Hugging Face's logo

lsnoo
/
russian_fairseq

Automatic Speech Recognition
Transformers
PyTorch
wav2vec2
Inference Endpoints
Model card Files Community
1
russian_fairseq
File size: 1,089 Bytes 
6fe790f
 
 
aa8ef91
0a61521
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
 
Wav2vec2.0-xlsr-53 model is fine-tuned on commonvoice russian dataset

Configs (yaml)

  checkpoint:
    save_interval: 1000
    save_interval_updates: 1000
    keep_interval_updates: 1
    no_epoch_ckechpoints: true
    best_checkpoint_metric: wer
 
  task:
    _name: audio_finetuning
    normalize: true
    labels: phn

  dataset:
    num_workers: 6
    max_tokens: 800000
    skip_invalid_size_inputs_valid_test: true
    valid_subset: valie

  distributed_training:
    ddp_backend: legacy_ddp
    distributed_world_size: 4

  criterion:
    _name: ctc
    zero_infinity: true

  optimization:
    max_update: 25000
    lr: [0.00001]
    sentence_avg: true
    update_freq: [4]

  optimizer:
    _name: adam
    adam_betas: (0.9, 0.98)
    adam_eps: 1e-8

  lr_scheduler:
    _name: tri_stage
    phase_ratio: [0.1, 0.4, 0.5]
    final_lr_scale: 0.05

  model:
    _name: wav2vec_ctc
    apply_mask: true
    mask_prob: 0.5
    mask_channel_prob: 0.1
    mask_channel_length: 64
    layerdrop: 0.1
    activation_dropout: 0.1
    feature_grad_mult: 0.0
    freeze_finetune_updates: 0