Hugging Face's logo

NbAiLab
/
whisper-tiny-nob

Automatic Speech Recognition
Transformers
PyTorch
TensorBoard
Safetensors
whisper
whisper-event
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Metrics Community
2
pere commited on
Commit
bf3841c
•
1 Parent(s): 6e47be8

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +137 -27
README.md CHANGED
@@ -1,65 +1,175 @@
1
  ---
2
- language:
3
- - no
4
  license: apache-2.0
5
  tags:
6
- - whisper-event
7
- - norwegian
8
  datasets:
9
- - NbAiLab/NCC_S
10
- - NbAiLab/NPSC
11
- - NbAiLab/NST
12
  metrics:
13
  - wer
14
  model-index:
15
- - name: Whisper Tiny Norwegian Bokmål
16
  results:
17
  - task:
18
  name: Automatic Speech Recognition
19
  type: automatic-speech-recognition
20
  dataset:
21
- name: FLEURS
22
- type: google/fleurs
23
- config: nb_no
24
  split: validation
25
- args: nb_no
26
  metrics:
27
  - name: Wer
28
  type: wer
29
- value: 32.76
30
  ---
31
 
32
- # Whisper Tiny Norwegian Bokmål
 
33
 
34
- This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) trained on several datasets.
35
 
36
- It is currently in the middle of a large training. Currently it achieves the following results on the evaluation set:
37
- - Loss: 0.604
38
- - Wer: 32.76
 
39
 
40
  ## Model description
41
 
42
- The model is trained on a large corpus of roughly 5.000 hours of voice. The sources are subtitles from the Norwegian broadcaster NRK, transcribed speeches from the Norwegian parliament and voice recordings from Norsk Språkteknologi.
43
 
44
  ## Intended uses & limitations
45
 
46
- The model will be free for everyone to use when it is finished.
 
 
 
 
 
 
47
 
48
  ### Training hyperparameters
49
 
50
  The following hyperparameters were used during training:
51
  - learning_rate: 3e-06
52
- - train_batch_size: 128
53
- - eval_batch_size: 32
54
  - seed: 42
55
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
56
- - lr_scheduler_type: linear
57
  - lr_scheduler_warmup_steps: 1000
58
- - training_steps: 100.000 (currently @5.000)
59
- - mixed_precision_training: fp16
 
 
60
 
61
- ### Live Training results
62
- See [Tensorboad Metrics](https://huggingface.co/NbAiLab/whisper-tiny-nob/tensorboard)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63
 
64
 
 
65
 
 
 
 
 
 
1
  ---
 
 
2
  license: apache-2.0
3
  tags:
4
+ - generated_from_trainer
 
5
  datasets:
6
+ - ncc_s
 
 
7
  metrics:
8
  - wer
9
  model-index:
10
+ - name: whisper-tiny-nob
11
  results:
12
  - task:
13
  name: Automatic Speech Recognition
14
  type: automatic-speech-recognition
15
  dataset:
16
+ name: ncc_s
17
+ type: ncc_s
18
+ config: 'no'
19
  split: validation
20
+ args: 'no'
21
  metrics:
22
  - name: Wer
23
  type: wer
24
+ value: 24.96954933008526
25
  ---
26
 
27
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
28
+ should probably proofread and complete it, then remove this comment. -->
29
 
30
+ # whisper-tiny-nob
31
 
32
+ This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the ncc_s dataset.
33
+ It achieves the following results on the evaluation set:
34
+ - Loss: 0.5128
35
+ - Wer: 24.9695
36
 
37
  ## Model description
38
 
39
+ More information needed
40
 
41
  ## Intended uses & limitations
42
 
43
+ More information needed
44
+
45
+ ## Training and evaluation data
46
+
47
+ More information needed
48
+
49
+ ## Training procedure
50
 
51
  ### Training hyperparameters
52
 
53
  The following hyperparameters were used during training:
54
  - learning_rate: 3e-06
55
+ - train_batch_size: 256
56
+ - eval_batch_size: 64
57
  - seed: 42
58
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
59
+ - lr_scheduler_type: constant_with_warmup
60
  - lr_scheduler_warmup_steps: 1000
61
+ - training_steps: 100000
62
+ - mixed_precision_training: Native AMP
63
+
64
+ ### Training results
65
 
66
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
67
+ |:-------------:|:-----:|:------:|:---------------:|:-------:|
68
+ | 1.8819 | 0.01 | 1000 | 1.1869 | 61.9671 |
69
+ | 1.6425 | 0.02 | 2000 | 0.9991 | 53.6541 |
70
+ | 1.548 | 0.03 | 3000 | 0.9147 | 50.2132 |
71
+ | 1.4636 | 0.04 | 4000 | 0.8605 | 47.0767 |
72
+ | 1.4113 | 0.05 | 5000 | 0.8253 | 45.7369 |
73
+ | 1.3484 | 0.01 | 6000 | 0.7946 | 43.4531 |
74
+ | 1.3127 | 0.02 | 7000 | 0.7740 | 42.2655 |
75
+ | 1.2994 | 0.03 | 8000 | 0.7551 | 40.8952 |
76
+ | 1.265 | 0.04 | 9000 | 0.7378 | 39.8599 |
77
+ | 1.2458 | 0.05 | 10000 | 0.7257 | 39.8904 |
78
+ | 1.2257 | 0.06 | 11000 | 0.7114 | 39.7990 |
79
+ | 1.2126 | 0.07 | 12000 | 0.6972 | 37.8806 |
80
+ | 1.1971 | 0.08 | 13000 | 0.6871 | 37.3021 |
81
+ | 1.1786 | 1.01 | 14000 | 0.6786 | 37.4239 |
82
+ | 1.1486 | 1.02 | 15000 | 0.6703 | 36.9976 |
83
+ | 1.1505 | 1.03 | 16000 | 0.6647 | 36.3581 |
84
+ | 1.1238 | 1.04 | 17000 | 0.6559 | 36.3886 |
85
+ | 1.1184 | 1.05 | 18000 | 0.6509 | 36.5104 |
86
+ | 1.115 | 1.06 | 19000 | 0.6452 | 35.9927 |
87
+ | 1.1013 | 1.07 | 20000 | 0.6382 | 34.5006 |
88
+ | 1.0969 | 1.08 | 21000 | 0.6331 | 34.3484 |
89
+ | 1.0784 | 2.0 | 22000 | 0.6304 | 34.2875 |
90
+ | 1.0774 | 2.01 | 23000 | 0.6249 | 34.1048 |
91
+ | 1.0719 | 2.02 | 24000 | 0.6194 | 33.8307 |
92
+ | 1.0638 | 2.03 | 25000 | 0.6158 | 32.9781 |
93
+ | 1.0592 | 2.04 | 26000 | 0.6105 | 32.6431 |
94
+ | 1.0493 | 2.05 | 27000 | 0.6041 | 32.7345 |
95
+ | 1.047 | 2.06 | 28000 | 0.6040 | 32.7649 |
96
+ | 1.0323 | 2.07 | 29000 | 0.5984 | 31.6078 |
97
+ | 1.0189 | 3.0 | 30000 | 0.5957 | 31.3033 |
98
+ | 1.0078 | 3.01 | 31000 | 0.5924 | 31.4251 |
99
+ | 1.0146 | 3.02 | 32000 | 0.5940 | 31.3033 |
100
+ | 1.0128 | 3.03 | 33000 | 0.5892 | 31.0292 |
101
+ | 1.0025 | 3.04 | 34000 | 0.5873 | 31.1815 |
102
+ | 0.999 | 3.05 | 35000 | 0.5838 | 30.6334 |
103
+ | 1.0045 | 3.06 | 36000 | 0.5799 | 30.4202 |
104
+ | 1.0005 | 3.07 | 37000 | 0.5770 | 30.1766 |
105
+ | 1.0017 | 3.08 | 38000 | 0.5733 | 29.6590 |
106
+ | 0.9878 | 4.01 | 39000 | 0.5745 | 30.2680 |
107
+ | 0.9854 | 4.02 | 40000 | 0.5720 | 30.0548 |
108
+ | 0.9624 | 4.03 | 41000 | 0.5703 | 29.5981 |
109
+ | 0.9639 | 4.04 | 42000 | 0.5681 | 29.5067 |
110
+ | 0.9569 | 4.05 | 43000 | 0.5679 | 29.6285 |
111
+ | 0.9682 | 4.06 | 44000 | 0.5643 | 29.5676 |
112
+ | 0.9539 | 4.07 | 45000 | 0.5601 | 29.5676 |
113
+ | 0.946 | 4.08 | 46000 | 0.5562 | 29.7199 |
114
+ | 0.9429 | 5.01 | 47000 | 0.5592 | 29.2935 |
115
+ | 0.9462 | 5.02 | 48000 | 0.5540 | 29.0804 |
116
+ | 0.9312 | 5.03 | 49000 | 0.5535 | 29.2935 |
117
+ | 0.9462 | 5.04 | 50000 | 0.5536 | 28.6845 |
118
+ | 0.922 | 5.05 | 51000 | 0.5539 | 28.7150 |
119
+ | 0.9253 | 5.06 | 52000 | 0.5510 | 28.8368 |
120
+ | 0.9065 | 0.01 | 53000 | 0.5493 | 28.5932 |
121
+ | 0.9096 | 0.02 | 54000 | 0.5490 | 28.5018 |
122
+ | 0.9329 | 0.03 | 55000 | 0.5483 | 28.2887 |
123
+ | 0.9181 | 0.04 | 56000 | 0.5471 | 27.9842 |
124
+ | 0.914 | 0.05 | 57000 | 0.5457 | 28.4105 |
125
+ | 0.9149 | 0.06 | 58000 | 0.5449 | 27.5883 |
126
+ | 0.9092 | 0.07 | 59000 | 0.5405 | 27.8319 |
127
+ | 0.9101 | 0.08 | 60000 | 0.5402 | 27.3447 |
128
+ | 0.9046 | 1.01 | 61000 | 0.5374 | 27.5579 |
129
+ | 0.8917 | 1.02 | 62000 | 0.5390 | 27.7406 |
130
+ | 0.8993 | 1.03 | 63000 | 0.5386 | 27.4056 |
131
+ | 0.8875 | 1.04 | 64000 | 0.5361 | 26.8575 |
132
+ | 0.8892 | 1.05 | 65000 | 0.5358 | 27.3447 |
133
+ | 0.8929 | 1.06 | 66000 | 0.5346 | 26.7357 |
134
+ | 0.8703 | 0.01 | 67000 | 0.5332 | 26.8270 |
135
+ | 0.8709 | 0.02 | 68000 | 0.5336 | 26.7052 |
136
+ | 0.8917 | 0.03 | 69000 | 0.5329 | 27.0706 |
137
+ | 0.8867 | 0.04 | 70000 | 0.5323 | 26.3398 |
138
+ | 0.8778 | 0.05 | 71000 | 0.5315 | 27.2838 |
139
+ | 0.8757 | 0.06 | 72000 | 0.5317 | 26.2485 |
140
+ | 0.8726 | 0.07 | 73000 | 0.5269 | 26.6443 |
141
+ | 0.8792 | 0.08 | 74000 | 0.5268 | 26.1571 |
142
+ | 0.8706 | 1.01 | 75000 | 0.5247 | 26.1571 |
143
+ | 0.8585 | 1.02 | 76000 | 0.5265 | 26.3703 |
144
+ | 0.8659 | 1.03 | 77000 | 0.5262 | 26.7357 |
145
+ | 0.8551 | 1.04 | 78000 | 0.5249 | 26.0658 |
146
+ | 0.8572 | 1.05 | 79000 | 0.5249 | 26.2789 |
147
+ | 0.8612 | 1.06 | 80000 | 0.5235 | 25.7613 |
148
+ | 0.8598 | 1.07 | 81000 | 0.5208 | 25.7004 |
149
+ | 0.8686 | 1.08 | 82000 | 0.5214 | 25.7004 |
150
+ | 0.8503 | 2.0 | 83000 | 0.5214 | 25.7004 |
151
+ | 0.8545 | 2.01 | 84000 | 0.5215 | 28.2278 |
152
+ | 0.8594 | 2.02 | 85000 | 0.5186 | 25.6699 |
153
+ | 0.86 | 2.03 | 86000 | 0.5196 | 25.5786 |
154
+ | 0.8514 | 2.04 | 87000 | 0.5203 | 25.1827 |
155
+ | 0.8505 | 2.05 | 88000 | 0.5164 | 28.0146 |
156
+ | 0.8512 | 2.06 | 89000 | 0.5174 | 25.0914 |
157
+ | 0.8495 | 2.07 | 90000 | 0.5141 | 25.5481 |
158
+ | 0.8381 | 3.0 | 91000 | 0.5130 | 24.9695 |
159
+ | 0.8253 | 3.01 | 92000 | 0.5147 | 25.5786 |
160
+ | 0.8387 | 3.02 | 93000 | 0.5168 | 24.9086 |
161
+ | 0.8425 | 3.03 | 94000 | 0.5135 | 25.2436 |
162
+ | 0.8339 | 3.04 | 95000 | 0.5162 | 25.6699 |
163
+ | 0.8402 | 3.05 | 96000 | 0.5147 | 25.7308 |
164
+ | 0.8396 | 3.06 | 97000 | 0.5143 | 25.6699 |
165
+ | 0.8432 | 3.07 | 98000 | 0.5100 | 24.8782 |
166
+ | 0.844 | 3.08 | 99000 | 0.5100 | 25.0609 |
167
+ | 0.8333 | 4.01 | 100000 | 0.5128 | 24.9695 |
168
 
169
 
170
+ ### Framework versions
171
 
172
+ - Transformers 4.26.0.dev0
173
+ - Pytorch 1.13.0+cu117
174
+ - Datasets 2.7.1.dev0
175
+ - Tokenizers 0.13.2