This is created by myself to solve the complicated of package.

conda create -n CHEN_MI python=3.9.18 pytorch torchvision torchaudio pytorch-cuda=11.8 cudatoolkit=11.8 -c pytorch -c nvidia

conda install conda-forge::cudnn

the best part of this method is created the environment and install package at the same time.

then using the -c conda-forge install the other package in need which will not introduce conflict.

the torch_optimizer is not included in conda, so only this one need to be install with pip

pip install torch_optimizer