[nemetron/mtp] fix nemotron mtp

[hanming-lu]

Summary

• Fix nemotron mtp
• Disable radix cache for this config first

python3 -m sglang.launch_server \
  --model-path nvidia/NVIDIA-Nemotron-Nano-9B-v2 \
  --speculative-algorithm STANDALONE \
  --speculative-num-steps 2\
  --speculative-eagle-topk 3 \
  --speculative-num-draft-tokens 5 \
  --speculative-draft-model-path meta-llama/Llama-3.2-1B \
  --speculative-draft-load-format dummy \
  --max-running-requests 8 \
  --max-total-tokens 2048 \
  --json-model-override-args '{"vocab_size": 131072}' \
  --device cuda \
  --host 127.0.0.1 \
  --port 21000

root@memx-crf-41-sr1:~/oss/sglang# python3 -m sglang.test.few_shot_gsm8k --num-questions 100 --port 21000
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 100/100 [02:58<00:00,  1.78s/it]
Accuracy: 0.910
Invalid: 0.010
Latency: 178.447 s
Output throughput: 66.137 token/s

[gemini-code-assist]

Summary of Changes

Hello @hanming-lu, 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 a specific bug within the Mamba attention layer implementation, focusing on the correct handling of intermediate states during sequence generation. By ensuring the appropriate indices are passed, it aims to stabilize and improve the functionality of the nemotron/mtp feature.

Highlights

• Mamba Attention Layer Fix: Corrected an issue in the Mamba attention layer by explicitly passing intermediate_state_indices to a function call within the forward method. This ensures proper state management during decoding processes, likely related to multi-token prediction (MTP) for Nemotron models.

🧠 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 correctly fixes a bug in the Mamba mixer layer that occurs during speculative decoding. The missing intermediate_state_indices parameter in the causal_conv1d_update_triton call would lead to a crash. The provided fix is accurate and essential for the functionality. I have included one minor suggestion to improve code clarity.

[hanming-lu]

/tag-and-rerun-ci