Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

New release for TF 2.17, JAX 4.31+? #1832

Open
jonas-eschle opened this issue Aug 13, 2024 · 3 comments
Open

New release for TF 2.17, JAX 4.31+? #1832

jonas-eschle opened this issue Aug 13, 2024 · 3 comments

Comments

@jonas-eschle
Copy link
Contributor

Hi all, is there a new release planned in the near future? TF 2.17 is out since a while, JAX also upgraded a few versions from the previous 0.4.25 to 0.4.31.

Or did the release strategy change to be synced with TF versions? (if so, apologies, I missed that)

@csuter
Copy link
Member

csuter commented Aug 13, 2024

For historic reasons, we've usually shipped a new stable version when TF releases one. I think the TF 2.17 aligned release just slipped through the cracks, as the usual releasers on the team have been attending to other things. We'll try and put one out soon. TBH, using the stable release is not that different from a nightly (tfp-nightly or tfp-nightly[jax]), if you pin to a particular dated version. Nightlies are tested with the most recent backends before they're pushed out.

@jonas-eschle
Copy link
Contributor Author

Sure, thanks! Would you need some help with this in general? (I am currently the active TFP release on conda-forge maintainer) I'd still prefer not to have the nightlies as a dependency, because it's conflicting with other dependencies deps, so we would need to convince packages to also use the nightlies (i.e. for a larger software build, for an environment exclusively with one package, or for developing, I agree)

@jonas-eschle
Copy link
Contributor Author

Hi @csuter , would it still possible to do this release? I understand that it may slipped, but late is better than never.

Nightlies aren't really an option to put as dependencies; this can easily screw up other dependencies that may use TensorFlow-Probability. More importantly, this solution won't work for the conda-forge release, which relies on github releases.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants