回答編集履歴
1
データ作成~表示までのサンプル追加
test
CHANGED
@@ -39,3 +39,119 @@
|
|
39
39
|
:(以下略)
|
40
40
|
|
41
41
|
```
|
42
|
+
|
43
|
+
|
44
|
+
|
45
|
+
2017/03/15追記 : データ作成~表示までの実行例
|
46
|
+
|
47
|
+
--
|
48
|
+
|
49
|
+
|
50
|
+
|
51
|
+
- 可視化したいデータ作成
|
52
|
+
|
53
|
+
参考:[TensorBoardで処理を可視化する](http://www.mwsoft.jp/programming/tensor/tutorial_tensorboad.html)
|
54
|
+
|
55
|
+
```Python
|
56
|
+
|
57
|
+
# -*- coding: utf-8 -*-
|
58
|
+
|
59
|
+
import tensorflow as tf
|
60
|
+
|
61
|
+
|
62
|
+
|
63
|
+
# 足し算
|
64
|
+
|
65
|
+
with tf.name_scope('add_scope'):
|
66
|
+
|
67
|
+
x = tf.constant(1, name='x')
|
68
|
+
|
69
|
+
y = tf.constant(2, name='y')
|
70
|
+
|
71
|
+
z = x + y
|
72
|
+
|
73
|
+
|
74
|
+
|
75
|
+
# 上の結果に掛け算
|
76
|
+
|
77
|
+
with tf.name_scope('multiply_scope'):
|
78
|
+
|
79
|
+
zz = y * z
|
80
|
+
|
81
|
+
|
82
|
+
|
83
|
+
with tf.Session() as sess:
|
84
|
+
|
85
|
+
with tf.name_scope('init_scope'):
|
86
|
+
|
87
|
+
# WARNING:tensorflow:From hoge.py initialize_all_variables (from tensorflow.python.ops.variables) is deprecated and will be removed after 2017-03-02.
|
88
|
+
|
89
|
+
# Instructions for updating:
|
90
|
+
|
91
|
+
# Use `tf.global_variables_initializer` instead.
|
92
|
+
|
93
|
+
#sess.run(tf.initialize_all_variables())
|
94
|
+
|
95
|
+
sess.run(tf.global_variables_initializer())
|
96
|
+
|
97
|
+
sess.run(zz)
|
98
|
+
|
99
|
+
|
100
|
+
|
101
|
+
# グラフを書こう
|
102
|
+
|
103
|
+
# AttributeError: module 'tensorflow.python.training.training' has no attribute 'SummaryWriter' #8164
|
104
|
+
|
105
|
+
# https://github.com/tensorflow/tensorflow/issues/8164
|
106
|
+
|
107
|
+
#summary_writer = tf.train.SummaryWriter('data', graph=sess.graph)
|
108
|
+
|
109
|
+
summary_writer = tf.summary.FileWriter('c:\\temp\\tf_log', graph=sess.graph)
|
110
|
+
|
111
|
+
|
112
|
+
|
113
|
+
# http://stackoverflow.com/questions/41066244/tensorflow-module-object-has-no-attribute-scalar-summary
|
114
|
+
|
115
|
+
#tf.scalar_summary('one_plus_one_summary', zz)
|
116
|
+
|
117
|
+
tf.summary.scalar('one_plus_one_summary', zz)
|
118
|
+
|
119
|
+
|
120
|
+
|
121
|
+
print("done.")
|
122
|
+
|
123
|
+
```
|
124
|
+
|
125
|
+
`c:\temp\tf_log`配下にデータ`events.out.tfevents.1489539251.hoge-PC`が作成される。
|
126
|
+
|
127
|
+
|
128
|
+
|
129
|
+
- tensorboard起動
|
130
|
+
|
131
|
+
コマンドプロンプト(cmd.exe)上で以下を実行。
|
132
|
+
|
133
|
+
```lang=DOS
|
134
|
+
|
135
|
+
C:\Windows\system32>activate tf3
|
136
|
+
|
137
|
+
(tf3) C:\Windows\system32>tensorboard --logdir=c:/temp/tf_log
|
138
|
+
|
139
|
+
Starting TensorBoard b'41' on port 6006
|
140
|
+
|
141
|
+
(You can navigate to http://172.xx.xx.xx:6006)
|
142
|
+
|
143
|
+
ERROR:tensorflow:Unable to get size of ~
|
144
|
+
|
145
|
+
など複数行のエラー、警告が表示されるがとりあえず無視
|
146
|
+
|
147
|
+
WARNING:tensorflow:Found more than one metagraph event per run. Overwriting the
|
148
|
+
|
149
|
+
metagraph with the newest event.
|
150
|
+
|
151
|
+
```
|
152
|
+
|
153
|
+
|
154
|
+
|
155
|
+
- ブラウザ(chrome)で`http://localhost:6006`を開く
|
156
|
+
|
157
|
+

|