Private AIM - Work Package 1

Welcome to the repository for Private AIM - Work Package 1 (WP1). This project focuses on developing federated learning methods for applications in AI-driven medicine. It is funded by the BMBF under grant number 01ZZ2316C.

---

Purpose

This repository serves as a testbed for implementing and evaluating workflows for federated medical analyses. Our goal is to address specific challenges outlined in the milestones and tasks of Private AIM WP1.

Once developed and validated, these methods will be integrated into the FLAME platform, making them accessible to all participating institutions.

---

Getting Started

To set up your environment and begin using the repository, follow these steps:

1. Install Conda: Follow the official guide to install Conda on your system.

2. Create a Conda Environment:
   a. Install via environment file:

   conda env create -f environment.yml
   

   b. Alternative installation:

   Install environment manually. This is necessary if the first version does not provide you with a GPU compatible pytorch version and you need to adapt it to your specific system requirements.

   Run the following command to create an environment with Python 3.10:

   conda create -n aim-wp1 python=3.10
   

   Install Dependencies:

   • Install PyTorch by following the instructions on the PyTorch website based on your system configuration.

   • Install additional packages:

     conda install -y lightning torchmetrics numpy pandas acsconv timm opacus hydra-core scipy matplotlib -c conda-forge && pip install flwr[simulation] wandb torchio faker medmnist plotly torchsummary
     

3. Configure the Project:
   Update the configuration file at src/config/config.yaml as needed. Refer to the configuration README for details.

4. Start Training:
   Launch the training script with:

   python src/train.py
   

---

If you encounter issues or have suggestions, feel free to open an issue or contact the project team. Happy coding! 🚀

---