SGLang-EAGLE3-Llama-4-Scout-17B-16E-Instruct-SpecForge

EAGLE3 For meta-llama/Llama-4-Scout-17B-16E-Instruct

About

SpecBundle is an open-source initiative, jointly driven by the community and industry, to democratize speculative decoding by providing high-performance speculative decoding draft weights for mainstream open-source models.

This checkpoint was trained by the SpecForge Team and released as the phase 1 of SpecBundle release. We regenerated the responses in the mlabonne/open-perfectblend and trained the model on 1.4M data samples for 2 epochs. This checkpoint was trained using the SpecForge framework.

Usage

You can use this checkpoint with the command below.

export SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN=1

python3 -m sglang.launch_server \
    --model meta-llama/Llama-4-Scout-17B-16E-Instruct \
    --speculative-algorithm EAGLE3 \
    --speculative-draft-model-path lmsys/SGLang-EAGLE3-Llama-4-Scout-17B-16E-Instruct-SpecForge \
    --speculative-num-steps 3 \
    --speculative-eagle-topk 1 \
    --speculative-num-draft-tokens 4 \
    --tp 4

Performance

This checkpoint exhibits superior performance on various benchmarks.

<table> <tr> <th>Throughput</th> <th>Acceptance Length</th> </tr> <tr> <td><img src="https://cdn-uploads.huggingface.co/production/uploads/63993d721fad4d6eb265d999/m35UETCjmjceOBB4ZnI76.png" /></td> <td><img src="https://cdn-uploads.huggingface.co/production/uploads/63993d721fad4d6eb265d999/VC7nziyWJPtSiWj9q3XOs.png" /></td> </tr> </table>

You can reproduce the performance with the command below:

# clone specforge
git clone https://github.com/sgl-project/SpecForge.git
cd SpecForge/benchmarks

# run benchmarks
python bench_eagle3.py \
        --model meta-llama/Llama-4-Scout-17B-16E-Instruct \
        --speculative-algorithm EAGLE3 \
        --speculative-draft-model-path lmsys/SGLang-EAGLE3-Llama-4-Scout-17B-16E-Instruct-SpecForge \
        --port 30000 \
        --config-list 8,0,0,0 8,3,1,4 8,5,1,6 8,5,3,6 8,7,1,8 8,7,4,8 \
        --benchmark-list gsm8k math500 mtbench humaneval livecodebench financeqa gpqa \
        --dtype bfloat16 \
        --tp 4 \
        --name llama4-scout-spec-bundle

Acknowledgement

We sincerely appreciate the collective efforts from both the developers in the open-source community and our industrial partners, especially Ant Group AQ Team, Meituan, Nex-AGI (Qiji Zhifeng), EigenAI for their invaluable contributions to the release of SpecBundle Phase 1.