You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
92 lines
5.1 KiB
92 lines
5.1 KiB
3 years ago
|
|
||
|
|
fs : 22050 # Hz, sample rate
|
||
|
|
n_fft : 1024 # fft frame size
|
||
|
|
win_length : 1024 # window size
|
||
|
|
n_shift : 256 # hop size between ajacent frame
|
||
|
|
fmin : 0 # Hz, min frequency when converting to mel
|
||
|
|
fmax : 8000 # Hz, max frequency when converting to mel
|
||
|
|
n_mels : 80 # mel bands
|
||
|
|
window: "hann" # Window function.
|
||
|
|
|
||
|
|
###########################################################
|
||
|
|
# DATA SETTING #
|
||
|
|
###########################################################
|
||
|
|
batch_size: 16
|
||
|
|
num_workers: 2
|
||
|
|
|
||
|
|
##########################################################
|
||
|
|
# TTS MODEL SETTING #
|
||
|
|
##########################################################
|
||
|
|
tts: transformertts # model architecture
|
||
|
|
model: # keyword arguments for the selected model
|
||
|
|
embed_dim: 0 # embedding dimension in encoder prenet
|
||
|
|
eprenet_conv_layers: 0 # number of conv layers in encoder prenet
|
||
|
|
# if set to 0, no encoder prenet will be used
|
||
|
|
eprenet_conv_filts: 0 # filter size of conv layers in encoder prenet
|
||
|
|
eprenet_conv_chans: 0 # number of channels of conv layers in encoder prenet
|
||
|
|
dprenet_layers: 2 # number of layers in decoder prenet
|
||
|
|
dprenet_units: 256 # number of units in decoder prenet
|
||
|
|
adim: 512 # attention dimension
|
||
|
|
aheads: 8 # number of attention heads
|
||
|
|
elayers: 6 # number of encoder layers
|
||
|
|
eunits: 1024 # number of encoder ff units
|
||
|
|
dlayers: 6 # number of decoder layers
|
||
|
|
dunits: 1024 # number of decoder ff units
|
||
|
|
positionwise_layer_type: conv1d # type of position-wise layer
|
||
|
|
positionwise_conv_kernel_size: 1 # kernel size of position wise conv layer
|
||
|
|
postnet_layers: 5 # number of layers of postnset
|
||
|
|
postnet_filts: 5 # filter size of conv layers in postnet
|
||
|
|
postnet_chans: 256 # number of channels of conv layers in postnet
|
||
|
|
use_scaled_pos_enc: True # whether to use scaled positional encoding
|
||
|
|
encoder_normalize_before: True # whether to perform layer normalization before the input
|
||
|
|
decoder_normalize_before: True # whether to perform layer normalization before the input
|
||
|
|
reduction_factor: 1 # reduction factor
|
||
|
|
init_type: xavier_uniform # initialization type
|
||
|
|
init_enc_alpha: 1.0 # initial value of alpha of encoder scaled position encoding
|
||
|
|
init_dec_alpha: 1.0 # initial value of alpha of decoder scaled position encoding
|
||
|
|
eprenet_dropout_rate: 0.0 # dropout rate for encoder prenet
|
||
|
|
dprenet_dropout_rate: 0.5 # dropout rate for decoder prenet
|
||
|
|
postnet_dropout_rate: 0.5 # dropout rate for postnet
|
||
|
|
transformer_enc_dropout_rate: 0.1 # dropout rate for transformer encoder layer
|
||
|
|
transformer_enc_positional_dropout_rate: 0.1 # dropout rate for transformer encoder positional encoding
|
||
|
|
transformer_enc_attn_dropout_rate: 0.1 # dropout rate for transformer encoder attention layer
|
||
|
|
transformer_dec_dropout_rate: 0.1 # dropout rate for transformer decoder layer
|
||
|
|
transformer_dec_positional_dropout_rate: 0.1 # dropout rate for transformer decoder positional encoding
|
||
|
|
transformer_dec_attn_dropout_rate: 0.1 # dropout rate for transformer decoder attention layer
|
||
|
|
transformer_enc_dec_attn_dropout_rate: 0.1 # dropout rate for transformer encoder-decoder attention layer
|
||
|
|
num_heads_applied_guided_attn: 2 # number of heads to apply guided attention loss
|
||
|
|
num_layers_applied_guided_attn: 2 # number of layers to apply guided attention loss
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
###########################################################
|
||
|
|
# UPDATER SETTING #
|
||
|
|
###########################################################
|
||
|
|
updater:
|
||
|
|
use_masking: true # whether to apply masking for padded part in loss calculation
|
||
|
|
loss_type: L1
|
||
|
|
use_guided_attn_loss: true # whether to use guided attention loss
|
||
|
|
guided_attn_loss_sigma: 0.4 # sigma in guided attention loss
|
||
|
|
guided_attn_loss_lambda: 10.0 # lambda in guided attention loss
|
||
|
|
modules_applied_guided_attn: ["encoder-decoder"] # modules to apply guided attention loss
|
||
|
|
bce_pos_weight: 5.0 # weight of positive sample in binary cross entropy calculation
|
||
|
|
|
||
|
|
|
||
|
|
##########################################################
|
||
|
|
# OPTIMIZER & SCHEDULER SETTING #
|
||
|
|
##########################################################
|
||
|
|
optimizer:
|
||
|
|
optim: adam # optimizer type
|
||
|
|
learning_rate: 0.001 # learning rate
|
||
|
|
|
||
|
|
###########################################################
|
||
|
|
# TRAINING SETTING #
|
||
|
|
###########################################################
|
||
|
|
max_epoch: 500
|
||
|
|
num_snapshots: 5
|
||
|
|
|
||
|
|
|
||
|
|
###########################################################
|
||
|
|
# OTHER SETTING #
|
||
|
|
###########################################################
|
||
|
|
seed: 10086
|