Pytorch CPU and conda environment conflicts
Closed this issue · 1 comments
Not a bug report per se, but I thought it would be useful to add this to the issues list in case other people have this problem.
When trying to set up the Conda environment I was having trouble installing a Cuda-compatible version of Pytorch. The CPU version was automatically installed when using the default conda_env.yaml file. Whenever I made changes to the .yaml file to try to install the GPU version the environment would take an extremely long time to solve and would contain conflicts. To get a working environment I was required to make the following changes to the conda_env.yaml file:
#$ conda env create --file conda-env.yaml
name: deepslide_env
channels:
- pytorch
- nvidia
- conda-forge
dependencies:
- python=3.11
- torchvision
- pytorch-cuda=11.7
- pytorch
- pandas
- matplotlib
- scikit-learn
- scikit-image
- pip
- pip:
- -r pip-requirements.txt
One consequence of this was using a newer version of Pandas, meaning the following line of code needed changing:
utils_model.py (line 63): cm.style.hide_index() ---> cm.style.hide(axis = 'index')
Stale issue message