some questions about initialization and retrain

[koilgg]

Hi, thank you for sharing your outstanding work! I have some questions about the settings in the paper, could you please support some more details about the experiment settings?

1. In the code, the J clients do not share the same initialization, and they are retrained before the first global round. Is there any difference of the settings between the retrain process at first and the later local-retrain?(like fedavg local retrain)
2. What is the batch-size of the dataset corresponding to the results in this paper?
3. Does the experiment in the paper use the following command?
   python main.py --model=moderate-cnn \ --dataset=cifar10 \ --lr=0.01 \ --retrain_lr=0.01 \ --batch-size=64 \ --epochs=150 \ --retrain_epochs=150 \ --n_nets=16 \ --partition=hetero-dir \ --comm_type=fedma \ --comm_round=10 \ --retrain=True \ --rematching=True

Thanks very much