TensorFlow 学习（二）—— tf Graph tf Session 与 tf Session run

• session：

  • with tf.Session() as sess:/ tf.InteractiveSession()

  • 初始化：

    • tf.global_variables_initializer()

    with tf.Session() as sess:
    	sess.run(tf.global_variables_initializer())
    

    • 1
    • 2

0. tf.Graph

• 命名空间与 operation name（oper.name 获取操作名）：

  c_0 = tf.constant(0, name="c")  # => operation named "c"
  print(c_0.name)
  # Already-used names will be "uniquified".
  c_1 = tf.constant(2, name="c")  # => operation named "c_1"
  
  # Name scopes add a prefix to all operations created in the same context.
  with tf.name_scope("outer"):
      c_2 = tf.constant(2, name="c")  # => operation named "outer/c"
      # Name scopes nest like paths in a hierarchical file system.
      with tf.name_scope("inner"):
          c_3 = tf.constant(3, name="c")  # => operation named "outer/inner/c"
  
      # Exiting a name scope context will return to the previous prefix.
      c_4 = tf.constant(4, name="c")  # => operation named "outer/c_1"
  
      # Already-used name scopes will be "uniquified".
      with tf.name_scope("inner"):
          c_5 = tf.constant(5, name="c")  # => operation named "outer/inner_1/c"
  

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18

1. 使用 tf.Session().run() 读取变量的值十分耗时

#CODING: UTF-8
import time
import tensorflow as tf

N = 100000
x = tf.constant([1.])
b = 1.

with tf.Session() as sess:
	sess.run(tf.initialize_all_variables()) 

	t1 = time.time()
	for _ in range(N):
		y = sess.run(x)
	print(‘使用sess.run() 读取变量数据耗时‘, time.time()-t1)

	t2 = time.time()
	for _ in range(N):
		a = b
	print(‘直接赋值耗时‘, time.time()-t2)

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

2. tf.Session().run() 与 Tensor.eval()

假设 x 为 tf 下的一个 Tensor 对象，t.eval() 执行的动作就是 tf.Session().run(t) 。

import tensorflow as tf
x = tf.constant([5.])

print(tf.Session().run(x))

with tf.Session():
	print(x.eval())

• 1
• 2
• 3
• 4
• 5
• 6
• 7

在第二个例子中，session的作用就象context manager，context manager在with块的生存期，将session作为默认的 session。对简单应用的情形（如单元测试），context manager的方法可以得到更简洁的代码；如果你的代码要处理多个graph和 session，更直白的方式可能是显式调用Session.run()。

• session.run()是非常耗时的,千万不要用session.run的方式去取数据
• Session.run()和Tensor.eval()区别是什么？