FedLab-benchmarks
FedLab-benchmarks copied to clipboard
Bug Fixed and Adjustment on fedavg_v1.2.0
- args.cuda is only be refered on model, but not be refered in params of SubsetSerialTrainer, which has a default value "False". This caused that pytorch wont train on gpus even args.cuda is setted to "True".
- data_indices can be replaced by Partitioner in fedlab.