This repository organizes your Federated Learning experiments on the NIH Chest X-ray dataset. We preserve your original notebook while providing a reproducible project layout and a script export.
- Original notebook:
notebooks/NIH_Chest_Xray_Federated_Learning.ipynb
- Exported notebook code:
src/notebook_export.py
- Minimal launcher:
src/train.py
- Default config:
configs/default.yaml
Note:
src/notebook_export.py
concatenates notebook cells in order. For long-term maintenance, consider refactoring into modules such asdatasets.py
,models.py
, andtrain.py
.
git clone https://github.com/<YOUR-USER-OR-ORG>/chestxray-federated-learning.git
cd chestxray-federated-learning
bash scripts/run_local.sh
Or manually:
python3 -m venv .venv && source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt
python src/train.py --workdir .
- The experiments use the NIH Chest X-ray (ChestX-ray14) dataset.
- Download/preprocessing follows the logic embedded in the notebook. (Some cells may rely on Google Colab; adjust or skip those when running locally.)
- If you reorganize the data into an
ImageFolder
layout, update paths inconfigs/default.yaml
accordingly.
requirements.txt
was generated based on actual imports detected in the notebook.
- Please install PyTorch / CUDA versions appropriate for your machine (e.g., CUDA 12.x).
If this code or its results are useful for your research, please cite this repository via CITATION.cff
.
MIT License (see LICENSE
)