DiffSpeech / egs /egs_bases /tts /ps_flow_small.yaml
RayeRen's picture
update
15e73a1
raw
history blame
747 Bytes
base_config: ./ps_flow.yaml
###########################
# models
###########################
# encoders
hidden_size: 128
ffn_hidden_size: 512
enc_ffn_kernel_size: 3
enc_layers: 3
word_enc_layers: 3
# dur predictor
dur_predictor_layers: 3
dur_predictor_kernel: 5
predictor_dropout: 0.2
## fvae
use_fvae: true
latent_size: 16
fvae_encoder_type: wn
fvae_decoder_type: wn
fvae_enc_dec_hidden: 128
fvae_kernel_size: 3
fvae_enc_n_layers: 8
fvae_dec_n_layers: 3
fvae_strides: 4
fvae_noise_scale: 1.0
# prior flow
use_prior_flow: true
prior_flow_hidden: 32
prior_flow_kernel_size: 3
prior_flow_n_blocks: 3
# post flow
post_glow_hidden: 128
post_glow_kernel_size: 3
post_glow_n_blocks: 8
post_glow_n_block_layers: 3
share_wn_layers: 4
noise_scale: 0.6