-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathnew_test_fash_group.py
144 lines (127 loc) · 7.63 KB
/
new_test_fash_group.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
import os
from fedbase.baselines import *
from fedbase.model.model import *
from fedbase.nodes.node import node
from fedbase.utils.tools import unpack_args
from fedbase.utils.data_loader import data_process
import torch
import torch.optim as optim
import torch.nn as nn
from functools import partial
import numpy as np
import multiprocessing as mp
import time
import torchvision.models as models
from joblib import Parallel, delayed
import argparse
from fedbase.utils.get_digit5 import generate_Digit5
from fedbase.utils.get_amazon_review import generate_AmazonReview
from fedbase.utils.get_domainnet import generate_DomainNet
os.chdir(os.path.dirname(os.path.abspath(__file__))) # set the current path as the working directory
global_rounds = 100
# num_nodes = 10
local_steps = 10
batch_size = 32
optimizer = partial(optim.SGD,lr=0.005, momentum=0.9)
# optimizer = partial(optim.SGD,lr=0.001)
# device = torch.device('cuda:2')
# device = torch.device('cuda') # Use GPU if available
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# print(device)
K = 2
H = 2
def main(seeds, dataset_splited, model, model_name, K=None,n_assign=None,cost_method='weighted',warm_up=False,warm_up_rounds=2):
np.random.seed(seeds)
# dataset_splited, model = dataset_splited_model[0], dataset_splited_model[1]
if model_name == 'FedAvg':
fedavg.run(dataset_splited, batch_size, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, device = device)
elif model_name == 'BayesFedAvg':
fedavg_bayes.run(dataset_splited,batch_size, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, device = device)
elif model_name == 'BayesPFedAvg':
fedavg_bayes.run(dataset_splited,batch_size, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, device = device, personalized=True)
elif model_name == 'Wecfl':
GNN.run(dataset_splited,batch_size,K, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, assign_method='wecfl',bayes=False, device = device)
elif model_name == 'Fesem':
fesem.run(dataset_splited,batch_size, K, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, reg_lam=0.001, device = device)
elif model_name == 'GNN':
GNN.run(dataset_splited,batch_size,K, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, assign_method='ifca',bayes=True, device = device,warm_up=warm_up)
elif model_name == 'JPDA':
jpda.run(dataset_splited, batch_size, K, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, bayes=True, num_assign=n_assign,device=device, cost_method=cost_method,warm_up=warm_up)
elif model_name == 'MHT':
mht.run(dataset_splited, batch_size, K, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, bayes=True, num_assign=n_assign,hypothesis=n_assign, device=device, cost_method=cost_method,warm_up=warm_up)
elif model_name == 'FedAMP':
fedamp.run(dataset_splited, batch_size, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, device = device)
elif model_name == 'FedSoft':
fedsoft.run(dataset_splited, batch_size, K, num_nodes, model, nn.CrossEntropyLoss, optimizer, global_rounds, local_steps, selection_size=10,num_classes=10,reg_lam=0.1,device = device)
if __name__ == '__main__':
dataset = 'fashion_mnist' #'amazon' #'digit5'
seeds = 1989 # 0,2020
if dataset == 'mnist':
model = CNNMnist
elif dataset == 'femnist':
model = CNNFemnist
elif dataset == 'fashion_mnist':
model = CNNFashion_Mnist
elif dataset == 'cifar10':
model = CNNCifar
elif dataset == 'digit5':
model = Digit5CNN # CNNDigit5
elif dataset == 'amazon':
model = AmazonMLP
elif dataset == 'domainnet':
model = AlexNet
client_group = 10
if dataset == 'digit5':
# for Digit5
domains=['mnistm', 'mnist', 'usps', 'svhn','syn'] #,
K = len(domains)
num_nodes = K*client_group #40
dataset_splited_list = [generate_Digit5(domains=domains, client_group=client_group, method='iid', alpha=10)]
elif dataset == 'amazon':
K = 4
num_nodes = K*client_group
dataset_splited_list = [generate_AmazonReview(client_group=client_group, method='iid', alpha=10)]
elif dataset == 'domainnet':
K = 6
num_nodes = K*client_group
dataset_splited_list = [generate_DomainNet(client_group=client_group, method='iid', alpha=10)]
else:
K = 4
num_nodes = 40 #40
noise = None #'rotation'
dataset_splited_list = [
# data_process(dataset).split_dataset_groupwise(K, 0.1, 'dirichlet', int(num_nodes/K), 5, 'dirichlet'),\
# data_process(dataset).split_dataset_groupwise(K, 3, 'class', int(num_nodes/K), 2, 'class'),\
# data_process(dataset).split_dataset(num_nodes, 3, 'class'),\
# data_process(dataset).split_dataset(num_nodes, 0.1, 'dirichlet'),\
# data_process(dataset).split_dataset_groupwise(K, 10, 'dirichlet', int(num_nodes/K), 0.1, 'dirichlet', noise),\
# data_process(dataset).split_dataset_groupwise(K, 10, 'dirichlet', int(num_nodes/K), 10, 'dirichlet', noise),\
# data_process(dataset).split_dataset_groupwise(K, 5, 'class', int(num_nodes/K), 2, 'class', noise) ,\
data_process(dataset).split_dataset_groupwise(K, 0.1, 'dirichlet', int(num_nodes/K), 10, 'dirichlet', noise)]
n_assign_list = [3,6]
model_name_list0 = ['BayesPFedAvg']#,] #,FedSoft，'BayesFedAvg','Fesem',,,,,'GNN','FedAvg','FedAvg','Wecfl',Fesem
model_name_list1 = ['GNN']
model_name_list2 = ['JPDA','MHT'] #,'MHT'
# cost_methods = ['weighted'] #,'average'
K_set = K
warm_ups = [False,True]
Parallel(n_jobs=2)(delayed(main)(seeds, dataset_splited, model, model_name, K_set) \
for dataset_splited in dataset_splited_list \
for model_name in model_name_list0)
# Parallel(n_jobs=2)(delayed(main)(seeds, dataset_splited, model, model_name, K_set, warm_up=warm_up) \
# for dataset_splited in dataset_splited_list \
# for model_name in model_name_list1 \
# for warm_up in warm_ups)
# Parallel(n_jobs=1)(delayed(main)(seeds, dataset_splited, model, model_name,K_set, n_assign, warm_up) \
# for dataset_splited in dataset_splited_list \
# for model_name in model_name_list2 \
# for n_assign in n_assign_list \
# for warm_up in warm_ups)
# main(seeds, dataset_splited_list[0], model, model_name_list2[0], K=K, n_assign=n_assign_list[0],warm_up=warm_ups[0])
# main(seeds, dataset_splited_list[0], model, model_name_list2[0], K=K, n_assign=n_assign_list[0],warm_up=warm_ups[1])
# main(seeds, dataset_splited_list[0], model, model_name_list2[0], K=K, n_assign=n_assign_list[1],warm_up=warm_ups[0])
# main(seeds, dataset_splited_list[0], model, model_name_list2[0], K=K, n_assign=n_assign_list[1],warm_up=warm_ups[1])
# main(seeds, dataset_splited_list[0], model, model_name_list2[1], K=K, n_assign=n_assign_list[0],warm_up=warm_ups[0])
# main(seeds, dataset_splited_list[0], model, model_name_list2[1], K=K, n_assign=n_assign_list[0],warm_up=warm_ups[1])
# main(seeds, dataset_splited_list[0], model, model_name_list2[1], K=K, n_assign=n_assign_list[1],warm_up=warm_ups[0])
# main(seeds, dataset_splited_list[0], model, model_name_list2[1], K=K, n_assign=n_assign_list[1],warm_up=warm_ups[1])