Skip to content
Other

AutoIHC

Automated Analysis of Breast Cancer Histology Slides

2 min readupdated

A set of pretrained weights for the AutoIHC framework from the OncoAI-ULBHUB research group, accompanying their 2026 paper in Technology in Cancer Research & Treatment. The model handles automated quantitative staining assessment on whole-slide immunohistochemistry (IHC) images for breast cancer. It separates tumor from stroma, segments cell nuclei, and calculates the H-score (a measure of staining intensity and the proportion of stained cells). Bioinformaticians and digital pathology researchers can use this tool to reproduce the paper's results or fine-tune it on their own data.

Key features

  • Modular architecture: weights are split across three analysis stages — TSM (tumor and stroma segmentation), NSM (nuclei segmentation), and HEM (H-score evaluation).
  • 5-fold cross-validation: five top checkpoints (from fold_0 to fold_4) are available for each method, making it easy to combine them into ensembles.
  • PyTorch support: models ship as standard .pth weight files and load easily via huggingface_hub.

What you can do

  • Reproduce the study: run the pipeline from the official GitHub repository on your own IHC slides to evaluate H-scores.
  • Build a model ensemble: download all five folds of a given method via snapshot_download for more stable predictions.
  • Fine-tune for your lab: use the checkpoints as a starting point to adapt the models to different scanners, staining protocols, or markers.

Example

from huggingface_hub import hf_hub_download
import torch

path = hf_hub_download(
    repo_id="OncoAI-ULBHUB/AutoIHC",
    filename="H-score model weights///best_model_fold_0.pth",
)
state_dict = torch.load(path, map_location="cpu")
Read moreDetails
Trained on a private in-house dataset of breast cancer IHC images (the dataset itself is not public).
Strictly for research purposes — this is not a medical device and is not suitable for clinical diagnosis.
Accuracy may drop on slides from different scanners or with other staining protocols and markers; the authors recommend validation and fine-tuning.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.5 GB · ready files1.1 GB
  • FP8 / INT8recommendedweights 0.3 GB0.8 GB
  • 4-bitweights 0.1 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 · BF16 — 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.

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

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

About the model

OncoAI-ULBHUB/AutoIHC
Author
OncoAI-ULBHUB
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments