HFL-DRL

For running HFL-DRL with GQUIC-256, model CNN on 1 server, 2 edges and 4 clients:

1. Install dependencies

• pip install -r requirements.txt

2. Start training

• Open 7 terminals
• First terminal: python server.py
• Second terminal: python head.py --edge_port 40001
• Third terminal: python head.py --edge_port 40002
• Terminal number 4,5,6,7: python distributed_client.py

You can config the training by modifying the parameters in options.py

For example:

• num_communication = 10
• num_edge_aggregation = 10
• num_local_update = 6