import tensorflow as tf import numpy as np tf.compat.v1.disable_eager_execution() x = tf.compat.v1.placeholder(shape = [1,10], dtype = tf.float32, name = ‘x‘) #placeholder 占位符没有初始值，图执行时临时赋值 W = tf.compat.v1.Variable(tf.compat.v1.random_uniform(shape = [10,5], minval = -0.1, maxval = 0.1, dtype = tf.float32),name = ‘W‘) b = tf.compat.v1.Variable(tf.zeros(shape = [5], dtype = tf.float32), name = ‘b‘) #variable 变量需要初始化random_uniform在minval-maxval之间均匀采样 h = tf.nn.sigmoid(tf.matmul(x,W)+b) with tf.compat.v1.Session() as sess: sess.run(tf.compat.v1.global_variables_initializer()) print(sess.run(h,feed_dict={x:np.random.rand(1,10)}))
[[0.52614367 0.5181997 0.5162186 0.48906413 0.46755365]]import tensorflow as tfimport numpy as np
tf.compat.v1.disable_eager_execution() graph = tf.Graph()#create a graph session = tf.compat.v1.InteractiveSession(graph=graph)#create a session,version 1 x = tf.compat.v1.placeholder(shape = [1,10], dtype = tf.float32, name = ‘x‘) #placeholder 占位符没有初始值，图执行时临时赋值 W = tf.compat.v1.Variable(tf.compat.v1.random_uniform(shape = [10,5], minval = -0.1, maxval = 0.1, dtype = tf.float32),name = ‘W‘) b = tf.compat.v1.Variable(tf.zeros(shape = [5], dtype = tf.float32), name = ‘b‘) #variable 变量需要初始化random_uniform在minval-maxval之间均匀采样 h = tf.nn.sigmoid(tf.matmul(x,W)+b)
初始化图中所有变量
tf.compat.v1.global_variables_initializer().run()
（等价于W.initializer.run() b.initializer.run()）
h_eval= session.run(h,feed_dict={x:np.random.rand(1,10)}) print(h_eval)
[[0.54421955 0.5040116 0.5199564 0.5172269 0.54251057]]
tf.Session() 与 tf.InteractiveSession()的区别
tf.InteractiveSession 把它自己当作默认会话，tf.Tensor.eval、tf.Operation.run 方法都可以在当前会话中进行操作运算。
举个栗子
sess = tf.InteractiveSession() a = tf.constant(5.0) b = tf.constant(6.0) c = a * b # We can just use ‘c.eval()‘ without passing ‘sess‘ print(c.eval()) sess.close()
等价于
a = tf.constant(5.0) b = tf.constant(6.0) c = a * b with tf.Session(): # We can also use ‘c.eval()‘ here. print(c.eval())
原文：https://www.cnblogs.com/jessyswing/p/12918222.html