warmestman commited on
Commit
1a6a1c3
1 Parent(s): 0d29bd0

End of training

Browse files
README.md CHANGED
@@ -9,35 +9,25 @@ tags:
9
  datasets:
10
  - mozilla-foundation/common_voice_16_1
11
  - google/fleurs
12
- metrics:
13
- - wer
14
  model-index:
15
- - name: Whisper Large MN - Ankhbayasgalan Davaadorj
16
- results:
17
- - task:
18
- name: Automatic Speech Recognition
19
- type: automatic-speech-recognition
20
- dataset:
21
- name: Common Voice 16.1 & FLEURS
22
- type: mozilla-foundation/common_voice_16_1
23
- config: mn
24
- split: None
25
- args: 'config: mn, split: test+validation'
26
- metrics:
27
- - name: Wer
28
- type: wer
29
- value: 30.806886309883946
30
  ---
31
 
32
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
33
  should probably proofread and complete it, then remove this comment. -->
34
 
35
- # Whisper Large MN - Ankhbayasgalan Davaadorj
36
 
37
  This model is a fine-tuned version of [openai/whisper-large-v2](https://huggingface.co/openai/whisper-large-v2) on the Common Voice 16.1 & FLEURS dataset.
38
  It achieves the following results on the evaluation set:
39
- - Loss: 0.6355
40
- - Wer: 30.8069
 
 
 
 
 
41
 
42
  ## Model description
43
 
@@ -57,44 +47,18 @@ More information needed
57
 
58
  The following hyperparameters were used during training:
59
  - learning_rate: 1e-05
60
- - train_batch_size: 16
61
- - eval_batch_size: 8
62
  - seed: 42
63
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
64
  - lr_scheduler_type: linear
65
  - lr_scheduler_warmup_steps: 500
66
- - training_steps: 20000
67
  - mixed_precision_training: Native AMP
68
 
69
- ### Training results
70
-
71
- | Training Loss | Epoch | Step | Validation Loss | Wer |
72
- |:-------------:|:------:|:-----:|:---------------:|:-------:|
73
- | 0.045 | 5.99 | 1000 | 0.4108 | 37.5062 |
74
- | 0.0081 | 11.98 | 2000 | 0.4551 | 34.9101 |
75
- | 0.0072 | 17.96 | 3000 | 0.4533 | 34.0080 |
76
- | 0.0036 | 23.95 | 4000 | 0.4884 | 33.9475 |
77
- | 0.0022 | 29.94 | 5000 | 0.4928 | 32.8365 |
78
- | 0.0029 | 35.93 | 6000 | 0.4907 | 32.9740 |
79
- | 0.0035 | 41.92 | 7000 | 0.4889 | 32.9245 |
80
- | 0.0005 | 47.9 | 8000 | 0.5133 | 31.7034 |
81
- | 0.0017 | 53.89 | 9000 | 0.5086 | 32.4900 |
82
- | 0.0005 | 59.88 | 10000 | 0.5206 | 31.5274 |
83
- | 0.001 | 65.87 | 11000 | 0.5128 | 31.8189 |
84
- | 0.0018 | 71.86 | 12000 | 0.5344 | 32.5560 |
85
- | 0.0 | 77.84 | 13000 | 0.5471 | 31.2414 |
86
- | 0.0 | 83.83 | 14000 | 0.5790 | 31.1314 |
87
- | 0.0 | 89.82 | 15000 | 0.5948 | 30.9609 |
88
- | 0.0 | 95.81 | 16000 | 0.6069 | 30.9334 |
89
- | 0.0 | 101.8 | 17000 | 0.6172 | 30.7849 |
90
- | 0.0 | 107.78 | 18000 | 0.6259 | 30.7189 |
91
- | 0.0 | 113.77 | 19000 | 0.6326 | 30.7189 |
92
- | 0.0 | 119.76 | 20000 | 0.6355 | 30.8069 |
93
-
94
-
95
  ### Framework versions
96
 
97
  - Transformers 4.38.0.dev0
98
  - Pytorch 2.2.0+cu121
99
- - Datasets 2.17.0
100
  - Tokenizers 0.15.2
 
9
  datasets:
10
  - mozilla-foundation/common_voice_16_1
11
  - google/fleurs
 
 
12
  model-index:
13
+ - name: Whisper Large V2 MN - Ankhbayasgalan Davaadorj
14
+ results: []
 
 
 
 
 
 
 
 
 
 
 
 
 
15
  ---
16
 
17
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
18
  should probably proofread and complete it, then remove this comment. -->
19
 
20
+ # Whisper Large V2 MN - Ankhbayasgalan Davaadorj
21
 
22
  This model is a fine-tuned version of [openai/whisper-large-v2](https://huggingface.co/openai/whisper-large-v2) on the Common Voice 16.1 & FLEURS dataset.
23
  It achieves the following results on the evaluation set:
24
+ - eval_loss: 0.2431
25
+ - eval_wer: 19.6462
26
+ - eval_runtime: 212.4702
27
+ - eval_samples_per_second: 0.48
28
+ - eval_steps_per_second: 0.122
29
+ - epoch: 10.99
30
+ - step: 1000
31
 
32
  ## Model description
33
 
 
47
 
48
  The following hyperparameters were used during training:
49
  - learning_rate: 1e-05
50
+ - train_batch_size: 8
51
+ - eval_batch_size: 4
52
  - seed: 42
53
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
54
  - lr_scheduler_type: linear
55
  - lr_scheduler_warmup_steps: 500
56
+ - training_steps: 2000
57
  - mixed_precision_training: Native AMP
58
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
59
  ### Framework versions
60
 
61
  - Transformers 4.38.0.dev0
62
  - Pytorch 2.2.0+cu121
63
+ - Datasets 2.17.1
64
  - Tokenizers 0.15.2
generation_config.json CHANGED
@@ -1,98 +1,4 @@
1
  {
2
- "alignment_heads": [
3
- [
4
- 10,
5
- 12
6
- ],
7
- [
8
- 13,
9
- 17
10
- ],
11
- [
12
- 16,
13
- 11
14
- ],
15
- [
16
- 16,
17
- 12
18
- ],
19
- [
20
- 16,
21
- 13
22
- ],
23
- [
24
- 17,
25
- 15
26
- ],
27
- [
28
- 17,
29
- 16
30
- ],
31
- [
32
- 18,
33
- 4
34
- ],
35
- [
36
- 18,
37
- 11
38
- ],
39
- [
40
- 18,
41
- 19
42
- ],
43
- [
44
- 19,
45
- 11
46
- ],
47
- [
48
- 21,
49
- 2
50
- ],
51
- [
52
- 21,
53
- 3
54
- ],
55
- [
56
- 22,
57
- 3
58
- ],
59
- [
60
- 22,
61
- 9
62
- ],
63
- [
64
- 22,
65
- 12
66
- ],
67
- [
68
- 23,
69
- 5
70
- ],
71
- [
72
- 23,
73
- 7
74
- ],
75
- [
76
- 23,
77
- 13
78
- ],
79
- [
80
- 25,
81
- 5
82
- ],
83
- [
84
- 26,
85
- 1
86
- ],
87
- [
88
- 26,
89
- 12
90
- ],
91
- [
92
- 27,
93
- 15
94
- ]
95
- ],
96
  "begin_suppress_tokens": [
97
  220,
98
  50257
@@ -100,218 +6,9 @@
100
  "bos_token_id": 50257,
101
  "decoder_start_token_id": 50258,
102
  "eos_token_id": 50257,
103
- "forced_decoder_ids": [
104
- [
105
- 1,
106
- null
107
- ],
108
- [
109
- 2,
110
- 50359
111
- ]
112
- ],
113
- "is_multilingual": true,
114
- "lang_to_id": {
115
- "<|af|>": 50327,
116
- "<|am|>": 50334,
117
- "<|ar|>": 50272,
118
- "<|as|>": 50350,
119
- "<|az|>": 50304,
120
- "<|ba|>": 50355,
121
- "<|be|>": 50330,
122
- "<|bg|>": 50292,
123
- "<|bn|>": 50302,
124
- "<|bo|>": 50347,
125
- "<|br|>": 50309,
126
- "<|bs|>": 50315,
127
- "<|ca|>": 50270,
128
- "<|cs|>": 50283,
129
- "<|cy|>": 50297,
130
- "<|da|>": 50285,
131
- "<|de|>": 50261,
132
- "<|el|>": 50281,
133
- "<|en|>": 50259,
134
- "<|es|>": 50262,
135
- "<|et|>": 50307,
136
- "<|eu|>": 50310,
137
- "<|fa|>": 50300,
138
- "<|fi|>": 50277,
139
- "<|fo|>": 50338,
140
- "<|fr|>": 50265,
141
- "<|gl|>": 50319,
142
- "<|gu|>": 50333,
143
- "<|haw|>": 50352,
144
- "<|ha|>": 50354,
145
- "<|he|>": 50279,
146
- "<|hi|>": 50276,
147
- "<|hr|>": 50291,
148
- "<|ht|>": 50339,
149
- "<|hu|>": 50286,
150
- "<|hy|>": 50312,
151
- "<|id|>": 50275,
152
- "<|is|>": 50311,
153
- "<|it|>": 50274,
154
- "<|ja|>": 50266,
155
- "<|jw|>": 50356,
156
- "<|ka|>": 50329,
157
- "<|kk|>": 50316,
158
- "<|km|>": 50323,
159
- "<|kn|>": 50306,
160
- "<|ko|>": 50264,
161
- "<|la|>": 50294,
162
- "<|lb|>": 50345,
163
- "<|ln|>": 50353,
164
- "<|lo|>": 50336,
165
- "<|lt|>": 50293,
166
- "<|lv|>": 50301,
167
- "<|mg|>": 50349,
168
- "<|mi|>": 50295,
169
- "<|mk|>": 50308,
170
- "<|ml|>": 50296,
171
- "<|mn|>": 50314,
172
- "<|mr|>": 50320,
173
- "<|ms|>": 50282,
174
- "<|mt|>": 50343,
175
- "<|my|>": 50346,
176
- "<|ne|>": 50313,
177
- "<|nl|>": 50271,
178
- "<|nn|>": 50342,
179
- "<|no|>": 50288,
180
- "<|oc|>": 50328,
181
- "<|pa|>": 50321,
182
- "<|pl|>": 50269,
183
- "<|ps|>": 50340,
184
- "<|pt|>": 50267,
185
- "<|ro|>": 50284,
186
- "<|ru|>": 50263,
187
- "<|sa|>": 50344,
188
- "<|sd|>": 50332,
189
- "<|si|>": 50322,
190
- "<|sk|>": 50298,
191
- "<|sl|>": 50305,
192
- "<|sn|>": 50324,
193
- "<|so|>": 50326,
194
- "<|sq|>": 50317,
195
- "<|sr|>": 50303,
196
- "<|su|>": 50357,
197
- "<|sv|>": 50273,
198
- "<|sw|>": 50318,
199
- "<|ta|>": 50287,
200
- "<|te|>": 50299,
201
- "<|tg|>": 50331,
202
- "<|th|>": 50289,
203
- "<|tk|>": 50341,
204
- "<|tl|>": 50348,
205
- "<|tr|>": 50268,
206
- "<|tt|>": 50351,
207
- "<|uk|>": 50280,
208
- "<|ur|>": 50290,
209
- "<|uz|>": 50337,
210
- "<|vi|>": 50278,
211
- "<|yi|>": 50335,
212
- "<|yo|>": 50325,
213
- "<|zh|>": 50260
214
- },
215
- "language": "mn",
216
- "max_initial_timestamp_index": 50,
217
  "max_length": 448,
218
- "no_timestamps_token_id": 50363,
219
  "pad_token_id": 50257,
220
- "prev_sot_token_id": 50361,
221
- "return_timestamps": false,
222
- "suppress_tokens": [
223
- 1,
224
- 2,
225
- 7,
226
- 8,
227
- 9,
228
- 10,
229
- 14,
230
- 25,
231
- 26,
232
- 27,
233
- 28,
234
- 29,
235
- 31,
236
- 58,
237
- 59,
238
- 60,
239
- 61,
240
- 62,
241
- 63,
242
- 90,
243
- 91,
244
- 92,
245
- 93,
246
- 359,
247
- 503,
248
- 522,
249
- 542,
250
- 873,
251
- 893,
252
- 902,
253
- 918,
254
- 922,
255
- 931,
256
- 1350,
257
- 1853,
258
- 1982,
259
- 2460,
260
- 2627,
261
- 3246,
262
- 3253,
263
- 3268,
264
- 3536,
265
- 3846,
266
- 3961,
267
- 4183,
268
- 4667,
269
- 6585,
270
- 6647,
271
- 7273,
272
- 9061,
273
- 9383,
274
- 10428,
275
- 10929,
276
- 11938,
277
- 12033,
278
- 12331,
279
- 12562,
280
- 13793,
281
- 14157,
282
- 14635,
283
- 15265,
284
- 15618,
285
- 16553,
286
- 16604,
287
- 18362,
288
- 18956,
289
- 20075,
290
- 21675,
291
- 22520,
292
- 26130,
293
- 26161,
294
- 26435,
295
- 28279,
296
- 29464,
297
- 31650,
298
- 32302,
299
- 32470,
300
- 36865,
301
- 42863,
302
- 47425,
303
- 49870,
304
- 50254,
305
- 50258,
306
- 50358,
307
- 50359,
308
- 50360,
309
- 50361,
310
- 50362
311
- ],
312
- "task_to_id": {
313
- "transcribe": 50359,
314
- "translate": 50358
315
- },
316
- "transformers_version": "4.38.0.dev0"
317
  }
 
1
  {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
  "begin_suppress_tokens": [
3
  220,
4
  50257
 
6
  "bos_token_id": 50257,
7
  "decoder_start_token_id": 50258,
8
  "eos_token_id": 50257,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9
  "max_length": 448,
 
10
  "pad_token_id": 50257,
11
+ "suppress_tokens": [],
12
+ "transformers_version": "4.38.0.dev0",
13
+ "use_cache": false
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  }
model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ab2935c969d23adee90ffbcf4590346b586828c5e0290325a0e668e5365f2326
3
  size 4992706480
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22d1b41fb225216c5b0dde466e02c726ff30c89d79c32fd32912811d6f0f7f03
3
  size 4992706480
model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:12427a78b5d9531028bd062332ef692449e8b799c7bdb874b70e208fa58dd743
3
  size 1180663192
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84c210bbbf696f0b1b7cfb5fb401c4cf84793b1f8f1f5abc0ffb908d738785c2
3
  size 1180663192
runs/Feb21_04-47-07_nrz8795syv/events.out.tfevents.1708490842.nrz8795syv.8876.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:034682ec35e5cfabf958d5757396ab01febd9d02db22812b0db72260cff2adcb
3
- size 14884
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:480d8378ce642da9003e56472031ee4aea94581683109cf2a66cd94300958766
3
+ size 15512