[megatron] support megatron embedding

[Jintao-Huang]

squash from #7630

[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 significantly extends the ms-swift framework by integrating full support for training embedding models with Megatron parallelism. The changes enable efficient distributed training of embedding models, such as Qwen3-Embedding, by adapting the core trainer logic, model architecture, loss calculation, and evaluation metrics to Megatron's parallelization strategies. This ensures that embedding models can leverage large-scale distributed computing resources effectively.

Highlights

• Megatron Embedding Support: Introduced comprehensive support for training embedding models using Megatron parallelism, including a new task_type='embedding' option and a dedicated MegatronEmbeddingTrainer.
• Documentation Updates: Updated README.md, README_CN.md, and various instruction and quick-start guides to reflect the new embedding capabilities, parameters, and Megatron parallelism support.
• New Embedding Examples: Added two new example scripts (qwen3_emb.sh and qwen3_vl_emb.sh) demonstrating how to train Qwen3 embedding models with Megatron.
• Loss and Metric Handling for Embedding: Enhanced the BaseLoss class to identify Megatron trainers, and updated InfonceLoss to correctly handle data parallel gathering. New EmbeddingMetricMixin and modifications to PairedMetrics and InfonceMetrics were added for robust evaluation.
• Megatron Model Bridge Enhancements: Modified the GPTBridge to manage padding for embedding layers during tensor parallelism splitting and to bypass output layer weight processing for embedding tasks.
• Trainer Refactoring and Renaming: Refactored the Megatron SFT pipeline to dispatch to the new MegatronEmbeddingTrainer for embedding tasks. Also, embedding.py and reranker.py in swift/trainers were renamed to embedding_trainer.py and reranker_trainer.py respectively.

🧠 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 Megatron support for embedding models, along with necessary updates to documentation, arguments, and trainer logic. The changes include adding embedding as a supported task type, implementing specific handling for embedding models in Megatron's GPT bridge and model, and creating a dedicated MegatronEmbeddingTrainer. Several documentation files have been updated to reflect these new capabilities. The refactoring of RerankerLoss to PointwiseRerankerLoss and the corresponding updates in loss/mapping.py are also noted. Overall, the changes are well-aligned with the goal of supporting Megatron embedding.

[gemini-code-assist]

The num_samples key is popped from res and then potentially re-added for Megatron. While this might be intended, ensure that num_samples is consistently handled across different scenarios (Megatron vs. non-Megatron, and various task types) to prevent unexpected behavior or incorrect batching/loss calculations.