-
Notifications
You must be signed in to change notification settings - Fork 10
/
shared_optim.py
176 lines (146 loc) · 6.41 KB
/
shared_optim.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
from __future__ import division
import math
import torch
import torch.optim as optim
from collections import defaultdict
class SharedRMSprop(optim.Optimizer):
"""Implements RMSprop algorithm with shared states.
"""
def __init__(self,
params,
lr=7e-4,
alpha=0.99,
eps=0.1,
weight_decay=0,
momentum=0,
centered=False):
defaults = defaultdict(lr=lr, alpha=alpha, eps=eps,
weight_decay=weight_decay, momentum=momentum, centered=centered)
super(SharedRMSprop, self).__init__(params, defaults)
for group in self.param_groups:
for p in group['params']:
state = self.state[p]
state['step'] = torch.zeros(1)
state['grad_avg'] = p.data.new().resize_as_(p.data).zero_()
state['square_avg'] = p.data.new().resize_as_(p.data).zero_()
state['momentum_buffer'] = p.data.new(
).resize_as_(p.data).zero_()
def share_memory(self):
for group in self.param_groups:
for p in group['params']:
state = self.state[p]
state['square_avg'].share_memory_()
state['step'].share_memory_()
state['grad_avg'].share_memory_()
state['momentum_buffer'].share_memory_()
def step(self, closure=None):
"""Performs a single optimization step.
Arguments:
closure (callable, optional): A closure that reevaluates the model
and returns the loss.
"""
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data
if grad.is_sparse:
raise RuntimeError(
'RMSprop does not support sparse gradients')
state = self.state[p]
square_avg = state['square_avg']
alpha = group['alpha']
state['step'] += 1
if group['weight_decay'] != 0:
grad = grad.add(group['weight_decay'], p.data)
square_avg.mul_(alpha).addcmul_(1 - alpha, grad, grad)
if group['centered']:
grad_avg = state['grad_avg']
grad_avg.mul_(alpha).add_(1 - alpha, grad)
avg = square_avg.addcmul(
-1, grad_avg, grad_avg).sqrt().add_(group['eps'])
else:
avg = square_avg.sqrt().add_(group['eps'])
if group['momentum'] > 0:
buf = state['momentum_buffer']
buf.mul_(group['momentum']).addcdiv_(grad, avg)
p.data.add_(-group['lr'], buf)
else:
p.data.addcdiv_(-group['lr'], grad, avg)
return loss
class SharedAdam(optim.Optimizer):
"""Implements Adam algorithm with shared states.
"""
def __init__(self,
params,
lr=1e-3,
betas=(0.9, 0.999),
eps=1e-3,
weight_decay=0, amsgrad=True):
defaults = defaultdict(lr=lr, betas=betas, eps=eps,
weight_decay=weight_decay, amsgrad=amsgrad)
super(SharedAdam, self).__init__(params, defaults)
for group in self.param_groups:
for p in group['params']:
state = self.state[p]
state['step'] = torch.zeros(1)
state['exp_avg'] = p.data.new().resize_as_(p.data).zero_()
state['exp_avg_sq'] = p.data.new().resize_as_(p.data).zero_()
state['max_exp_avg_sq'] = p.data.new(
).resize_as_(p.data).zero_()
def share_memory(self):
for group in self.param_groups:
for p in group['params']:
state = self.state[p]
state['step'].share_memory_()
state['exp_avg'].share_memory_()
state['exp_avg_sq'].share_memory_()
state['max_exp_avg_sq'].share_memory_()
def step(self, closure=None):
"""Performs a single optimization step.
Arguments:
closure (callable, optional): A closure that reevaluates the model
and returns the loss.
"""
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group['params']:
if p.grad is None:
continue
grad = p.grad.data
if grad.is_sparse:
raise RuntimeError(
'Adam does not support sparse gradients, please consider SparseAdam instead')
amsgrad = group['amsgrad']
state = self.state[p]
exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
if amsgrad:
max_exp_avg_sq = state['max_exp_avg_sq']
beta1, beta2 = group['betas']
state['step'] += 1
if group['weight_decay'] != 0:
grad = grad.add(group['weight_decay'], p.data)
# Decay the first and second moment running average coefficient
exp_avg.mul_(beta1).add_(1 - beta1, grad)
exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)
if amsgrad:
# Maintains the maximum of all 2nd moment running avg. till
# now
torch.max(max_exp_avg_sq, exp_avg_sq, out=max_exp_avg_sq)
# Use the max. for normalizing running avg. of gradient
denom = max_exp_avg_sq.sqrt().add_(group['eps'])
else:
denom = exp_avg_sq.sqrt().add_(group['eps'])
# bias_correction1 = 1 - beta1**state['step'][0]
# bias_correction2 = 1 - beta2**state['step'][0]
bias_correction1 = 1 - beta1**state['step'].item()
bias_correction2 = 1 - beta2**state['step'].item()
step_size = group['lr'] * \
math.sqrt(bias_correction2) / bias_correction1
p.data.addcdiv_(-step_size, exp_avg, denom)
return loss