Spandrel gives your project support for various PyTorch architectures meant for AI Super-Resolution, restoration, and inpainting. Based on the model support implemented in chaiNNer.
My current guess is that our CI fails because we cache too many models. GH only gives us 10GB, so we were almost hitting that limit previously.
This PR changes two things:
Don't cache pip installs. We don't really get a huge speedup for that, but we do have to pay 250MB per runner.
Remove model files that were downloaded from GitHub. We mainly use the cache to speed up tests, but there's no difference in download speed between GitHub CI downloading a model from cache vs downloading it from a GitHub release page.
My current guess is that our CI fails because we cache too many models. GH only gives us 10GB, so we were almost hitting that limit previously.
This PR changes two things:
I also formatted our YAML files.