def init_server(args, device, comm, rank, size, model, train_data_num, train_data_global, test_data_global):
# aggregator
client_num = size - 1
aggregator = FedAVGAggregator(train_data_global, test_data_global, train_data_num, client_num, device, model, args)
# start the distributed training
server_manager = FedAVGServerManager(args, comm, rank, size, aggregator)
server_manager.run()
def init_server(args, device, comm, rank, size, model, train_data_num, train_data_global, test_data_global,
train_data_local_dict, test_data_local_dict, train_data_local_num_dict):
# aggregator
worker_num = size - 1
aggregator = FedAVGAggregator(train_data_global, test_data_global, train_data_num,
train_data_local_dict, test_data_local_dict, train_data_local_num_dict, worker_num, device, model, args)
# start the distributed training
server_manager = FedAVGServerManager(args, aggregator, comm, rank, size)
server_manager.send_init_msg()
server_manager.run()
config=args
)
# Set the random seed. The np.random seed determines the dataset partition.
# The torch_manual_seed determines the initial weight.
# We fix these two, so that we can reproduce the result.
np.random.seed(0)
torch.manual_seed(10)
# GPU 0
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# 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])
aggregator = FedAVGAggregator(train_data_global, test_data_global, train_data_num,
train_data_local_dict, test_data_local_dict, train_data_local_num_dict,
args.client_num_per_round, device, model, args)
size = args.client_num_per_round + 1
server_manager = FedAVGServerManager(args, aggregator, rank=0, size=size, backend="MQTT")
server_manager.run()
app.run(host='127.0.0.1', port=5000, debug=False)