详解tf.device()指定tensorflow运行的GPU或CPU设备实现

import tensorflow as tf
 
with tf.device('/gpu:1'):
  v1 = tf.constant([1.0, 2.0, 3.0], shape=[3], name='v1')
  v2 = tf.constant([1.0, 2.0, 3.0], shape=[3], name='v2')
  sumV12 = v1 + v2
 
  with tf.Session(config=tf.ConfigProto(log_device_placement=True)) as sess:
    print sess.run(sumV12)

import tensorflow as tf
 
v1 = tf.constant([1.0, 2.0, 3.0], shape=[3], name='v1')
v2 = tf.constant([1.0, 2.0, 3.0], shape=[3], name='v2')
sumV12 = v1 + v2
 
with tf.Session(config=tf.ConfigProto(log_device_placement=True)) as sess:
  print sess.run(sumV12)

import tensorflow as tf
 
with tf.device('/cpu:0'):
  v1 = tf.constant([1.0, 2.0, 3.0], shape=[3], name='v1')
  v2 = tf.constant([1.0, 2.0, 3.0], shape=[3], name='v2')
  sumV12 = v1 + v2
 
  with tf.Session(config=tf.ConfigProto(log_device_placement=True)) as sess:
    print sess.run(sumV12)