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

Update data_utils.py

Browse Source
This commit is contained in:
Stardust·减
2023-09-23 20:17:19 +08:00
committed by GitHub
parent 855d5b8111
commit e95292dfd6
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
data_utils.py
View File

@@ -154,13 +154,13 @@ class TextAudioSpeakerLoader(torch.utils.data.Dataset):
if language_str == "ZH": if language_str == "ZH":
bert = bert bert = bert
ja_bert = torch.zeros(768, len(phone)) ja_bert = torch.zeros(1024, len(phone))
elif language_str == "JA": elif language_str == "JA":
ja_bert = bert ja_bert = bert
bert = torch.zeros(1024, len(phone)) bert = torch.zeros(1024, len(phone))
else: else:
bert = torch.zeros(1024, len(phone)) bert = torch.zeros(1024, len(phone))
ja_bert = torch.zeros(768, len(phone)) ja_bert = torch.zeros(1024, len(phone))
assert bert.shape[-1] == len(phone), ( assert bert.shape[-1] == len(phone), (
bert.shape, bert.shape,
len(phone), len(phone),
@@ -221,7 +221,7 @@ class TextAudioSpeakerCollate:
tone_padded = torch.LongTensor(len(batch), max_text_len) tone_padded = torch.LongTensor(len(batch), max_text_len)
language_padded = torch.LongTensor(len(batch), max_text_len) language_padded = torch.LongTensor(len(batch), max_text_len)
bert_padded = torch.FloatTensor(len(batch), 1024, max_text_len) bert_padded = torch.FloatTensor(len(batch), 1024, max_text_len)
ja_bert_padded = torch.FloatTensor(len(batch), 768, max_text_len) ja_bert_padded = torch.FloatTensor(len(batch), 1024, max_text_len)
spec_padded = torch.FloatTensor(len(batch), batch[0][1].size(0), max_spec_len) spec_padded = torch.FloatTensor(len(batch), batch[0][1].size(0), max_spec_len)
wav_padded = torch.FloatTensor(len(batch), 1, max_wav_len) wav_padded = torch.FloatTensor(len(batch), 1, max_wav_len)