[v1] add batch generator

[hiyouga]

What does this PR do?

Fixes # (issue)

Before submitting

• Did you read the contributor guideline?
• Did you write any new necessary tests?

[gemini-code-assist]

Summary of Changes

Hello @hiyouga, 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 significantly refactors the data loading and batching infrastructure by introducing a unified BatchGenerator. This change streamlines how data is prepared and fed to the model, offering more control over batching strategies and improving the processing of diverse sample types for training. It centralizes configuration related to batching and sequence length, and enhances the rendering pipeline to support advanced features like DPO-specific token types and tool call handling.

Highlights

• New Batching System: Introduced a new BatchGenerator class to unify and manage data loading and batching, replacing the previous DataLoader and BatchingQueue implementations. This new system supports stateful dataloading and various batching strategies.
• Configuration Updates: The TrainingArguments now includes batching_strategy and batching_workers parameters, and the global_batch_size field has been made more flexible. The cutoff_len argument has been moved from DataArguments to TrainingArguments for better logical grouping.
• Enhanced Sample Processing: The Renderer class gained a new process_samples method, centralizing the conversion of raw samples into model-ready inputs for both Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) tasks. This includes handling token_type_ids for DPO and robust tool call parsing.
• Data Conversion Improvements: Updated data converters (sharegpt_converter, pair_converter) to more robustly handle tools information and structure messages, ensuring compatibility with the new rendering and batching pipeline.

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 introduces a significant and well-structured refactoring of the data loading pipeline by adding a new BatchGenerator class. This new class centralizes batching logic and leverages torchdata.StatefulDataLoader for efficient, distributed data loading, which is a great improvement. The removal of the old DataLoader and BatchingQueue simplifies the codebase.

My review includes a critical fix for a bug in the new padding logic that would have led to incorrect training. I've also provided a suggestion to make the drop_last requirement more flexible for single-device training. Overall, this is a solid contribution that modernizes the data handling in the project.