TensorFlow的计算都可以通过计算图的模型来建立,而计算图上的每一个节点代表了一个计算,计算的结果就保存在张量中,所以张量和计算图上节点所代表的计算结果是对应的,这样张量的命名就可以通过“node:src_output"的形式来给出,其中node为节点的名称,src_output表示当前张量来自节点的第几个输出。
详见示例:
import tensorflow as tf
g = tf.Graph()
with g.as_default():
a = tf.Variable(initial_value=tf.ones(shape=[1,2], dtype=tf.float32),trainable=False,name='a')
b = tf.constant([1.0, 2.0],name = 'b')
result = tf.add(a, b, name = 'add')
result1 = tf.add(a, b, name = 'add')
result2 = tf.add(a, b, name = 'add')
# 或者写为(二者是等价的)
'''
result = a + b
result1 = a + b
result2 = a + b
'''
with tf.Session(graph=g) as sess:
sess.run(tf.global_variables_initializer())
print(sess.run(result)) # [2. 3.]]
print(result) # Tensor("add:0", shape=(1, 2), dtype=float32)
print(result1) # Tensor("add_1:0", shape=(1, 2), dtype=float32)
print(result2) # Tensor("add_2:0", shape=(1, 2), dtype=float32)
f = tf.Graph()
with f.as_default():
a = tf.Variable(initial_value=tf.ones(shape=[1,2], dtype=tf.float32),trainable=False,name='a')
b = tf.constant([1.0, 2.0],name = 'b')
result = tf.add(a, b, name = 'add')
result1 = tf.add(a, b, name = 'add1')
result2 = tf.add(a, b, name = 'add2')
with tf.Session(graph=f) as sess:
sess.run(tf.global_variables_initializer())
print(sess.run(result)) # [2. 3.]]
print(result) # Tensor("add:0", shape=(1, 2), dtype=float32)
print(result1) # Tensor("add1:0", shape=(1, 2), dtype=float32)
print(result2) # Tensor("add2:0", shape=(1, 2), dtype=float32)
Reference
- 郑泽宇等.TensorFLow实战Google深度学习框架(第2版),电子工业出版社,2018.