-
Notifications
You must be signed in to change notification settings - Fork 1
/
utils.py
185 lines (158 loc) · 7.58 KB
/
utils.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
import torch
import Node
from torch.optim.lr_scheduler import _LRScheduler,ReduceLROnPlateau
class GradualWarmupScheduler(_LRScheduler):
""" Gradually warm-up(increasing) learning rate in optimizer.
Proposed in 'Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour'.
Args:
optimizer (Optimizer): Wrapped optimizer.
multiplier: target learning rate = base lr * multiplier
total_epoch: target learning rate is reached at total_epoch, gradually
after_scheduler: after target_epoch, use this scheduler(eg. ReduceLROnPlateau)
"""
def __init__(self, optimizer, total_epoch, init_lr=1e-7, after_scheduler=None):
self.init_lr = init_lr
assert init_lr > 0, 'Initial LR should be greater than 0.'
self.total_epoch = total_epoch
self.after_scheduler = after_scheduler
self.finished = False
super().__init__(optimizer)
def get_lr(self):
if self.last_epoch > self.total_epoch:
if self.after_scheduler:
if not self.finished:
self.finished = True
return self.after_scheduler.get_lr()
return self.base_lrs
return [(((base_lr - self.init_lr) / self.total_epoch) * self.last_epoch + self.init_lr) for base_lr in
self.base_lrs]
def step_ReduceLROnPlateau(self, metrics, epoch=None):
if epoch is None:
epoch = self.last_epoch + 1
self.last_epoch = epoch if epoch != 0 else 1 # ReduceLROnPlateau is called at the end of epoch, whereas others are called at beginning
if self.last_epoch <= self.total_epoch:
warmup_lr = [(((base_lr - self.init_lr) / self.total_epoch) * self.last_epoch + self.init_lr) for base_lr in
self.base_lrs]
for param_group, lr in zip(self.optimizer.param_groups, warmup_lr):
param_group['lr'] = lr
else:
if epoch is None:
self.after_scheduler.step(metrics, None)
else:
self.after_scheduler.step(metrics, epoch - self.total_epoch)
def step(self, epoch=None, metrics=None):
if type(self.after_scheduler) != ReduceLROnPlateau:
if (self.finished and self.after_scheduler) or self.total_epoch == 0:
if epoch is None:
self.after_scheduler.step(None)
else:
self.after_scheduler.step(epoch - self.total_epoch)
else:
return super(GradualWarmupScheduler, self).step(epoch)
else:
self.step_ReduceLROnPlateau(metrics, epoch)
class Recorder(object):
def __init__(self, args, logger):
self.args = args
self.counter = 0
self.tra_loss = {}
self.tra_acc = {}
self.val_loss = {}
self.val_acc = {}
self.logger = logger
for i in range(self.args.node_num + 1):
self.val_loss[str(i)] = []
self.val_acc[str(i)] = []
self.val_loss[str(i)] = []
self.val_acc[str(i)] = []
self.acc_best = torch.zeros(self.args.node_num + 1)
self.get_a_better = torch.zeros(self.args.node_num + 1)
def validate(self, node):
self.counter += 1
node.model.to(node.device).eval()
total_loss = 0.0
correct = 0.0
with torch.no_grad():
for idx, (data, target) in enumerate(node.test_data):
data, target = data.to(node.device), target.to(node.device)
output = node.model(data)
total_loss += torch.nn.CrossEntropyLoss()(output, target)
pred = output.argmax(dim=1)
correct += pred.eq(target.view_as(pred)).sum().item()
total_loss = total_loss / (idx + 1)
acc = correct / len(node.test_data.dataset) * 100
self.val_loss[str(node.num)].append(total_loss)
self.val_acc[str(node.num)].append(acc)
if self.val_acc[str(node.num)][-1] > self.acc_best[node.num]:
self.get_a_better[node.num] = 1
self.acc_best[node.num] = self.val_acc[str(node.num)][-1]
torch.save(node.model.state_dict(),
'save/model/Node{:d}_{:s}.pt'.format(node.num, node.args.local_model))
# add warm_up lr
if self.args.warm_up == True and str(node.num) != '0':
node.sche_local.step(metrics=self.val_acc[str(node.num)][-1])
node.sche_meme.step(metrics=self.val_acc[str(node.num)][-1])
if self.val_acc[str(node.num)][-1] <= self.acc_best[node.num]:
print('##### Node{:d}: Not better Accuracy: {:.2f}%'.format(node.num, self.val_acc[str(node.num)][-1]))
node.meme.to(node.device).eval()
total_loss = 0.0
correct = 0.0
with torch.no_grad():
for idx, (data, target) in enumerate(node.test_data):
data, target = data.to(node.device), target.to(node.device)
output = node.meme(data)
total_loss += torch.nn.CrossEntropyLoss()(output, target)
pred = output.argmax(dim=1)
correct += pred.eq(target.view_as(pred)).sum().item()
total_loss = total_loss / (idx + 1)
acc = correct / len(node.test_data.dataset) * 100
def log(self, node):
return self.val_acc[str(node.num)][-1], self.val_loss[str(node.num)][-1]
def printer(self, node):
if self.get_a_better[node.num] == 1 and node.num == 0:
print('Node{:d}: A Better Accuracy: {:.2f}%! Model Saved!'.format(node.num, self.acc_best[node.num]))
self.get_a_better[node.num] = 0
elif self.get_a_better[node.num] == 1:
self.get_a_better[node.num] = 0
def finish(self):
torch.save([self.val_loss, self.val_acc],
'save/record/loss_acc_{:s}_{:s}.pt'.format(self.args.algorithm, self.args.notes))
print('Finished!\n')
for i in range(self.args.node_num + 1):
print('Node{}: Best Accuracy = {:.2f}%'.format(i, self.acc_best[i]))
def Catfish(Node_List, args):
if args.catfish is None:
pass
else:
Node_List[0].model = Node.init_model(args.catfish)
Node_List[0].optimizer = Node.init_optimizer(Node_List[0].model, args)
def LR_scheduler(rounds, Node_List, args, Global_node = None):
# trigger = 7
if rounds > 15 and rounds <=30:
trigger = 15
elif rounds > 30 and rounds <=45:
trigger = 25
elif rounds > 45 and rounds <=50:
trigger = 40
else:
trigger = 51
if rounds != 0 and rounds % trigger == 0 and rounds < args.stop_decay:
args.lr *= 0.5
for i in range(len(Node_List)):
Node_List[i].args.lr = args.lr
Node_List[i].args.alpha = args.alpha
Node_List[i].args.beta = args.beta
Node_List[i].optimizer.param_groups[0]['lr'] = args.lr
Node_List[i].meme_optimizer.param_groups[0]['lr'] = args.lr
if Global_node !=None:
Global_node.args.lr = args.lr
Global_node.model_optimizer.param_groups[0]['lr'] = args.lr
print('Learning rate={:.10f}'.format(args.lr))
def Summary(args):
print("Summary:\n")
print("algorithm:{}\n".format(args.algorithm))
print("dataset:{}\tbatchsize:{}\n".format(args.dataset, args.batch_size))
print("node_num:{},\tsplit:{}\n".format(args.node_num, args.split))
# print("iid:{},\tequal:{},\n".format(args.iid == 1, args.unequal == 0))
print("global epochs:{},\tlocal epochs:{},\n".format(args.R, args.E))
print("global_model:{},\tlocal model:{},\n".format(args.global_model, args.local_model))