Deploy instructions

If necessary: update the minimum PyTorch/Python version, and the minimum linear_operator version:
- In setup.py
- In README.md
- In .github/workflows/*.yml
- In .conda/meta.yaml
On Github, go to https://github.com/cornellius-gp/gpytorch/releases/new to create a new release.
- Under the Tag: Select Tag dropdown, click Create New Tag. Name the tag v<semantic_version_number> (e.g. v1.15.2).
- Click Generate Release Notes to automatically populate the release name and description.
- Click Publish Release

That should be it! The deploy actions should take care of the rest.

If the deploy script fails...

Go to the RTD admin page for GPyTorch and click on the "Versions" tab
Find the latest tag under "Activate a Version", and click the "Activate" button. Check the "active" check box.
Under "Admin > Advanced Settings", change the default version to the latest version.

Locally create a packaged version (python setup.py sdist)
Upload the latest package to PyPI (twine upload dist/gpytorch-<version_number>.tar.gz)

Make sure you have conda-build installed: conda install conda-build
Let the meta.yaml file know about the pytorch channel conda config --append channels pytorch
Change in to the .conda directory: cd .conda
Run conda-build . in the .conda directory
Upload the latest package to Anaconda Cloud using the command printed out at the end of conda-build.