[人工智能]TensorFlow 从入门到精通（0）—— tensorflow编程基础

import tensorflow as tf

一、初体验

node1 = tf.constant([[1,2,3],[4,5,6]])
node2 = tf.constant([[1,2,3],[4,5,6]])
node3 = tf.add(node1,node2)

node3

<tf.Tensor: shape=(2, 3), dtype=int32, numpy=
array([[ 2,  4,  6],
       [ 8, 10, 12]], dtype=int32)>

# 通过.numpy()获取值
node3.numpy()

array([[ 2,  4,  6],
       [ 8, 10, 12]], dtype=int32)

node3.dtype

tf.int32

node3.shape

TensorShape([2, 3])

node3.get_shape()

TensorShape([2, 3])

二、常量

a = tf.constant(1)
a

<tf.Tensor: shape=(), dtype=int32, numpy=1>

b = tf.constant(2.0)
b

<tf.Tensor: shape=(), dtype=float32, numpy=2.0>

# 强制类型转换
a = tf.cast(a,tf.float32)

tf.add(a,b).numpy()

3.0

三、变量

v1 = tf.Variable([1,2])
v1

<tf.Variable 'Variable:0' shape=(2,) dtype=int32, numpy=array([1, 2], dtype=int32)>

v2 = tf.Variable([3,4],tf.float32)

v2

<tf.Variable 'Variable:0' shape=(2,) dtype=int32, numpy=array([3, 4], dtype=int32)>

tf.Variable(a)

<tf.Variable 'Variable:0' shape=() dtype=float32, numpy=1.0>

v2

<tf.Variable 'Variable:0' shape=(2,) dtype=int32, numpy=array([3, 4], dtype=int32)>

# 更改变量的值
v2.assign(v2+1)

<tf.Variable 'UnreadVariable' shape=(2,) dtype=int32, numpy=array([4, 5], dtype=int32)>

v2.assign_add([2,1])

<tf.Variable 'UnreadVariable' shape=(2,) dtype=int32, numpy=array([6, 6], dtype=int32)>

v2.assign_sub([1,2])

<tf.Variable 'UnreadVariable' shape=(2,) dtype=int32, numpy=array([5, 4], dtype=int32)>

四、tf2执行tf1代码

import tensorflow.compat.v1 as tf

tf.disable_eager_execution()

node1 = tf.constant(3)
node2 = tf.constant(4)
node3 = node1 + node2
node3

<tf.Tensor 'add:0' shape=() dtype=int32>

sess = tf.Session()
sess.run(node3)

7

sess.close()

五、会话模式

5.1三种对话模式

# 1.会话模式一
tens1 = tf.constant([1])
sess = tf.Session()
print(sess.run(tens1))
sess.close()

[1]

# 2.会话模式二
tens1 = tf.constant(1)
sess = tf.Session()
try:
    print(sess.run(tens1))
except:
    ...
finally:
    sess.close()

1

# 3.会话模式三
tens1 = tf.constant(1)

with tf.Session() as sess:
    print(sess.run(tens1))

1

5.2指定默认的对话

• 指定了默认会话可以通过tf.Tensor.eval函数来计算一个张量的取值

tens1 = tf.constant(1)
sess = tf.Session()
with sess.as_default():
    print(tens1.eval())

1

# 下面代码和上面同理
tens1 = tf.constant(1)
sess = tf.Session()
print(sess.run(tens1))
print(tens1.eval(session=sess))

1
1

# tf.InteractiveSession() 这个函数会自动将生成的会话注册为默认会话
tens1 = tf.constant(1)
sess = tf.InteractiveSession()
print(tens1.eval())

1

六、变量的初始化

node1 = tf.Variable(3.0)
node2 = tf.Variable(4.0)
result = node1+node2 # tf.add(node1,node2)

node1_init = node1.initializer
node2_init = node2.initializer

sess = tf.Session()
sess.run(node1_init)
sess.run(node2_init)
print(sess.run([result,node1,node2]))

[7.0, 3.0, 4.0]

# 可以对变量进行批量初始化
node1 = tf.Variable(3.0)
node2 = tf.Variable(4.0)
result = node1+node2
init = tf.global_variables_initializer()
sess.run(init)
print(sess.run(result))

7.0

七、占位符 placeholder

• 类似于python中的%s

x = tf.placeholder(tf.float32)

x_value = float(input())

 123

with tf.Session() as sess:
    result = sess.run(x,feed_dict={x:x_value})
    print(result)

123.0