-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathmain.py
245 lines (200 loc) · 12 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
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
from typing import Optional, Any, Callable
import numpy as np
import argparse
import math
import time
import torch
import copy
import learning_flow
import train_script
from Nets import CNNMnist, MLP
import scipy.io as sio
def initial():
# network parameters
setup = argparse.ArgumentParser()
setup.add_argument('--K', type=int, default=20, help='total # of devices')
setup.add_argument('--N', type=int, default=1, help='# of relays')
setup.add_argument('--PL', type=float, default=3.0, help='path loss exponent')
# simulation parameters
setup.add_argument('--trial', type=int, default=50, help='# of Monte Carlo Trials')
setup.add_argument('--SNR', type=float, default=100, help='-noise variance in dB')
setup.add_argument('--P_r', type=float, default=0.1, help='relay transmit power budget 0.1W')
setup.add_argument('--verbose', type=int, default=1, help=r'whether output or not')
setup.add_argument('--seed', type=int, default=1, help='random seed (default: 1)')
# learning parameters
setup.add_argument('--gpu', type=int, default=0, help=r'Use which gpu')
setup.add_argument('--local_ep', type=int, default=1, help="the number of local epochs, E")
setup.add_argument('--local_bs', type=int, default=0, help="0 for no effect, local bath size, B")
setup.add_argument('--lr', type=float, default=0.05, help="learning rate, lambda")
setup.add_argument('--low_lr', type=float, default=1e-5, help="learning rate lower bound, bar_lambda")
setup.add_argument('--gamma', type=float, default=0.9, help="learning rate decrease ratio, gamma")
setup.add_argument('--step', type=int, default=50, help="learning rate decrease step, bar_T")
setup.add_argument('--momentum', type=float, default=0.99,
help="SGD momentum, used only for multiple local updates")
setup.add_argument('--epochs', type=int, default=500, help="rounds of training, T")
setup.add_argument('--iid', type=int, default=1, help="1 for iid, 0 for non-iid")
setup.add_argument('--noniid_level', type=int, default=2, help="number of classes at each device for non-iid")
setup.add_argument('--V_idx', type=int, default=0, help="Variable index")
args = setup.parse_args()
return args
if __name__ == '__main__':
setup = initial()
np.random.seed(setup.seed)
torch.manual_seed(setup.seed)
setup.init_lr = copy.deepcopy(setup.lr)
print(setup)
data = sio.loadmat('matlab/DATA/trial_50_K_20_N_1_PL_3_Pr.mat')
Pr_set = [0.01, 0.1, 0.3, 0.5, 1]
V_idx = setup.V_idx
setup.P_r = Pr_set[V_idx]
store_filename = 'store/trial_{}_K_{}_N_{}_B_{}_E_{}_lr_{}_SNR_{}_PL_{}_Pr_{}.npz'.format(setup.trial, setup.K,
setup.N, setup.local_bs,
setup.local_ep, setup.lr,
setup.SNR, setup.PL,
setup.P_r)
print(store_filename)
setup.sigma = np.power(10, -setup.SNR / 10)
channel_U = data['channel_U']
channel_R = data['channel_R']
channel_UR = data['channel_UR']
Proposed_a_k1 = data['Proposed_a_k1']
Proposed_a_k2 = data['Proposed_a_k2']
Proposed_b_n = data['Proposed_b_n']
Proposed_c_1 = data['Proposed_c_1']
Proposed_c_2 = data['Proposed_c_2']
Single_a_k1 = data['Single_a_k1']
Single_c_1 = data['Single_c_1']
Xu_a_k1 = data['Xu_a_k1']
Xu_b_n = data['Xu_b_n']
Xu_eta = data['Xu_eta']
MSE_1 = np.zeros([setup.trial, 2 * setup.epochs])
MSE_2 = np.zeros([setup.trial, setup.epochs])
MSE_3 = np.zeros([setup.trial, 2 * setup.epochs])
MSE_4 = np.zeros([setup.trial, setup.epochs])
MSE_5 = np.zeros([setup.trial, setup.epochs])
MSE2_1 = np.zeros([setup.trial, 2 * setup.epochs])
MSE2_2 = np.zeros([setup.trial, setup.epochs])
MSE2_3 = np.zeros([setup.trial, 2 * setup.epochs])
MSE2_4 = np.zeros([setup.trial, setup.epochs])
MSE2_5 = np.zeros([setup.trial, setup.epochs])
result_store = []
result_set = []
result_CNN_set = []
result_CNN_MB_set = []
print(torch.__version__)
print(torch.version.cuda)
print(torch.backends.cudnn.version())
setup.device = torch.device(
'cuda:{}'.format(setup.gpu) if torch.cuda.is_available() and setup.gpu != -1 else 'cpu')
print(setup.device)
for i in range(setup.trial):
print('This is the {0}-th trial'.format(i))
setup.h_k = channel_U[: setup.K, i]
setup.f_n = channel_R[: setup.N, i]
setup.g_kn = channel_UR[: setup.K, : setup.N, i]
p_a_k1 = Proposed_a_k1[V_idx, : setup.K, i]
p_a_k2 = Proposed_a_k2[V_idx, : setup.K, i]
p_b_n = Proposed_b_n[V_idx, : setup.N, i]
p_c_1 = Proposed_c_1[V_idx, i]
p_c_2 = Proposed_c_2[V_idx, i]
s_a_k1 = Single_a_k1[V_idx, : setup.K, i]
s_c_1 = Single_c_1[V_idx, i]
x_a_k1 = Xu_a_k1[V_idx, : setup.K, i]
x_b_n = Xu_b_n[V_idx, : setup.N, i]
x_eta = Xu_eta[V_idx, i]
Error_free = 1
Proposed = 1
Single_slot = 1
Xu_scheme = 1
result = {}
if setup.iid:
train_images, train_labels, test_images, test_labels, size = train_script.load_fmnist_iid(setup.K)
else:
train_images, train_labels, test_images, test_labels, size = train_script.load_fmnist_noniid(setup.K,
setup.non_iid_level)
net_glob = CNNMnist(num_classes=10, num_channels=1, batch_norm=True).to(setup.device)
# net_glob = MLP(784, 64, 10).to(setup.device)
setup.size = size
setup.rho = np.ones(setup.K, dtype=float) * (setup.size / np.sum(setup.size))
if setup.verbose:
print(net_glob)
w_glob = net_glob.state_dict()
w_0 = copy.deepcopy(w_glob)
d = 0
for item in w_glob.keys():
d = d + int(np.prod(w_glob[item].shape))
print('Total Number of Parameters={}'.format(d))
net_glob.load_state_dict(w_glob)
idxs_users = np.asarray(range(setup.N))
if Error_free:
print('Error_Free Channel is running')
loss_train1, accuracy_test1, loss_test1, mse_1, mse2_1 = learning_flow.learning_iter(setup, d, net_glob,
w_glob, train_images,
train_labels,
test_images,
test_labels, 1, None,
None, None,
None, None)
result['loss_train1'] = np.asarray(loss_train1)
result['accuracy_test1'] = np.asarray(accuracy_test1)
result['loss_test1'] = np.asarray(loss_test1)
print('result {}'.format(result['accuracy_test1'][len(result['accuracy_test1']) - 1]))
MSE_1[i, :] = mse_1
MSE2_1[i, :] = mse2_1
if Proposed:
print('Proposed Scheme is running')
w_glob = copy.deepcopy(w_0)
net_glob.load_state_dict(w_glob)
loss_train2, accuracy_test2, loss_test2, mse_2, mse2_2 = learning_flow.learning_iter(setup, d, net_glob,
w_glob, train_images,
train_labels,
test_images,
test_labels, 2, p_a_k1,
p_a_k2, p_b_n,
p_c_1, p_c_2)
result['loss_train2'] = np.asarray(loss_train2)
result['accuracy_test2'] = np.asarray(accuracy_test2)
result['loss_test2'] = np.asarray(loss_test2)
print('result {}'.format(result['accuracy_test2'][len(result['accuracy_test2']) - 1]))
MSE_2[i, :] = mse_2
MSE2_2[i, :] = mse2_2
if Single_slot:
print('Conventional Scheme is running')
w_glob = copy.deepcopy(w_0)
net_glob.load_state_dict(w_glob)
loss_train3, accuracy_test3, loss_test3, mse_3, mse2_3 = learning_flow.learning_iter(setup, d, net_glob,
w_glob, train_images,
train_labels,
test_images,
test_labels, 3,
s_a_k1, None,
None, s_c_1, None)
result['loss_train3'] = np.asarray(loss_train3)
result['accuracy_test3'] = np.asarray(accuracy_test3)
result['loss_test3'] = np.asarray(loss_test3)
print('result {}'.format(result['accuracy_test3'][len(result['accuracy_test3']) - 1]))
MSE_3[i, :] = mse_3
MSE2_3[i, :] = mse2_3
if Xu_scheme:
print('Existing Scheme is running')
w_glob = copy.deepcopy(w_0)
net_glob.load_state_dict(w_glob)
loss_train5, accuracy_test5, loss_test5, mse_5, mse2_5 = learning_flow.learning_iter(setup, d, net_glob,
w_glob, train_images,
train_labels,
test_images,
test_labels, 5,
x_a_k1,
None, x_b_n, None,
x_eta)
result['loss_train5'] = np.asarray(loss_train5)
result['accuracy_test5'] = np.asarray(accuracy_test5)
result['loss_test5'] = np.asarray(loss_test5)
print('result {}'.format(result['accuracy_test5'][len(result['accuracy_test5']) - 1]))
MSE_5[i, :] = mse_5
MSE2_5[i, :] = mse2_5
result_store.append(result)
np.savez(store_filename, vars(setup), result_store, MSE_1, MSE_2, MSE_3, MSE_4, MSE_5, MSE2_1, MSE2_2, MSE2_3,
MSE2_4, MSE2_5)
np.savez(store_filename, vars(setup), result_store, MSE_1, MSE_2, MSE_3, MSE_4, MSE_5, MSE2_1, MSE2_2, MSE2_3,
MSE2_4, MSE2_5)