[Feature] Provide a trained eagle3 model for Qwen/Qwen3-Next-80B-A3B-Instruct

[huang3eng]

Checklist

• 1. If the issue you raised is not a feature but a question, please raise a discussion at https://github.com/sgl-project/SpecForge/discussions/new/choose Otherwise, it will be closed.
• 2. Please use English, otherwise it will be closed.

Motivation

Great work! Currently, the eagle3 model for Qwen/Qwen3-Next-80B-A3B-Instruct is missing in SpecBundle. I have successfully trained it using 8 b200 GPUs, and the acceleration ratio in benchmark tests is improved compared to lmsys/SGLang-EAGLE3-Qwen3-Next-80B-A3B-Instruct-FP8-SpecForge-Meituan. Can I contribute this model to https://huggingface.co/collections/lmsys/specbundle?

Related resources

No response

[jiapingW]

Sure. You can open source your model and prepare its docs and give us your url. We'll add it to specbundle.

[FrankLeeeee]

Hi @huang3eng can you share with me the link? I will add it the specbundle collection.

[huang3eng]

@FrankLeeeee Due to compliance mechanisms, the company's computers cannot upload related files to Hugging Face. I am currently using a personal computer to train the draft model on a GPU rental platform, and it has already been trained. It is expected that it will be released this week.

[FrankLeeeee]

Ok, I see, thanks!

[huang3eng]

@FrankLeeeee @jiapingW The draft model link as follows, https://huggingface.co/huang3eng/SGLang-EAGLE3-Qwen3-Next-80B-A3B-Instruct-SpecForge

[huang3eng]

@FrankLeeeee Is there some problem with the draft model? It seems that it hasn't been added yet.

[jiapingW]

Thanks. We'll handle it soon.