numpy - 如何在 PyTorch 中选择二维索引？

Question

给定 a = torch.randn(3, 2, 4, 5) ，我如何选择像 (2, :, 0, :), (1, :, 1, :), (2, :, 2, :), (0, :, 3, :) 这样的子张量(大小为 (2, 4, 5) 或 (4, 2, 5) 的结果张量？

虽然 a[2, :, 0, :]给

 0.5580 -0.0337  1.0048 -0.5044  0.6784
-1.6117  1.0084  1.1886  0.1278  0.3739
[torch.FloatTensor of size 2x5]

然而， a[[2, 1, 2, 0], :, [0, 1, 2, 3], :]给

TypeError: Performing basic indexing on a tensor and encountered an error indexing dim 0 with an object of type list. The only supported types are integers, slices, numpy scalars, or if indexing with a torch.LongTensor or torch.ByteTensor only a single Tensor may be passed.

虽然 numpy返回 (4, 2, 5)张量成功。

Answer 1

对你起作用吗？

import torch

a = torch.randn(3, 2, 4, 5)
print(a.size())

b = [a[2, :, 0, :], a[1, :, 1, :], a[2, :, 2, :], a[0, :, 3, :]]
b = torch.stack(b, 0)

print(b.size()) # torch.Size([4, 2, 5])