Skip to content
Other

lmd-vitl-p8-affinity-lvd

3D-сегментация нейронов по аффинностям

2 min read

Модель от ai-hhmi предсказывает воксельные аффинности для сегментации нейронов в 3D-данных электронной и expansion-микроскопии с шагом 8 нм. Это файнтюн энкодера DINOv3 ViT-L с 8³-воксельными патчами (303 млн параметров) и UNETR-декодером на 1,85 млн параметров. Чекпоинт после 100 тыс. шагов обучения на 41 размеченном объёме.

Ключевые особенности

  • Энкодер стартовал с lmd-vitl-p8-simmim-lvd: DINOv3 LVD-1689M, затем SimMIM на LMD.
  • Выход — 6 каналов аффинностей со смещениями (1,0,0), (0,1,0), (0,0,1), (10,0,0), (0,10,0), (0,0,10).
  • Вход — один канал (x, y, z), изотропные воксели 8 нм, интенсивности [0, 1], каждая сторона кратна 8.
  • Обучение end-to-end с UNETR-декодером: 100 тыс. шагов, 41 размеченный объём, BCE по 6 логитам.
  • Веса fp32, 1.22 ГБ; энкодер лежит под encoder. и его можно переиспользовать отдельно.
  • Лицензия DINOv3: модель — производная от весов Meta DINOv3, публикации должны ссылаться на DINOv3.

Пример

from pathlib import Path
from huggingface_hub import snapshot_download
from safetensors.torch import load_file
root = Path(snapshot_download("ai-hhmi/lmd-vitl-p8-affinity-lvd"))
config = json.loads((root / "config.json").read_text())
net = AffinityNet(config)
net.load_state_dict(load_file(root / "model.safetensors"))
net = net.to("cuda").eval()
ПодробнееPQ: 0.300 на crop512, 0.277 на whole volumes.
Веса fp32, 1.22 ГБ; энкодер 303 млн + декодер 1,85 млн параметров.
Нужны PyTorch, safetensors, huggingface_hub и mia-train (коммит 0fe3872); VRAM не указана.
Вход: 8 нм, [0,1], стороны кратны 8; другие размеры не тестировали.
Лицензия DINOv3 — производная от весов Meta DINOv3.
Discuss Telegram 0 22

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 1.1 GB · ready files1.7 GB
  • BF16weights 0.6 GB1.1 GB
  • FP8 / INT8recommendedweights 0.3 GB0.8 GB
  • 4-bitweights 0.2 GB0.7 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 2060 · F32 — from RTX 2060

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.

    # ai-hhmi/lmd-vitl-p8-affinity-lvd
    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 ai-hhmi/lmd-vitl-p8-affinity-lvd --local-dir ~/models/lmd-vitl-p8-affinity-lvd
    # 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

  1. dinov3-vit7b16-pretrain-lvd1689mfamily root
  2. dinov3-vitl16-pretrain-lvd1689m
  3. lmd-vitl-p8-simmim-lvd
  4. lmd-vitl-p8-affinity-lvd

New posts about versions of dinov3-vit7b16-pretrain-lvd1689m (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

ai-hhmi/lmd-vitl-p8-affinity-lvd
Author
ai-hhmi
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

220

Ask AI about this model

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

Comments