最近看《面向机器智能tensorflow实践》这本书,里面提到在一份代码中实现多个graph。
下面是一种实现方式,适合较短代码段
import tensorflow as tf
#c = tf.add(a, b, "add")
g1 = tf.Graph()
g2 = tf.Graph()
with g1.as_default():
sess = tf.Session()
a = tf.constant([5], name="a")
b = tf.constant([3], name="b")
c = tf.add(a, b, "add_in_g1")
summery_writer = tf.summary.FileWriter("./train", sess.graph)
summery_writer.close()
print(sess.run(c))
sess.close()
with g2.as_default():
sess = tf.Session()
a = tf.constant([5], name="a")
b = tf.constant([3], name="b")
c = tf.add(a, b, "add_in_g2")
summery_writer = tf.summary.FileWriter("./test", sess.graph)
summery_writer.close()
print(sess.run(c))
sess.close()
下面是另一种方式实现,每次声明默认计算图,使用完毕后关掉计算图
import tensorflow as tf
a = tf.constant([5], name="a")
b = tf.constant([3], name="b")
#c = tf.add(a, b, "add")
g1 = tf.Graph().as_default()
sess = tf.Session()
c = tf.add(a, b, "add")
summery_writer = tf.summary.FileWriter("./train", sess.graph)
summery_writer.close()
print(sess.run(c))
sess.close()
g2 = tf.Graph().as_default()
sess = tf.Session()
d = tf.multiply(a, b, name="multiply")
summery_writer = tf.summary.FileWriter("./test", sess.graph)
print(sess.run(d))
summery_writer.close()
sess.close()
在第二幅图中出现了第一幅图中的add操作,我认为是因为这些op本身是在默认计算图中声明的,当把当前计算图作为默认图时,上述的op就也跟着显示了。