[megatron] support megatron fsdp

[Jintao-Huang]

No description provided.

[gemini-code-assist]

Summary of Changes

Hello @Jintao-Huang, 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 introduces foundational support for Megatron FSDP, a technique designed to optimize memory usage and scalability in large-scale distributed training. It provides new configuration options that allow users to enable FSDP, choose between Megatron's implementation or PyTorch's FSDP2, and specify different sharding strategies for data parallelism. These changes are reflected in both the core argument definition and the user-facing documentation.

Highlights

• Megatron FSDP Support: Introduced support for Megatron Fully Sharded Data Parallel (FSDP) to enhance distributed training capabilities.
• New Command-Line Parameters: Added three new command-line arguments: use_megatron_fsdp to enable Megatron FSDP, use_torch_fsdp2 for PyTorch's FSDP2 implementation, and data_parallel_sharding_strategy to define how data parallelism sharding is applied.
• Argument Class Integration: Integrated the new FSDP-related parameters into the MegatronArguments class within swift/megatron/argument/megatron_args.py.
• Documentation Updates: Updated both Chinese and English documentation files (docs/source/Megatron-SWIFT/Command-line-parameters.md and docs/source_en/Megatron-SWIFT/Command-line-parameters.md) to reflect the newly added FSDP command-line parameters.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

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 support for Megatron FSDP by adding new command-line parameters and their corresponding fields in the MegatronArguments dataclass. The documentation has been updated in both Chinese and English, which is crucial for user understanding. The changes are well-aligned with the pull request's objective.

[gemini-code-assist]

It's good practice to add metadata={'help': '...'} to dataclass fields that represent command-line arguments. This makes the arguments self-documenting and improves clarity when generating help messages or inspecting the arguments programmatically. Please consider adding descriptive help strings for these new FSDP-related parameters.

Suggested change

	use_megatron_fsdp: bool = False
	use_torch_fsdp2: bool = False
	data_parallel_sharding_strategy: Literal['no_shard', 'optim', 'optim_grads', 'optim_grads_params'] = 'no_shard'
	use_megatron_fsdp: bool = field(default=False, metadata={'help': 'Use Megatron FSDP in DDP.'})
	use_torch_fsdp2: bool = field(default=False, metadata={'help': 'Use torch FSDP2 implementation (recommend using `--use_megatron_fsdp` instead).'})
	data_parallel_sharding_strategy: Literal['no_shard', 'optim', 'optim_grads', 'optim_grads_params'] = field(default='no_shard', metadata={'help': "Sharding strategy for data parallelism. Options are 'no_shard', 'optim', 'optim_grads', 'optim_grads_params'."})