介绍 TensorFlow 中启动图: tf.Session(),tf.InteractivesSession(),tf.train.Supervisor().managed_session() 用法的区别:
(1)tf.Session()
构造阶段完成后, 才能启动图. 启动图的第一步是创建一个 Session
对象, 如果无任何创建参数, 会话构造器将启动默认图.
具体代码:
- #coding:utf-8
- import tensorflow as tf
- matrix1 = tf.constant([[3., 3.]])
- matrix2 = tf.constant([[2.], [2.]])
- preduct = tf.matmul(matrix1, matrix2)
- # 使用 "with" 代码块来自动完成关闭动作.
- with tf.Session() as sess:
- sess.run(tf.global_variables_initializer())
- print sess.run(preduct)
#coding:utf-8
import tensorflow as tf
matrix1 = tf.constant([[3., 3.]])
matrix2 = tf.constant([[2.], [2.]])
preduct = tf.matmul(matrix1, matrix2)
# 使用 "with" 代码块来自动完成关闭动作.
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
print sess.run(preduct)
(2)tf.InteractivesSession()
为了便于使用诸如 IPython之类的 Python 交互环境, 可以使用InteractiveSession 代替
Session
类, 使用 Tensor.eval()
和
Operation.run()
方法代替
Session.run()
. 这样可以避免使用一个变量来持有会话。
具体代码:
- #coding:utf-8
- import tensorflow as tf
- matrix1 = tf.constant([[3., 3.]])
- matrix2 = tf.constant([[2.], [2.]])
- preduct = tf.matmul(matrix1, matrix2)
- sess_ = tf.InteractiveSession()
- tf.global_variables_initializer().run()
- print preduct.eval()
- sess_.close()
#coding:utf-8
import tensorflow as tf
matrix1 = tf.constant([[3., 3.]])
matrix2 = tf.constant([[2.], [2.]])
preduct = tf.matmul(matrix1, matrix2)
sess_ = tf.InteractiveSession()
tf.global_variables_initializer().run()
print preduct.eval()
sess_.close()
(3) tf.train.Supervisor().managed_session()
与上面两种启动图相比较来说,Supervisor() 帮助我们处理一些事情:
(a) 自动去 checkpoint 加载数据或者初始化数据
(b) 自动有一个 Saver ,可以用来保存 checkpoint
eg: sv.saver.save(sess, save_path)
(c) 有一个 summary_computed 用来保存 Summary
因此我们可以省略了以下内容:
(a)手动初始化或者从 checkpoint 中加载数据
(b)不需要创建 Saver 类, 使用 sv 内部的就可以
(c)不需要创建 Summary_Writer()
具体代码:
- import tensorflow as tf
- matrix1 = tf.constant([[3., 3.]])
- matrix2 = tf.constant([[2.], [2.]])
- preduct = tf.matmul(matrix1, matrix2)
- sv = tf.train.Supervisor(logdir=None, init_op=tf.global_variables_initializer())
- with sv.managed_session() as sess:
- print sess.run(preduct)
import tensorflow as tf
matrix1 = tf.constant([[3., 3.]])
matrix2 = tf.constant([[2.], [2.]])
preduct = tf.matmul(matrix1, matrix2)
sv = tf.train.Supervisor(logdir=None, init_op=tf.global_variables_initializer())
with sv.managed_session() as sess:
print sess.run(preduct)
另外一个栗子:
- #coding:utf-8
- import tensorflow as tf
- a = tf.Variable(1)
- b = tf.Variable(2)
- c = tf.add(a, b)
- update = tf.assign(a, c)
- init = tf.global_variables_initializer()
- sv = tf.train.Supervisor(logdir="./tmp/", init_op=init)
- saver = sv.saver
- with sv.managed_session() as sess:
- for i in range(1000):
- update_ = sess.run(update)
- #print("11111", update)
- if i % 100 == 0:
- sv.saver.save(sess, "./tmp/", global_step=i)
#coding:utf-8
import tensorflow as tf
a = tf.Variable(1)
b = tf.Variable(2)
c = tf.add(a, b)
update = tf.assign(a, c)
init = tf.global_variables_initializer()
sv = tf.train.Supervisor(logdir="./tmp/", init_op=init)
saver = sv.saver
with sv.managed_session() as sess:
for i in range(1000):
update_ = sess.run(update)
#print("11111", update)
if i % 100 == 0:
sv.saver.save(sess, "./tmp/", global_step=i)
如何使用 Supervisor() 来启动图和保存训练参数