如何重塑具有多个“无"维度的张量? [英] How to reshape a tensor with multiple `None` dimensions?

问题描述

我遇到了将中间 4D 张量流张量 X 重塑为 3D 张量 Y 的问题，其中

I encountered a problem to reshape an intermediate 4D tensorflow tensor X to a 3D tensor Y, where

• X 的形状是 ( batch_size, nb_rows, nb_cols, nb_filters )
• Y 的形状是 ( batch_size, nb_rows*nb_cols, nb_filters )
• batch_size = None

• X is of shape ( batch_size, nb_rows, nb_cols, nb_filters )
• Y is of shape ( batch_size, nb_rows*nb_cols, nb_filters )
• batch_size = None

当然，当nb_rows 和nb_cols 是已知整数时，我可以毫无问题地重塑X.但是，在我的应用程序中，我需要处理这种情况

Of course, when nb_rows and nb_cols are known integers, I can reshape X without any problem. However, in my application I need to deal with the case

nb_rows = nb_cols = None

我该怎么办?我尝试了 Y = tf.reshape( X, (-1, -1, nb_filters)) 但它显然无法正常工作.

What should I do? I tried Y = tf.reshape( X, (-1, -1, nb_filters)) but it clearly fails to work.

对我来说，这个操作是确定性的，因为它总是将两个中间轴压缩成一个，同时保持第一个轴和最后一个轴不变.有人可以帮我吗?

For me, this operation is deterministic because it always squeezes the two middle axes into a single one while keeping the first axis and the last axis unchanged. Can anyone help me?

推荐答案

此时可以通过tf.shape(X)访问X的动态形状:

In this case you can access to the dynamic shape of X through tf.shape(X):

shape = [tf.shape(X)[k] for k in range(4)]
Y = tf.reshape(X, [shape[0], shape[1]*shape[2], shape[3]])

这篇关于如何重塑具有多个“无"维度的张量?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！