Hugging Face's logo

sammy786
/
wav2vec2-xlsr-tatar

Automatic Speech Recognition
Transformers
PyTorch
Tatar
wav2vec2
Generated from Trainer
hf-asr-leaderboard
model_for_talk
mozilla-foundation/common_voice_8_0
robust-speech-event
Eval Results
Inference Endpoints
Model card Files Community
1
sammy786 commited on
Commit
37077d8
•
1 Parent(s): 49d3c29

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +153 -0
README.md ADDED
@@ -0,0 +1,153 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - tt
4
+ license: apache-2.0
5
+ tags:
6
+ - automatic-speech-recognition
7
+ - mozilla-foundation/common_voice_8_0
8
+ - generated_from_trainer
9
+ - tt
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-tatar
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: tt
25
+ metrics:
26
+ - name: Test WER
27
+ type: wer
28
+ value: 16.87
29
+ - name: Test CER
30
+ type: cer
31
+ value: 3.64
32
+ ---
33
+ # sammy786/wav2vec2-xlsr-tatar
34
+
35
+ 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 - tt dataset.
36
+ It achieves the following results on evaluation set (which is 10 percent of train data set merged with other and dev datasets):
37
+ - Loss: 7.66
38
+ - Wer: 7.08
39
+
40
+ ## Model description
41
+ "facebook/wav2vec2-xls-r-1b" was finetuned.
42
+
43
+ ## Intended uses & limitations
44
+ More information needed
45
+ ## Training and evaluation data
46
+ Training data -
47
+ Common voice Finnish train.tsv, dev.tsv and other.tsv
48
+
49
+ ## Training procedure
50
+ For creating the train dataset, all possible datasets were appended and 90-10 split was used.
51
+
52
+ ### Training hyperparameters
53
+
54
+ The following hyperparameters were used during training:
55
+
56
+ - learning_rate: 0.000045637994662983496
57
+ - train_batch_size: 16
58
+ - eval_batch_size: 16
59
+ - seed: 13
60
+ - gradient_accumulation_steps: 2
61
+ - total_train_batch_size: 32
62
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
63
+ - lr_scheduler_type: cosine_with_restarts
64
+ - lr_scheduler_warmup_steps: 500
65
+ - num_epochs: 40
66
+ - mixed_precision_training: Native AMP
67
+
68
+
69
+ ### Training results
70
+
71
+
72
+ | Step | Training Loss | Validation Loss | Wer |
73
+ |-------|---------------|-----------------|----------|
74
+ | 200 | 4.849400 | 1.874908 | 0.995232 |
75
+ | 400 | 1.105700 | 0.257292 | 0.367658 |
76
+ | 600 | 0.723000 | 0.181150 | 0.250513 |
77
+ | 800 | 0.660600 | 0.167009 | 0.226078 |
78
+ | 1000 | 0.568000 | 0.135090 | 0.177339 |
79
+ | 1200 | 0.721200 | 0.117469 | 0.166413 |
80
+ | 1400 | 0.416300 | 0.115142 | 0.153765 |
81
+ | 1600 | 0.346000 | 0.105782 | 0.153963 |
82
+ | 1800 | 0.279700 | 0.102452 | 0.146149 |
83
+ | 2000 | 0.273800 | 0.095818 | 0.128468 |
84
+ | 2200 | 0.252900 | 0.102302 | 0.133766 |
85
+ | 2400 | 0.255100 | 0.096592 | 0.121316 |
86
+ | 2600 | 0.229600 | 0.091263 | 0.124561 |
87
+ | 2800 | 0.213900 | 0.097748 | 0.125687 |
88
+ | 3000 | 0.210700 | 0.091244 | 0.125422 |
89
+ | 3200 | 0.202600 | 0.084076 | 0.106284 |
90
+ | 3400 | 0.200900 | 0.093809 | 0.113238 |
91
+ | 3600 | 0.192700 | 0.082918 | 0.108139 |
92
+ | 3800 | 0.182000 | 0.084487 | 0.103371 |
93
+ | 4000 | 0.167700 | 0.091847 | 0.104960 |
94
+ | 4200 | 0.183700 | 0.085223 | 0.103040 |
95
+ | 4400 | 0.174400 | 0.083862 | 0.100589 |
96
+ | 4600 | 0.163100 | 0.086493 | 0.099728 |
97
+ | 4800 | 0.162000 | 0.081734 | 0.097543 |
98
+ | 5000 | 0.153600 | 0.077223 | 0.092974 |
99
+ | 5200 | 0.153700 | 0.086217 | 0.090789 |
100
+ | 5400 | 0.140200 | 0.093256 | 0.100457 |
101
+ | 5600 | 0.142900 | 0.086903 | 0.097742 |
102
+ | 5800 | 0.131400 | 0.083068 | 0.095225 |
103
+ | 6000 | 0.126000 | 0.086642 | 0.091252 |
104
+ | 6200 | 0.135300 | 0.083387 | 0.091186 |
105
+ | 6400 | 0.126100 | 0.076479 | 0.086352 |
106
+ | 6600 | 0.127100 | 0.077868 | 0.086153 |
107
+ | 6800 | 0.118000 | 0.083878 | 0.087676 |
108
+ | 7000 | 0.117600 | 0.085779 | 0.091054 |
109
+ | 7200 | 0.113600 | 0.084197 | 0.084233 |
110
+ | 7400 | 0.112000 | 0.078688 | 0.081319 |
111
+ | 7600 | 0.110200 | 0.082534 | 0.086087 |
112
+ | 7800 | 0.106400 | 0.077245 | 0.080988 |
113
+ | 8000 | 0.102300 | 0.077497 | 0.079332 |
114
+ | 8200 | 0.109500 | 0.079083 | 0.088339 |
115
+ | 8400 | 0.095900 | 0.079721 | 0.077809 |
116
+ | 8600 | 0.094700 | 0.079078 | 0.079730 |
117
+ | 8800 | 0.097400 | 0.078785 | 0.079200 |
118
+ | 9000 | 0.093200 | 0.077445 | 0.077015 |
119
+ | 9200 | 0.088700 | 0.078207 | 0.076617 |
120
+ | 9400 | 0.087200 | 0.078982 | 0.076485 |
121
+ | 9600 | 0.089900 | 0.081209 | 0.076021 |
122
+ | 9800 | 0.081900 | 0.078158 | 0.075757 |
123
+ | 10000 | 0.080200 | 0.078074 | 0.074498 |
124
+ | 10200 | 0.085000 | 0.078830 | 0.073373 |
125
+ | 10400 | 0.080400 | 0.078144 | 0.073373 |
126
+ | 10600 | 0.078200 | 0.077163 | 0.073902 |
127
+ | 10800 | 0.080900 | 0.076394 | 0.072446 |
128
+ | 11000 | 0.080700 | 0.075955 | 0.071585 |
129
+ | 11200 | 0.076800 | 0.077031 | 0.072313 |
130
+ | 11400 | 0.076300 | 0.077401 | 0.072777 |
131
+ | 11600 | 0.076700 | 0.076613 | 0.071916 |
132
+ | 11800 | 0.076000 | 0.076672 | 0.071916 |
133
+ | 12000 | 0.077200 | 0.076490 | 0.070989 |
134
+ | 12200 | 0.076200 | 0.076688 | 0.070856 |
135
+ | 12400 | 0.074400 | 0.076780 | 0.071055 |
136
+ | 12600 | 0.076300 | 0.076768 | 0.071320 |
137
+ | 12800 | 0.077600 | 0.076727 | 0.071055 |
138
+ | 13000 | 0.077700 | 0.076714 | 0.071254 |
139
+
140
+
141
+ ### Framework versions
142
+ - Transformers 4.16.0.dev0
143
+ - Pytorch 1.10.0+cu102
144
+ - Datasets 1.17.1.dev0
145
+ - Tokenizers 0.10.3
146
+
147
+ #### Evaluation Commands
148
+
149
+ 1. To evaluate on `mozilla-foundation/common_voice_8_0` with split `test`
150
+
151
+ ```bash
152
+ python eval.py --model_id sammy786/wav2vec2-xlsr-tatar --dataset mozilla-foundation/common_voice_8_0 --config tt --split test
153
+ ```