deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-04 17:39:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

add instruct

Browse Source
This commit is contained in:
lyuxiang.lx
2025-12-11 09:43:25 +00:00
parent 3298d6f3e3
commit ebef63066f
5 changed files with 36 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
tools/make_parquet_list.py
View File

@@ -37,6 +37,8 @@ def job(utt_list, parquet_file, utt2parquet_file, spk2parquet_file):
speech_token_list = [utt2speech_token.get(utt, []) for utt in utt_list]
if args.dpo:
reject_speech_token_list = [utt2reject_speech_token[utt] for utt in utt_list]
if args.instruct:
instruct_list = [utt2instruct[utt] for utt in utt_list]
# 保存到parquet,utt2parquet_file,spk2parquet_file
df = pd.DataFrame()
@@ -50,6 +52,8 @@ def job(utt_list, parquet_file, utt2parquet_file, spk2parquet_file):
df['speech_token'] = speech_token_list
if args.dpo:
df['reject_speech_token'] = reject_speech_token_list
if args.instruct:
df['instruct'] = instruct_list
df.to_parquet(parquet_file)
with open(utt2parquet_file, 'w') as f:
json.dump({k: parquet_file for k in utt_list}, f, ensure_ascii=False, indent=2)
@@ -68,6 +72,10 @@ if __name__ == "__main__":
type=int,
default=1,
help='num processes for make parquets')
parser.add_argument('--instruct',
action='store_true',
default=False,
help='has instruct file or not')
parser.add_argument('--src_dir',
type=str)
parser.add_argument('--des_dir',
@@ -91,6 +99,11 @@ if __name__ == "__main__":
for l in f:
l = l.replace('\n', '').split()
utt2spk[l[0]] = l[1]
if args.instruct is True:
with open('{}/instruct'.format(args.src_dir)) as f:
for l in f:
l = l.replace('\n', '').split()
utt2instruct[l[0]] = ' '.join(l[1:])
utt2embedding = torch.load('{}/utt2embedding.pt'.format(args.src_dir))
spk2embedding = torch.load('{}/spk2embedding.pt'.format(args.src_dir))
utt2speech_token = torch.load('{}/utt2speech_token.pt'.format(args.src_dir))