Keep the version of the dependency library consistent with that of sglang.

[sleepcoo]

No description provided.

[gemini-code-assist]

Summary of Changes

Hello @sleepcoo, 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 primarily focuses on upgrading core Python dependencies, specifically torch and transformers, to maintain consistency with the sglang library. This dependency update necessitated corresponding code modifications within the specforge modeling components to adapt to API changes in the newer transformers library, particularly concerning how model keyword arguments are handled. Additionally, the PR includes a minor but important enhancement to the wandb argument validation logic for improved robustness in distributed environments.

Highlights

• Dependency Version Updates: The requirements.txt file has been updated to specify newer versions for torch (from 2.6.0 to 2.7.1) and transformers (from 4.53.2 to 4.54.0). Additionally, torchaudio==2.7.1 and torchvision==0.22.1 have been added as new dependencies, likely to align with the sglang library's requirements.
• Transformers API Adaptation: The model implementations for Llama4TextCausalLM and Qwen3MoeForCausalLM have been adjusted to reflect changes in the updated transformers library. This includes removing the KwargsForCausalLM class and its associated Unpack type hints, as well as the LossKwargs import, indicating a streamlined approach to handling keyword arguments in the newer transformers version.
• Wandb Argument Validation Improvement: The validate_wandb_args utility function has been refined. It now includes an early exit if wandb is not enabled, preventing unnecessary validation. Furthermore, the check for a missing wandb_key now applies to all distributed ranks, ensuring consistent error reporting across the system.

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 or fill out our survey to provide feedback.

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 updates the project's dependencies, including torch and transformers, and adapts the codebase to the new versions. The changes in the model files appear to correctly handle API changes from the transformers library. I've made one suggestion in specforge/utils.py to improve the error handling in the distributed setting for a cleaner user experience.

[sleepcoo]

llama3 is training ShareGPT for one Epoch Acceptance length: 2.42

python3 benchmark/mtbench/bench_sglang_eagle.py    

[FrankLeeeee]

Hi @sleepcoo can you resolve the conflict?

[FrankLeeeee]

didn't call pre-commit?

[sleepcoo]

didn't call pre-commit?

Done,Caused by existing code

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!