FedOps Custom Silo Setting

This guide provides step-by-step instructions on executing FedOps Silo, a Federated Learning Lifecycle Management Operations framework.

Baseline

- Baseline
   - client_main.py
   - client_manager_main.py
   - server_main.py
   - models.py
   - data_preparation.py
   - requirementst.txt (server requirement)
   - conf
      -  config.yaml

Steps

1. Start by fork the FedOps-Silo.

2. Customize the FedOps example code.

   • In accordance with the baseline form, write the code with reference to the example and store it in the your own git remote storage.

   • Data & Model:

     • Prepare your data in data_preparation.py and define your model in models.py.
     • Enter your data and model in conf/config.yaml

   • Client:

     • Configure settings in conf/config.yaml for task ID, data, and WandB information.
     • Implement data_preparation.py for data handling.
     • Build models.py for local model specifications.
     • Register data and model, and run the client using client_main.py.

   • Server:

     • Configure settings in conf/config.yaml for FL/Aggregation hyperparameters and data information.
     • Implement models.py to initialize the global model.
     • Register data (for evaluating the global model) and initialize the global model in server_main.py.
       
       

3. Create FL task on FedOps web interface.

   • Use the FedOps web interface to create your FL task.
   • Specify the Git repository address for your FL server code.
   • Refer FedOps Silo Guide
     
     

4. Run the clients.

   • Run client_main.py and client_manager_main.py
   • Or choose either Docker or shell(localhost) to run the clients.
     
     

5. Initiate the FL task.

   • Select the desired clients on the FedOps web interface and initiate the FL task by clicking the "FL start" button.
     
     

6. Monitor and manage the FL task

   • Monitor the performance of local and global models, and manage/download the global model as the FL task administrator through the FedOps web interface.
     
     

7. Monitor data and local model performance

   • Monitor the health of data and local model performance as the device administrator through the designated WandB.
     
     

Support

For any questions or issues, please contact the FedOps support team at gyom1204@gachon.ac.kr