使用tensorflow js从2D张量获取数据 [英] Get data from 2D tensor with tensorflow js

问题描述

我想通过 tensorflow.js 从2D张量中获取数据。我尝试使用 data（）方法，如下所示：

I would like to get the data from a 2D tensor with tensorflow.js. I tried to use the data() method like this:

const X = tf.tensor2d([1, 2, 3, 4], [2, 2, 5, 3]);
X.data().then(X => console.log(X)};

但结果是一个扁平的1D数组：

But the result is a flatten 1D array:

Float32Array(8) [1, 2, 3, 4, 2, 2, 5, 3]

有没有办法保持阵列的形状？

Is there a way to keep the shape of the array?

推荐答案

为了提高速度，Tensor中的数据总是被平铺为类型1维数组。

Data in the Tensor is always stored flattened as types 1 dimensional array, for speed.

你给出的例子不起作用，因为 tensor2d 的第二个参数是 shape 。工作你需要包装另一个数组：

The example you gave will not work, because 2nd parameter to tensor2d is shape. To make it work you either need to wrap it another array:

const x = tf.tensor2d([[1, 2, 3, 4], [2, 2, 5, 3]]); //shape inferred as [2, 4]

或者你可以明确地提供形状：

or you could explicitly provide shape:

const x = tf.tensor2d([1, 2, 3, 4, 2, 2, 5, 3], [2, 4]); // shape explicitly passed

正如您所建议的，当您检查数据时，您将始终获得一维数组，无论原始形状如何

as you suggested though, when you inspect data you will always get 1 dimensional array, regardless of original shape

await x.data() // Float32Array(8) [1, 2, 3, 4, 2, 2, 5, 3]
x.shape // [2, 4]

但是如果你 print（）你的张量，形状会被考虑在内，它会显示为

if however you print() your tensor, shape is taken into account and it will appear as

Tensor
    [[1, 2, 3, 4],
     [2, 2, 5, 3]]

这篇关于使用tensorflow js从2D张量获取数据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！