-
Notifications
You must be signed in to change notification settings - Fork 9
/
_beta
277 lines (227 loc) · 9.97 KB
/
_beta
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
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
#!/usr/bin/env python
# -*- coding: utf-8 -*-
from __future__ import print_function, absolute_import
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn import init
import math
import torch.utils.model_zoo as model_zoo
import pdb
import numpy as np
from collections import OrderedDict
__all__ = ['ResNet', 'resnet20_ste_6']
model_urls = {
'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
}
import torch
import torch.nn as nn
import torch.nn.functional as F
bit_num = 1
range_bit = 2 ** bit_num
class RoundWithGradient(torch.autograd.Function):
@staticmethod
def forward(ctx, x):
return x.round()
@staticmethod
def backward(ctx, g):
return g
class GradientScale(torch.autograd.Function):
@staticmethod
def forward(ctx, x):
return x
@staticmethod
def backward(ctx, g):
return g / 0.001
class ste(torch.autograd.Function):
@staticmethod
def forward(ctx, x):
x_abs = torch.abs(x) + 0.5
x_abs = torch.clamp(x_abs, min=0.5+(1e-4), max=(range_bit/2)+0.5-(1e-4))
x_sign = torch.sign(x)
x_sign = torch.sign(x_sign - (1e-4))
output = x_sign * torch.round(x_abs)
return output
@staticmethod
def backward(ctx, g):
return g
class DSQConv(nn.Conv2d):
def __init__(self, in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True,
momentum = 0.1,
num_bit = bit_num, QInput = True, bSetQ = True):
super(DSQConv, self).__init__(in_channels, out_channels, kernel_size, stride, padding, dilation, groups, bias)
self.num_bit = num_bit
self.quan_input = QInput
self.bit_range = 2**self.num_bit - 1
self.is_quan = bSetQ
self.temp = -1
self.sig_w = 5 # T:10 - 5, T:1 - 1
self.sig_a = 5 # T:10 - 5, T:1 - 1
self.q_value = torch.from_numpy(np.linspace(0,self.bit_range,self.bit_range+1))
self.q_value = self.q_value.reshape(len(self.q_value),1,1,1,1).float().cuda()
if self.is_quan:
# using int32 max/min as init and backprogation to optimization
# Weight
self.uW = nn.Parameter(data = torch.tensor(2 **31 - 1).float().cuda())
self.lW = nn.Parameter(data = torch.tensor((-1) * (2**32)).float().cuda())
self.register_buffer('init', torch.tensor(1).float().cuda())
self.beta_w = nn.Parameter(data = torch.tensor(0.2).float().cuda())
# Bias
if self.bias is not None:
self.uB = nn.Parameter(data = torch.tensor(2 **31 - 1).float())
self.lB = nn.Parameter(data = torch.tensor((-1) * (2**32)).float())
self.register_buffer('running_uB', torch.tensor([self.uB.data]))# init with ub
self.register_buffer('running_lB', torch.tensor([self.lB.data]))# init with lb
self.alphaB = nn.Parameter(data = torch.tensor(1).float())
# Activation input
if self.quan_input:
self.uA = nn.Parameter(data = torch.tensor(2 **31 - 1).float().cuda())
self.lA = nn.Parameter(data = torch.tensor((-1) * (2**32)).float().cuda())
self.beta_a = nn.Parameter(data = torch.tensor(0.2).float().cuda())
def clipping(self, x, upper, lower):
# clip lower
x = x + F.relu(lower - x)
# clip upper
x = x - F.relu(x - upper)
return x
def step(self, x):
if self.num_bit == 1:
output = (x+1) / 2
output = RoundWithGradient.apply(output)
return 2 * output - 1
else:
return RoundWithGradient.apply(x)
def w_quan(self, x, u, l):
delta = (u - l) / (self.bit_range)
interval = (x - l) / delta
output = 2 * RoundWithGradient.apply(interval) - self.bit_range
return output
def a_quan(self, x, u, l):
delta = (u - l) / (self.bit_range)
interval = (x - l) / delta
output = RoundWithGradient.apply(interval)
return output
def sigmoid(self, x, T=2):
output = 1 / (1+torch.exp(-(x)*T))
return output
def forward(self, x):
if self.is_quan:
if self.init:
print(self.init)
self.init = torch.tensor(0)
self.lW.data = -self.weight.std() * 3
self.uW.data = self.weight.std() * 3
self.lA.data = -x.std() * 3
self.uA.data = x.std() * 3
self.beta_w.data = torch.mean(torch.abs(self.weight)) / self.bit_range
self.beta_a.data = torch.mean(torch.abs(x)) / self.bit_range
self.lW.data = torch.clamp(self.lW.data, min=-1e+1, max=-1e-6)
self.uW.data = torch.clamp(self.uW.data, min=1e-6, max=1e+1)
self.lA.data = torch.clamp(self.lA.data, min=-1e+1, max=-1e-6)
self.uA.data = torch.clamp(self.uA.data, min=1e-6, max=1e+1)
curr_running_lw = self.lW
curr_running_uw = self.uW
curr_running_la = 0
curr_running_ua = self.uA
# Weight kernel_soft_argmax
Qweight = self.clipping(self.weight, curr_running_uw, curr_running_lw)
Qweight = self.w_quan(Qweight, curr_running_uw, curr_running_lw)
Qweight = torch.abs(self.beta_w) * Qweight
Qbias = self.bias
# Input(Activation)
Qactivation = x
if self.quan_input:
Qactivation = self.clipping(x, curr_running_ua, curr_running_la)
Qactivation = self.a_quan(Qactivation, curr_running_ua, curr_running_la)
Qactivation = torch.abs(self.beta_a) * Qactivation
output = F.conv2d(Qactivation, Qweight, Qbias, self.stride, self.padding, self.dilation, self.groups)
else:
output = F.conv2d(x, self.weight, self.bias, self.stride,
self.padding, self.dilation, self.groups)
return output
class LambdaLayer(nn.Module):
def __init__(self, lambd):
super(LambdaLayer, self).__init__()
self.lambd = lambd
def forward(self, x):
return self.lambd(x)
class BasicBlock(nn.Module):
expansion = 1
def __init__(self, inplanes, planes, stride=1, downsample=None):
super(BasicBlock, self).__init__()
self.conv1 = DSQConv(inplanes, planes, kernel_size=3, stride=stride,
padding=1, bias=False)
self.bn1 = nn.BatchNorm2d(planes)
self.relu = nn.ReLU(inplace=True)
self.conv2 = DSQConv(planes, planes, kernel_size=3, stride=1,
padding=1, bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.shortcut = nn.Sequential()
if stride != 1 or inplanes != planes:
self.shortcut = LambdaLayer(lambda x:
F.pad(x[:, :, ::2, ::2], (0, 0, 0, 0, planes//4, planes//4), "constant", 0))
def forward(self, x):
conv1_out = F.relu(self.bn1(self.conv1(x)))
conv2_out = self.bn2(self.conv2(conv1_out))
out = conv2_out + self.shortcut(x)
out = F.relu(out)
return out, conv1_out, conv2_out
# conv1_out = self.conv1(self.bn1(x))
# conv2_out = self.conv2(self.bn2(F.relu(conv1_out)))
# out = conv2_out + self.shortcut(x)
# out = F.relu(out)
# return out, conv1_out, conv2_out
class ResNet(nn.Module):
def __init__(self, block, num_blocks, num_classes=10):
super(ResNet, self).__init__()
self.in_planes = 16
self.conv1 = nn.Conv2d(3, 16, kernel_size=3, stride=1, padding=1,
bias=False)
self.bn1 = nn.BatchNorm2d(16)
self.layer1 = self._make_layer(block, 16, num_blocks[0], stride=1)
self.layer2 = self._make_layer(block, 32, num_blocks[1], stride=2)
self.layer3 = self._make_layer(block, 64, num_blocks[2], stride=2)
self.bn2 = nn.BatchNorm1d(64)
# self.bn3 = nn.BatchNorm1d(num_classes)
self.linear = nn.Linear(64, num_classes)
def _make_layer(self, block, planes, num_blocks, stride):
strides = [stride] + [1]*(num_blocks-1)
ret_dict = dict()
for i, stride in enumerate(strides):
layers = []
layers.append(block(self.in_planes, planes, stride))
ret_dict['block_{}'.format(i)] = nn.Sequential(*layers)
self.in_planes = planes * block.expansion
return nn.Sequential(OrderedDict(ret_dict))
def forward(self, x):
ret_dict = dict()
# out = F.relu(self.bn1(self.conv1(x)))
# out = F.hardtanh(self.bn1(self.conv1(x)))
out = F.relu(self.conv1(x))
layer_names = self.layer1._modules.keys()
for i, layer_name in enumerate(layer_names):
out, conv1_out, conv2_out = self.layer1._modules[layer_name](out)
ret_dict['layer1_{}_conv1'.format(i)] = conv1_out
ret_dict['layer1_{}_conv2'.format(i)] = conv2_out
layer_names = self.layer2._modules.keys()
for i, layer_name in enumerate(layer_names):
out, conv1_out, conv2_out = self.layer2._modules[layer_name](out)
ret_dict['layer2_{}_conv1'.format(i)] = conv1_out
ret_dict['layer2_{}_conv2'.format(i)] = conv2_out
layer_names = self.layer3._modules.keys()
for i, layer_name in enumerate(layer_names):
out, conv1_out, conv2_out = self.layer3._modules[layer_name](out)
ret_dict['layer3_{}_conv1'.format(i)] = conv1_out
ret_dict['layer3_{}_conv2'.format(i)] = conv2_out
out = F.avg_pool2d(out, out.size()[3])
out = out.view(out.size(0), -1)
out = self.bn2(out)
out = self.linear(out)
ret_dict['out'] = out
return ret_dict
def resnet20_ste_6(pretrained=True, **kwargs):
"""Constructs a ResNet-18 model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
"""
return ResNet(BasicBlock, [3, 3, 3], **kwargs)