-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathTrainer.py
193 lines (163 loc) · 8.14 KB
/
Trainer.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
from ast import arg
from tqdm import tqdm
import torch.nn as nn
import numpy as np
import torch
KL_Loss = nn.KLDivLoss(reduction='batchmean')
Softmax = nn.Softmax(dim=1)
LogSoftmax = nn.LogSoftmax(dim=1)
CE_Loss = nn.CrossEntropyLoss()
def train_normal(node,args):
node.model.to(node.device).train()
train_loader = node.train_data
total_loss = 0.0
avg_loss = 0.0
correct = 0.0
acc = 0.0
description = "Training (the {:d}-batch): tra_Loss = {:.4f} tra_Accuracy = {:.2f}%"
with tqdm(train_loader) as epochs:
for idx, (data, target) in enumerate(epochs):
node.optimizer.zero_grad()
epochs.set_description(description.format(idx + 1, avg_loss, acc))
data, target = data.to(node.device), target.to(node.device)
output = node.model(data)
loss = CE_Loss(output, target)
loss.backward()
node.optimizer.step()
total_loss += loss
avg_loss = total_loss / (idx + 1)
pred = output.argmax(dim=1)
correct += pred.eq(target.view_as(pred)).sum()
acc = correct / len(train_loader.dataset) * 100
def train_avg(node,args):
node.meme.to(node.device).train()
train_loader = node.train_data
total_loss = 0.0
avg_loss = 0.0
correct = 0.0
acc = 0.0
description = "Node{:d}: loss={:.4f} acc={:.2f}%"
with tqdm(train_loader) as epochs:
for idx, (data, target) in enumerate(epochs):
node.meme_optimizer.zero_grad()
epochs.set_description(description.format(node.num, avg_loss, acc))
data, target = data.to(node.device), target.to(node.device)
output = node.meme(data)
loss = CE_Loss(output, target)
loss.backward()
node.meme_optimizer.step()
total_loss += loss
avg_loss = total_loss / (idx + 1)
pred = output.argmax(dim=1)
correct += pred.eq(target.view_as(pred)).sum()
acc = correct / len(train_loader.dataset) * 100
node.model = node.meme
def train_mutual(node,args,logger):
node.model.to(node.device).train()
node.meme.to(node.device).train()
train_loader = node.train_data
total_local_loss = 0.0
avg_local_loss = 0.0
correct_local = 0.0
acc_local = 0.0
total_meme_loss = 0.0
avg_meme_loss = 0.0
correct_meme = 0.0
acc_meme = 0.0
train_index = 0
total_global_kl_loss = 0.0
total_local_kl_loss = 0.0
avg_global_kl_loss = 0.0
avg_local_kl_loss = 0.0
description = 'Node{:d}: loss_model={:.4f} acc_model={:.2f}% loss_meme={:.4f} acc_meme={:.2f}%'
with tqdm(train_loader) as epochs:
for idx, (data, target) in enumerate(epochs):
train_index = train_index + 1
node.optimizer.zero_grad()
node.meme_optimizer.zero_grad()
epochs.set_description(description.format(node.num, avg_local_loss, acc_local, avg_meme_loss, acc_meme))
data, target = data.to(node.device), target.to(node.device)
output_local = node.model(data)
output_meme = node.meme(data)
kl_local = KL_Loss(LogSoftmax(output_local), Softmax(output_meme.detach()))
kl_meme = KL_Loss(LogSoftmax(output_meme), Softmax(output_local.detach()))
total_local_kl_loss += kl_local
total_global_kl_loss += kl_meme
_output_local = nn.Softmax(dim=1)(output_local)
_, src_idx = torch.sort(_output_local, 1, descending=True)
_output_meme = nn.Softmax(dim=1)(output_meme)
_, src_idx_meme = torch.sort(_output_meme, 1, descending=True)
if args.topk > 0:
topk = np.min([args.topk, args.classes])
for i in range( _output_local.size()[0]):
output_local[i, src_idx[i, topk:]] = (1.0 - _output_local[i, src_idx[i, :topk]].sum())/ ( _output_local.size()[1] - topk)
output_meme[i, src_idx[i, topk:]] = (1.0 - _output_meme[i, src_idx[i, :topk]].sum())/ ( _output_meme.size()[1] - topk)
ce_local = CE_Loss(output_local, target)
ce_meme = CE_Loss(output_meme, target)
loss_local = node.args.alpha * ce_local + (1 - node.args.alpha) * kl_local
loss_meme = node.args.beta * ce_meme + (1 - node.args.beta) * kl_meme
loss_local.backward()
loss_meme.backward()
node.optimizer.step()
node.meme_optimizer.step()
total_local_loss += loss_local
avg_local_loss = total_local_loss / (idx + 1)
pred_local = output_local.argmax(dim=1)
correct_local += pred_local.eq(target.view_as(pred_local)).sum()
acc_local = correct_local / len(train_loader.dataset) * 100
total_meme_loss += loss_meme
avg_meme_loss = total_meme_loss / (idx + 1)
pred_meme = output_meme.argmax(dim=1)
correct_meme += pred_meme.eq(target.view_as(pred_meme)).sum()
acc_meme = correct_meme / len(train_loader.dataset) * 100
avg_global_kl_loss = total_global_kl_loss / (idx + 1)
avg_local_kl_loss = total_local_kl_loss / (idx + 1)
if args.mix > 0:
mixed_total_local_loss = 0.0
mixed_avg_local_loss = 0.0
mixed_correct_local = 0.0
mixed_acc_local = 0.0
mixed_total_meme_loss = 0.0
mixed_avg_meme_loss = 0.0
mixed_correct_meme = 0.0
mixed_acc_meme = 0.0
alpha = 0.3
lam = np.random.beta(alpha, alpha)
index = torch.randperm(data.size()[0]).cuda()
other_data, other_target = data[index, :],target[index]
other_data, other_target = other_data.to(node.device), other_target.to(node.device)
mixed_input = lam * data + (1 - lam) * other_data
mixed_label = lam * target + (1 - lam) * other_target
mixed_output_local = node.model(mixed_input)
mixed_output_meme = node.meme(mixed_input)
# mixed_output_local,mixed_output_meme = mixed_output_local.to(node.device),mixed_output_meme.to(node.device)
mixed_ce_local = CE_Loss(mixed_output_local, mixed_label.long())
mixed_kl_local = KL_Loss(LogSoftmax(mixed_output_local), Softmax(mixed_output_meme.detach()))
mixed_ce_meme = CE_Loss(mixed_output_meme, mixed_label.long())
mixed_kl_meme = KL_Loss(LogSoftmax(mixed_output_meme), Softmax(mixed_output_local.detach()))
mixed_loss_local = node.args.alpha * mixed_ce_local + (1 - node.args.alpha) * mixed_kl_local
mixed_loss_meme = node.args.beta * mixed_ce_meme + (1 - node.args.beta) * mixed_kl_meme
(args.mix * mixed_loss_local).backward()
(args.mix * mixed_loss_meme).backward()
node.optimizer.step()
node.meme_optimizer.step()
mixed_total_local_loss += mixed_loss_local
mixed_avg_local_loss = mixed_total_local_loss / (idx + 1)
mixed_pred_local = mixed_output_local.argmax(dim=1)
mixed_correct_local += mixed_pred_local.eq(target.view_as(pred_local)).sum()
mixed_acc_local = mixed_correct_local / len(train_loader.dataset) * 100
mixed_total_meme_loss += mixed_loss_meme
mixed_avg_meme_loss = mixed_total_meme_loss / (idx + 1)
mixed_pred_meme = mixed_output_meme.argmax(dim=1)
mixed_correct_meme += mixed_pred_meme.eq(mixed_label.view_as(mixed_pred_meme)).sum()
mixed_acc_meme = mixed_correct_meme / len(train_loader.dataset) * 100
class Trainer(object):
def __init__(self, args, logger=None):
if args.algorithm == 'fed_mutual':
self.train = train_mutual
elif args.algorithm == 'fed_avg':
self.train = train_avg
elif args.algorithm == 'normal':
self.train = train_normal
def __call__(self, node,args,logger):
self.train(node,args,logger)