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

fix(bug).when generating text that contains only punctuation marks or whitespace characters, the CPU usage reaches 100%, and the process crashes.

Browse Source
This commit is contained in:
0xCAFEBABE0
2024-12-30 10:48:43 +08:00
parent bcc58cb4cb
commit b60c37b31a
4 changed files with 12 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
cosyvoice/cli/frontend.py
View File

@@ -31,7 +31,7 @@ except ImportError:
from tn.chinese.normalizer import Normalizer as ZhNormalizer
from tn.english.normalizer import Normalizer as EnNormalizer
use_ttsfrd = False
from cosyvoice.utils.frontend_utils import contains_chinese, replace_blank, replace_corner_mark, remove_bracket, spell_out_number, split_paragraph
from cosyvoice.utils.frontend_utils import contains_chinese, replace_blank, replace_corner_mark, remove_bracket, spell_out_number, split_paragraph, is_only_punctuation
class CosyVoiceFrontEnd:
@@ -109,6 +109,10 @@ class CosyVoiceFrontEnd:
def text_normalize(self, text, split=True):
text = text.strip()
# When generating text that contains only punctuation marks or whitespace characters
# - Returning empty texts ensures consistent processing logic.
if is_only_punctuation(text):
return []
if contains_chinese(text):
if self.use_ttsfrd:
texts = [i["text"] for i in json.loads(self.frd.do_voicegen_frd(text))["sentences"]]

5
cosyvoice/cli/model.py
View File

@@ -19,7 +19,6 @@ from torch.nn import functional as F
from contextlib import nullcontext
import uuid
from cosyvoice.utils.common import fade_in_out
from cosyvoice.utils.common import is_only_punctuation
class CosyVoiceModel:
@@ -146,10 +145,6 @@ class CosyVoiceModel:
llm_prompt_speech_token=torch.zeros(1, 0, dtype=torch.int32),
flow_prompt_speech_token=torch.zeros(1, 0, dtype=torch.int32),
prompt_speech_feat=torch.zeros(1, 0, 80), stream=False, speed=1.0, **kwargs):
# When generating text that contains only punctuation marks or whitespace characters
# - Returning 10ms of silence ensures consistent processing logic.
if is_only_punctuation(text):
return {'tts_speech': torch.zeros(1, int(0.01 * 22050))}
# this_uuid is used to track variables related to this inference thread
this_uuid = str(uuid.uuid1())
with self.lock: