CLI reference

All commands use:

python -m client.submit_job --server http://<server-ip>:8000 <command>

Replace <server-ip> with your coordinator’s Tailscale IP (e.g. 100.64.0.1).

Commands

start

Start a new FedAvg training run. The coordinator shards CIFAR-10 across all registered workers and begins round 1.

python -m client.submit_job --server http://<server-ip>:8000 start

status

Check the current training status — which round is active, how many workers have submitted updates, and whether training is complete.

python -m client.submit_job --server http://<server-ip>:8000 status

monitor

Watch round-by-round progress in real-time. Prints updates as each round completes with loss and accuracy metrics.

python -m client.submit_job --server http://<server-ip>:8000 monitor

results

Fetch final training results after all rounds are complete. Returns aggregated metrics (loss, accuracy) across all rounds.

python -m client.submit_job --server http://<server-ip>:8000 results

workers

List all registered workers, their hardware info, GPU type, and current status (idle, training, disconnected).

python -m client.submit_job --server http://<server-ip>:8000 workers

Common flags

Examples

Full workflow:

# Start training
python -m client.submit_job --server http://100.64.0.1:8000 start

# Monitor progress
python -m client.submit_job --server http://100.64.0.1:8000 monitor

# Get results when done
python -m client.submit_job --server http://100.64.0.1:8000 results