sammy786 commited on
Commit
480a0cf
1 Parent(s): 57da202

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +103 -0
README.md ADDED
@@ -0,0 +1,103 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - et
4
+ license: apache-2.0
5
+ tags:
6
+ - automatic-speech-recognition
7
+ - mozilla-foundation/common_voice_8_0
8
+ - generated_from_trainer
9
+ - et
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-estonian
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: et
25
+ metrics:
26
+ - name: Test WER
27
+ type: wer
28
+ value: 23.61
29
+ - name: Test CER
30
+ type: cer
31
+ value: 4.60
32
+ ---
33
+ # sammy786/wav2vec2-xlsr-estonian
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 - et 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: 17.94
38
+ - Wer: 30.38
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: 8
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: 30
66
+ - mixed_precision_training: Native AMP
67
+
68
+
69
+ ### Training results
70
+
71
+
72
+ | Step | Training Loss | Validation Loss | Wer |
73
+ |------|---------------|-----------------|----------|
74
+ | 200 | 3.729100 | 1.096018 | 0.959867 |
75
+ | 400 | 0.996900 | 0.310228 | 0.443600 |
76
+ | 600 | 0.762900 | 0.210873 | 0.346117 |
77
+ | 800 | 0.621400 | 0.200381 | 0.331513 |
78
+ | 1000 | 0.408000 | 0.196382 | 0.322014 |
79
+ | 1200 | 0.320200 | 0.176281 | 0.312515 |
80
+ | 1400 | 0.315300 | 0.179433 | 0.303847 |
81
+ | 1600 | 0.445800 | 0.420985 | 0.315839 |
82
+ | 1800 | 0.644600 | 0.433833 | 0.354904 |
83
+ | 2000 | 0.550900 | 0.327117 | 0.336500 |
84
+ | 2200 | 0.498600 | 0.289830 | 0.325457 |
85
+ | 2400 | 0.488300 | 0.294309 | 0.314177 |
86
+ | 2600 | 0.491700 | 0.311175 | 0.318689 |
87
+ | 2800 | 0.508500 | 0.314744 | 0.320470 |
88
+ | 3000 | 0.499900 | 0.314834 | 0.320589 |
89
+
90
+
91
+ ### Framework versions
92
+ - Transformers 4.16.0.dev0
93
+ - Pytorch 1.10.0+cu102
94
+ - Datasets 1.17.1.dev0
95
+ - Tokenizers 0.10.3
96
+
97
+ #### Evaluation Commands
98
+
99
+ 1. To evaluate on `mozilla-foundation/common_voice_8_0` with split `test`
100
+
101
+ ```bash
102
+ python eval.py --model_id sammy786/wav2vec2-xlsr-estonian --dataset mozilla-foundation/common_voice_8_0 --config et --split test
103
+ ```