-
Notifications
You must be signed in to change notification settings - Fork 9
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
030a126
commit 48cead2
Showing
12 changed files
with
396 additions
and
45 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
74 changes: 74 additions & 0 deletions
74
configs/libritts_conformerppg_uttdvec_ppgvcf0_gradtts_ppgvchifigan.yaml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,74 @@ | ||
# experiment | ||
dataset: libritts | ||
train_meta: data/libritts/train_nodev_clean/metadata.csv | ||
dev_meta: data/libritts/dev_clean/metadata.csv | ||
train_set: train_nodev_clean | ||
dev_set: dev_clean | ||
|
||
|
||
# encoder-decoder | ||
ling_enc: conformerppg | ||
spk_enc: utt_dvec | ||
pros_enc: ppgvc_f0 | ||
decoder: GradTTS | ||
mel_type: ppgvc_mel | ||
vocoder: ppgvc_hifigan | ||
|
||
# training | ||
fp16_run: !!bool False | ||
epochs: 2000 | ||
save_freq: 1 # save ckpt frequency | ||
show_freq: 100 # show training information frequency | ||
load_only_params: !!bool False | ||
seed: !!int 1234 | ||
trainer: GradTTSTrainer | ||
ngpu: 2 | ||
|
||
#dataloader | ||
dataset_class: Dataset | ||
mel_segment_length: !!int 128 | ||
sort: !!bool False | ||
dump_dir: dump | ||
num_workers: !!int 8 | ||
batch_size: 32 | ||
drop_last: !!bool True | ||
rm_long_utt: !!bool True # remove too long utterances from metadata | ||
max_utt_duration: !!float 10.0 # max utterance duration (seconds) | ||
|
||
|
||
# decoder params | ||
decoder_params: | ||
use_prior_loss: !!bool False | ||
n_feats: !!int 80 | ||
input_dim: !!int 144 | ||
spk_emb_dim: !!int 256 | ||
prosodic_rep_type: continuous | ||
prosodic_net: | ||
hidden_dim: !!int 80 | ||
n_enc_channels: !!int 192 | ||
filter_channels: !!int 768 | ||
filter_channels_dp: !!int 256 | ||
n_enc_layers: !!int 6 | ||
enc_kernel: !!int 3 | ||
enc_dropout: !!float 0.1 | ||
n_heads: !!int 2 | ||
window_size: !!int 4 | ||
dec_dim: !!int 64 | ||
beta_min: !!float 0.05 | ||
beta_max: !!float 20.0 | ||
pe_scale: !!int 1000 # 1 for `grad-tts-old.pt` checkpoint | ||
|
||
#optimizer & scheduler | ||
optimizer: | ||
lr: !!float 1e-4 | ||
|
||
# loss hyper-parameters | ||
losses: | ||
alpha: 1. | ||
|
||
|
||
|
||
|
||
|
||
|
||
|
100 changes: 100 additions & 0 deletions
100
configs/libritts_conformerppg_uttdvec_ppgvcf0_vits_none.yaml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,100 @@ | ||
# experiment | ||
dataset: libritts | ||
train_meta: data/libritts/train_nodev_clean/metadata.csv | ||
dev_meta: data/libritts/dev_clean/metadata.csv | ||
train_set: train_nodev_clean | ||
dev_set: dev_clean | ||
|
||
|
||
# encoder-decoder | ||
ling_enc: conformer_ppg | ||
spk_enc: utt_dvec | ||
pros_enc: ppgvc_f0 | ||
decoder: VITS | ||
mel_type: vits_spec | ||
|
||
|
||
# training | ||
fp16_run: !!bool False | ||
epochs: 1000 | ||
save_freq: 1 # save ckpt frequency | ||
show_freq: 100 # show training information frequency | ||
load_only_params: !!bool False | ||
seed: !!int 1234 | ||
trainer: VITSTrainer | ||
ngpu: 2 | ||
|
||
#dataloader | ||
dataset_class: VITSDataset | ||
sampling_rate: !!int 24000 | ||
vits_hop_size: !!int 240 | ||
spec_max_len: !!int 480 | ||
sort: !!bool True | ||
dump_dir: dump | ||
num_workers: !!int 8 | ||
batch_size: !!int 48 | ||
drop_last: !!bool True | ||
rm_long_utt: !!bool False # remove too long utterances from metadata | ||
max_utt_duration: !!float 10.0 # max utterance duration (seconds) | ||
|
||
|
||
# decoder params | ||
decoder_params: | ||
spk_emb_dim: 256 | ||
prosodic_rep_type: continuous | ||
prosodic_net: | ||
hidden_dim: 192 | ||
input_dim: !!int 144 | ||
spec_channels: !!int 513 | ||
inter_channels: !!int 192 | ||
hidden_channels: !!int 192 | ||
filter_channels: !!int 768 | ||
n_heads: !!int 2 | ||
n_layers: !!int 6 | ||
kernel_size: !!int 3 | ||
p_dropout: !!float 0.1 | ||
resblock : 1 | ||
resblock_kernel_sizes: [3,7,11] | ||
resblock_dilation_sizes: [[1,3,5], [1,3,5], [1,3,5]] | ||
upsample_rates: [8,5,3,2] | ||
upsample_initial_channel: !!int 512 | ||
upsample_kernel_sizes: [15, 15, 5, 5] | ||
n_layers_q: !!int 3 | ||
use_spectral_norm: !!bool False | ||
filter_length: !!int 1024 | ||
n_mels_channels: !!int 80 | ||
win_length: !!int 1024 | ||
hop_length: !!int 240 | ||
sampling_rate: !!int 24000 | ||
segment_size: !!int 9600 | ||
|
||
|
||
|
||
|
||
#optimizer & scheduler | ||
optimizer: | ||
generator: | ||
lr: !!float 1e-4 | ||
betas: [0.8,0.99] | ||
eps: !!float 1e-9 | ||
discriminator: | ||
lr: !!float 1e-4 | ||
betas: [0.8,0.99] | ||
eps: !!float 1e-9 | ||
scheduler: | ||
generator: | ||
lr_decay: !!float 0.999875 | ||
discriminator: | ||
lr_decay: !!float 0.999875 | ||
|
||
# loss hyper-parameters | ||
losses: | ||
mel: !!int 45 | ||
kl: !!int 1 | ||
|
||
|
||
|
||
|
||
|
||
|
||
|
74 changes: 74 additions & 0 deletions
74
configs/libritts_vqwav2vec_uttdvec_ppgvcf0_gradtts_ppgvchifigan.yaml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,74 @@ | ||
# experiment | ||
dataset: libritts | ||
train_meta: data/libritts/train_nodev_clean/metadata.csv | ||
dev_meta: data/libritts/dev_clean/metadata.csv | ||
train_set: train_nodev_clean | ||
dev_set: dev_clean | ||
|
||
|
||
# encoder-decoder | ||
ling_enc: vqwav2vec | ||
spk_enc: utt_dvec | ||
pros_enc: ppgvc_f0 | ||
decoder: GradTTS | ||
mel_type: ppgvc_mel | ||
vocoder: ppgvc_hifigan | ||
|
||
# training | ||
fp16_run: !!bool False | ||
epochs: 2000 | ||
save_freq: 1 # save ckpt frequency | ||
show_freq: 100 # show training information frequency | ||
load_only_params: !!bool False | ||
seed: !!int 1234 | ||
trainer: GradTTSTrainer | ||
ngpu: 2 | ||
|
||
#dataloader | ||
dataset_class: Dataset | ||
mel_segment_length: !!int 128 | ||
sort: !!bool False | ||
dump_dir: dump | ||
num_workers: !!int 8 | ||
batch_size: 32 | ||
drop_last: !!bool True | ||
rm_long_utt: !!bool True # remove too long utterances from metadata | ||
max_utt_duration: !!float 10.0 # max utterance duration (seconds) | ||
|
||
|
||
# decoder params | ||
decoder_params: | ||
use_prior_loss: !!bool False | ||
n_feats: !!int 80 | ||
input_dim: !!int 512 | ||
spk_emb_dim: !!int 256 | ||
prosodic_rep_type: continuous | ||
prosodic_net: | ||
hidden_dim: !!int 80 | ||
n_enc_channels: !!int 192 | ||
filter_channels: !!int 768 | ||
filter_channels_dp: !!int 256 | ||
n_enc_layers: !!int 6 | ||
enc_kernel: !!int 3 | ||
enc_dropout: !!float 0.1 | ||
n_heads: !!int 2 | ||
window_size: !!int 4 | ||
dec_dim: !!int 64 | ||
beta_min: !!float 0.05 | ||
beta_max: !!float 20.0 | ||
pe_scale: !!int 1000 # 1 for `grad-tts-old.pt` checkpoint | ||
|
||
#optimizer & scheduler | ||
optimizer: | ||
lr: !!float 1e-4 | ||
|
||
# loss hyper-parameters | ||
losses: | ||
alpha: 1. | ||
|
||
|
||
|
||
|
||
|
||
|
||
|
Oops, something went wrong.