Hugging Face's logo

artyomboyko
/
whisper-small-fine_tuned-ru

Automatic Speech Recognition
Transformers
PyTorch
TensorBoard
Safetensors
whisper
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
3
artyomboyko commited on
Commit
1015518
•
1 Parent(s): 884afc1

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +21 -129
README.md CHANGED
@@ -1,28 +1,12 @@
1
  ---
2
- language:
3
- - ru
4
  license: apache-2.0
5
  tags:
6
- - hf-asr-leaderboard
7
  - generated_from_trainer
8
- datasets:
9
- - mozilla-foundation/common_voice_13_0
10
  metrics:
11
  - wer
12
  model-index:
13
  - name: whisper-small-fine_tuned-ru
14
- results:
15
- - task:
16
- name: Automatic Speech Recognition
17
- type: automatic-speech-recognition
18
- dataset:
19
- name: common_voice_13_0
20
- type: mozilla-foundation/common_voice_13_0
21
- args: 'config: ru, split: test'
22
- metrics:
23
- - name: Wer
24
- type: wer
25
- value: 17.724332
26
  ---
27
 
28
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -30,18 +14,18 @@ should probably proofread and complete it, then remove this comment. -->
30
 
31
  # whisper-small-fine_tuned-ru
32
 
33
- This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the [common_voice_13_0](https://huggingface.co/datasets/mozilla-foundation/common_voice_13_0) dataset.
34
  It achieves the following results on the evaluation set:
35
- - Loss: 0.22031
36
- - Wer: 17.724332
37
 
38
  ## Model description
39
 
40
- Same as original model (see [whisper-small](https://huggingface.co/openai/whisper-small)). ***But! This model has been fine-tuned for the task of transcribing the Russian language.***
41
 
42
  ## Intended uses & limitations
43
 
44
- Same as original model (see [whisper-small](https://huggingface.co/openai/whisper-small)).
45
 
46
  ## Training and evaluation data
47
 
@@ -49,124 +33,32 @@ More information needed
49
 
50
  ## Training procedure
51
 
52
- The model is fine-tuned using the following notebook (available only in the Russian version): https://github.com/blademoon/Whisper_Train
53
-
54
  ### Training hyperparameters
55
 
56
  The following hyperparameters were used during training:
57
- - learning_rate: 1e-06
58
  - train_batch_size: 16
59
  - eval_batch_size: 8
60
  - seed: 42
61
- - optimizer: Pytorch Adam with betas=(0.9,0.999) and epsilon=1e-08
62
  - lr_scheduler_type: linear
63
  - lr_scheduler_warmup_steps: 250
64
- - training_steps: 50000
65
 
66
  ### Training results
67
 
68
- | Training Loss | Epoch | Step | Validation Loss | Wer |
69
- |:-------------:|:-----:|:-----:|:---------------:|:-------:|
70
- | 0.344 | 0.22 | 500 | 0.3936 | 58.4474 |
71
- | 0.1948 | 0.44 | 1000 | 0.2391 | 57.0232 |
72
- | 0.1853 | 0.66 | 1500 | 0.2255 | 66.1826 |
73
- | 0.186 | 0.88 | 2000 | 0.2180 | 65.3833 |
74
- | 0.1532 | 1.1 | 2500 | 0.2135 | 50.6050 |
75
- | 0.1374 | 1.32 | 3000 | 0.2107 | 47.9428 |
76
- | 0.1359 | 1.54 | 3500 | 0.2082 | 60.0693 |
77
- | 0.1387 | 1.76 | 4000 | 0.2052 | 58.8674 |
78
- | 0.1212 | 1.97 | 4500 | 0.2027 | 51.9571 |
79
- | 0.111 | 2.19 | 5000 | 0.2027 | 50.0780 |
80
- | 0.1108 | 2.41 | 5500 | 0.2013 | 42.9664 |
81
- | 0.1148 | 2.63 | 6000 | 0.2000 | 40.7882 |
82
- | 0.114 | 2.85 | 6500 | 0.2002 | 32.6050 |
83
- | 0.092 | 3.07 | 7000 | 0.2000 | 32.9307 |
84
- | 0.0783 | 3.29 | 7500 | 0.2001 | 33.1413 |
85
- | 0.0989 | 3.51 | 8000 | 0.1986 | 32.0313 |
86
- | 0.0919 | 3.73 | 8500 | 0.1991 | 28.7199 |
87
- | 0.0928 | 3.95 | 9000 | 0.1982 | 26.1798 |
88
- | 0.0721 | 4.17 | 9500 | 0.2007 | 22.4960 |
89
- | 0.078 | 4.39 | 10000 | 0.2012 | 26.0774 |
90
- | 0.0764 | 4.61 | 10500 | 0.2004 | 24.7906 |
91
- | 0.0812 | 4.83 | 11000 | 0.2003 | 24.8022 |
92
- | 0.0531 | 5.05 | 11500 | 0.2022 | 21.3837 |
93
- | 0.0587 | 5.27 | 12000 | 0.2038 | 21.1638 |
94
- | 0.0553 | 5.48 | 12500 | 0.2039 | 21.9224 |
95
- | 0.0537 | 5.7 | 13000 | 0.2042 | 20.9671 |
96
- | 0.0608 | 5.92 | 13500 | 0.2049 | 21.1068 |
97
- | 0.0467 | 6.14 | 14000 | 0.2073 | 18.6528 |
98
- | 0.0533 | 6.36 | 14500 | 0.2088 | 18.7843 |
99
- | 0.048 | 6.58 | 15000 | 0.2092 | 18.5609 |
100
- | 0.0479 | 6.8 | 15500 | 0.2101 | 19.1648 |
101
- | 0.0383 | 7.02 | 16000 | 0.2105 | 18.9379 |
102
- | 0.0384 | 7.24 | 16500 | 0.2147 | 18.8018 |
103
- | 0.0451 | 7.46 | 17000 | 0.2156 | 18.9170 |
104
- | 0.0399 | 7.68 | 17500 | 0.2163 | 18.3806 |
105
- | 0.0387 | 7.9 | 18000 | 0.2159 | 17.9605 |
106
- | ***0.0347*** | ***8.12*** | ***18500*** | ***0.2203*** | ***17.7243*** |
107
- | 0.0324 | 8.34 | 19000 | 0.2231 | 17.8163 |
108
- | 0.035 | 8.56 | 19500 | 0.2231 | 17.8954 |
109
- | 0.0338 | 8.78 | 20000 | 0.2234 | 17.7371 |
110
- | 0.0305 | 9.0 | 20500 | 0.2244 | 17.8035 |
111
- | 0.0244 | 9.21 | 21000 | 0.2305 | 17.8942 |
112
- | 0.0249 | 9.43 | 21500 | 0.2321 | 17.9024 |
113
- | 0.0242 | 9.65 | 22000 | 0.2328 | 18.2212 |
114
- | 0.0269 | 9.87 | 22500 | 0.2327 | 17.8104 |
115
- | 0.0198 | 10.09 | 23000 | 0.2380 | 17.7301 |
116
- | 0.0191 | 10.31 | 23500 | 0.2396 | 17.8861 |
117
- | 0.0218 | 10.53 | 24000 | 0.2412 | 17.7464 |
118
- | 0.0219 | 10.75 | 24500 | 0.2406 | 17.7453 |
119
- | 0.0206 | 10.97 | 25000 | 0.2427 | 17.9128 |
120
- | 0.0182 | 11.19 | 25500 | 0.2482 | 18.0676 |
121
- | 0.0143 | 11.41 | 26000 | 0.2506 | 17.9245 |
122
- | 0.0162 | 11.63 | 26500 | 0.2501 | 18.1572 |
123
- | 0.0172 | 11.85 | 27000 | 0.2535 | 18.1164 |
124
- | 0.0148 | 12.07 | 27500 | 0.2558 | 18.1130 |
125
- | 0.0123 | 12.29 | 28000 | 0.2573 | 18.4085 |
126
- | 0.0129 | 12.51 | 28500 | 0.2603 | 18.0978 |
127
- | 0.0136 | 12.72 | 29000 | 0.2615 | 18.1793 |
128
- | 0.011 | 12.94 | 29500 | 0.2617 | 18.2247 |
129
- | 0.0096 | 13.16 | 30000 | 0.2666 | 18.2712 |
130
- | 0.01 | 13.38 | 30500 | 0.2667 | 18.4457 |
131
- | 0.0122 | 13.6 | 31000 | 0.2690 | 18.1095 |
132
- | 0.0121 | 13.82 | 31500 | 0.2700 | 18.1653 |
133
- | 0.0088 | 14.04 | 32000 | 0.2720 | 18.4539 |
134
- | 0.0076 | 14.26 | 32500 | 0.2746 | 18.2956 |
135
- | 0.0086 | 14.48 | 33000 | 0.2764 | 18.5644 |
136
- | 0.0086 | 14.7 | 33500 | 0.2771 | 18.5260 |
137
- | 0.0085 | 14.92 | 34000 | 0.2788 | 18.4481 |
138
- | 0.008 | 15.14 | 34500 | 0.2803 | 18.4923 |
139
- | 0.0074 | 15.36 | 35000 | 0.2824 | 18.6028 |
140
- | 0.0069 | 15.58 | 35500 | 0.2838 | 18.7692 |
141
- | 0.008 | 15.8 | 36000 | 0.2848 | 18.6901 |
142
- | 0.0065 | 16.02 | 36500 | 0.2864 | 18.7413 |
143
- | 0.006 | 16.24 | 37000 | 0.2885 | 18.5458 |
144
- | 0.0061 | 16.45 | 37500 | 0.2885 | 18.6470 |
145
- | 0.0056 | 16.67 | 38000 | 0.2898 | 18.3736 |
146
- | 0.0061 | 16.89 | 38500 | 0.2912 | 18.8064 |
147
- | 0.0048 | 17.11 | 39000 | 0.2933 | 18.9018 |
148
- | 0.0053 | 17.33 | 39500 | 0.2939 | 18.6168 |
149
- | 0.006 | 17.55 | 40000 | 0.2954 | 18.7238 |
150
- | 0.0045 | 17.77 | 40500 | 0.2952 | 18.8099 |
151
- | 0.0059 | 17.99 | 41000 | 0.2964 | 18.5551 |
152
- | 0.0053 | 18.21 | 41500 | 0.2980 | 18.7157 |
153
- | 0.004 | 18.43 | 42000 | 0.2988 | 18.6412 |
154
- | 0.0049 | 18.65 | 42500 | 0.2990 | 18.7099 |
155
- | 0.0048 | 18.87 | 43000 | 0.3004 | 18.7552 |
156
- | 0.0041 | 19.09 | 43500 | 0.3015 | 18.8169 |
157
- | 0.0048 | 19.31 | 44000 | 0.3018 | 18.8518 |
158
- | 0.0039 | 19.53 | 44500 | 0.3022 | 18.9437 |
159
- | 0.0041 | 19.75 | 45000 | 0.3029 | 18.8239 |
160
- | 0.0041 | 19.96 | 45500 | 0.3036 | 18.8169 |
161
- | 0.004 | 20.18 | 46000 | 0.3045 | 18.8274 |
162
- | 0.0044 | 20.4 | 46500 | 0.3048 | 18.8867 |
163
- | 0.0042 | 20.62 | 47000 | 0.3054 | 18.8425 |
164
- | 0.0044 | 20.84 | 47500 | 0.3058 | 18.8448 |
165
- | 0.004 | 21.06 | 48000 | 0.3057 | 18.8425 |
166
- | 0.0038 | 21.28 | 48500 | 0.3062 | 18.7029 |
167
- | 0.0038 | 21.5 | 49000 | 0.3063 | 18.8413 |
168
- | 0.0046 | 21.72 | 49500 | 0.3063 | 18.8227 |
169
- | 0.0036 | 21.94 | 50000 | 0.3064 | 18.8483 |
170
 
171
 
172
  ### Framework versions
 
1
  ---
 
 
2
  license: apache-2.0
3
  tags:
 
4
  - generated_from_trainer
 
 
5
  metrics:
6
  - wer
7
  model-index:
8
  - name: whisper-small-fine_tuned-ru
9
+ results: []
 
 
 
 
 
 
 
 
 
 
 
10
  ---
11
 
12
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 
14
 
15
  # whisper-small-fine_tuned-ru
16
 
17
+ This model is a fine-tuned version of [artyomboyko/whisper-small-fine_tuned-ru](https://huggingface.co/artyomboyko/whisper-small-fine_tuned-ru) on the None dataset.
18
  It achieves the following results on the evaluation set:
19
+ - Loss: 0.2298
20
+ - Wer: 17.8930
21
 
22
  ## Model description
23
 
24
+ More information needed
25
 
26
  ## Intended uses & limitations
27
 
28
+ More information needed
29
 
30
  ## Training and evaluation data
31
 
 
33
 
34
  ## Training procedure
35
 
 
 
36
  ### Training hyperparameters
37
 
38
  The following hyperparameters were used during training:
39
+ - learning_rate: 5e-07
40
  - train_batch_size: 16
41
  - eval_batch_size: 8
42
  - seed: 42
43
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
44
  - lr_scheduler_type: linear
45
  - lr_scheduler_warmup_steps: 250
46
+ - training_steps: 5000
47
 
48
  ### Training results
49
 
50
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
51
+ |:-------------:|:-----:|:----:|:---------------:|:-------:|
52
+ | 0.0286 | 0.22 | 500 | 0.2225 | 18.0199 |
53
+ | 0.0287 | 0.44 | 1000 | 0.2235 | 18.0455 |
54
+ | 0.0334 | 0.66 | 1500 | 0.2243 | 18.1956 |
55
+ | 0.0373 | 0.88 | 2000 | 0.2239 | 17.9803 |
56
+ | 0.0261 | 1.1 | 2500 | 0.2268 | 17.7674 |
57
+ | 0.0252 | 1.32 | 3000 | 0.2277 | 17.8221 |
58
+ | 0.0265 | 1.54 | 3500 | 0.2290 | 17.6336 |
59
+ | 0.0271 | 1.76 | 4000 | 0.2293 | 17.8279 |
60
+ | 0.0252 | 1.97 | 4500 | 0.2293 | 17.8744 |
61
+ | 0.023 | 2.19 | 5000 | 0.2298 | 17.8930 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
62
 
63
 
64
  ### Framework versions