pytorch version 2.1.2 struggling to be found - not clear what python version needed

[chrisaharden]

When running pip install -r requirements.txt, I'm getting the following error:

ERROR: Could not find a version that satisfies the requirement torch==2.1.2 (from versions: 2.2.0, 2.2.1, 2.2.2, 2.3.0, 2.3.1, 2.4.0, 2.4.1) ERROR: No matching distribution found for torch==2.1.2

I'm assuming I have the wrong version of python installed. I have 3.12.7 installed. It is not clear what python version is needed from the README or requirements.txt. According to https://pypi.org/project/torch/2.1.2/, the version supported should be Python >=3.8.0.

Any ideas on what I may need to do to get the pip install -r requirements.txt to install successfully? I am about to try installing the latest pytorch, but usually that starts down a path of version mismatches I'm hoping to avoid.

[vvuk]

The following worked for me (linux):

conda create -n pyramid python=3.11.9
conda activate pyramid
conda config --set pip_interop_enabled True

conda install pytorch==2.1.2 torchvision=0.16.2 torchaudio==2.1.2 pytorch-cuda==12.1 -c pytorch -c nvidia
pip3 install decord
pip3 install contexttimer

.. edit requirements.txt and change torch==2.1.2 to pytorch==2.1.2 and comment out decord

conda install --file requirements.txt -c pytorch -c nvidia -c fastai

Also make sure you:

export IMAGEIO_FFMPEG_EXE=/usr/bin/ffmpeg

as something installs a very broken ffmpeg binary in the env, and you won't know that it's broken until the very end of the generation process.

[chrisaharden]

Thanks!!