122、TensorFlow多设备运行

# 如果你想使你的程序运行在不同的设备上
# tf.device函数提供了一个方便的方法来实现
# 所有在特定上下文中的操作都放置在相同的设备上面
# A device specification has the following form:
#              /job:<JOB_NAME>/task:<TASK_INDEX>/device:<DEVICE_TYPE>:<DEVICE_INDEX>
# <JOB_NAME> 是一个字母数字字符串，并且不以数字开头
# <DEVICE_TYPE> GPU or CPU
# <TASK_INDEX>是一个非负值，表示在一个job中的第几个task
# <DEVICE_INDEX> 表示第几个设备
# operation created outside either context will run on the "best possible" device
# For example , if you have a GPU and a CPU avaiable , and the operation
# has a GPU implementation , TensorFlow will choose the GPU
import tensorflow as tf
weights = tf.random_normal([2, 2, 3])
with tf.device("/device:CPU:0"):
    # operations created in this context will be pinned to the CPU
    img = tf.cast(tf.image.decode_jpeg(tf.read_file("img.jpg")), dtype=tf.float32)

with tf.device("/device:GPU:0"):
    # Operations created in this context will be pinned to the GPU
    result = tf.matmul(weights, img)