Skip to content
Computer vision

MedChain-R1-9B

клиническая цепочка рассуждений для медицинской локализации

2 min readupdated

MedChain-R1-9B — 9B vision-language модель для sequential medical image grounding. На вход она получает последовательность медицинских изображений и клинический вопрос, сначала пишет rationale в <think>…</think>, а затем возвращает один bounding box в пиксельных координатах. Построена на Qwen3.5-9B: dCoT-SFT на корпусе MedChain, затем GRPO с наградой за IoU и формат.

Ключевые особенности

  • Принимает серию медицинских изображений и вопрос, выдаёт рассуждение и bounding box в формате <answer>(x1,y1),(x2,y2)</answer>.
  • Обучена на MedChain: 182 573 отфильтрованных rationale для 8 задач MedSG-Bench и 10 модальностей визуализации.
  • Сначала dCoT-SFT, затем GRPO с наградой за IoU и формат; 9B-чекпоинт рекомендован как самый точный в семействе.
  • На MedSG-Bench этот чекпоинт набирает 77.54 mIoU и 84.87 Acc@0.5.
  • Есть 4B-вариант для сред с ограниченными ресурсами.

Пример

from transformers import AutoModelForImageTextToText, AutoProcessor
repo = "adidukrembzuai/MedChain-R1-9B"
model = AutoModelForImageTextToText.from_pretrained(repo, torch_dtype=torch.bfloat16, device_map="auto").eval()
processor = AutoProcessor.from_pretrained(repo)
question = "Compare these two images and give the coordinates of their difference."
prompt = question + "\nFirst output <think>...</think>, then <answer>(x1,y1),(x2,y2)</answer>."
# передайте prompt и изображения в processor.apply_chat_template и model.generate
ПодробнееНа MedSG-Bench: 77.54 mIoU и 84.87 Acc@0.5 у этого 9B-чекпоинта.
Обучение: dCoT-SFT на 182 573 rationale, затем GRPO с наградой за IoU и формат.
Это исследовательский артефакт, не для автономной диагностики; rationale обусловлены размеченным box и могут быть post-hoc.
Оценка ограничена MedSG-Bench; 4B-версия даёт 75.97 mIoU / 83.67 Acc@0.5.
Discuss Telegram 0 35

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 18 GB · ready files20 GB
  • FP8 / INT8recommendedweights 8.8 GB10 GB
  • 4-bitweights 4.9 GB5.9 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.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2080 Ti · BF16 — from RTX 4090

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

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.

    # adidukrembzuai/MedChain-R1-9B
    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 adidukrembzuai/MedChain-R1-9B --local-dir ~/models/MedChain-R1-9B
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to MedChain-R1-9B
  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

  1. Qwen3.5-9B-Basefamily root
  2. Qwen3.5-9B
  3. MedChain-R1-9B

New posts about versions of Qwen3.5-9B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 6All on Hugging Face
Merges 1All on Hugging Face

About the model

adidukrembzuai/MedChain-R1-9B
Author
adidukrembzuai
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

350

Ask AI about this model

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

Comments