-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
187 lines (171 loc) · 9.07 KB
/
main.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
#!/usr/bin/env python
import argparse
from FLAlgorithms.servers.serveravg import FedAvg
from FLAlgorithms.servers.serverlocal import Fedlocal
from FLAlgorithms.servers.serverFedProx import FedProx
from FLAlgorithms.servers.serverFedCluster import FedCluster
from FLAlgorithms.servers.serverIFCA import FedIFCA
from FLAlgorithms.servers.serverFedKLtopk import FedKLtopk
from FLAlgorithms.servers.serverFedKL import FedKL
from FLAlgorithms.servers.serverFedKLEM import FedKLEM
from FLAlgorithms.servers.serverFedSEM import FedSEM
from FLAlgorithms.servers.serverperavg import PerAvg
from FLAlgorithms.servers.serverpFedMe import pFedMe
from FLAlgorithms.servers.serverFedJSEM import FedJSEM
from FLAlgorithms.servers.serverFedfuzzyC import FedfuzzyC
from FLAlgorithms.servers.serverFedKLEM_PerFedavg import FedKLEM_PerFedAvg
from FLAlgorithms.servers.serverFedKLEM_intra_cluster import FedKLEM_intra_cluster
from FLAlgorithms.servers.serverFedKLEM_intra_cluster_PerFedavg import FedKLEM_intra_cluster_PerFedavg
from FLAlgorithms.servers.serverFedKLEM_PerFedavg_ratio import FedKLEM_PerFedAvg_ratio
from FLAlgorithms.servers.serverFedKLEM_PerFedavg_ratio_e import FedKLEM_PerFedAvg_ratio_e
from utils.model_utils import *
from utils.plot_utils import *
from utils.data_utils import *
from utils.constants import *
import torch
from multiprocessing import Pool
def init_data(args, mode):
#除了FEMNIST和Shakespeare都会返回类似 ('data/Cifar10/u80-alpha0.5-ratio1.0', '80')
data_dir_u = get_data_dir(args)
if len(data_dir_u) == 2: data_dir = data_dir_u[0]
else: data_dir = data_dir_u
root_path=os.path.join(data_dir , mode)
print(root_path)
#root_path=os.path.join(data_dir , "participate")
#root_path=os.path.join(data_dir , "unseen")
u = 0 #用在label_swapped的时候，是participate的user数目
if 'label_swapped' in args.dataset.lower():
split = 'label_swapped'
u = int(data_dir_u[1])
else:
split = '' #null表示其他
if 'femnist' in args.dataset.lower(): #Cifar100-alpha0.5-ratio1.0-u100
experiment = 'femnist'
elif 'emnist' in args.dataset.lower():
experiment = 'emnist'
print('yea')
elif 'mnist' in args.dataset.lower():
experiment = 'mnist'
print('wuwuu')
elif 'cifar100' in args.dataset.lower():
print("cifar100")
experiment = 'cifar100'
elif 'cifar10' in args.dataset.lower():
print("cifar10")
experiment = 'cifar10'
elif 'shakespeare' in args.dataset.lower():
experiment = 'shakespeare'
print("===> Building data iterators..")
train_iterators, val_iterators, test_iterators , public_iterator, len_trains, len_tests, len_public=\
get_loaders(
args,
split,
u,
type_=LOADER_TYPE[experiment], #数据类型
root_path=root_path, # 参与训练的数据目录
data_dir=data_dir, # 总数据目录
batch_size=args.batch_size,
is_validation=False
)
return train_iterators, val_iterators, test_iterators, public_iterator, len_trains, len_tests, len_public
def create_server_n_user(args, i):
args.device = torch.device("cuda:{}".format(args.gpu) if torch.cuda.is_available() and args.gpu != -1 else "cpu")
#args.device = server_device = "cpu"
data_participate = init_data(args, mode = "participate")
data_unseen = init_data(args, mode = "unseen")
model = create_model_new(args)
if ('FedKLEM_PerFedAvg_ratio_e' in args.algorithm):
server = FedKLEM_PerFedAvg_ratio_e(args, model, data_participate, data_unseen, i)
elif ('FedKLEM_PerFedAvg_ratio' in args.algorithm):
server = FedKLEM_PerFedAvg_ratio(args, model, data_participate, data_unseen, i)
elif('FedKLEM_intra_cluster_PerFedavg' in args.algorithm):
server = FedKLEM_intra_cluster_PerFedavg(args, model, data_participate, data_unseen, i)
elif('FedKLEM_IntraC' in args.algorithm):
server = FedKLEM_intra_cluster(args, model, data_participate, data_unseen, i)
elif('FedKLEM_PerFedAvg' in args.algorithm):
server = FedKLEM_PerFedAvg(args, model, data_participate, data_unseen, i)
elif ('FedAvg' in args.algorithm):
server=FedAvg(args, model, data_participate, data_unseen, i)
elif ('FedProx' in args.algorithm):
server = FedProx(args, model, data_participate, data_unseen, i)
elif ('FedKLEM' in args.algorithm):
server = FedKLEM(args, model, data_participate, data_unseen, i)
elif ('FedCluster' in args.algorithm):
server = FedCluster(args, model, data_participate, data_unseen, i)
elif ('FedIFCA' in args.algorithm):
server = FedIFCA(args, model, data_participate, data_unseen, i)
elif ('FedKLtopk' in args.algorithm):
server = FedKLtopk(args, model, data_participate, data_unseen, i)
elif ('FedKL' in args.algorithm):
server = FedKL(args, model, data_participate, data_unseen, i)
elif('Fedlocal' in args.algorithm):
server = Fedlocal(args, model, data_participate, data_unseen, i)
elif('FedSEM' in args.algorithm):
server = FedSEM(args, model, data_participate, data_unseen, i)
elif('PerFedavg' in args.algorithm):
server = PerAvg(args, model, data_participate, data_unseen, i)
elif('pFedME' in args.algorithm):
server = pFedMe(args, model, data_participate, data_unseen, i)
elif('FedJSEM' in args.algorithm):
server = FedJSEM(args, model, data_participate, data_unseen, i)
elif('FedfuzzyC' in args.algorithm):
server = FedfuzzyC(args, model, data_participate, data_unseen, i)
else:
print("Algorithm {} has not been implemented.".format(args.algorithm))
exit()
return server
def run_job(args, i):
torch.manual_seed(i)
print("\n\n [ Start training iteration {} ] \n\n".format(i))
# Generate model
server = create_server_n_user(args, i)
if args.train:
server.train(args)
#server.test()
def main(args):
for i in range(args.times):
run_job(args, i)
print("Finished training.")
if __name__ == "__main__":
#torch.backends.cudnn.benchmark = False
#torch.backends.cudnn.enabled = False
parser = argparse.ArgumentParser()
parser.add_argument("--dataset", type=str, default="Mnist")
parser.add_argument("--model", type=str, default="cnn")
parser.add_argument("--train", type=int, default=1, choices=[0,1])
parser.add_argument("--algorithm", type=str, default="pFedMe")
parser.add_argument("--batch_size", type=int, default=32)
parser.add_argument("--gen_batch_size", type=int, default=32, help='number of samples from generator')
parser.add_argument("--learning_rate", type=float, default=0.01, help="Local learning rate")
parser.add_argument("--personal_learning_rate", type=float, default=0.01, help="Personalized learning rate to caculate theta aproximately using K steps")
parser.add_argument("--ensemble_lr", type=float, default=1e-4, help="Ensemble learning rate.")
parser.add_argument("--beta", type=float, default=1.0, help="Average moving parameter for pFedMe, or Second learning rate of Per-FedAvg")
parser.add_argument("--lamda", type=int, default=1, help="Regularization term")
parser.add_argument("--mix_lambda", type=float, default=0.1, help="Mix lambda for FedMXI baseline")
parser.add_argument("--embedding", type=int, default=0, help="Use embedding layer in generator network")
parser.add_argument("--num_glob_iters", type=int, default=200)
parser.add_argument("--local_epochs", type=int, default=0)
parser.add_argument("--num_users", type=int, default=20, help="Number of Users per round")
parser.add_argument("--K", type=int, default=1, help="Computation steps")
parser.add_argument("--times", type=int, default=3, help="running time")
parser.add_argument("--p", type=int, default=3, help="CLuster numbers")
parser.add_argument("--load_model", type=bool, default=False, help="Load model or train from start")
parser.add_argument("--result_path", type=str, default="results", help="directory path to save results")
parser.add_argument("--E", type=int, default="0", help="fit_epochs or fit_batchs")
parser.add_argument("--gpu", type=int, default=0, help="Which GPU to run the experiments, -1 mean CPU, 0,1,2 for GPU")
parser.add_argument("--test_unseen", type=bool, default=False)
args = parser.parse_args()
print("=" * 80)
print("Summary of training process:")
print("Algorithm: {}".format(args.algorithm))
print("Batch size: {}".format(args.batch_size))
print("Learing rate : {}".format(args.learning_rate))
print("Ensemble learing rate : {}".format(args.ensemble_lr))
print("Average Moving : {}".format(args.beta))
print("Subset of users : {}".format(args.num_users))
print("Number of global rounds : {}".format(args.num_glob_iters))
print("Number of local rounds : {}".format(args.local_epochs))
print("Dataset : {}".format(args.dataset))
print("Local Model : {}".format(args.model))
print("=" * 80)
main(args)