feat(layers): Migrate RandomChoice and RandomApply layers from keras-cv to keras #20752

harshaljanjani · 2025-01-13T08:37:58Z

Added RandomChoice and RandomApply layers to keras.layers.preprocessing.image_preprocessing
Implemented necessary transform methods (transform_images, transform_labels, transform_bounding_boxes, transform_segmentation_masks) to comply with BaseImagePreprocessingLayer
Updated tests to ensure compatibility with keras and fix failing test cases
Added support for batchwise processing, auto-vectorization, and random seed control
Addresses add RandomApply and RandomChoice #20727

- Added RandomChoice and RandomApply layers to keras.layers.preprocessing.image_preprocessing. - Implemented necessary transform methods (transform_images, transform_labels, transform_bounding_boxes, transform_segmentation_masks) to comply with BaseImagePreprocessingLayer. - Updated tests to ensure compatibility with keras and fix failing test cases. - Added support for batchwise processing, auto-vectorization, and random seed control.

codecov-commenter · 2025-01-13T08:44:49Z

Codecov Report

Attention: Patch coverage is 63.90977% with 48 lines in your changes missing coverage. Please review.

Project coverage is 81.96%. Comparing base (e010829) to head (f4c3f16).
Report is 32 commits behind head on master.

Files with missing lines	Patch %	Lines
.../preprocessing/image_preprocessing/random_apply.py	51.85%	38 Missing and 1 partial ⚠️
...preprocessing/image_preprocessing/random_choice.py	80.43%	6 Missing and 3 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #20752      +/-   ##
==========================================
+ Coverage   81.94%   81.96%   +0.02%     
==========================================
  Files         553      561       +8     
  Lines       51536    52424     +888     
  Branches     7978     8104     +126     
==========================================
+ Hits        42231    42970     +739     
- Misses       7363     7475     +112     
- Partials     1942     1979      +37

Flag	Coverage Δ
keras	`81.77% <63.15%> (+0.01%)`	⬆️
keras-jax	`64.27% <62.40%> (+0.28%)`	⬆️
keras-numpy	`59.00% <62.40%> (+0.15%)`	⬆️
keras-openvino	`29.82% <24.06%> (+<0.01%)`	⬆️
keras-tensorflow	`64.80% <62.40%> (+0.12%)`	⬆️
keras-torch	`64.18% <62.40%> (+0.16%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

keras/src/layers/preprocessing/image_preprocessing/random_apply.py

…andomApply and RandomChoice layers - Updated layers to handle JAX backend-specific behavior and avoid tracer leaks. - Updated tests to use backend-agnostic Keras operations. - Skipped JAX-specific tests with TODO to investigate further. - Removed Tensorflow-specific dependencies for better backend compatibility.

keras/src/layers/preprocessing/image_preprocessing/random_apply.py

keras/src/layers/preprocessing/image_preprocessing/random_apply_test.py

- Add handling for layers with get_random_transformation method in both RandomApply and RandomChoice preprocessing layers - Fix seed handling in RandomChoice to ensure consistent behavior across backends - Remove JAX-specific test skipmarks after addressing tracer errors

keras/src/layers/preprocessing/image_preprocessing/random_apply.py

- Simplified docstrings. - Removed copyright headers from files. - Refactored RandomApply and RandomChoice. - Streamlined transformation logic and improved overall code organization. - Eliminated redundant backend-specific code paths.

harshaljanjani · 2025-01-18T02:11:23Z

@fchollet, would love your thoughts on these changes whenever you get a chance, thanks! Also please do look into the comment I've left unresolved regarding whether it should inherit from keras.layers.Layer instead.

fchollet

Thanks for the PR!

@innat do you want to review?

keras/src/layers/preprocessing/image_preprocessing/random_apply.py

innat · 2025-01-18T19:00:21Z

Sure, I'll review it. Thanks for tagging me.

keras/src/layers/preprocessing/image_preprocessing/random_choice.py

- removed unused auto_vectorize parameter from both layers - shortened docstring headers to fit on one line - added backticks around boolean keywords (True, False) in docstrings - fixed incorrect transform methods (bounding_boxes, labels, segmentation_masks) to return inputs unchanged - removed unnecessary transformation methods from RandomChoice for consistency with original implementation

harshaljanjani · 2025-01-18T20:56:04Z

Thanks for the reviews, @fchollet and @innat. I've tried to incorporate the suggested changes, will check the results of the test cases as soon as I have the opportunity!
Edit: The test cases have passed. Please take a look when you have the chance. Thanks!

innat · 2025-01-19T13:40:20Z

@fchollet
My review is pending, it might require approval.

harshaljanjani · 2025-01-22T17:42:52Z

@fchollet please do look into the changes, thanks!

fchollet · 2025-01-24T22:23:54Z

My review is pending, it might require approval.

I am not seeing anything to approve? I was under the impression anyone could add a review.

@fchollet please do look into the changes, thanks!

Looking now!

harshaljanjani · 2025-01-27T04:08:19Z

Hi @fchollet,
Just wanted to kindly follow up to see if you had a chance to review the changes, please let me know if there's anything else I need to address, thanks!

fchollet · 2025-01-27T20:55:25Z

Tests / Check the code format (pull_request) Failing after 1m

Please fix the code format by running sh shell/format.sh. If any errors remain, fix them manually.

keras/src/layers/preprocessing/image_preprocessing/random_apply.py

fchollet · 2025-01-27T20:57:51Z

I am not even sure we should be brining over these layers in the first place, to be honest.

harshaljanjani · 2025-01-28T02:34:38Z

I am not even sure we should be brining over these layers in the first place, to be honest.

I'd love to keep working on these layers if they are needed in the codebase; if not, I guess closing the PR without merging it is, but I've proceeded to implement the fixes in any case.

Please fix the code format by running sh shell/format.sh. If any errors remain, fix them manually.

Seems the changes need to be made in both src/__init__ and api/__init__ files, as the failing checks are related to API validation rather than linting.

- Add proper transformation of images, labels, and bounding boxes - Extract _should_augment method for cleaner random decision logic and move utility function to test file

harshaljanjani · 2025-02-04T16:50:49Z

Hi, is there a reason this PR is not moving forward? Please do clarify, thanks!

google-ml-butler bot added the size:L label Jan 13, 2025

google-ml-butler bot assigned gbaned Jan 13, 2025

innat reviewed Jan 13, 2025

View reviewed changes

keras/src/layers/preprocessing/image_preprocessing/random_apply.py Outdated Show resolved Hide resolved

harshaljanjani added 2 commits January 13, 2025 22:13

fix: Make the linter happy :)

9b3dda3

harshaljanjani requested a review from innat January 14, 2025 03:44

google-ml-butler bot added the awaiting review label Jan 14, 2025

innat reviewed Jan 14, 2025

View reviewed changes

keras/src/layers/preprocessing/image_preprocessing/random_apply.py Show resolved Hide resolved

innat reviewed Jan 14, 2025

View reviewed changes

keras/src/layers/preprocessing/image_preprocessing/random_apply_test.py Outdated Show resolved Hide resolved

harshaljanjani requested a review from innat January 14, 2025 13:34

fchollet reviewed Jan 16, 2025

View reviewed changes

fix: Refactor RandomApply and RandomChoice layers

9fd3bd2

- Simplified docstrings. - Removed copyright headers from files. - Refactored RandomApply and RandomChoice. - Streamlined transformation logic and improved overall code organization. - Eliminated redundant backend-specific code paths.

harshaljanjani requested a review from fchollet January 16, 2025 07:23

fchollet reviewed Jan 18, 2025

View reviewed changes

innat reviewed Jan 18, 2025

View reviewed changes

keras/src/layers/preprocessing/image_preprocessing/random_choice.py Outdated Show resolved Hide resolved

harshaljanjani requested review from innat and fchollet January 19, 2025 08:04

fchollet reviewed Jan 27, 2025

View reviewed changes

harshaljanjani added 2 commits January 28, 2025 13:04

feat: Refactor RandomApply layer completely to stay true to the original

4c73a5d

- Add proper transformation of images, labels, and bounding boxes - Extract _should_augment method for cleaner random decision logic and move utility function to test file

refactor: Remove dead code and edit docstring

f4c3f16

harshaljanjani requested a review from fchollet January 28, 2025 12:20

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(layers): Migrate RandomChoice and RandomApply layers from keras-cv to keras #20752

feat(layers): Migrate RandomChoice and RandomApply layers from keras-cv to keras #20752

harshaljanjani commented Jan 13, 2025

codecov-commenter commented Jan 13, 2025 •

edited

Loading

harshaljanjani commented Jan 18, 2025 •

edited

Loading

fchollet left a comment

innat commented Jan 18, 2025

harshaljanjani commented Jan 18, 2025 •

edited

Loading

innat commented Jan 19, 2025

harshaljanjani commented Jan 22, 2025

fchollet commented Jan 24, 2025

harshaljanjani commented Jan 27, 2025

fchollet commented Jan 27, 2025

fchollet commented Jan 27, 2025

harshaljanjani commented Jan 28, 2025 •

edited

Loading

harshaljanjani commented Feb 4, 2025

feat(layers): Migrate RandomChoice and RandomApply layers from keras-cv to keras #20752

Are you sure you want to change the base?

feat(layers): Migrate RandomChoice and RandomApply layers from keras-cv to keras #20752

Conversation

harshaljanjani commented Jan 13, 2025

codecov-commenter commented Jan 13, 2025 • edited Loading

Codecov Report

harshaljanjani commented Jan 18, 2025 • edited Loading

fchollet left a comment

Choose a reason for hiding this comment

innat commented Jan 18, 2025

harshaljanjani commented Jan 18, 2025 • edited Loading

innat commented Jan 19, 2025

harshaljanjani commented Jan 22, 2025

fchollet commented Jan 24, 2025

harshaljanjani commented Jan 27, 2025

fchollet commented Jan 27, 2025

fchollet commented Jan 27, 2025

harshaljanjani commented Jan 28, 2025 • edited Loading

harshaljanjani commented Feb 4, 2025

codecov-commenter commented Jan 13, 2025 •

edited

Loading

harshaljanjani commented Jan 18, 2025 •

edited

Loading

harshaljanjani commented Jan 18, 2025 •

edited

Loading

harshaljanjani commented Jan 28, 2025 •

edited

Loading