如何在 Tensorflow 中增加一个变量? [英] How to Increment a Variable in Tensorflow?

问题描述

当我尝试在 Tensorflow 中使用主管时，我意识到:

When trying to use the supervisor in Tensorflow I was made aware that :

您的训练操作负责增加全局步长值.

your training op is responsible for incrementing the global step value.

(参考)

那么如何在 Tensorflow 中的图表中增加一个变量?

So how do you increment a variable in a graph in Tensorflow?

推荐答案

非常简单的解决方案:

global_step = tf.Variable(1, name='global_step', trainable=False, dtype=tf.int32)
increment_global_step_op = tf.assign(global_step, global_step+1)

然后当你想增加它时，只需在当前 tf.Session sess 下运行该操作.

Then when you want to increment it, just run that op under the current tf.Session sess.

step = sess.run(increment_global_step_op)

放在step中的结果是自增后的自增变量的值.在这种情况下，global_step 的值增加后.所以2.

The result placed in step is the value of the incremented variable after the increment. In this case, the value of global_step after being incremented. So 2.

如果您像我一样将其用于 global_step，请将其与您的 training_op 一起运行.

If you're using this for global_step like me, run it along with your training_op.

result = sess.run([out, increment_global_step_op], {x: [i]})

这篇关于如何在 Tensorflow 中增加一个变量?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！