evaluation_input.txt

setting = FedEM ## 0 Method of training ##
## FedEM, FedAvg, or local##

experiment = cifar10 ## 1 Dataset name ##
decentralized = False ## 2 Decentralized ##
sampling_rate = 1.0 ## 3 Sampling rate ##
input_dimension = None ## 4 Input dimension ##
output_dimension = None ## 5 Output dimension ##
n_rounds = 10 ## 6 Number of rounds training takes place ##
bz = 128 ## 7 bz ##
local_steps = 1 ## 8 Local steps ##
lr_lambda = 0 ## 9 Learning rate lambda ##
lr = 0.03 ## 10 Learning rate ##
lr_scheduler = multi_step ## 11 Learning rate scheduler ##
log_freq = 10 ## 12 Log Frequency ##
device = cuda ## 13 Device ##
optimizer = sgd ## 14 Device ##
mu = 0 ## 15 mu ##
communication_probability = 0.1 ## 16 Communication probability ##
q = 1 ## 17 q ##
locally_tune_clients = False ## 18 Locally tune clients ##
seed = 1234 ## 19 Seed ##
verbose = 1 ## 20 Verbose ##
weight_save_path = weights/cifar10/pFedDef/fedem ## 21 Weight save path##
validation = False ## 22 Validation ##