Skip to content
Other

Aster-JEV-27B

мультимодальная модель выбора ответа

1 min read

Aster-JEV-27B — полнопараметрическая мультимодальная decision-модель на базе Qwen/Qwen3.8-27B (26,09 млрд параметров). Она выбирает ответ из набора вариантов и возвращает вероятности; в карточке описаны vision-бенчмарк и latency на A800 80GB.

Ключевые особенности

  • Полный файнтюн всех параметров на базе Qwen/Qwen3.8-27B.
  • Мультимодальный vision-сценарий: один forward pass возвращает выбор и вероятности для всех вариантов.
  • Вопросы с выбором из 2–26 вариантов, до 262 144 входных токенов.
  • Изображения до 1 638 400 пикселей на изображение; жёсткого лимита на число изображений нет.
  • Веса под CC BY-NC 4.0, базовая Qwen — Apache-2.0.

Пример запуска

hf download k2021211255/Aster-JEV-27B --revision single-a800-v1 --local-dir Aster-JEV-27B
python -m pip install -r Aster-JEV-27B/requirements.txt
CUDA_VISIBLE_DEVICES=0 PYTHONPATH=Aster-JEV-27B decision-index vision pipeline --dir work/vision --engine aster_jev.engine:AsterJEVEngine --option model=Aster-JEV-27B --workers 1 --out runs/Aster-JEV-27B
ПодробнееПубличный self-scored Vision index: 76.7057, 9 867 / 12 210 верных.
Latency на A800 80GB, batch size 1: mean 507,2 мс, median 361,5 мс, P80 581,9 мс.
Модель целиком работает на cuda:0; проверка на RTX PRO 6000 (mean/median/P80 ≤1 000 мс) ожидается.
Ограничение: весь вход должен помещаться в память устройства; ошибки ёмкости сообщаются явно.
Discuss Telegram 0 57

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 49 GB · ready files54 GB
  • FP8 / INT8recommendedweights 24 GB27 GB
  • 4-bitweights 14 GB16 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

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.

    # k2021211255/Aster-JEV-27B
    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 k2021211255/Aster-JEV-27B --local-dir ~/models/Aster-JEV-27B
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Aster-JEV-27B
  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 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 2All on Hugging Face

About the model

k2021211255/Aster-JEV-27B
Author
k2021211255
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

570

Ask AI about this model

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

Comments