search

MahmoudAshraf97 / whisper-diarization

Automatic Speech Recognition with Speaker Diarization based on OpenAI Whisper
BSD 2-Clause "Simplified" License
3.53k stars 301 forks source link

Requirements Packages are Outdated #82

Closed RickNeedsMoreCoffee closed 5 months ago

RickNeedsMoreCoffee commented 1 year ago

pip install -r requirements.txt causes several failures when setting up with venv. Have not tested conda env. Recommend modifying requirements to the following to solve dependency related issues:

wget cython wheel onnx cmake nemo_toolkit faster-whisper git+https://github.com/m-bain/whisperX.git@4cb167a225c0ebaea127fd6049abfaa3af9f8bb4 git+https://github.com/facebookresearch/demucs#egg=demucs deepmultilingualpunctuation

I'm still unable to get audio to work as I keep getting modulenotfound errors for wget which I have not solved despite quite a bit of troubleshooting against it.

MahmoudAshraf97 commented 1 year ago

cython should be installed before the requirements as mentioned in the instructions wheel, cmake are either installed along with pip or are bundled with almost all OS installations nowadays so no need to include them as for onnx I never stumbled upon a use case where it's required

RickNeedsMoreCoffee commented 1 year ago

yeah i just started manually adding libraries that were causing hangs when establishing the venv and this combination seemed to get past it hence the recommendation

Barron-Brock commented 1 year ago

I am also running into this issue. Tried following another thread, #73 Fix Cython 3.0.0 brokes PyYAML==5.4.1 (from NeMo requirements) but I haven't had any luck over there either.

dahifi commented 1 year ago

Just want to add that install fails on WSL unless you use python 3.10.12 per the nemo installer.