Skip to content
Computer vision

RSI-Jev v7.0-VL 3B

VLM, которая отвечает вероятностями за один проход

1 min read

Это урезанная до 20 слоёв версия v7.0-VL 4B без переобучения: 3,30 млрд параметров, первые 20 из 32 decoder-слоёв, главная голова layer-20 и early exit на layer-16. Модель отвечает на вопросы типов choice / noul / score по документу и до четырёх изображений.

Ключевые особенности

  • System One: один forward pass, без генерации reasoning, вероятность для каждого варианта.
  • Все веса — от v7.0-VL 4B; на 4 001 строке Decision Index ответы совпадают бит-в-бит.
  • Самодостаточный чекпоинт: 20 decoder-слоёв, эмбеддинги, vision tower и две головы.
  • Decision Index 0.3: 60,01 против 60,61 у 4B; медианная задержка 19,6 мс против 29,4 мс.
  • Jev HTTP API: до 32 768 токенов, до 5 120 вариантов, вход не обрезается.
  • effort задаёт число слоёв (16 или 20), а не число токенов; токены не генерируются.

Пример

pip install "rsi-jev[fast,vision] @ git+https://github.com/Shanghua-Gao/RSI-Jev"
rsi-jev serve v7.0-vl-3b
ПодробнееDecision Index 0.3: 60,01 (у 4B — 60,61); 15-бенчмарковый suite — 0,791; held-out — 0,797.
Запуск на одной RTX PRO 6000 Blackwell: медиана 19,6 мс, 95-й перцентиль 151 мс; пакет — 6,8 ГБ.
Ограничения: нет layer 32, знания проседают (MMLU-Pro 0,469 против 0,502), пороги не переобучали под два выхода.
Discuss Telegram 0 45

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.

    # shgao/rsi-jev-v7.0-vl-3b
    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 shgao/rsi-jev-v7.0-vl-3b --local-dir ~/models/rsi-jev-v7.0-vl-3b
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to rsi-jev-v7.0-vl-3b
  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 rsi-jev-v7.0-vl-3b (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

shgao/rsi-jev-v7.0-vl-3b
Author
shgao
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

450

Ask AI about this model

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

Comments

Computer vision

ALWM

ALWM