Based on this announcement from PyTorch, it sounds like we will have to switch to pip for torch-related installations instead of the conda channel at some point or another. It may also be worth investigating whether or not alternatives for conda, such as uv or poetry, are worth adopting for package and environment management.
This will be addressed in #333, which does the following:
Removes usage of the conda channels for PyTorch-related installations. This is in light of the following announcement that suggests, without the change in this PR, issues may occur in the future: https://github.com/pytorch/pytorch/issues/138506
Adds uv as an option to manage an environment for the toolbox.
Based on this announcement from PyTorch, it sounds like we will have to switch to pip for torch-related installations instead of the conda channel at some point or another. It may also be worth investigating whether or not alternatives for conda, such as uv or poetry, are worth adopting for package and environment management.
poetry: https://python-poetry.org/docs/
uv: https://docs.astral.sh/uv/