Tensorflow 学习-图的概念

1
2
3
4
5
6
7
8
9
import tensorflow as tf  # tf是tensorflow别名
g1 = tf.Graph()  # 声明图
g2 = tf.get_default_graph()  # 获取当前默认的图

a = tf.constant([1.0, 2.0], name='a')  # 定义一个常量使用tf.constant方法
g3 = a.graph  # 通过a.graph可以查看张量所属的计算图，如果没有特别指定，则属于当前默认的计算图

print(g1, 'n', g2, 'n', g3)

1
2
3
4
<tensorflow.python.framework.ops.Graph object at 0x000001F6F2BCAEF0> 
 <tensorflow.python.framework.ops.Graph object at 0x000001F6F29F8C18> 
 <tensorflow.python.framework.ops.Graph object at 0x000001F6F29F8C18>

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
import tensorflow as tf

# 如果不加下面这段，会出现 The TensorFlow library wasn’t compiled to use SSE instructions, but these are available on your machine and could speed up CPU computations 警告
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

g1 = tf.Graph()
with g1.as_default():  # 声明图的作用域
    # 在计算图g1中定义常量'v1',并设置初始值为1.0。
    v1 = tf.constant(1.0, name="v1")

g2 = tf.Graph()
with g2.as_default():
    # 在计算图g2中定义常量'v2',并设置初始值微2.0。
    v2= tf.constant(2.0, name="v2")

with tf.Session(graph=g2) as sess:
    list = tf.import_graph_def(g1.as_graph_def(), return_elements=["v1:0"], name = '')
    print(sess.run(list[0] + v2))

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
import tensorflow as tf

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

g1 = tf.Graph()
with g1.as_default():
    # 在计算图g1中定义变量'v',并设置初始值为0。
    v = tf.get_variable('v', initializer=tf.zeros_initializer()(shape=[1]))

g2 = tf.Graph()
with g2.as_default():
    # 在计算图g2中定义变量'v',并设置初始值微1。
    v = tf.get_variable('v', initializer=tf.ones_initializer()(shape=[1]))

# 在计算图g1中读取变量'v'的取值
with tf.Session(graph=g1) as sess:
    tf.global_variables_initializer().run()
    with tf.variable_scope('', reuse=True):
        # 在计算图g1中，变量'v'的取值应该为0，下一行代码会输出[0.]。
        print(sess.run(tf.get_variable('v')))

# 在计算图g2中读取变量'v'的取值
with tf.Session(graph=g2) as sess:
    tf.global_variables_initializer().run()
    with tf.variable_scope('', reuse=True):
        # 在计算图g2中，变量'v'的取值应该为1，下一行代码会输出[1.]。
        print(sess.run(tf.get_variable('v')))

1
2
3
4
g1 = tf.Graph()

tf.train.write_graph(g1.as_graph_def(), '.', 'graph.pb', False)

1
2
3
4
5
6
7
8
9
with tf.gfile.FastGFile("graph.pb", 'rb') as f:
    graph_def = tf.GraphDef()
    graph_def.ParseFromString(f.read())
    tf.import_graph_def(graph_def, name='')

sess = tf.Session()
cl_tensor = sess.graph.get_tensor_by_name("cl:0")
cl = sess.run(cl_tensor)

1
2
3
4
5
g=tf.Graph()
# 指定计算运行的设备。
with g.device('/gpu:0'):
    result=a+b

1
2
3
4
5
6
7
8
tf.constant(
    value, 
    dtype=None,
    shape=None,
    name='Const',
    verify_shape=False
)

1
2
3
4
sess=tf.Session()
with sess.as_default():
    print(tensor.eval())

1
2
3
4
5
6
tensor=tf.constant([2, 4])
sess=tf.Session()
with sess.as_default():
    print( tensor.eval())

1
2
3
4
5
tensor=tf.constant([1, 2], dtype=tf.float32)
sess=tf.Session()
with sess.as_default():
    print(tensor.eval())

1
2
3
4
5
tensor=tf.constant(1, shape=[2, 3])
sess=tf.Session()
with sess.as_default():
    print(tensor.eval())

1
2
tensor=tf.constant([1, 2, 3, 4, 5, 6, 7], shape=[2, 3])

1
2
3
4
5
tensor=tf.constant([1, 2], shape=[1, 4, 3])
sess=tf.Session()
with sess.as_default():
    print(tensor.eval())

1
2
3
tensor=tf.constant([1, 2])
print(tensor)

1
2
3
tensor=tf.constant([1, 2], name="jiayu")
print(tensor)

1
2
tensor=tf.constant([[1, 2, 3], [4, 5, 6]], shape=[2, 3], verify_shape=True)

1
2
tensor=tf.constant([1, 2], shape=[3, 2], verify_shape=True)