Open
Description
确认清单
- 我已经阅读过 README.md 和 dependencies.md 文件
- 我已经确认之前没有 issue 或 discussion 涉及此 BUG
- 我已经确认问题发生在最新代码或稳定版本中
Forge Commit 或者 Tag
Python 版本
3.10
PyTorch 版本
'2.4.0+cu121'
操作系统信息
ubuntu
BUG 描述
4个请求同时触发，几乎必现。
BUG 端点
v2/tts
复现参数
const payload = {
text,
spk: { from_spk_name: voice },
encoder: { format: 'raw' },
adjust: { speed_rate: 1, normalize: true, headroom: 0.05},
infer: { stream_chunk_size: 32, no_cache: true, stream: true, seed:42, batch_size:10, spliter_threshold:64},
tts: { style: 'chat', temperature:0.1, top_p: 0.75,top_k:10}
};
const res = await fetch('/api/v2/tts',
期望结果
正常播放
实际结果
播放异常
错误信息
2025-05-14 05:11:03,614 - modules.api.api_setup - ERROR - Uncaught exception
Traceback (most recent call last):
File "/usr/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
self.run()
File "/usr/lib/python3.10/threading.py", line 953, in run
self._target(*self._args, **self._kwargs)
File "/app/modules/core/pipeline/generate/BatchGenerate.py", line 55, in generate
self.generate_batch_stream(batch)
File "/app/modules/core/pipeline/generate/BatchGenerate.py", line 98, in generate_batch_stream
for results in model.generate_batch_stream(
File "/app/modules/core/models/tts/ChatTtsModel.py", line 166, in _gen
for results in infer.generate_audio_stream(
File "/app/modules/core/models/tts/ChatTTS/ChatTTSInfer.py", line 404, in _generator
for audio_arr in gen:
File "/home/venv/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 36, in generator_context
response = gen.send(None)
File "/app/modules/core/models/tts/ChatTTS/ChatTTSInfer.py", line 152, in _infer
for result in self.instance._infer_code(
File "/home/venv/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 36, in generator_context
response = gen.send(None)
File "/app/modules/repos_static/ChatTTS/ChatTTS/model/gpt.py", line 598, in generate
idx_next = torch.multinomial(scores, num_samples=1).to(finish.device)
RuntimeError: probability tensor contains either `inf`, `nan` or element < 0