Skip to content
Other

MAMAGuard-Net

25 nnU-Net v2 checkpoints for MRI tumor segmentation

2 min readupdated

A research weight collection for breast tumor segmentation on dynamic contrast-enhanced MRI (DCE-MRI). Released by the paper's authors, it includes 25 final checkpoints built on the nnU-Net v2 ResEncL architecture. The models were trained using 5-fold cross-validation (splitting data into 5 parts for evaluation) across five different data augmentation strategies, including phase-consistent intensity randomization. Given a three-channel input (pre-contrast, early, and late post-contrast phases), the model outputs a tumor segmentation mask. This set is useful for medical imaging researchers looking to reproduce experiments and benchmark augmentation methods.

Key features

  • nnU-Net v2 ResEncL architecture: A proven medical image segmentation framework featuring a residual encoder.
  • Five augmentation strategies: Ranging from no augmentation (A0) and standard nnU-Net transforms (A1) to affine transformations (A3) and phase-order-constrained augmentation (A4).
  • Full reproducibility: 5 folds per configuration (seed 20261031), plans.json files, dataset descriptions, and integrity verification via SHA256SUMS.
  • Three-phase MRI input: Models expect scans in a strict channel order — pre-contrast, early post-contrast, and late post-contrast.

What you can do

  • Benchmark augmentation effects: Evaluate how different intensity and geometric distortions impact tumor segmentation accuracy on multi-center MAMA-MIA data.
  • Run inference on a holdout set: Use the nnUNetv2_predict_from_modelfolder command with the corresponding fold and configuration metadata to generate segmentation masks.
  • Use as a baseline: Take the pre-trained weights and training plans as a starting point for your own dynamic MRI analysis research.

Example

hf download harshitrk07/MAMAGuard-Net \
  --revision 4e8d231c3ad821be90be8b5c30b329a658c26d59 \
  --include 'A3_shared_affine/**' 'SHA256SUMS' \
  --local-dir mamaguard-weights

cd mamaguard-weights
awk '$2 ~ /^A3_shared_affine\// {print}' SHA256SUMS | sha256sum -c -
Read moreDetails
The total size of all 25 checkpoints is roughly 26.35 GiB (configurations can be downloaded selectively).
Running inference and registering custom trainers requires access to the private v0.1.0-paper code repository (by invitation).
The MAMA-MIA dataset is distributed separately under its own access and citation terms.
These models are intended solely for internal cross-validation research: clinical use and generalization to the external Yunnan cohort have not been validated, and no separate license is specified for the weights.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 1.1 GB · ready files1.7 GB
  • FP8 / INT8recommendedweights 0.5 GB1.1 GB
  • 4-bitweights 0.3 GB0.8 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.

    # harshitrk07/MAMAGuard-Net
    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 harshitrk07/MAMAGuard-Net --local-dir ~/models/MAMAGuard-Net
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to MAMAGuard-Net
  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 MAMAGuard-Net (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

harshitrk07/MAMAGuard-Net
Author
harshitrk07
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