pyannote uninstalling torch cuda and install torch cpu

[versus666jzx]

Tested versions

• Reproducible in Version 3.1.1

System information

win11, pyannote.audio 3.1.1

Issue description

If I install pyannote.audio in an env in which torch with cuda is already installed, then during installation the torch with cuda is removed and torch cpu is installed. Even though the torch versions satisfy requirements.txt.

Minimal reproduction example (MRE)

pip install torch torchaudio --index-url https://download.pytorch.org/whl/cu118 pip install pyannote.audio

[Valeronich]

Same problem omg

[gdk00]

Struggled with that too...

[beton18]

same issue

[hbredin]

I'd suggest using the same --index-url when installing pyannote.audio.

pip install pyannote.audio --index-url https://download.pytorch.org/whl/cu118

[versus666jzx]

I'd suggest using the same --index-url when installing pyannote.audio.

pip install pyannote.audio --index-url https://download.pytorch.org/whl/cu118

Expectedly doesn't work.

[versus666jzx]

But if use this

pip install pyannote.audio --extra-index-url https://download.pytorch.org/whl/cu118

then torch and torchaudio packages with cuda support will be installed

It would be cool if this was somehow reflected in the installation documentation or if the requirements.txt or setup files were corrected.

[hbredin]

It would be cool if you could open a PR fixing this now that you have found a solution.

[ThatJeffGuy]

But if use this

pip install pyannote.audio --extra-index-url https://download.pytorch.org/whl/cu118

then torch and torchaudio packages with cuda support will be installed

It would be cool if this was somehow reflected in the installation documentation or if the requirements.txt or setup files were corrected.

Hey friend, am trying to get this to work myself, but after running this, it yells at me about torchvision. Error importing pyannote.audio, torch, torchvision, or torchaudio: No module named 'torchvision'

I use this script that our lovely AI overlord generated for me to check my environment:

import torch
import torchaudio
import torchvision
import speechbrain

print("PyTorch Version:", torch.__version__)
print("CUDA Available:", torch.cuda.is_available())
print("CUDA Version:", torch.version.cuda)
print("torchaudio version:", torchaudio.__version__)
print("torchvision version:", torchvision.__version__)
print("speechbrain version:", speechbrain.__version__)

# Check if CUDA is available
print("CUDA available:", torch.cuda.is_available())
if torch.cuda.is_available():
    print("CUDA device name:", torch.cuda.get_device_name(0))