用火炬随机选择? [英] Random Choice with Pytorch?

问题描述

我有一张张量的图片，想从中随机选择.我正在寻找np.random.choice()的等效项.

I have a tensor of pictures, and would like to randomly select from it. I'm looking for the equivalent of np.random.choice().

import torch

pictures = torch.randint(0, 256, (1000, 28, 28, 3))

假设我要这些照片中的10张.

Let's say I want 10 of these pictures.

推荐答案

torch没有等效的np.random.choice()实现.另一种方法是使用改组后的索引或随机整数进行索引.

torch has no equivalent implementation of np.random.choice(). The alternative is indexing with a shuffled index or random integers.

1. 生成 n 个随机索引
2. 使用这些索引来索引原始张量

1. Generate n random indices
2. Index your original tensor with these indices

pictures[torch.randint(len(pictures), (10,))]  

要做到这一点，无需替换:

To do it without replacement:

1. 随机播放索引
2. 采用 n 首个元素

1. Shuffle the index
2. Take the n first elements

indices = torch.randperm(len(pictures))[:10]

pictures[indices]

详细了解 torch.randint 和 torch.randperm .

Read more about torch.randint and torch.randperm.

这篇关于用火炬随机选择?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！