akera commited on
Commit
b0a816c
1 Parent(s): 8959226

whisper-medium-sb-lug-eng-v2

Browse files
Files changed (2) hide show
  1. README.md +25 -75
  2. generation_config.json +5 -0
README.md CHANGED
@@ -1,29 +1,29 @@
1
  ---
2
- base_model: openai/whisper-medium
3
- datasets:
4
- - generator
5
  license: apache-2.0
 
6
  tags:
7
  - generated_from_trainer
 
 
8
  model-index:
9
- - name: whisper-medium-sb-lug-eng
10
  results: []
11
  ---
12
 
13
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
14
  should probably proofread and complete it, then remove this comment. -->
15
 
16
- # whisper-medium-sb-lug-eng
17
 
18
  This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the generator dataset.
19
  It achieves the following results on the evaluation set:
20
- - Loss: 0.1720
21
- - Wer Lug: 0.81
22
- - Wer Eng: 0.068
23
- - Wer Mean: 0.439
24
- - Cer Lug: 0.494
25
- - Cer Eng: 0.039
26
- - Cer Mean: 0.267
27
 
28
  ## Model description
29
 
@@ -49,73 +49,23 @@ The following hyperparameters were used during training:
49
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
  - lr_scheduler_type: linear
51
  - lr_scheduler_warmup_steps: 500
52
- - training_steps: 30000
53
  - mixed_precision_training: Native AMP
54
 
55
  ### Training results
56
 
57
- | Training Loss | Epoch | Step | Validation Loss | Wer Lug | Wer Eng | Wer Mean | Cer Lug | Cer Eng | Cer Mean |
58
- |:-------------:|:------:|:-----:|:---------------:|:-------:|:-------:|:--------:|:-------:|:-------:|:--------:|
59
- | 0.9804 | 0.0167 | 500 | 0.3683 | 0.692 | 0.043 | 0.368 | 0.203 | 0.019 | 0.111 |
60
- | 0.7775 | 0.0333 | 1000 | 0.2594 | 0.725 | 0.044 | 0.385 | 0.395 | 0.019 | 0.207 |
61
- | 0.6492 | 0.05 | 1500 | 0.2316 | 0.649 | 0.041 | 0.345 | 0.263 | 0.02 | 0.142 |
62
- | 0.6128 | 0.0667 | 2000 | 0.2111 | 0.513 | 0.04 | 0.277 | 0.197 | 0.018 | 0.108 |
63
- | 0.543 | 0.0833 | 2500 | 0.2023 | 0.579 | 0.043 | 0.311 | 0.239 | 0.018 | 0.129 |
64
- | 0.5461 | 0.1 | 3000 | 0.1932 | 0.425 | 0.04 | 0.233 | 0.138 | 0.019 | 0.078 |
65
- | 0.5545 | 0.1167 | 3500 | 0.1836 | 0.624 | 0.043 | 0.334 | 0.381 | 0.021 | 0.201 |
66
- | 0.4895 | 0.1333 | 4000 | 0.1802 | 0.407 | 0.043 | 0.225 | 0.156 | 0.022 | 0.089 |
67
- | 0.4922 | 0.15 | 4500 | 0.1771 | 0.377 | 0.051 | 0.214 | 0.136 | 0.033 | 0.084 |
68
- | 0.521 | 0.1667 | 5000 | 0.1817 | 0.316 | 0.049 | 0.183 | 0.097 | 0.028 | 0.062 |
69
- | 0.3948 | 1.0153 | 5500 | 0.1724 | 0.422 | 0.079 | 0.251 | 0.17 | 0.057 | 0.113 |
70
- | 0.3914 | 1.032 | 6000 | 0.1727 | 0.744 | 0.04 | 0.392 | 0.651 | 0.018 | 0.334 |
71
- | 0.3807 | 1.0487 | 6500 | 0.1730 | 0.585 | 0.053 | 0.319 | 0.428 | 0.028 | 0.228 |
72
- | 0.395 | 1.0653 | 7000 | 0.1701 | 0.737 | 0.043 | 0.39 | 0.635 | 0.024 | 0.329 |
73
- | 0.3774 | 1.082 | 7500 | 0.1654 | 0.545 | 0.046 | 0.296 | 0.396 | 0.024 | 0.21 |
74
- | 0.4017 | 1.0987 | 8000 | 0.1626 | 0.465 | 0.046 | 0.256 | 0.28 | 0.024 | 0.152 |
75
- | 0.3901 | 1.1153 | 8500 | 0.1593 | 0.516 | 0.051 | 0.283 | 0.25 | 0.026 | 0.138 |
76
- | 0.3829 | 1.1320 | 9000 | 0.1608 | 0.48 | 0.049 | 0.264 | 0.247 | 0.024 | 0.135 |
77
- | 0.3536 | 1.1487 | 9500 | 0.1657 | 0.37 | 0.043 | 0.207 | 0.143 | 0.021 | 0.082 |
78
- | 0.3506 | 1.1653 | 10000 | 0.1606 | 0.395 | 0.041 | 0.218 | 0.172 | 0.021 | 0.097 |
79
- | 0.2737 | 2.014 | 10500 | 0.1604 | 0.457 | 0.07 | 0.263 | 0.235 | 0.044 | 0.139 |
80
- | 0.3073 | 2.0307 | 11000 | 0.1626 | 0.458 | 0.046 | 0.252 | 0.243 | 0.022 | 0.132 |
81
- | 0.2906 | 2.0473 | 11500 | 0.1581 | 0.444 | 0.062 | 0.253 | 0.222 | 0.038 | 0.13 |
82
- | 0.2882 | 2.064 | 12000 | 0.1591 | 0.519 | 0.053 | 0.286 | 0.3 | 0.024 | 0.162 |
83
- | 0.2642 | 2.0807 | 12500 | 0.1630 | 0.547 | 0.05 | 0.299 | 0.293 | 0.029 | 0.161 |
84
- | 0.2848 | 2.0973 | 13000 | 0.1627 | 0.509 | 0.055 | 0.282 | 0.244 | 0.03 | 0.137 |
85
- | 0.2887 | 2.114 | 13500 | 0.1585 | 0.524 | 0.067 | 0.296 | 0.28 | 0.047 | 0.163 |
86
- | 0.2879 | 2.1307 | 14000 | 0.1593 | 0.646 | 0.065 | 0.356 | 0.355 | 0.045 | 0.2 |
87
- | 0.2955 | 2.1473 | 14500 | 0.1581 | 0.873 | 0.062 | 0.468 | 0.512 | 0.038 | 0.275 |
88
- | 0.2639 | 2.164 | 15000 | 0.1533 | 0.772 | 0.057 | 0.414 | 0.454 | 0.037 | 0.245 |
89
- | 0.2111 | 3.0127 | 15500 | 0.1622 | 0.776 | 0.074 | 0.425 | 0.518 | 0.046 | 0.282 |
90
- | 0.2299 | 3.0293 | 16000 | 0.1628 | 0.849 | 0.061 | 0.455 | 0.559 | 0.036 | 0.297 |
91
- | 0.2279 | 3.046 | 16500 | 0.1633 | 0.803 | 0.064 | 0.434 | 0.632 | 0.036 | 0.334 |
92
- | 0.2339 | 3.0627 | 17000 | 0.1617 | 0.845 | 0.045 | 0.445 | 0.553 | 0.022 | 0.288 |
93
- | 0.2387 | 3.0793 | 17500 | 0.1599 | 0.773 | 0.055 | 0.414 | 0.436 | 0.029 | 0.232 |
94
- | 0.2098 | 3.096 | 18000 | 0.1616 | 0.675 | 0.059 | 0.367 | 0.45 | 0.037 | 0.243 |
95
- | 0.2201 | 3.1127 | 18500 | 0.1619 | 0.713 | 0.066 | 0.389 | 0.476 | 0.039 | 0.257 |
96
- | 0.2312 | 3.1293 | 19000 | 0.1603 | 0.994 | 0.053 | 0.524 | 0.605 | 0.03 | 0.318 |
97
- | 0.2389 | 3.146 | 19500 | 0.1572 | 0.751 | 0.054 | 0.403 | 0.455 | 0.032 | 0.244 |
98
- | 0.2183 | 3.1627 | 20000 | 0.1635 | 0.667 | 0.056 | 0.362 | 0.42 | 0.034 | 0.227 |
99
- | 0.1707 | 4.0113 | 20500 | 0.1654 | 0.682 | 0.05 | 0.366 | 0.433 | 0.026 | 0.23 |
100
- | 0.1874 | 4.028 | 21000 | 0.1641 | 0.744 | 0.054 | 0.399 | 0.425 | 0.03 | 0.228 |
101
- | 0.1836 | 4.0447 | 21500 | 0.1666 | 0.651 | 0.063 | 0.357 | 0.397 | 0.039 | 0.218 |
102
- | 0.1847 | 4.0613 | 22000 | 0.1635 | 0.788 | 0.069 | 0.429 | 0.502 | 0.044 | 0.273 |
103
- | 0.1742 | 4.078 | 22500 | 0.1651 | 0.695 | 0.051 | 0.373 | 0.4 | 0.027 | 0.214 |
104
- | 0.1733 | 4.0947 | 23000 | 0.1652 | 0.678 | 0.064 | 0.371 | 0.427 | 0.039 | 0.233 |
105
- | 0.1651 | 4.1113 | 23500 | 0.1659 | 0.666 | 0.071 | 0.369 | 0.458 | 0.046 | 0.252 |
106
- | 0.1924 | 4.128 | 24000 | 0.1664 | 0.792 | 0.069 | 0.431 | 0.486 | 0.046 | 0.266 |
107
- | 0.1828 | 4.1447 | 24500 | 0.1670 | 0.746 | 0.068 | 0.407 | 0.538 | 0.043 | 0.291 |
108
- | 0.165 | 4.1613 | 25000 | 0.1675 | 0.746 | 0.072 | 0.409 | 0.469 | 0.047 | 0.258 |
109
- | 0.1437 | 5.01 | 25500 | 0.1706 | 0.728 | 0.066 | 0.397 | 0.481 | 0.04 | 0.261 |
110
- | 0.148 | 5.0267 | 26000 | 0.1700 | 0.755 | 0.069 | 0.412 | 0.457 | 0.041 | 0.249 |
111
- | 0.1509 | 5.0433 | 26500 | 0.1700 | 0.787 | 0.068 | 0.427 | 0.497 | 0.039 | 0.268 |
112
- | 0.1442 | 5.06 | 27000 | 0.1715 | 0.762 | 0.068 | 0.415 | 0.47 | 0.039 | 0.254 |
113
- | 0.1282 | 5.0767 | 27500 | 0.1698 | 0.796 | 0.064 | 0.43 | 0.477 | 0.037 | 0.257 |
114
- | 0.1377 | 5.0933 | 28000 | 0.1710 | 0.796 | 0.068 | 0.432 | 0.481 | 0.04 | 0.261 |
115
- | 0.1456 | 5.11 | 28500 | 0.1719 | 0.758 | 0.07 | 0.414 | 0.481 | 0.04 | 0.26 |
116
- | 0.143 | 5.1267 | 29000 | 0.1716 | 0.795 | 0.07 | 0.433 | 0.488 | 0.04 | 0.264 |
117
- | 0.1484 | 5.1433 | 29500 | 0.1719 | 0.812 | 0.069 | 0.44 | 0.492 | 0.04 | 0.266 |
118
- | 0.1463 | 5.16 | 30000 | 0.1720 | 0.81 | 0.068 | 0.439 | 0.494 | 0.039 | 0.267 |
119
 
120
 
121
  ### Framework versions
 
1
  ---
 
 
 
2
  license: apache-2.0
3
+ base_model: openai/whisper-medium
4
  tags:
5
  - generated_from_trainer
6
+ datasets:
7
+ - generator
8
  model-index:
9
+ - name: whisper-medium-sb-lug-eng-v2
10
  results: []
11
  ---
12
 
13
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
14
  should probably proofread and complete it, then remove this comment. -->
15
 
16
+ # whisper-medium-sb-lug-eng-v2
17
 
18
  This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the generator dataset.
19
  It achieves the following results on the evaluation set:
20
+ - Loss: 0.1255
21
+ - Wer Lug: 0.124
22
+ - Wer Eng: 0.171
23
+ - Wer Mean: 0.147
24
+ - Cer Lug: 0.028
25
+ - Cer Eng: 0.164
26
+ - Cer Mean: 0.096
27
 
28
  ## Model description
29
 
 
49
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
  - lr_scheduler_type: linear
51
  - lr_scheduler_warmup_steps: 500
52
+ - training_steps: 5000
53
  - mixed_precision_training: Native AMP
54
 
55
  ### Training results
56
 
57
+ | Training Loss | Epoch | Step | Validation Loss | Wer Lug | Wer Eng | Wer Mean | Cer Lug | Cer Eng | Cer Mean |
58
+ |:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|:--------:|:-------:|:-------:|:--------:|
59
+ | 0.7222 | 0.1 | 500 | 0.2662 | 0.391 | 0.146 | 0.268 | 0.085 | 0.104 | 0.095 |
60
+ | 0.5452 | 0.2 | 1000 | 0.1894 | 0.223 | 0.058 | 0.141 | 0.05 | 0.034 | 0.042 |
61
+ | 0.4172 | 0.3 | 1500 | 0.1764 | 0.199 | 0.276 | 0.238 | 0.043 | 0.269 | 0.156 |
62
+ | 0.3949 | 0.4 | 2000 | 0.1583 | 0.177 | 0.035 | 0.106 | 0.039 | 0.017 | 0.028 |
63
+ | 0.3626 | 0.5 | 2500 | 0.1508 | 0.153 | 0.157 | 0.155 | 0.035 | 0.122 | 0.078 |
64
+ | 0.3467 | 0.6 | 3000 | 0.1397 | 0.14 | 0.258 | 0.199 | 0.033 | 0.213 | 0.123 |
65
+ | 0.3443 | 0.7 | 3500 | 0.1333 | 0.139 | 0.044 | 0.092 | 0.032 | 0.027 | 0.029 |
66
+ | 0.3169 | 0.8 | 4000 | 0.1297 | 0.129 | 0.027 | 0.078 | 0.029 | 0.011 | 0.02 |
67
+ | 0.3276 | 0.9 | 4500 | 0.1264 | 0.124 | 0.086 | 0.105 | 0.028 | 0.058 | 0.043 |
68
+ | 0.317 | 1.0 | 5000 | 0.1255 | 0.124 | 0.171 | 0.147 | 0.028 | 0.164 | 0.096 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
69
 
70
 
71
  ### Framework versions
generation_config.json CHANGED
@@ -44,6 +44,7 @@
44
  ],
45
  "is_multilingual": true,
46
  "lang_to_id": {
 
47
  "<|af|>": 50327,
48
  "<|am|>": 50334,
49
  "<|ar|>": 50272,
@@ -92,9 +93,11 @@
92
  "<|ko|>": 50264,
93
  "<|la|>": 50294,
94
  "<|lb|>": 50345,
 
95
  "<|ln|>": 50353,
96
  "<|lo|>": 50336,
97
  "<|lt|>": 50293,
 
98
  "<|lv|>": 50301,
99
  "<|mg|>": 50349,
100
  "<|mi|>": 50295,
@@ -109,6 +112,7 @@
109
  "<|nl|>": 50271,
110
  "<|nn|>": 50342,
111
  "<|no|>": 50288,
 
112
  "<|oc|>": 50328,
113
  "<|pa|>": 50321,
114
  "<|pl|>": 50269,
@@ -129,6 +133,7 @@
129
  "<|sv|>": 50273,
130
  "<|sw|>": 50318,
131
  "<|ta|>": 50287,
 
132
  "<|te|>": 50299,
133
  "<|tg|>": 50331,
134
  "<|th|>": 50289,
 
44
  ],
45
  "is_multilingual": true,
46
  "lang_to_id": {
47
+ "<|ach|>": 51866,
48
  "<|af|>": 50327,
49
  "<|am|>": 50334,
50
  "<|ar|>": 50272,
 
93
  "<|ko|>": 50264,
94
  "<|la|>": 50294,
95
  "<|lb|>": 50345,
96
+ "<|lgg|>": 51867,
97
  "<|ln|>": 50353,
98
  "<|lo|>": 50336,
99
  "<|lt|>": 50293,
100
+ "<|lug|>": 51865,
101
  "<|lv|>": 50301,
102
  "<|mg|>": 50349,
103
  "<|mi|>": 50295,
 
112
  "<|nl|>": 50271,
113
  "<|nn|>": 50342,
114
  "<|no|>": 50288,
115
+ "<|nyn|>": 51868,
116
  "<|oc|>": 50328,
117
  "<|pa|>": 50321,
118
  "<|pl|>": 50269,
 
133
  "<|sv|>": 50273,
134
  "<|sw|>": 50318,
135
  "<|ta|>": 50287,
136
+ "<|teo|>": 51869,
137
  "<|te|>": 50299,
138
  "<|tg|>": 50331,
139
  "<|th|>": 50289,