What is the best method to install PyTorch?

[ryan-minato]

I have tried using the index and find_links options in the [python-repos] section of the pants.toml file, but they prevent the generation of the lock file.
The build process gets stuck indefinitely and eventually times out. As a workaround, I decided not to specify a CUDA version, and PyTorch installation was successful. It worked.However, I am unable to utilize the 'export' and 'package' goals..

What is the best method to install PyTorch in Pants and export the virtual environment for IDE usage (in my case, PyCharm)? Is this even possible, and if so, what would be the recommended approach?
This is almost the final (yet most unavoidable) hurdle I face in migrating my ML project to Pants.