Skip to content

fix: use local token_hop_len in streaming loop to avoid concurrent st… #1849

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Caxson wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix: use local token_hop_len in streaming loop to avoid concurrent st… #1849

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 4 additions & 3 deletions cosyvoice/cli/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -342,10 +342,11 @@ def tts(self, text=torch.zeros(1, 0, dtype=torch.int32), flow_embedding=torch.ze
p.start()
if stream is True:
token_offset = 0
prompt_token_pad = int(np.ceil(flow_prompt_speech_token.shape[1] / self.token_hop_len) * self.token_hop_len - flow_prompt_speech_token.shape[1])
token_hop_len = self.token_hop_len
prompt_token_pad = int(np.ceil(flow_prompt_speech_token.shape[1] / token_hop_len) * token_hop_len - flow_prompt_speech_token.shape[1])
while True:
time.sleep(0.1)
this_token_hop_len = self.token_hop_len + prompt_token_pad if token_offset == 0 else self.token_hop_len
this_token_hop_len = token_hop_len + prompt_token_pad if token_offset == 0 else token_hop_len
if len(self.tts_speech_token_dict[this_uuid]) - token_offset >= this_token_hop_len + self.flow.pre_lookahead_len:
this_tts_speech_token = torch.tensor(self.tts_speech_token_dict[this_uuid][:token_offset + this_token_hop_len + self.flow.pre_lookahead_len]).unsqueeze(dim=0)
this_tts_speech = self.token2wav(token=this_tts_speech_token,
Expand All @@ -357,7 +358,7 @@ def tts(self, text=torch.zeros(1, 0, dtype=torch.int32), flow_embedding=torch.ze
stream=stream,
finalize=False)
token_offset += this_token_hop_len
self.token_hop_len = min(self.token_max_hop_len, self.token_hop_len * self.stream_scale_factor)
token_hop_len = min(self.token_max_hop_len, token_hop_len * self.stream_scale_factor)
yield {'tts_speech': this_tts_speech.cpu()}
if self.llm_end_dict[this_uuid] is True and len(self.tts_speech_token_dict[this_uuid]) - token_offset < this_token_hop_len + self.flow.pre_lookahead_len:
break
Expand Down