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

tensorflow-lite: Add v2.15.0 #21763

Merged
merged 1 commit into from
Aug 21, 2024

Conversation

talyz
Copy link
Contributor

@talyz talyz commented Dec 14, 2023

Specify library name and version: tensorflow-lite/v2.15.0

Add 2.15.0 and update a few dependencies which were causing crashes when running on M1 macs under rosetta. Also, remove 2.12.0, which was just a copy of 2.10.0.

Depends on #21762, #21761, #21760 and #21759, which should be merged before this.


@conan-center-bot

This comment has been minimized.

@minus7
Copy link

minus7 commented Dec 20, 2023

Removal of 2.12.0 closes #20170

I tested this and it works (in contrast to the currently published versions 2.10 and "2.12", which fail to build)

@talyz
Copy link
Contributor Author

talyz commented Feb 28, 2024

Rebased on master, now that the dependent package updates are merged.

@conan-center-bot

This comment has been minimized.

@talyz talyz force-pushed the tensorflow-lite-2.15.0 branch 2 times, most recently from fe6974e to e8483a2 Compare February 28, 2024 15:30
@conan-center-bot

This comment has been minimized.

@conan-center-bot

This comment has been minimized.

@conan-center-bot

This comment has been minimized.

@talyz
Copy link
Contributor Author

talyz commented Mar 1, 2024

All build failures and linter issues should be resolved now and this should be ready for review. We're using this package in production since a few months back (static build, windows and mac).

@mologie
Copy link
Contributor

mologie commented Mar 1, 2024

Likewise using this on macOS, iOS, and Linux without modification. Cross-compiling to Android requires working around a problem in xnnpack/clang, but that's an upstream issue and I'll provide a separate PR to the xnnpack recipe once this one is merged. LGTM, thank you @talyz for building this integration!

(I am not associated with JFrog, still needs maintainer approval.)

@talyz
Copy link
Contributor Author

talyz commented Mar 18, 2024

Just rebased on master to get the separate dependency update in. @uilianries is there anything more that needs to happen for this to be reviewed / merged?

@conan-center-bot

This comment has been minimized.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can be simplified or even avoided by using CMakeDeps.set_property() on the cmake_file_name and cmake_target_name properties.

@ghost ghost mentioned this pull request May 9, 2024
@mologie
Copy link
Contributor

mologie commented Jun 7, 2024

Authentic Vision is now using this recipe as-is in production since April with no known issues on Ubuntu 22.04, 24.04, iOS 13-17, and Android 6-14. The XNNPACK change at #22951 was recently merged to allow building for Android -- if anyone following this PR wants to do that, bump your caches :).

We'd eventually want to bump TensorFlow and its dependencies to 2.15.1 and 2.16.X, would appreciate if this could be reviewed again and merged in preparation for that.

@AbrilRBS
Copy link
Member

AbrilRBS commented Jun 8, 2024

We'd eventually want to bump TensorFlow and its dependencies to 2.15.1 and 2.16.X, would appreciate if this could be reviewed again and merged in preparation for that.

We'll push this review forward this next week, thanks a lot for taking the time to create the PR in the first place. Could you fix the conandata conflicts meanwhile? Thanks!

@mologie
Copy link
Contributor

mologie commented Jun 9, 2024

We'll push this review forward this next week, thanks a lot for taking the time to create the PR in the first place. Could you fix the conandata conflicts meanwhile? Thanks!

Thanks for the update @RubenRBS, the original author is @talyz though and probably received your comment with this thread anyway. I have just been using this recipe from this branch and worked on Android dependencies a bit.

@conan-center-bot

This comment has been minimized.

@conan-center-bot

This comment has been minimized.

@conan-center-bot

This comment has been minimized.

@mologie
Copy link
Contributor

mologie commented Jul 17, 2024

The bot is right @talyz, patches/disable_fetch_content.patch referenced in conandata.yml does not exist in the source tree. (It's unclear to me why this does not fail with a FileNotFoundError at some point, did not dig into Conan for that.)

Other patches have a version prefix and order in their filenames, would thus suggest to match that style for consistency.

Once fixed, this should pass CI.

@conan-center-bot
Copy link
Collaborator

Conan v1 pipeline ✔️

All green in build 5 (0354bf31bf89a9ab189a706c64df75a613ae599f):

  • tensorflow-lite/2.10.0:
    All packages built successfully! (All logs)

  • tensorflow-lite/2.15.0:
    All packages built successfully! (All logs)

  • tensorflow-lite/2.12.0:
    All packages built successfully! (All logs)


Conan v2 pipeline ✔️

Note: Conan v2 builds are now mandatory. Please read our discussion about it.

All green in build 5 (0354bf31bf89a9ab189a706c64df75a613ae599f):

  • tensorflow-lite/2.15.0:
    All packages built successfully! (All logs)

  • tensorflow-lite/2.12.0:
    All packages built successfully! (All logs)

  • tensorflow-lite/2.10.0:
    All packages built successfully! (All logs)

Copy link
Member

@AbrilRBS AbrilRBS left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

10/10 :)

@conan-center-bot conan-center-bot merged commit 461310a into conan-io:master Aug 21, 2024
20 checks passed
@talyz talyz deleted the tensorflow-lite-2.15.0 branch August 23, 2024 08:46
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

Successfully merging this pull request may close these issues.

8 participants