tuna2134/sbv2-v2
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
rcell
2023-07-23 15:37:50 +08:00
parent 8fc674fe2c
commit 1fa3ce1d03
4 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
models.py
View File

@@ -515,7 +515,7 @@ class SynthesizerTrn(nn.Module):
if n_speakers > 1: if n_speakers > 1:
self.emb_g = nn.Embedding(n_speakers, gin_channels) self.emb_g = nn.Embedding(n_speakers, gin_channels)
else: else:
self.ref_enc = ReferenceEncoder(spec_channels) self.ref_enc = ReferenceEncoder(spec_channels, gin_channels)
def forward(self, x, x_lengths, y, y_lengths, sid, tone, language, bert): def forward(self, x, x_lengths, y, y_lengths, sid, tone, language, bert):

2
preprocess_text.py
View File

@@ -10,7 +10,7 @@ transcription_path = 'filelists/aishell.list'
train_path = 'filelists/train.list' train_path = 'filelists/train.list'
val_path = 'filelists/val.list' val_path = 'filelists/val.list'
config_path = "configs/config.json" config_path = "configs/config.json"
val_per_spk = 2 val_per_spk = 4
max_val_total = 8 max_val_total = 8
if 1 in stage: if 1 in stage:

6
spec_gen.py
View File

@@ -2,11 +2,11 @@ import utils
from data_utils import TextAudioSpeakerLoader from data_utils import TextAudioSpeakerLoader
from tqdm import tqdm from tqdm import tqdm
config_path = 'configs/config.json' config_path = 'configs/fzh.json'
hps = utils.get_hparams_from_file(config_path) hps = utils.get_hparams_from_file(config_path)
train_dataset = TextAudioSpeakerLoader("filelists/train.list", hps.data) train_dataset = TextAudioSpeakerLoader(hps.data.training_files, hps.data)
eval_dataset = TextAudioSpeakerLoader("filelists/val.list", hps.data) eval_dataset = TextAudioSpeakerLoader(hps.data.validation_files, hps.data)
for _ in tqdm(train_dataset): for _ in tqdm(train_dataset):
pass pass

2
train_ms.py
View File

@@ -279,7 +279,7 @@ def evaluate(hps, generator, eval_loader, writer_eval):
bert = bert.cuda() bert = bert.cuda()
tone = tone.cuda() tone = tone.cuda()
language = language.cuda() language = language.cuda()
for use_sdp in [False, True]: for use_sdp in [True, False]:
y_hat, attn, mask, *_ = generator.module.infer(x, x_lengths, speakers, tone, language, bert, y=spec, max_len=1000, sdp_ratio=0.0 if not use_sdp else 1.0) y_hat, attn, mask, *_ = generator.module.infer(x, x_lengths, speakers, tone, language, bert, y=spec, max_len=1000, sdp_ratio=0.0 if not use_sdp else 1.0)
y_hat_lengths = mask.sum([1, 2]).long() * hps.data.hop_length y_hat_lengths = mask.sum([1, 2]).long() * hps.data.hop_length