Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This makes several improvements to the release process, in particular Docker builds. The relevant changes are:
We now build the CUDA images on CI (thanks to Reduce CUDA images size #2724). Also, since installing CUDA directly is now straightforward, I removed the
docker/
build scripts and custom base images, in favour of a single Dockerfile withVARIANT
build argument.Instead of building
edge
Docker images on every commit, we now buildnightly
images daily (or when explicitly triggered), this includes CUDA images. The workflow also dispatches Desktop builds, so Docker and Desktop nightly is aligned. I will remove the schedule trigger on Desktop CI right after merging this.For context, now that we build CUDA images on CI, it is very useful to have those available for changes on
main
. In particular, so that it is possible to run onmain
and properly use Fly runtime and Livebook Apps Docker deployments with CUDA images (also with changes frommain
). I don't think it's necessary to build all of these on every commit, especially that we can trigger the nightly release whenever needed.Changing the base image of the default Docker image from Debian slim to Ubuntu. Those have basically the same size, and I don't think there's any particular reason for using Debian. This is mostly for consistency with the CUDA images, so that everything has the same base. For most purposes, Debian and Ubuntu are compatible (installing dependencies, etc), so this is a non-intrusive change, even if some users build on top of the Livebook image.
I removed the CUDA 11.8 image, so now we will only build
-cuda12
. Currently it still uses cuDNN 8, but I will bump it to cuDNN 9, as soon as we release new nx/xla using cuDNN 9.