Fix shape mismatch in Keras Attention layer during masking #21595
Conversation
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Summary of Changes
Hello @sonali-kumari1, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
This pull request resolves a critical RuntimeError within the Keras Attention layer that occurred when using masks. The issue stemmed from an incompatibility in shape between the attention scores and the provided masks, which prevented proper broadcasting. The fix ensures that the mask's dimensions are correctly aligned with the attention scores, allowing the masking operation to proceed without errors and improving the robustness of the Attention layer.
Highlights
- Keras Attention Layer Bug Fix: Addresses a
RuntimeErrorencountered in the Keras Attention layer. This error was caused by a shape mismatch between the attention scores and the mask, leading to broadcasting issues during training. - Mask Dimension Handling: Introduces a conditional check within the
_apply_scoresmethod to ensure thepadding_maskhas the correct dimensions. If the mask is 2D, it is expanded to 3D by adding a new dimension ataxis=-2, resolving the broadcasting conflict.
Using Gemini Code Assist
The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.
Invoking Gemini
You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
Customization
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.
You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
There was a problem hiding this comment.
Code Review
This pull request addresses a RuntimeError in the Keras Attention layer that occurs when using a 2D mask. The error is due to a shape mismatch between the 2D mask and the 3D attention scores, which prevents proper broadcasting. The proposed fix correctly identifies when the padding mask is 2D and expands its dimensions to make it compatible for broadcasting with the scores tensor. The change is concise, well-targeted, and effectively resolves the bug. The implementation is correct.
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #21595 +/- ##
==========================================
- Coverage 82.84% 82.71% -0.13%
==========================================
Files 565 568 +3
Lines 55656 56897 +1241
Branches 8685 8890 +205
==========================================
+ Hits 46108 47063 +955
- Misses 7433 7640 +207
- Partials 2115 2194 +79
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
Hi @fchollet @hertschuh - |
| @@ -88,11 +88,13 @@ def test_attention_with_mask(self): | |||
|
|
|||
| def test_attention_2D_mask_shape_mismatch(self): | |||
| layer = layers.Attention() | |||
| batch_size, Tq, Tv, dim = 2, 3, 3, 4 | |||
| batch_size, Tq, Tv, dim = 2, 3, 4, 4 | |||
There was a problem hiding this comment.
Sorry to be bother you again about this, but can you make all values unique? Like use 5 for dim.
Again, we want to make sure dimensions are matched up correctly in unit tests.
There was a problem hiding this comment.
@hertschuh, I updated the unit test to use distinct values for batch_size, Tq, Tv and dim to properly detect shape mismatches. Thank you!
google-ml-butler
bot
added
kokoro:force-run
ready to pull
github-project-automation
bot
moved this from Assigned Reviewer
to Approved by Reviewer
in PR Queue
This PR fixes the
RuntimeErrorencountered in the Attention layer when using masks. The error arises from a shape mismatch between the mask and attention scores, resulting in a broadcasting issue during training.Fixes: #21483