Question

对于以下代码，我可以对n使用占位符，并通过进给来传递n吗？我是tensorflow的新手。

n = int(input("Enter an integer: "))

one=tf.constant(1)
#increase
increasing_value=tf.Variable(0,name="increasing_value")
increasing_op=tf.assign_add(increasing_value,one)

#sum
sumvalue=tf.Variable(0,name="sumvalue")
sum_op=tf.assign_add(sumvalue,increasing_value)

init=tf.global_variables_initializer()

with tf.Session() as session:
    session.run(init)
    for _ in range (n):
        session.run(increasing_op)
        session.run(sum_op)
    print(session.run(sumvalue))

Answer 1

你可以试试吗？

n = tf.placeholder(tf.int32, name='n')

fedvalue = session.run( n , feed_dict = { n : 10 })
for _ in range ( fedvalue ):
    session.run(increasing_op)
    session.run(sum_op)

Answer 2

您可以使用tf.while_loop在TensorFlow中复制代码。

import tensorflow as tf

n = tf.placeholder(tf.int32, [])
increasing_value = tf.constant(0, dtype=tf.int32)
sum_value = tf.constant(0, dtype=tf.int32)

def loop_body(i, increasing_value, sum_value):
    increased_value = increasing_value + 1
    return i + 1, increased_value, sum_value + increased_value

i = tf.constant(0, dtype=tf.int32)
_, increasing_value, sum_value = tf.while_loop(
    lambda i, _, __: i < n,
    loop_body,
    [i, increasing_value, sum_value])

with tf.Session() as session:
    for x in range(10):
        print(session.run(sum_value, feed_dict={n: x}))

输出：

但是，TensorFlow循环通常很慢，您应该尝试找到进行计算的矢量方式（也就是说，将其表示为对值和归约数组的操作）。在您的特定情况下，您的代码仅在计算1 + 2 + 3 + 4 + ⋯，这就是(n * (n + 1)) / 2：

import tensorflow as tf

n = tf.placeholder(tf.int32, [])
sum_value = (n * n + n) // 2

with tf.Session() as session:
    for x in range(10):
        print(session.run(sum_value, feed_dict={n: x}))

输出：

在以下情况下如何在张量流中使用占位符

2 个答案: