[2noise/ChatTTS]torchaudio.save报错

最新的代码中使用torchaudio.save(audioFile, torch.from_numpy(wavs[0]), 24000)时，在实体机上运行正常，但是在docker中运行的时候就会报错： File "/data/ChatTTS/examples/web/infura.py", line 82, in customAudio torchaudio.save(audioFile, torch.from_numpy(wavs[0]), 24000) File "/root/anaconda3/lib/python3.10/site-packages/torchaudio/_backend/utils.py", line 315, in save backend = dispatcher(uri, format, backend) File "/root/anaconda3/lib/python3.10/site-packages/torchaudio/_backend/utils.py", line 223, in dispatcher raise RuntimeError(f"Couldn't find appropriate backend to handle uri {uri} and format {format}.")

DExpress-dev

For Linux: pip install sox Windows & Mac: pip install soundfile

menphix-watanabe

多谢回答，但是安装了sox，问题依旧！

DExpress-dev

可以用tools/audio下相关函数保存，用法详见examples/web和examples/cmd。

fumiama

谢谢回复，但是使用了cmd中的类似代码后也是报错，只是报错的原因是：

File "/data/ChatTTS/examples/web/infura.py", line 56, in saveMp3 data = pcm_arr_to_mp3_view(wav) File "/data/ChatTTS/ttstools/audio/pcm.py", line 16, in pcm_arr_to_mp3_view wf.writeframes(float_to_int16(wav)) File "/root/anaconda3/lib/python3.10/site-packages/numba/core/dispatcher.py", line 423, in _compile_for_args error_rewrite(e, 'typing') File "/root/anaconda3/lib/python3.10/site-packages/numba/core/dispatcher.py", line 364, in error_rewrite raise e.with_traceback(None) numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend) Unknown attribute 'max' of type float32

DExpress-dev

打印一下wav，看看是什么。pcm_arr_to_mp3_view接受的参数是一维numpy数组，可通过wavs[0]获得。

fumiama

Make sure that the dependency libraries required for audio processing are installed in the container,print it first

import ChatTTS
import torch
import torchaudio

print("[audio list]",str(torchaudio.list_audio_backends()))
# when result is  [] ，I think so you must install libsox-dev sox

just like this and try again

apt-get update && apt-get install -y libsox-dev sox

# ['sox', 'soundfile']

Of-qinhai