def init_client(args, device, comm, process_id, size, model, train_data_num, local_data_num, train_data_local):
# trainer
client_ID = process_id - 1
trainer = FedAVGTrainer(client_ID, train_data_local, local_data_num, train_data_num, device, model, args)
client_manager = FedAVGClientManager(args, comm, process_id, size, trainer)
client_manager.run()
def init_client(args, device, comm, process_id, size, model, train_data_num, train_data_local_num_dict, train_data_local_dict):
# trainer
client_index = process_id - 1
trainer = FedAVGTrainer(client_index, train_data_local_dict, train_data_local_num_dict, train_data_num, device, model, args)
client_manager = FedAVGClientManager(args, trainer, comm, process_id, size)
client_manager.run()
4)
# load data
dataset = load_data(args, args.dataset)
[
train_data_num, test_data_num, train_data_global, test_data_global,
train_data_local_num_dict, train_data_local_dict, test_data_local_dict,
class_num
] = dataset
# create model.
# Note if the model is DNN (e.g., ResNet), the training will be very slow.
# In this case, please use our FedML distributed version (./fedml_experiments/distributed_fedavg)
model = create_model(args, model_name=args.model, output_dim=dataset[7])
client_index = client_ID - 1
trainer = FedAVGTrainer(client_index, train_data_local_dict,
train_data_local_num_dict, train_data_num, device,
model, args)
size = args.client_num_per_round + 1
client_manager = FedAVGClientManager(args,
trainer,
rank=client_ID,
size=size,
backend="MQTT")
client_manager.run()
client_manager.start_training()
time.sleep(100000)