📦️ Fix macOS packaging #838
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The TensorFlow Text package requires use of a matching minor version of TensorFlow. In order to support Apple Silicon macs, Tensorflow Text chooses between the `tensorflow` and `tensorflow-macos` packages automatically based on system architecture. However, I believe that commit a7e1b5b accidentally wiped out this logic. Additionally, this commit failed to increment the minor version of the `tensorflow-macos` dependency. This typically isn't a problem for users installing TensorFlow Text using `pip` on non-Apple Silicon devices, since `pip` just tries to find the best set of packages for the current system. However, this _is_ a problem for package managers like Poetry, which create a deterministic lock file allowing a similar set of dependencies to be installed across different systems, modulo any architectural differences. Since Poetry needs to resolve a combination of dependencies that will work for _any_ system, it eagerly tries to identify a version of `tensorflow-macos` that will work on Apple Silicon devices. However, since the minor versions of the `tensorflow` and `tensorflow-macos` dependencies don't match, it's unable to find an appropriate version of `keras` that both can use. As a result, installing `tensorflow-text (>=2.8.0)` using Poetry is currently impossible. These changes restore the logic to automatically select between `tensorflow` and `tensorflow-macos` based on platform and architecture identifiers, and bump the version of the `tensorflow-macos` dependency to match the minor version of `tensorflow` and `tensorflow-text`.
copybara-service bot
pushed a commit
that referenced
this pull request
Feb 17, 2022
PiperOrigin-RevId: 429177751
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
The TensorFlow Text package requires use of a matching minor version of TensorFlow. In order to support Apple Silicon macs, Tensorflow Text chooses between the
tensorflowandtensorflow-macospackages automatically based on system architecture. However, I believe that commit a7e1b5b accidentally wiped out this logic. Additionally, this commit failed to increment the minor version of thetensorflow-macosdependency.This typically isn't a problem for users installing TensorFlow Text using
pipon non-Apple Silicon devices, sincepipjust tries to find the best set of packages for the current system. However, this is a problem for package managers like Poetry, which create a deterministic lock file allowing a similar set of dependencies to be installed across different systems, modulo any architectural differences. Since Poetry needs to resolve a combination of dependencies that will work for any system, it eagerly tries to identify a version oftensorflow-macosthat will work on Apple Silicon devices. However, since the minor versions of thetensorflowandtensorflow-macosdependencies don't match, it's unable to find an appropriate version ofkerasthat both can use. As a result, installingtensorflow-text (>=2.8.0)using Poetry is currently impossible.These changes restore the logic to automatically select between
tensorflowandtensorflow-macosbased on platform and architecture identifiers, and bump the version of thetensorflow-macosdependency to match the minor version oftensorflowandtensorflow-text.