siuze commited on
Commit
0d55394
1 Parent(s): 1144e58
exp/asr_train_asr_transformer_raw_can_char/config.yaml CHANGED
@@ -10,7 +10,12 @@ num_workers: 1
10
  num_att_plot: 3
11
  dist_backend: nccl
12
  dist_init_method: env://
 
 
13
  local_rank: 0
 
 
 
14
  multiprocessing_distributed: false
15
  unused_parameters: false
16
  sharded_ddp: false
@@ -20,6 +25,7 @@ cudnn_deterministic: true
20
  collect_stats: false
21
  write_collected_feats: false
22
  max_epoch: 30
 
23
  val_scheduler_criterion:
24
  - valid
25
  - loss
@@ -41,17 +47,26 @@ no_forward_run: false
41
  resume: true
42
  train_dtype: float32
43
  use_amp: false
 
44
  use_matplotlib: true
45
  use_tensorboard: true
46
  create_graph_in_tensorboard: false
47
  use_wandb: false
 
 
 
 
48
  wandb_model_log_interval: -1
49
  detect_anomaly: false
 
50
  init_param: []
51
  ignore_init_mismatch: false
52
  freeze_param: []
 
53
  batch_size: 16
 
54
  batch_bins: 1000000
 
55
  train_shape_file:
56
  - exp/asr_stats_raw_can_char/train/speech_shape
57
  - exp/asr_stats_raw_can_char/train/text_shape.char
@@ -59,6 +74,7 @@ valid_shape_file:
59
  - exp/asr_stats_raw_can_char/valid/speech_shape
60
  - exp/asr_stats_raw_can_char/valid/text_shape.char
61
  batch_type: folded
 
62
  fold_length:
63
  - 80000
64
  - 150
@@ -86,6 +102,7 @@ valid_data_path_and_name_and_type:
86
  allow_variable_data_keys: false
87
  max_cache_size: 0.0
88
  max_cache_fd: 32
 
89
  exclude_weight_decay: false
90
  exclude_weight_decay_conf: {}
91
  optim: adam
@@ -131,13 +148,22 @@ token_list:
131
  - q
132
  - <sos/eos>
133
  init: xavier_uniform
 
134
  ctc_conf:
135
  dropout_rate: 0.0
136
  ctc_type: builtin
137
  reduce: true
 
138
  use_preprocessor: true
139
  token_type: char
 
 
 
 
 
 
140
  rir_apply_prob: 1.0
 
141
  noise_apply_prob: 1.0
142
  noise_db_range: '13_15'
143
  short_noise_thres: 0.5
@@ -145,6 +171,7 @@ aux_ctc_tasks: []
145
  frontend: default
146
  frontend_conf:
147
  fs: 16k
 
148
  specaug_conf: {}
149
  normalize: global_mvn
150
  normalize_conf:
@@ -154,6 +181,7 @@ model_conf:
154
  ctc_weight: 0.3
155
  lsm_weight: 0.1
156
  length_normalized_loss: false
 
157
  preencoder_conf: {}
158
  encoder: transformer
159
  encoder_conf:
@@ -166,6 +194,7 @@ encoder_conf:
166
  attention_dropout_rate: 0.0
167
  input_layer: conv2d
168
  normalize_before: true
 
169
  postencoder_conf: {}
170
  decoder: transformer
171
  decoder_conf:
 
10
  num_att_plot: 3
11
  dist_backend: nccl
12
  dist_init_method: env://
13
+ dist_world_size: null
14
+ dist_rank: null
15
  local_rank: 0
16
+ dist_master_addr: null
17
+ dist_master_port: null
18
+ dist_launcher: null
19
  multiprocessing_distributed: false
20
  unused_parameters: false
21
  sharded_ddp: false
 
25
  collect_stats: false
26
  write_collected_feats: false
27
  max_epoch: 30
28
+ patience: null
29
  val_scheduler_criterion:
30
  - valid
31
  - loss
 
47
  resume: true
48
  train_dtype: float32
49
  use_amp: false
50
+ log_interval: null
51
  use_matplotlib: true
52
  use_tensorboard: true
53
  create_graph_in_tensorboard: false
54
  use_wandb: false
55
+ wandb_project: null
56
+ wandb_id: null
57
+ wandb_entity: null
58
+ wandb_name: null
59
  wandb_model_log_interval: -1
60
  detect_anomaly: false
61
+ pretrain_path: null
62
  init_param: []
63
  ignore_init_mismatch: false
64
  freeze_param: []
65
+ num_iters_per_epoch: null
66
  batch_size: 16
67
+ valid_batch_size: null
68
  batch_bins: 1000000
69
+ valid_batch_bins: null
70
  train_shape_file:
71
  - exp/asr_stats_raw_can_char/train/speech_shape
72
  - exp/asr_stats_raw_can_char/train/text_shape.char
 
74
  - exp/asr_stats_raw_can_char/valid/speech_shape
75
  - exp/asr_stats_raw_can_char/valid/text_shape.char
76
  batch_type: folded
77
+ valid_batch_type: null
78
  fold_length:
79
  - 80000
80
  - 150
 
102
  allow_variable_data_keys: false
103
  max_cache_size: 0.0
104
  max_cache_fd: 32
105
+ valid_max_cache_size: null
106
  exclude_weight_decay: false
107
  exclude_weight_decay_conf: {}
108
  optim: adam
 
148
  - q
149
  - <sos/eos>
150
  init: xavier_uniform
151
+ input_size: null
152
  ctc_conf:
153
  dropout_rate: 0.0
154
  ctc_type: builtin
155
  reduce: true
156
+ joint_net_conf: null
157
  use_preprocessor: true
158
  token_type: char
159
+ bpemodel: null
160
+ non_linguistic_symbols: null
161
+ cleaner: null
162
+ g2p: null
163
+ speech_volume_normalize: null
164
+ rir_scp: null
165
  rir_apply_prob: 1.0
166
+ noise_scp: null
167
  noise_apply_prob: 1.0
168
  noise_db_range: '13_15'
169
  short_noise_thres: 0.5
 
171
  frontend: default
172
  frontend_conf:
173
  fs: 16k
174
+ specaug: null
175
  specaug_conf: {}
176
  normalize: global_mvn
177
  normalize_conf:
 
181
  ctc_weight: 0.3
182
  lsm_weight: 0.1
183
  length_normalized_loss: false
184
+ preencoder: null
185
  preencoder_conf: {}
186
  encoder: transformer
187
  encoder_conf:
 
194
  attention_dropout_rate: 0.0
195
  input_layer: conv2d
196
  normalize_before: true
197
+ postencoder: null
198
  postencoder_conf: {}
199
  decoder: transformer
200
  decoder_conf: