removing tf dep - Githubissues

conda-forge / keras-feedstock

A conda-smithy repository for keras.

BSD 3-Clause "New" or "Revised" License

6 stars 28 forks source link

removing tf dep #58

Closed ngam closed 2 years ago

ngam commented 2 years ago

running with constrained tensorflow is more of a hassle than anything else here. The constraining should happen in tensorflow, not here, imo. Removing. Consult upstream: https://github.com/keras-team/keras/blob/d8fcb9d4d4dad45080ecfdd575483653028f8eda/keras/tools/pip_package/setup.py#L35-L38

Checklist

[ ] Used a personal fork of the feedstock to propose changes
[ ] Bumped the build number (if the version is unchanged)
[ ] Reset the build number to 0 (if the version changed)
[ ] Re-rendered with the latest conda-smithy (Use the phrase code>@<space/conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
[ ] Ensured the license file is being packaged.

conda-forge-linter commented 2 years ago

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

ngam commented 2 years ago

@xhochy could you please review and merge? I believe it makes no sense to have tensorflow explicitly here. We should just constrain stuff carefully in the tensorflow-feedstock. Of course, a workaround is just to build this tiny package in the tensorflow feedstock but that would be super redundant?

xhochy commented 2 years ago

I'm not sure what the issue is? From my understanding keras and tensorflow are released together and are meant to be only used in this combination.

ngam commented 2 years ago

I'm not sure what the issue is? From my understanding keras and tensorflow are released together and are meant to be only used in this combination.

The issue is that building tensorflow 2.8.0 (which explicitly depends on keras >=2.8.0,<2.9.0) will necessitate pulling in keras 2.8.0, but keras 2.8.0 will want tensorflow 2.8.0 which is not yet built.

https://github.com/conda-forge/tensorflow-feedstock/pull/203

I am not sure how this didn't come up yet though... makes no sense that it worked fine with 2.7.0, so could it be something else somewhere?

ngam commented 2 years ago

Actually, never mind. I think I was mistaken --- I thought it was getting errors because of keras, but it actually wasn't catching it earlier (not enough time between merging here and running there). Converting to draft for now. Let's see how things develop.

ngam commented 2 years ago

Okay, confirmed. Closing. The other problem was the confusing errors from python3.10 that also completely disoriented me!