Hugging Face's logo

patriotyk
/
vocos-mel-hifigan-compat-44100khz

Text-to-Speech
PyTorch
TensorBoard
vocos
hifigan
tts
melspectrogram
vocoder
mel
Model card Files Metrics Community
Serhiy Stetskovych commited on
Commit
8219f7d
•
1 Parent(s): 79906cb

Add configs

Browse files
Files changed (2) hide show
  1. logs/version_0/config.yaml +151 -0
  2. logs/version_1/config.yaml +151 -0
logs/version_0/config.yaml ADDED
@@ -0,0 +1,151 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # pytorch_lightning==1.8.6
2
+ seed_everything: 4444
3
+ trainer:
4
+ logger:
5
+ class_path: pytorch_lightning.loggers.TensorBoardLogger
6
+ init_args:
7
+ save_dir: /home/patriotyk/vocos/logs
8
+ name: lightning_logs
9
+ version: null
10
+ log_graph: false
11
+ default_hp_metric: true
12
+ prefix: ''
13
+ sub_dir: null
14
+ logdir: null
15
+ comment: ''
16
+ purge_step: null
17
+ max_queue: 10
18
+ flush_secs: 120
19
+ filename_suffix: ''
20
+ write_to_disk: true
21
+ comet_config:
22
+ disabled: true
23
+ enable_checkpointing: true
24
+ callbacks:
25
+ - class_path: pytorch_lightning.callbacks.LearningRateMonitor
26
+ init_args:
27
+ logging_interval: null
28
+ log_momentum: false
29
+ - class_path: pytorch_lightning.callbacks.ModelSummary
30
+ init_args:
31
+ max_depth: 2
32
+ - class_path: pytorch_lightning.callbacks.ModelCheckpoint
33
+ init_args:
34
+ dirpath: null
35
+ filename: vocos_checkpoint_{epoch}_{step}_{val_loss:.4f}
36
+ monitor: val_loss
37
+ verbose: false
38
+ save_last: true
39
+ save_top_k: 3
40
+ save_weights_only: false
41
+ mode: min
42
+ auto_insert_metric_name: true
43
+ every_n_train_steps: null
44
+ train_time_interval: null
45
+ every_n_epochs: null
46
+ save_on_train_epoch_end: null
47
+ - class_path: vocos.helpers.GradNormCallback
48
+ default_root_dir: null
49
+ gradient_clip_val: null
50
+ gradient_clip_algorithm: null
51
+ num_nodes: 1
52
+ num_processes: null
53
+ devices: -1
54
+ gpus: null
55
+ auto_select_gpus: false
56
+ tpu_cores: null
57
+ ipus: null
58
+ enable_progress_bar: true
59
+ overfit_batches: 0.0
60
+ track_grad_norm: -1
61
+ check_val_every_n_epoch: 1
62
+ fast_dev_run: false
63
+ accumulate_grad_batches: null
64
+ max_epochs: null
65
+ min_epochs: null
66
+ max_steps: -1
67
+ min_steps: null
68
+ max_time: null
69
+ limit_train_batches: null
70
+ limit_val_batches: 100
71
+ limit_test_batches: null
72
+ limit_predict_batches: null
73
+ val_check_interval: null
74
+ log_every_n_steps: 100
75
+ accelerator: gpu
76
+ strategy: ddp
77
+ sync_batchnorm: false
78
+ precision: 32
79
+ enable_model_summary: true
80
+ num_sanity_val_steps: 2
81
+ resume_from_checkpoint: null
82
+ profiler: null
83
+ benchmark: null
84
+ deterministic: null
85
+ reload_dataloaders_every_n_epochs: 0
86
+ auto_lr_find: false
87
+ replace_sampler_ddp: true
88
+ detect_anomaly: false
89
+ auto_scale_batch_size: false
90
+ plugins: null
91
+ amp_backend: native
92
+ amp_level: null
93
+ move_metrics_to_cpu: false
94
+ multiple_trainloader_mode: max_size_cycle
95
+ inference_mode: true
96
+ data:
97
+ class_path: vocos.dataset.VocosDataModule
98
+ init_args:
99
+ train_params:
100
+ filelist_path: /home/patriotyk/tts_corpus_44100/train_vocos.txt
101
+ sampling_rate: 44100
102
+ num_samples: 32768
103
+ batch_size: 20
104
+ num_workers: 24
105
+ val_params:
106
+ filelist_path: /home/patriotyk/tts_corpus_44100/val_vocos.txt
107
+ sampling_rate: 44100
108
+ num_samples: 96768
109
+ batch_size: 20
110
+ num_workers: 24
111
+ model:
112
+ class_path: vocos.experiment.VocosExp
113
+ init_args:
114
+ feature_extractor:
115
+ class_path: vocos.feature_extractors.MelSpectrogramFeatures
116
+ init_args:
117
+ sample_rate: 44100
118
+ n_fft: 2048
119
+ hop_length: 512
120
+ n_mels: 80
121
+ padding: same
122
+ f_min: 0
123
+ f_max: 8000
124
+ norm: slaney
125
+ mel_scale: slaney
126
+ backbone:
127
+ class_path: vocos.models.VocosBackbone
128
+ init_args:
129
+ input_channels: 80
130
+ dim: 512
131
+ intermediate_dim: 1536
132
+ num_layers: 8
133
+ layer_scale_init_value: null
134
+ adanorm_num_embeddings: null
135
+ head:
136
+ class_path: vocos.heads.ISTFTHead
137
+ init_args:
138
+ dim: 512
139
+ n_fft: 2048
140
+ hop_length: 512
141
+ padding: same
142
+ sample_rate: 44100
143
+ initial_learning_rate: 0.0003
144
+ num_warmup_steps: 0
145
+ mel_loss_coeff: 45.0
146
+ mrd_loss_coeff: 1.0
147
+ pretrain_mel_steps: 0
148
+ decay_mel_coeff: false
149
+ evaluate_utmos: true
150
+ evaluate_pesq: true
151
+ evaluate_periodicty: true
logs/version_1/config.yaml ADDED
@@ -0,0 +1,151 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # pytorch_lightning==1.8.6
2
+ seed_everything: 4444
3
+ trainer:
4
+ logger:
5
+ class_path: pytorch_lightning.loggers.TensorBoardLogger
6
+ init_args:
7
+ save_dir: /home/patriotyk/vocos/logs
8
+ name: lightning_logs
9
+ version: null
10
+ log_graph: false
11
+ default_hp_metric: true
12
+ prefix: ''
13
+ sub_dir: null
14
+ logdir: null
15
+ comment: ''
16
+ purge_step: null
17
+ max_queue: 10
18
+ flush_secs: 120
19
+ filename_suffix: ''
20
+ write_to_disk: true
21
+ comet_config:
22
+ disabled: true
23
+ enable_checkpointing: true
24
+ callbacks:
25
+ - class_path: pytorch_lightning.callbacks.LearningRateMonitor
26
+ init_args:
27
+ logging_interval: null
28
+ log_momentum: false
29
+ - class_path: pytorch_lightning.callbacks.ModelSummary
30
+ init_args:
31
+ max_depth: 2
32
+ - class_path: pytorch_lightning.callbacks.ModelCheckpoint
33
+ init_args:
34
+ dirpath: null
35
+ filename: vocos_checkpoint_{epoch}_{step}_{val_loss:.4f}
36
+ monitor: val_loss
37
+ verbose: false
38
+ save_last: true
39
+ save_top_k: 3
40
+ save_weights_only: false
41
+ mode: min
42
+ auto_insert_metric_name: true
43
+ every_n_train_steps: null
44
+ train_time_interval: null
45
+ every_n_epochs: null
46
+ save_on_train_epoch_end: null
47
+ - class_path: vocos.helpers.GradNormCallback
48
+ default_root_dir: null
49
+ gradient_clip_val: null
50
+ gradient_clip_algorithm: null
51
+ num_nodes: 1
52
+ num_processes: null
53
+ devices: -1
54
+ gpus: null
55
+ auto_select_gpus: false
56
+ tpu_cores: null
57
+ ipus: null
58
+ enable_progress_bar: true
59
+ overfit_batches: 0.0
60
+ track_grad_norm: -1
61
+ check_val_every_n_epoch: 1
62
+ fast_dev_run: false
63
+ accumulate_grad_batches: null
64
+ max_epochs: null
65
+ min_epochs: null
66
+ max_steps: 4000000
67
+ min_steps: null
68
+ max_time: null
69
+ limit_train_batches: null
70
+ limit_val_batches: 100
71
+ limit_test_batches: null
72
+ limit_predict_batches: null
73
+ val_check_interval: null
74
+ log_every_n_steps: 100
75
+ accelerator: gpu
76
+ strategy: ddp
77
+ sync_batchnorm: false
78
+ precision: 32
79
+ enable_model_summary: true
80
+ num_sanity_val_steps: 2
81
+ resume_from_checkpoint: ../vocos/logs/lightning_logs/version_10/checkpoints/last.ckpt
82
+ profiler: null
83
+ benchmark: null
84
+ deterministic: null
85
+ reload_dataloaders_every_n_epochs: 0
86
+ auto_lr_find: false
87
+ replace_sampler_ddp: true
88
+ detect_anomaly: false
89
+ auto_scale_batch_size: false
90
+ plugins: null
91
+ amp_backend: native
92
+ amp_level: null
93
+ move_metrics_to_cpu: false
94
+ multiple_trainloader_mode: max_size_cycle
95
+ inference_mode: true
96
+ data:
97
+ class_path: vocos.dataset.VocosDataModule
98
+ init_args:
99
+ train_params:
100
+ filelist_path: /home/patriotyk/tts_corpus_44100/train_vocos.txt
101
+ sampling_rate: 44100
102
+ num_samples: 32768
103
+ batch_size: 20
104
+ num_workers: 24
105
+ val_params:
106
+ filelist_path: /home/patriotyk/tts_corpus_44100/val_vocos.txt
107
+ sampling_rate: 44100
108
+ num_samples: 96768
109
+ batch_size: 20
110
+ num_workers: 24
111
+ model:
112
+ class_path: vocos.experiment.VocosExp
113
+ init_args:
114
+ feature_extractor:
115
+ class_path: vocos.feature_extractors.MelSpectrogramFeatures
116
+ init_args:
117
+ sample_rate: 44100
118
+ n_fft: 2048
119
+ hop_length: 512
120
+ n_mels: 80
121
+ padding: same
122
+ f_min: 0
123
+ f_max: 8000
124
+ norm: slaney
125
+ mel_scale: slaney
126
+ backbone:
127
+ class_path: vocos.models.VocosBackbone
128
+ init_args:
129
+ input_channels: 80
130
+ dim: 512
131
+ intermediate_dim: 1536
132
+ num_layers: 8
133
+ layer_scale_init_value: null
134
+ adanorm_num_embeddings: null
135
+ head:
136
+ class_path: vocos.heads.ISTFTHead
137
+ init_args:
138
+ dim: 512
139
+ n_fft: 2048
140
+ hop_length: 512
141
+ padding: same
142
+ sample_rate: 44100
143
+ initial_learning_rate: 0.0003
144
+ num_warmup_steps: 0
145
+ mel_loss_coeff: 45.0
146
+ mrd_loss_coeff: 1.0
147
+ pretrain_mel_steps: 0
148
+ decay_mel_coeff: false
149
+ evaluate_utmos: true
150
+ evaluate_pesq: true
151
+ evaluate_periodicty: true