Skip to content
Other

Qwen3.8-27B-Quark-DFlash2

draft-модель для быстрой генерации

1 min readupdated

Это спекулятивная draft-модель для Qwen3.8-27B, обученная в AMD Quark. Она добавляет в vLLM метод dflash и предлагает до 7 токенов на шаг, чтобы целевая LLM генерировала быстрее. Размер — около 1.27B параметров в bfloat16.

Ключевые особенности

  • Обучена на полном Tulu-3 SFT mixture с отключённым thinking, 3 эпохи (17 118 шагов).
  • Архитектура draft: 5 non-causal слоёв, hidden 5120, 40 heads / 8 KV heads, block size 8.
  • Использует grouped conv с kernel 2 и candidate selector (rank 256, top-k 16).
  • Для работы нужен vLLM с поддержкой DFlash2 (DFlash2DraftModel).

Пример запуска

vllm serve Qwen/Qwen3.8-27B --language-model-only \
  --speculative-config '{"method": "dflash", "model": "ZachWangAMD/Qwen3.8-27B-Quark-DFlash2", "num_speculative_tokens": 7}'
ПодробнееSpec-Bench (K=7): acceptance length — 3.67, ускорение против целевой модели — 2.31x.
Acceptance по позициям: 0.781 / 0.568 / 0.415 / 0.312.
Нужен vLLM с DFlash2; целевая модель подаётся с --language-model-only.
Не меняйте conv_kernel_size: при значении 1 acceptance length падает до 1.79.
Discuss Telegram 0 36

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # amd/Qwen3.8-27B-Quark-DFlash2
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download amd/Qwen3.8-27B-Quark-DFlash2 --local-dir ~/models/Qwen3.8-27B-Quark-DFlash2
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of Qwen3.8-27B-Quark-DFlash2 (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

amd/Qwen3.8-27B-Quark-DFlash2
Author
amd
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

360

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments