-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlenet5_once_numpy.py
131 lines (112 loc) · 3.8 KB
/
lenet5_once_numpy.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
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import sys
from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
from lib_round import *
FLAGS = None
save_file='./ckptlenet_mul/last.ckpt'
#使用自己搭建的卷积和乘累加验证正确率单一测试
#输入：32x32
#第1层：卷积 6 5x5 -> 28x28x6
#第2层：池化 2x2 -> 14x14x6
#第3层：卷积 16 5x5 -> 10x10x16
#第4层：池化 2x2 -> 5x5x16
#第5层：全连接 120 -> 120
#第6层：全连接 84 ->84
#第7层：全连接 10 ->10
#6位整数，10位小数
#定义一堆函数
def write_file(fout,string,array):
fout.write(">>> "+ string + ":\n")
sh = np.shape(array)
fout.write(str(np.shape(array)))
fout.write("\n")
#write data
transfer = array.reshape(-1)
temp = []
for i in transfer:
temp.append("%04x" % i)
retransfer = np.array(temp)
#retransfer = retransfer.reshape(sh)
if len(sh) == 4: #如果是卷积结果
for ch in range(sh[3]):
fout.write("Channel %d:\n" % ch)
for ver in range(sh[2]):
fout.write("%02d:" % ver)
for hor in range(sh[1]):
fout.write(" " + retransfer[ver*sh[2]*sh[3] +hor*sh[3] +ch])
fout.write("\n")
else:
fout.write(str(retransfer))
fout.write("\n\n")
fout = open('result.txt','w')
print("import data")
# Import data
mnist = input_data.read_data_sets("../MNIST_data", one_hot=True)
print("start predict")
save_file='./ckptlenet/last.ckpt'
reader=tf.train.NewCheckpointReader(save_file)
# Create the model
# 定义输入层
with open('input_origin784.txt','r') as fin:
lines = fin.readlines()
xin = []
for line in lines:
xin.append(int(line.rstrip(),16))
a= []
a.append(xin)
xin = np.array(a)
xin = xin.reshape(1,28,28,1).astype(np.int16)
ximg = xin.reshape(28,28)
#write_file(fout,'input',xin)
#plt.imshow(ximg,cmap='gray')
#plt.axis('off')
#plt.show()
#定义第一层卷积
#第1层：卷积 6 5x5 -> 28x28x6
W_conv1 = change_int(reader.get_tensor('Layer1/L1_weight'))
b_conv1 = change_int(reader.get_tensor('Layer1/L1_bias'))
conv1_mul = conv2d_same(xin, W_conv1)
conv1_add = conv_add(conv1_mul,b_conv1)
result_l1 = relu(conv1_add)
write_file(fout,'result Layer 1',result_l1)
#第2层：池化 2x2 -> 14x14x6
result_l2 = max_pool_2x2(result_l1)
write_file(fout,'result Layer 2',result_l2)
#第3层：卷积 16 5x5 -> 10x10x16
W_conv2 = change_int(reader.get_tensor('Layer3/L3_weight'))
b_conv2 = change_int(reader.get_tensor('Layer3/L3_bias'))
conv2_mul = conv2d(result_l2, W_conv2)
conv2_add = conv_add(conv2_mul,b_conv2)
result_l3 = relu(conv2_add)
write_file(fout,'result Layer 3',result_l3)
#第4层：池化 2x2 -> 5x5x16
result_l4 = max_pool_2x2(result_l3)
write_file(fout,'result Layer 4',result_l4)
#第5层：全连接 120 -> 120
result_l4_flat = result_l4.reshape([-1, 5*5*16])
W_fc1 = change_int(reader.get_tensor('Layer5/F1_weight'))
b_fc1 = change_int(reader.get_tensor('Layer5/F1_bias'))
fc1_mul = matmul(result_l4_flat, W_fc1)
fc1_result_relu = relu(fc1_mul + b_fc1)
result_l5 = fc1_result_relu
write_file(fout,'result Layer 5',result_l5)
#第6层：全连接 84 ->84
W_fc2 = change_int(reader.get_tensor('Layer6/F2_weight'))
b_fc2 = change_int(reader.get_tensor('Layer6/F2_bias'))
fc2_mul = matmul(result_l5, W_fc2)
fc2_result_relu = relu(fc2_mul + b_fc2)
result_l6 = fc2_result_relu
write_file(fout,'result Layer 6',result_l6)
#第7层：全连接 10 ->10
W_fc3 = change_int(reader.get_tensor('Layer7/F3_weight'))
b_fc3 = change_int(reader.get_tensor('Layer7/F3_bias'))
fc3_mul = matmul(result_l6, W_fc3)
result_l7 = relu(fc3_mul + b_fc3)
write_file(fout,'result Layer 7',result_l7)
fout.close()