fix:integerlookup one_hot shape inference for 2D inputs

[maitry63]

This PR fixes a regression in IntegerLookup with output_mode="one_hot" where 2D inputs (batch, sequence_length) were incorrectly producing output shapes in symbolic mode.
Fixes: #22520

[gemini-code-assist]

Code Review

This pull request addresses a shape collapse issue in the IndexLookup layer when using one_hot output mode with 2D symbolic inputs. The changes update compute_output_shape and compute_output_spec to correctly preserve input dimensions and add regression tests for these scenarios. The review feedback identifies significant code duplication between these two methods regarding depth calculation and one_hot logic, recommending a refactor into a shared helper method to improve code maintainability.

[gemini-code-assist]

There is some code duplication between compute_output_spec and compute_output_shape that could be refactored to improve maintainability.

1. The calculation of depth is identical in both methods (here and in lines 561-564). This could be extracted to a private helper method to avoid redundancy.
2. The logic for one_hot output shape calculation is also duplicated. It seems compute_output_shape is now correct for all cases. If possible, compute_output_spec could be simplified to call compute_output_shape for all modes except 'int', which would remove the duplication.

If this duplication is intentional and necessary for the fix to work correctly in all execution modes, adding a comment explaining the reason would be very helpful for future maintenance.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 91.66667% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.30%. Comparing base (8f5ef11) to head (e38e26d).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
keras/src/layers/preprocessing/index_lookup.py	91.66%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #22592   +/-   ##
=======================================
  Coverage   83.30%   83.30%           
=======================================
  Files         596      596           
  Lines       67962    67960    -2     
  Branches    10580    10578    -2     
=======================================
  Hits        56615    56615           
+ Misses       8600     8599    -1     
+ Partials     2747     2746    -1     

Flag	Coverage Δ
keras	83.11% <91.66%> (+<0.01%)	⬆️
keras-jax	59.72% <91.66%> (+<0.01%)	⬆️
keras-numpy	54.34% <87.50%> (+<0.01%)	⬆️
keras-openvino	51.80% <4.16%> (+<0.01%)	⬆️
keras-tensorflow	61.03% <91.66%> (+<0.01%)	⬆️
keras-torch	59.91% <91.66%> (+<0.01%)	⬆️

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.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[keerthanakadiri]

Code Review

This pull request addresses a shape collapse issue in the IndexLookup layer when using one_hot output mode with 2D symbolic inputs. The changes update compute_output_shape and compute_output_spec to correctly preserve input dimensions and add regression tests for these scenarios. The review feedback identifies significant code duplication between these two methods regarding depth calculation and one_hot logic, recommending a refactor into a shared helper method to improve code maintainability.

Hi @maitry63, Can you check with this once, looks like outdated. Thanks !

[hertschuh]

Most changes are spurious formatting changes. Please undo all of the these to keep only the actual changes.

Please also rebase.

Thanks!