38 lines
553 B
YAML
38 lines
553 B
YAML
|
audio:
|
||
|
|
num_mels: 80
|
||
|
|
n_fft: 2048
|
||
|
|
sr: 22050
|
||
|
|
preemphasis: 0.97
|
||
|
|
hop_length: 256 #275
|
||
|
|
win_length: 1024 #1102
|
||
|
|
power: 1.2
|
||
|
|
min_level_db: -100
|
||
|
|
ref_level_db: 20
|
||
|
|
|
||
|
|
network:
|
||
|
|
hidden_size: 256
|
||
|
|
embedding_size: 512
|
||
|
|
encoder_num_head: 4
|
||
|
|
encoder_n_layers: 3
|
||
|
|
decoder_num_head: 4
|
||
|
|
decoder_n_layers: 3
|
||
|
|
outputs_per_step: 1
|
||
|
|
stop_token: False
|
||
|
|
|
||
|
|
vocoder:
|
||
|
|
hidden_size: 256
|
||
|
|
|
||
|
|
train:
|
||
|
|
batch_size: 32
|
||
|
|
learning_rate: 0.001
|
||
|
|
warm_up_step: 4000
|
||
|
|
grad_clip_thresh: 1.0
|
||
|
|
|
||
|
|
checkpoint_interval: 1000
|
||
|
|
image_interval: 2000
|
||
|
|
|
||
|
|
max_epochs: 10000
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
|