Hugging Face's logo

sammy786
/
wav2vec2-xlsr-chuvash

Automatic Speech Recognition
Transformers
PyTorch
Chuvash
wav2vec2
mozilla-foundation/common_voice_8_0
Generated from Trainer
robust-speech-event
model_for_talk
hf-asr-leaderboard
Eval Results
Inference Endpoints
Model card Files Community
sammy786 commited on
Commit
1fb170a
•
1 Parent(s): f23cd0e

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +121 -0
README.md ADDED
@@ -0,0 +1,121 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - cv
4
+ license: apache-2.0
5
+ tags:
6
+ - automatic-speech-recognition
7
+ - mozilla-foundation/common_voice_8_0
8
+ - generated_from_trainer
9
+ - cv
10
+ - robust-speech-event
11
+ - model_for_talk
12
+ datasets:
13
+ - mozilla-foundation/common_voice_8_0
14
+
15
+ model-index:
16
+ - name: sammy786/wav2vec2-xlsr-chuvash
17
+ results:
18
+ - task:
19
+ name: Automatic Speech Recognition
20
+ type: automatic-speech-recognition
21
+ dataset:
22
+ name: Common Voice 8
23
+ type: mozilla-foundation/common_voice_8_0
24
+ args: cv
25
+ metrics:
26
+ - name: Test WER
27
+ type: wer
28
+ value: 32.11
29
+ - name: Test CER
30
+ type: cer
31
+ value: 7.49
32
+ -
33
+ ---
34
+ # sammy786/wav2vec2-xlsr-chuvash
35
+
36
+ This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - cv dataset.
37
+ It achieves the following results on evaluation set (which is 10 percent of train data set merged with other and dev datasets):
38
+ - Loss: 18.02
39
+ - Wer: 29.22
40
+
41
+ ## Model description
42
+ "facebook/wav2vec2-xls-r-1b" was finetuned.
43
+
44
+ ## Intended uses & limitations
45
+ More information needed
46
+ ## Training and evaluation data
47
+ Training data -
48
+ Common voice Finnish train.tsv, dev.tsv and other.tsv
49
+
50
+ ## Training procedure
51
+ For creating the train dataset, all possible datasets were appended and 90-10 split was used.
52
+
53
+ ### Training hyperparameters
54
+
55
+ The following hyperparameters were used during training:
56
+
57
+ - learning_rate: 0.000045637994662983496
58
+ - train_batch_size: 8
59
+ - eval_batch_size: 16
60
+ - seed: 13
61
+ - gradient_accumulation_steps: 4
62
+ - total_train_batch_size: 32
63
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
64
+ - lr_scheduler_type: cosine_with_restarts
65
+ - lr_scheduler_warmup_steps: 500
66
+ - num_epochs: 30
67
+ - mixed_precision_training: Native AMP
68
+
69
+
70
+ ### Training results
71
+
72
+
73
+ | Step | Training Loss | Validation Loss | Wer |
74
+ |:----:|:-------------:|:---------------:|:--------:|
75
+ | 200 | 6.559100 | 2.274687 | 1.000000 |
76
+ | 400 | 1.346100 | 0.508268 | 0.681995 |
77
+ | 600 | 0.797500 | 0.391174 | 0.572876 |
78
+ | 800 | 0.556300 | 0.308620 | 0.489283 |
79
+ | 1000 | 0.435800 | 0.273956 | 0.454014 |
80
+ | 1200 | 0.388700 | 0.311027 | 0.499415 |
81
+ | 1400 | 0.338300 | 0.243977 | 0.413874 |
82
+ | 1600 | 0.294000 | 0.214134 | 0.385230 |
83
+ | 1800 | 0.276000 | 0.245991 | 0.397311 |
84
+ | 2000 | 0.253900 | 0.208324 | 0.363016 |
85
+ | 2200 | 0.233600 | 0.222156 | 0.370811 |
86
+ | 2400 | 0.219700 | 0.202602 | 0.364186 |
87
+ | 2600 | 0.205000 | 0.241339 | 0.384451 |
88
+ | 2800 | 0.176000 | 0.263558 | 0.384061 |
89
+ | 3000 | 0.166700 | 0.211768 | 0.333398 |
90
+ | 3200 | 0.160600 | 0.198677 | 0.321512 |
91
+ | 3400 | 0.154600 | 0.208655 | 0.328722 |
92
+ | 3600 | 0.146800 | 0.188022 | 0.317810 |
93
+ | 3800 | 0.133200 | 0.181083 | 0.313133 |
94
+ | 4000 | 0.134200 | 0.190084 | 0.316251 |
95
+ | 4200 | 0.114200 | 0.193034 | 0.312159 |
96
+ | 4400 | 0.117300 | 0.194122 | 0.312354 |
97
+ | 4600 | 0.112300 | 0.191111 | 0.305534 |
98
+ | 4800 | 0.107800 | 0.185930 | 0.302611 |
99
+ | 5000 | 0.100400 | 0.178625 | 0.299883 |
100
+ | 5200 | 0.099800 | 0.176442 | 0.294622 |
101
+ | 5400 | 0.100800 | 0.177935 | 0.294427 |
102
+ | 5600 | 0.096300 | 0.182903 | 0.293843 |
103
+ | 5800 | 0.094200 | 0.181041 | 0.293453 |
104
+ | 6000 | 0.097600 | 0.179865 | 0.290725 |
105
+ | 6200 | 0.091600 | 0.180327 | 0.292868 |
106
+ | 6400 | 0.093100 | 0.180275 | 0.292284 |
107
+
108
+
109
+ ### Framework versions
110
+ - Transformers 4.16.0.dev0
111
+ - Pytorch 1.10.0+cu102
112
+ - Datasets 1.17.1.dev0
113
+ - Tokenizers 0.10.3
114
+
115
+ #### Evaluation Commands
116
+
117
+ 1. To evaluate on `mozilla-foundation/common_voice_8_0` with split `test`
118
+
119
+ ```bash
120
+ python eval.py --model_id sammy786/wav2vec2-xlsr-chuvash --dataset mozilla-foundation/common_voice_8_0 --config cv --split test
121
+ ```