fix: remove attention mask shift in flex attention implementation

[jialefu]

Motivation

The flex_attention implementation of EAGLE-3 uses an unnecessary attention mask shift. Here's an example when lck(current ttt_length)=2.

Modifications

Remove the attention shift:

Related Issues

Accuracy Test

Benchmark & Profiling

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://sgl-fru7574.slack.com/archives/C09784E3EN6 to discuss your PR.

[gemini-code-assist]

Summary of Changes

Hello @jialefu, 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 addresses and corrects an issue within the EAGLE-3 attention mechanism where an extraneous shift was being applied to the attention mask. By eliminating this unnecessary shift and refining the padding logic, the changes ensure the attention mask functions as intended, providing accurate causal and padding mask application within the flex attention implementation. This fix aims to improve the correctness of the model's attention behavior.

Highlights

• Attention Mask Logic Refinement: The generate_eagle3_mask function has been updated to remove an unnecessary attention mask shift. The causal_mask calculation no longer subtracts a shift_left value, and the padding_mask now explicitly checks q_idx < seq_lengths[b] to ensure correct masking behavior. The parameter shift_left has also been renamed to lck for clarity.
• Parameter Renaming and Usage: The shift_left parameter in generate_eagle3_mask has been consistently renamed to lck across its definition, internal usage for mask naming, and calls from llama3_eagle.py and test_flex_attention.py.
• Test Case Alignment: The test_eagle3_flex_mask unit test has been updated to reflect the parameter name change, ensuring that the test suite correctly validates the modified generate_eagle3_mask function.

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 by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

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 pull request 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

1. 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. ↩

[gemini-code-assist]

Code Review

This pull request correctly removes an unnecessary attention mask shift in the EAGLE-3 flex attention implementation. The change simplifies the causal mask generation and also correctly fixes a bug related to padding by ensuring that queries from padding tokens do not attend to any keys. My review includes one suggestion to further improve maintainability and performance by removing a now-unused parameter from the generate_eagle3_mask function.

[yubofredwang]

Thanks for making the change. Please fix the unit tests as well! Linking the related discussion here for more context: SafeAILab/EAGLE#247

[jialefu]

Thanks for making the change. Please fix the unit tests as well! Linking the related discussion here for more context: SafeAILab/EAGLE#247

Hi, I've removed the attention mask shift in the unit tests, the SDPA implementation, and benchmark_flex_attention.py.

[yubofredwang]

@jialefu please fix lint