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
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
| # network architecture
| # encoder related
| encoder: data2vec_encoder
| encoder_conf:
| extractor_mode: layer_norm
| encoder_layerdrop: 0.1
| dropout_input: 0.0
| dropout_features: 0.0
| feature_grad_mult: 0.0
| encoder_embed_dim: 768
|
| mask_prob: 0.65
| mask_length: 10
|
| loss_beta: 0
| loss_scale: null
|
| instance_norm_target_layer: true
| average_top_k_layers: 8
|
| pos_conv_depth: 5
| conv_pos: 95
|
| ema_decay: 0.999
| ema_end_decay: 0.9999
| ema_anneal_end_step: 30000
| ema_transformer_only: true
| ema_layers_only: true
|
| require_same_masks: true
| mask_dropout: 0
|
| # frontend related
| frontend: wav_frontend
| frontend_conf:
| fs: 16000
| window: hamming
| n_mels: 80
| frame_length: 25
| frame_shift: 10
| lfr_m: 1
| lfr_n: 1
|
| # hybrid CTC/attention
| model_conf:
| ctc_weight: 1.0
| lsm_weight: 0.1 # label smoothing option
| length_normalized_loss: false
|
| # optimization related
| accum_grad: 1
| grad_clip: 5
| patience: none
| max_epoch: 150
| val_scheduler_criterion:
| - valid
| - acc
| best_model_criterion:
| - - valid
| - cer_ctc
| - min
| keep_nbest_models: 10
|
| # NoamLR is deprecated. Use WarmupLR.
| # The following is equivalent setting for NoamLR:
| #
| # optim: adam
| # optim_conf:
| # lr: 10.
| # scheduler: noamlr
| # scheduler_conf:
| # model_size: 256
| # warmup_steps: 25000
| #
| optim: adam
| optim_conf:
| lr: 0.00005
| scheduler: warmuplr # pytorch v1.1.0+ required
| scheduler_conf:
| warmup_steps: 25000
|
| specaug: specaug
| specaug_conf:
| apply_time_warp: true
| time_warp_window: 5
| time_warp_mode: bicubic
| apply_freq_mask: true
| freq_mask_width_range:
| - 0
| - 30
| num_freq_mask: 2
| apply_time_mask: true
| time_mask_width_range:
| - 0
| - 40
| num_time_mask: 2
|
| dataset_conf:
| data_names: speech,text
| data_types: sound,text
| shuffle: True
| shuffle_conf:
| shuffle_size: 2048
| sort_size: 500
| batch_conf:
| batch_type: token
| batch_size: 25000
| num_workers: 8
|
| log_interval: 50
| unused_parameters: true
| normalize: None
|
|
