Hugging Face's logo

marinone94
/
xls-r-300m-sv-robust

Automatic Speech Recognition
Transformers
PyTorch
Swedish
wav2vec2
mozilla-foundation/common_voice_9_0
generated_from_trainer
Eval Results
Inference Endpoints
Model card Files Community
1
marinone94 commited on
Commit
6caeebe
•
1 Parent(s): be4e905

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +21 -80
README.md CHANGED
@@ -1,73 +1,23 @@
1
  ---
2
- language:
3
- - sv-SE
4
  license: cc0-1.0
5
  tags:
6
- - automatic-speech-recognition
7
  - generated_from_trainer
8
- - hf-asr-leaderboard
9
- - model_for_talk
10
- - mozilla-foundation/common_voice_8_0
11
- - robust-speech-event
12
- - sv
13
  datasets:
14
- - mozilla-foundation/common_voice_8_0
15
  model-index:
16
- - name: XLS-R-300M - Swedish
17
- results:
18
- - task:
19
- name: Automatic Speech Recognition
20
- type: automatic-speech-recognition
21
- dataset:
22
- name: mozilla-foundation/common_voice_8_0
23
- type: mozilla-foundation/common_voice_8_0
24
- args: sv-SE
25
- metrics:
26
- - name: Test WER
27
- type: wer
28
- value: 8.72
29
- - name: Test CER
30
- type: cer
31
- value: 3.05
32
- - task:
33
- name: Automatic Speech Recognition
34
- type: automatic-speech-recognition
35
- dataset:
36
- name: speech-recognition-community-v2/eval_data
37
- type: speech-recognition-community-v2/eval_data
38
- args: sv
39
- metrics:
40
- - name: Validation WER
41
- type: wer
42
- value: 19.67
43
- - name: Validation CER
44
- type: cer
45
- value: 8.94
46
- - task:
47
- name: Automatic Speech Recognition
48
- type: automatic-speech-recognition
49
- dataset:
50
- name: speech-recognition-community-v2/eval_data
51
- type: speech-recognition-community-v2/eval_data
52
- args: sv
53
- metrics:
54
- - name: Test WER
55
- type: wer
56
- value: 15.94
57
- - name: Test CER
58
- type: cer
59
- value: 7.71
60
- widget:
61
- - example_title: Swedish
62
- src: https://cdn-media.huggingface.co/speech_samples/cv_swedish_1.mp3
63
  ---
64
 
 
 
 
65
  #
66
 
67
- This model is a fine-tuned version of [KBLab/wav2vec2-large-voxrex](https://huggingface.co/KBLab/wav2vec2-large-voxrex) on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - SV-SE dataset.
68
  It achieves the following results on the evaluation set:
69
- - Loss: 0.1595
70
- - Wer: 0.1200
71
 
72
  ## Model description
73
 
@@ -86,7 +36,7 @@ More information needed
86
  ### Training hyperparameters
87
 
88
  The following hyperparameters were used during training:
89
- - learning_rate: 0.00025
90
  - train_batch_size: 32
91
  - eval_batch_size: 32
92
  - seed: 42
@@ -94,7 +44,7 @@ The following hyperparameters were used during training:
94
  - total_train_batch_size: 128
95
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
96
  - lr_scheduler_type: linear
97
- - lr_scheduler_warmup_ratio: 0.25
98
  - num_epochs: 100.0
99
  - mixed_precision_training: Native AMP
100
 
@@ -102,29 +52,20 @@ The following hyperparameters were used during training:
102
 
103
  | Training Loss | Epoch | Step | Validation Loss | Wer |
104
  |:-------------:|:-----:|:----:|:---------------:|:------:|
105
- | 3.0418 | 5.49 | 500 | 3.0176 | 1.0 |
106
- | 1.1819 | 10.98 | 1000 | 0.2562 | 0.2168 |
107
- | 1.0032 | 16.48 | 1500 | 0.1746 | 0.1546 |
108
- | 0.9077 | 21.97 | 2000 | 0.1600 | 0.1339 |
109
- | 0.8687 | 27.47 | 2500 | 0.1647 | 0.1378 |
110
- | 0.8081 | 32.96 | 3000 | 0.1608 | 0.1353 |
111
- | 0.7923 | 38.46 | 3500 | 0.1534 | 0.1277 |
112
- | 0.7349 | 43.95 | 4000 | 0.1546 | 0.1303 |
113
- | 0.7199 | 49.45 | 4500 | 0.1617 | 0.1277 |
114
- | 0.7028 | 54.94 | 5000 | 0.1572 | 0.1287 |
115
- | 0.6912 | 60.44 | 5500 | 0.1560 | 0.1249 |
116
- | 0.6492 | 65.93 | 6000 | 0.1542 | 0.1260 |
117
- | 0.6407 | 71.43 | 6500 | 0.1605 | 0.1240 |
118
- | 0.6222 | 76.92 | 7000 | 0.1577 | 0.1219 |
119
- | 0.6039 | 82.42 | 7500 | 0.1645 | 0.1249 |
120
- | 0.5928 | 87.91 | 8000 | 0.1590 | 0.1214 |
121
- | 0.6022 | 93.4 | 8500 | 0.1597 | 0.1213 |
122
- | 0.5814 | 98.9 | 9000 | 0.1599 | 0.1199 |
123
 
124
 
125
  ### Framework versions
126
 
127
  - Transformers 4.17.0.dev0
128
  - Pytorch 1.10.2+cu102
129
- - Datasets 1.18.3
130
  - Tokenizers 0.11.0
 
1
  ---
 
 
2
  license: cc0-1.0
3
  tags:
 
4
  - generated_from_trainer
 
 
 
 
 
5
  datasets:
6
+ - common_voice
7
  model-index:
8
+ - name: ''
9
+ results: []
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  ---
11
 
12
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
13
+ should probably proofread and complete it, then remove this comment. -->
14
+
15
  #
16
 
17
+ This model is a fine-tuned version of [KBLab/wav2vec2-large-voxrex](https://huggingface.co/KBLab/wav2vec2-large-voxrex) on the common_voice dataset.
18
  It achieves the following results on the evaluation set:
19
+ - Loss: 0.1321
20
+ - Wer: 0.1115
21
 
22
  ## Model description
23
 
 
36
  ### Training hyperparameters
37
 
38
  The following hyperparameters were used during training:
39
+ - learning_rate: 7.5e-05
40
  - train_batch_size: 32
41
  - eval_batch_size: 32
42
  - seed: 42
 
44
  - total_train_batch_size: 128
45
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
46
  - lr_scheduler_type: linear
47
+ - lr_scheduler_warmup_ratio: 0.2
48
  - num_epochs: 100.0
49
  - mixed_precision_training: Native AMP
50
 
 
52
 
53
  | Training Loss | Epoch | Step | Validation Loss | Wer |
54
  |:-------------:|:-----:|:----:|:---------------:|:------:|
55
+ | 2.9099 | 10.42 | 1000 | 2.8369 | 1.0 |
56
+ | 1.0745 | 20.83 | 2000 | 0.1957 | 0.1673 |
57
+ | 0.934 | 31.25 | 3000 | 0.1579 | 0.1389 |
58
+ | 0.8691 | 41.66 | 4000 | 0.1457 | 0.1290 |
59
+ | 0.8328 | 52.08 | 5000 | 0.1435 | 0.1205 |
60
+ | 0.8068 | 62.5 | 6000 | 0.1350 | 0.1191 |
61
+ | 0.7822 | 72.91 | 7000 | 0.1347 | 0.1155 |
62
+ | 0.7769 | 83.33 | 8000 | 0.1321 | 0.1131 |
63
+ | 0.7678 | 93.75 | 9000 | 0.1321 | 0.1115 |
 
 
 
 
 
 
 
 
 
64
 
65
 
66
  ### Framework versions
67
 
68
  - Transformers 4.17.0.dev0
69
  - Pytorch 1.10.2+cu102
70
+ - Datasets 2.2.2
71
  - Tokenizers 0.11.0