Support for macOS Monterey

[yarri-oss]

Support for macOS Monterey

Description

Adding support for macOS Monterey M1 local build.

Brief Description of the PR:

Adding changes to bazel configuration and scripts to enable building from source for macOS Monterey M1.
Fixes # (issue)

Type of change

• Bug fix
• New Tutorial
• Updated or additional documentation
• Additional Testing
• New Activation and the changes conform to the activation contribution guidelines
• New Callback and the changes conform to the callback contribution guidelines
• New Image addition and the changes conform to the image op contribution guidelines
• New Layer and the changes conform to the layer contribution guidelines
• New Loss and the changes conform to the loss contribution guidelines
• New Metric and the changes conform to the metric contribution guidelines
• New Optimizer and the changes conform to the optimizer contribution guidelines
• New RNN Cell and the changes conform to the rnn contribution guidelines
• New Seq2seq addition and the changes conform to the seq2seq contribution guidelines
• New Text addition and the changes conform to the text op contribution guidelines

Checklist:

• I've properly formatted my code according to the guidelines
  • By running Black + Flake8
  • By running pre-commit hooks
• This PR addresses an already submitted issue for TensorFlow Addons
• I have made corresponding changes to the documentation
• I have added tests that prove my fix is effective or that my feature works
• This PR contains modifications to C++ custom-ops

How Has This Been Tested?

Locally

If you're adding a bugfix or new feature please describe the tests that you ran to verify your changes:
*

[bhack]

Any side effect about diverging version from https://github.com/tensorflow/tensorflow/blob/master/.bazelversion ?

[yarri-oss]

Potentially, and if so this might justify a new approach to handle macOS TFA builds perhaps with a macos_m1 branch.

Created an issue to continue the discussion here: #2616

[bhack]

If we don't have a new rollback we are at 4.2.1 in https://github.com/tensorflow/tensorflow/blob/master/.bazelversion

[bhack]

Are we not going to distribute macosx_11 packages anymore?

I see that tensorflow-macosx wheels are still on macosx_11_0:
https://pypi.org/project/tensorflow-macos/2.6.0/#files

[yarri-oss]

This is build from source issue, so only local wheel files. We seem to have only a minimal macOS version support, but if I understand the platform correctly, we may need to change the build_pip_pkg.sh file to instead check for either x86 or M1 and OS version.

[bhack]

I don't think it is only local as we are using the same in Action:

https://github.com/tensorflow/addons/blob/master/.github/workflows/make_wheel_macOS_arm64.sh#L27

addons/BUILD

Line 3 in 41eaa27

srcs = ["build_deps/build_pip_pkg.sh"],

[seanpmorgan]

Next Steps:

• Drop bazel version increase (already done)
• Increment the minimum build version in make_wheel_macOS_arm64 to 12
• Increase the plat-name (as is done in this PR)

[seanpmorgan]

@kulinseth we're looking to publish macos ARM64 wheels for TFA, but have heard that the package will not resolve for Monterey users since it's plat-name is for macos-11. Given that M1 chips have better performance on macos-12 (I think?) we wanted to make sure that majority of users will be able to resolve a TFA for macos-12 built for ARM.

However, I noticed tensorflow-macos is still publishing macos-11 plat-names:
https://pypi.org/project/tensorflow-macos/2.7.0/#files

1. Is there a plan to increase the OS of the tensorflow-macos build?
2. Also is it true that macos 12 users will not be able to resolve/install an earlier tensorflow-macos if it's labeled in pypi with a prior macos version? Thinking about this a bit more I'm confused since we publish our TFA x86_64 wheels for macos-10 and haven't heard this complaint.

[kulinseth]

@seanpmorgan , thanks for the issue:

@kulinseth we're looking to publish macos ARM64 wheels for TFA, but have heard that the package will not resolve for Monterey users since it's plat-name is for macos-11. Given that M1 chips have better performance on macos-12 (I think?) we wanted to make sure that majority of users will be able to resolve a TFA for macos-12 built for ARM.

These wheels should work on MacOS 12 as well. To make sure macos-11 users don't regress, we were releasing with plat-name macos-11.

However, I noticed tensorflow-macos is still publishing macos-11 plat-names: https://pypi.org/project/tensorflow-macos/2.7.0/#files

1. Is there a plan to increase the OS of the tensorflow-macos build?

https://pypi.org/project/tensorflow-macos/2.11.0/#files
I think this has been addressed.

2. Also is it true that macos 12 users will not be able to resolve/install an earlier tensorflow-macos if it's labeled in pypi with a prior macos version? Thinking about this a bit more I'm confused since we publish our TFA x86_64 wheels for macos-10 and haven't heard this complaint.

These should work for macos-12 users. When releasing your wheels you want to release for the minimum deployment target for MacOS which is supported. The arm64 support has been added since macos-11, and some users had requested support on Macos-11, so we were releasing with that plat-name. But since then macos-12 and macos-13 stable versions have been released, so we have bumped it up to macos-12. Please let me know if there are any concerns.