "eval_runtime": 212.7231,
513
+ "eval_samples_per_second": 10.469,
514
+ "eval_steps_per_second": 1.312,
515
+ "step": 6500
516
+ },
517
+ {
518
+ "epoch": 10.54,
519
+ "learning_rate": 0.00015416666666666663,
520
+ "loss": 0.0648,
521
+ "step": 6600
522
+ },
523
+ {
524
+ "epoch": 10.7,
525
+ "learning_rate": 0.0001515625,
526
+ "loss": 0.0612,
527
+ "step": 6700
528
+ },
529
+ {
530
+ "epoch": 10.86,
531
+ "learning_rate": 0.000148984375,
532
+ "loss": 0.0617,
533
+ "step": 6800
534
+ },
535
+ {
536
+ "epoch": 11.02,
537
+ "learning_rate": 0.00014638020833333332,
538
+ "loss": 0.0614,
539
+ "step": 6900
540
+ },
541
+ {
542
+ "epoch": 11.18,
543
+ "learning_rate": 0.00014377604166666665,
544
+ "loss": 0.0548,
545
+ "step": 7000
546
+ },
547
+ {
548
+ "epoch": 11.18,
549
+ "eval_cer": 0.04586640545095723,
550
+ "eval_loss": 0.09844804555177689,
551
+ "eval_runtime": 216.1335,
552
+ "eval_samples_per_second": 10.304,
553
+ "eval_steps_per_second": 1.291,
554
+ "step": 7000
555
+ },
556
+ {
557
+ "epoch": 11.34,
558
+ "learning_rate": 0.000141171875,
559
+ "loss": 0.0559,
560
+ "step": 7100
561
+ },
562
+ {
563
+ "epoch": 11.5,
564
+ "learning_rate": 0.00013856770833333332,
565
+ "loss": 0.0549,
566
+ "step": 7200
567
+ },
568
+ {
569
+ "epoch": 11.66,
570
+ "learning_rate": 0.00013596354166666667,
571
+ "loss": 0.0571,
572
+ "step": 7300
573
+ },
574
+ {
575
+ "epoch": 11.82,
576
+ "learning_rate": 0.000133359375,
577
+ "loss": 0.0589,
578
+ "step": 7400
579
+ },
580
+ {
581
+ "epoch": 11.98,
582
+ "learning_rate": 0.00013075520833333334,
583
+ "loss": 0.0537,
584
+ "step": 7500
585
+ },
586
+ {
587
+ "epoch": 11.98,
588
+ "eval_cer": 0.04361918206830983,
589
+ "eval_loss": 0.10360600054264069,
590
+ "eval_runtime": 277.1197,
591
+ "eval_samples_per_second": 8.036,
592
+ "eval_steps_per_second": 1.007,
593
+ "step": 7500
594
+ },
595
+ {
596
+ "epoch": 12.14,
597
+ "learning_rate": 0.00012815104166666666,
598
+ "loss": 0.052,
599
+ "step": 7600
600
+ },
601
+ {
602
+ "epoch": 12.3,
603
+ "learning_rate": 0.000125546875,
604
+ "loss": 0.051,
605
+ "step": 7700
606
+ },
607
+ {
608
+ "epoch": 12.46,
609
+ "learning_rate": 0.00012294270833333333,
610
+ "loss": 0.0505,
611
+ "step": 7800
612
+ },
613
+ {
614
+ "epoch": 12.62,
615
+ "learning_rate": 0.00012033854166666665,
616
+ "loss": 0.0507,
617
+ "step": 7900
618
+ },
619
+ {
620
+ "epoch": 12.78,
621
+ "learning_rate": 0.00011773437499999999,
622
+ "loss": 0.0491,
623
+ "step": 8000
624
+ },
625
+ {
626
+ "epoch": 12.78,
627
+ "eval_cer": 0.04207781731226321,
628
+ "eval_loss": 0.09535224735736847,
629
+ "eval_runtime": 213.4424,
630
+ "eval_samples_per_second": 10.434,
631
+ "eval_steps_per_second": 1.307,
632
+ "step": 8000
633
+ }
634
+ ],
635
+ "max_steps": 12520,
636
+ "num_train_epochs": 20,
637
+ "total_flos": 6.6069952716691825e+19,
638
+ "trial_name": null,
639
+ "trial_params": null
640
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:626e41080316a8d71c09c424912b5499166206ed73bd40113ba9f0dae71225c6
3
+ size 2863
vocab.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"n": 1, "t͡ɕ͈": 2, "ʌ": 3, "t": 4, "wʌ": 5, "m": 6, "k͈": 7, "ju": 8, "tʰ": 9, "l": 10, "d": 11, "s͈": 12, "t͡ɕ": 13, "p̚": 14, "h": 15, "dʑ": 16, "i": 17, "ɛ": 18, "ɰi": 19, "ja": 20, "p͈": 21, "jʌ": 22, "wɛ": 23, "k̚": 24, "a": 25, "b": 26, "jo": 27, "o": 28, "ɾ": 29, "s": 30, "t͡ɕʰ": 31, "ŋ": 32, "wa": 33, "g": 34, "t̚": 35, "kʰ": 36, "wi": 37, "ɯ": 38, "p": 39, "pʰ": 40, "t͈": 41, "k": 42, "jɛ": 43, "u": 44, "|": 0, "[UNK]": 45, "[PAD]": 46}