Based on the flower pytorch quickstart example.
Additions:
- wandb integration
- experiment counting
- SplitLRFedAvg strategy that splits the clients in two groups, one with a constant lr and one with an adaptive one
> pip install -r requirements.txt
> mkdir .exp-count
> ./run-server <num-rounds>
> ./run-clients <num-clients>
> wandb init
> (paste your API key)