fish-speech icon indicating copy to clipboard operation
fish-speech copied to clipboard

Crash on startup: AttributeError: torchaudio.list_audio_backends (TorchAudio 2.9 removal)

Open nuttyorchid opened this issue 2 months ago • 4 comments

Self Checks

  • [x] This template is only for bug reports. For questions, please visit Discussions.
  • [x] I have thoroughly reviewed the project documentation (installation, training, inference) but couldn't find information to solve my problem. English 中文 日本語 Portuguese (Brazil)
  • [x] I have searched for existing issues, including closed ones. Search issues
  • [x] I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
  • [x] [FOR CHINESE USERS] 请务必使用英文提交 Issue,否则会被关闭。谢谢!:)
  • [x] Please do not modify this template and fill in all required fields.

Cloud or Self Hosted

Self Hosted (Source)

Environment Details

ubuntu24.04 / cuda128 / torchaudio==2.9.0

Steps to Reproduce

Running Fish Audio with TorchAudio 2.9.0 fails due to removal of torchaudio.list_audio_backends(), which is still called in fish-agent/tools/api.py. The function was deprecated in 2.8 and fully removed in 2.9 (released Oct 2025).

The project’s pyproject.toml declares:

cu128 = [ "torch>=2.5.1", "torchaudio", ]

so torchaudio is unpinned, allowing 2.9+ to install automatically and break compatibility.


Reproduction Steps

Create and activate a clean conda environment:

conda create -n fish python=3.11 conda activate fish

Install Fish Audio in editable mode with the CUDA 12.8 extras:

pip install -e .[cu128]

Run the Fish Audio agent or any module initializing audio I/O.

Observe the crash:

AttributeError: module 'torchaudio' has no attribute 'list_audio_backends'

✔️ Expected Behavior

No response

❌ Actual Behavior

No response

nuttyorchid avatar Oct 17 '25 23:10 nuttyorchid

python -m tools.run_webui or python -m tools.api_server

Both will Observe the crash:

backends = torchaudio.list_audio_backends() ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ AttributeError: module 'torchaudio' has no attribute 'list_audio_backends'

henrylu avatar Oct 22 '25 15:10 henrylu

pip install torch==2.8.0 torchvision==0.23.0 torchaudio==2.8.0 --index-url https://download.pytorch.org/whl/cu129 should fix it.

That worked for me.

MrEdwards007 avatar Nov 05 '25 23:11 MrEdwards007

Yes, It works

pip install torch==2.8.0 torchvision==0.23.0 torchaudio==2.8.0 --index-url https://download.pytorch.org/whl/cu129 should fix it.

po4920826 avatar Nov 22 '25 19:11 po4920826