TensorFlow 中张量值的条件赋值 [英] Conditional assignment of tensor values in TensorFlow

问题描述

我想在 tensorflow 中复制以下 numpy 代码.例如，我想为之前值为 1 的所有张量索引分配一个 0.

I want to replicate the following numpy code in tensorflow. For example, I want to assign a 0 to all tensor indices that previously had a value of 1.

a = np.array([1, 2, 3, 1])
a[a==1] = 0

# a should be [0, 2, 3, 0]

如果我在 tensorflow 中编写类似的代码，我会收到以下错误.

If I write similar code in tensorflow I get the following error.

TypeError: 'Tensor' object does not support item assignment

方括号中的条件应该是任意的，如a[a<1] = 0.

The condition in the square brackets should be arbitrary as in a[a<1] = 0.

有没有办法在tensorflow中实现这种条件赋值"(因为没有更好的名字)?

Is there a way to realize this "conditional assignment" (for lack of a better name) in tensorflow?

推荐答案

比较运算符，例如 大于 在 TensorFlow API 中可用.

Comparison operators such as greater than are available within TensorFlow API.

然而，在直接操作张量方面，没有什么能与简洁的 NumPy 语法等效.您必须使用单独的 comparison、where 和 assign 运算符来执行相同的操作.

However, there is nothing equivalent to the concise NumPy syntax when it comes to manipulating the tensors directly. You have to make use of individual comparison, where and assign operators to perform the same action.

与您的 NumPy 示例等效的代码如下:

Equivalent code to your NumPy example is this:

import tensorflow as tf

a = tf.Variable( [1,2,3,1] )    
start_op = tf.global_variables_initializer()    
comparison = tf.equal( a, tf.constant( 1 ) )    
conditional_assignment_op = a.assign( tf.where (comparison, tf.zeros_like(a), a) )

with tf.Session() as session:
    # Equivalent to: a = np.array( [1, 2, 3, 1] )
    session.run( start_op )
    print( a.eval() )    
    # Equivalent to: a[a==1] = 0
    session.run( conditional_assignment_op )
    print( a.eval() )

# Output is:
# [1 2 3 1]
# [0 2 3 0]

打印语句当然是可选的，它们只是为了证明代码正确执行.

The print statements are of course optional, they are just there to demonstrate the code is performing correctly.

这篇关于TensorFlow 中张量值的条件赋值的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！