Skip to content
Other

BioHCL-SSSM

a lightweight hybrid architecture for pneumonia detection

2 min readupdated

BioHCL-SSSM is a hybrid computer vision model by unknownBrain that combines convolutional networks with state-space models (Hierarchical Convolutional State-Space Hybrid Architecture) to detect pediatric pneumonia from chest X-rays. The repository bundles a complete workflow: 10 trained cross-validation checkpoints, modular PyTorch code, evaluation scripts, heatmap generators, and benchmark tables comparing it against other architectures. At just 14.01M parameters, it outperforms much heavier transformers like ViT-Base (86.57M parameters). This project is useful for medical AI researchers, developers building X-ray analysis systems, and anyone exploring compact alternatives to Vision Transformers.

Key features

  • Compact and fast: The model packs 14.01M parameters and requires 8.56 GFLOPs, yet beats ViT-Base on the official test set (96.47% vs. 92.47% accuracy) with 83.8% fewer parameters, while also outperforming Swin-T at half the size.
  • High cross-validation accuracy: Across 10 folds, the model achieved a mean accuracy of 99.37% (±0.33%), an F1-score of 0.9957, and a ROC-AUC of 0.9929, making only 33 errors across 5,232 images.
  • Explainability (XAI) tools: The project includes the build_xai.py script to generate attention maps and explain model predictions using Attention, GradCAM, LIME, and SHAP.
  • Full reproducibility: Alongside weights for all 10 folds (54 MB each) and the best checkpoint best_student_cv.pt, the repo provides raw logs from 64 ablation runs, CSV metric tables, and ready-made ROC, PR, and confusion matrix plots.

What you can do

  • Use pretrained weights for inference and fine-tuning: Load the best_student_cv.pt checkpoint (99.81% validation accuracy, 100% recall) or any of the 10 folds for chest X-ray classification experiments.
  • Visualize pathological regions: Run build_xai.py to generate GradCAM, LIME, or SHAP heatmaps and inspect which lung areas the model relies on for its diagnosis.
  • Reproduce ablation experiments: Use the standalone run_colab_remaining_ablations.py script to run the remaining sweeps with different loss weights (lambda_bio) and benchmark against the baseline MHSA attention mechanism.

Example

git clone [https://huggingface.co/unknownBrain/BioHCL-SSSM-Complete-Workspace](https://huggingface.co/unknownBrain/BioHCL-SSSM-Complete-Workspace)
cd BioHCL-SSSM-Complete-Workspace
pip install torch torchvision timm albumentations huggingface_hub scikit-learn pandas pillow
python run_colab_remaining_ablations.py
Read moreDetails
Official test set (624 images from the Guangzhou Women and Children’s Medical Center dataset, Kermany et al., 2018): 96.47% Accuracy, 97.22% F1, 99.37% ROC-AUC.
Best cross-validation fold (Fold 2): 99.81% Accuracy, 0.9987 F1, 100% Recall, 99.26% Specificity.
Hardware and runtime: Running the remaining experiments script takes about 10 minutes on an RTX 3080 Ti-class GPU or in Google Colab.
License: MIT.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 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.

    # unknownBrain/BioHCL-SSSM-Complete-Workspace
    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 unknownBrain/BioHCL-SSSM-Complete-Workspace --local-dir ~/models/BioHCL-SSSM-Complete-Workspace
    # 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

New posts about versions of BioHCL-SSSM-Complete-Workspace (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

unknownBrain/BioHCL-SSSM-Complete-Workspace
Author
unknownBrain
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