hparams = HParams(
# model
freq=8,
dim_neck=8,
freq_2=8,
dim_neck_2=1,
freq_3=8,
dim_neck_3=32,
dim_enc=512,
dim_enc_2=128,
dim_enc_3=256,
dim_freq=80,
dim_spk_emb=512, # 82,
dim_f0=257,
dim_dec=512,
len_raw=128,
chs_grp=16,
# interp
min_len_seg=19,
max_len_seg=32,
min_len_seq=64,
max_len_seq=128,
max_len_pad=192,
# data loader
root_dir='assets/spmel',
feat_dir='assets/raptf0',
batch_size=16,
mode='train',
shuffle=True,
num_workers=0,
samplier=8,
)
hparams = HParams(
name='wavenet_vocoder',
builder='wavenet',
input_type='raw',
quantize_channels=65536,
sample_rate=16000,
silence_threshold=2,
num_mels=80,
fmin=125,
fmax=7600,
fft_size=1024,
hop_size=256,
frame_shift_ms=None,
min_level_db=-100,
ref_level_db=20,
rescaling=True,
rescaling_max=0.999,
allow_clipping_in_normalization=True,
log_scale_min=-32.23619130191664,
out_channels=30,
layers=24,
stacks=4,
residual_channels=512,
gate_channels=512,
skip_out_channels=256,
dropout=0.050000000000000044,
kernel_size=3,
weight_normalization=True,
legacy=True,
cin_channels=80,
upsample_conditional_features=True,
upsample_scales=[4, 4, 4, 4],
freq_axis_kernel_size=3,
gin_channels=-1,
n_speakers=-1,
pin_memory=True,
test_size=0.0441,
test_num_samples=None,
random_state=1234,
adam_beta1=0.9,
adam_beta2=0.999,
adam_eps=1e-08,
amsgrad=False,
initial_learning_rate=0.001,
lr_schedule='noam_learning_rate_decay',
lr_schedule_kwargs={},
nepochs=2000,
weight_decay=0.0,
clip_thresh=-1,
max_time_sec=None,
max_time_steps=8000,
exponential_moving_average=True,
ema_decay=0.9999,
checkpoint_interval=10000,
train_eval_interval=10000,
test_eval_epoch_interval=5,
save_optimizer_state=True,
# model
freq=8,
dim_neck=8,
freq_2=8,
dim_neck_2=1,
freq_3=8,
dim_neck_3=32,
dim_enc=512,
dim_enc_2=128,
dim_enc_3=256,
dim_freq=80,
dim_spk_emb=82,
dim_f0=257,
dim_dec=512,
len_raw=128,
chs_grp=16,
# interp
min_len_seg=19,
max_len_seg=32,
min_len_seq=64,
max_len_seq=128,
max_len_pad=192,
# data loader
root_dir='assets/spmel',
feat_dir='assets/raptf0',
batch_size=16,
mode='train',
shuffle=True,
num_workers=0,
samplier=8,
)
hparams = HParams(
loss_reconstruction_w=9,
loss_disentanglement_w=1,
# threshold=4e-4,
# model
freq=8, # content codes降采样
dim_neck=8, # (blstm dim)content_encoder
freq_2=8, # rhythm codes降采样 rhythm_encoder(encoder_2,encoder_t)
dim_neck_2=1, # (blstm dim)rhythm_encoder(encoder_2,encoder_t)
freq_3=8, # pitch codes降采样
dim_neck_3=32, # (blstm dim)pitch_encoder
dim_enc=512, # (conv dim) content
dim_enc_2=128, # (conv dim) rhythm (encoder_2,encoder_t)
dim_enc_3=256, # (conv dim) pitch
# Decoder parameters
n_frames_per_step=1, # currently only 1 is supported
decoder_rnn_dim=1024,
prenet_dim=256,
max_decoder_steps=1000,
gate_threshold=0.5,
p_attention_dropout=0.1,
p_decoder_dropout=0.1,
# Attention parameters
attention_rnn_dim=1024,
attention_dim=128,
# Location Layer parameters
attention_location_n_filters=32,
attention_location_kernel_size=31,
# Mel-post processing network parameters
postnet_embedding_dim=512,
postnet_kernel_size=5,
postnet_n_convolutions=5,
n_mel_channels=80,
# x(mel) : 80维
dim_freq=80,
dim_spk_emb=107, # 82 at first
embedding_spk=64,
# f0 ：257维
dim_f0=257,
dim_dec=512,
dim_ortho=1024,
layer_norm_eps=1e-12,
len_raw=128,
chs_grp=16,
# interp
# 为了random resampling，先分段segment
# 每个段segment长度：19帧～32帧
min_len_seg=19,
max_len_seg=32,
# min_len_seq = 64,
min_len_seq=32,
# max_len_seq = 128,
max_len_seq=48,
max_len_pad=408, #192 at first,
# data loader
root_dir='assets/spmel',
feat_dir='assets/raptf0',
batch_size=16,
mode='train',
shuffle=True,
num_workers=0,
samplier=8,
#MBV
enc_mbv_size=7,
)
hparams = HParams(
# model
freq=8,
dim_neck=8,
freq_2=8,
dim_neck_2=1,
freq_3=8,
dim_neck_3=32,
out_channels=10 * 3,
layers=24,
stacks=4,
residual_channels=512,
gate_channels=512, # split into 2 groups internally for gated activation
skip_out_channels=256,
cin_channels=80,
gin_channels=-1, # i.e., speaker embedding dim
weight_normalization=True,
n_speakers=-1,
dropout=1 - 0.95,
kernel_size=3,
upsample_conditional_features=True,
upsample_scales=[4, 4, 4, 4],
freq_axis_kernel_size=3,
legacy=True,
dim_enc=512,
dim_enc_2=128,
dim_enc_3=256,
dim_freq=80,
dim_spk_emb=82,
dim_f0=257,
dim_dec=512,
len_raw=128,
chs_grp=16,
# interp
min_len_seg=19,
max_len_seg=32,
min_len_seq=64,
max_len_seq=128,
max_len_pad=192,
# data loader
root_dir='assets/spmel',
feat_dir='assets/raptf0',
batch_size=16,
mode='train',
shuffle=True,
num_workers=0,
samplier=8,
# Convenient model builder
builder="wavenet",
hop_size=256,
log_scale_min=float(-32.23619130191664),
)
hparams = HParams(
# model
freq=8,
dim_neck=8,
freq_2=8,
dim_neck_2=1,
freq_3=8,
dim_neck_3=32,
dim_enc=512, # content encoder
dim_enc_2=128, # rhythm encoder
dim_enc_3=256, # pitch encoder
dim_freq=80,
dim_spk_emb=20,
dim_f0=257,
dim_dec=512,
len_raw=128,
chs_grp=16,
# interp
min_len_seg=19,
max_len_seg=32,
min_len_seq=64,
max_len_seq=128,
max_len_pad=192,
# data loader
root_dir='/hd0/speechsplit/preprocessed/spmel',
feat_dir='/hd0/speechsplit/preprocessed/raptf0',
batch_size=16,
mode='train',
shuffle=True,
num_workers=0,
samplier=8,
)
hparams = HParams(
name="olr",
sample_rate=16000,
# num_mels=80,
n_fft=int(0.04 * 16000),
# n_fft=2048,
hop_length=int(0.02 * 16000),
win_length=int(0.04 * 16000),
deltas=False,
# training testing evaluating
model_type='Cnn_9layers_AvgPooling',
use_cuda=True,
max_epoch=100,
batch_size=32,
lang=[
"Kazak", "Tibet", "Uyghu", "ct_cn", "id_id", "ja_jp", "ko_kr", "ru_ru",
"vi_vn", "zh_cn"
],
#'Kazak',
#'TE_IN',
#'Tibet',
#'Uyghu',
#'ca_es',
#'ct_cn',
#'el_gr',
#'id_id',
#'ja_jp',
##'ko_kr',
#'ru_ru',
#'shanghai',
#'sichuan',
#'minnan',
#'vi_vn',
#'zh_cn',
#],
)