deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-05 01:49:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix triton token2wav model cache thread unsafety

Browse Source
This commit is contained in:
김의진
2025-10-27 17:20:14 +09:00
parent 6e01309e01
commit cd26dd1932
1 changed files with 4 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
runtime/triton_trtllm/model_repo/token2wav/1/model.py
View File

@@ -28,6 +28,7 @@ import json
import os
import logging
from uuid import uuid4
import torch
from torch.utils.dlpack import to_dlpack
@@ -235,17 +236,17 @@ class TritonPythonModel:
stream = True
else:
stream = False
request_id = request.request_id()
uuid = uuid4().hex
audio_hat = self.token2wav_model.model.token2wav(token=target_speech_tokens,
prompt_token=prompt_speech_tokens,
prompt_feat=prompt_speech_feat,
embedding=prompt_spk_embedding,
token_offset=token_offset,
uuid=request_id,
uuid=uuid,
stream=stream,
finalize=finalize)
if finalize:
self.token2wav_model.model.hift_cache_dict.pop(request_id)
self.token2wav_model.model.hift_cache_dict.pop(uuid)
else:
tts_mel, _ = self.token2wav_model.model.flow.inference(