按 2d 张量中的值索引 pytorch 4d 张量

dir*_*lik 2 python pytorch

我有两个 pytorch 张量:

  • X 有形状 (A, B, C, D)
  • I 有形状 (A, B)

中的值I是范围内的整数[0, C)


获得Yshape张量的最有效方法是什么(A, B, D),例如:

Y[i][j][k] = X[i][j][ I[i][j] ][k]
Run Code Online (Sandbox Code Playgroud)

Sha*_*hai 6

您可能想要torch.gather用于索引并expand调整I到所需的大小:

eI = I[..., None, None].expand(-1, -1, 1, X.size(3))  # make eI the same for the last dimension
Y = torch.gather(X, dim=2, index=eI).squeeze()
Run Code Online (Sandbox Code Playgroud)

测试代码:

A = 3 
B = 4 
C = 5 
D = 7

X = torch.rand(A, B, C, D)
I = torch.randint(0, C, (A, B), dtype=torch.long)

eI = I[..., None, None].expand(-1, -1, 1, X.size(3))
Y = torch.gather(X, dim=2, index=eI).squeeze()

# manually gather
refY = torch.empty(A, B, D)
for i in range(A):
    for j in range(B):
        refY[i, j, :] = X[i, j, I[i,j], :]

(refY == Y).all()
# Out[]: tensor(1, dtype=torch.uint8)
Run Code Online (Sandbox Code Playgroud)