torch.transpose — PyTorch 2.7 documentation (original) (raw)

torch.transpose(input, dim0, dim1) → Tensor¶

Returns a tensor that is a transposed version of input. The given dimensions dim0 and dim1 are swapped.

If input is a strided tensor then the resulting outtensor shares its underlying storage with the input tensor, so changing the content of one would change the content of the other.

If input is a sparse tensor then the resulting out tensor does not share the underlying storage with the input tensor.

If input is a sparse tensor with compressed layout (SparseCSR, SparseBSR, SparseCSC or SparseBSC) the argumentsdim0 and dim1 must be both batch dimensions, or must both be sparse dimensions. The batch dimensions of a sparse tensor are the dimensions preceding the sparse dimensions.

Note

Transpositions which interchange the sparse dimensions of a SparseCSRor SparseCSC layout tensor will result in the layout changing between the two options. Transposition of the sparse dimensions of a ` SparseBSR` or SparseBSC layout tensor will likewise generate a result with the opposite layout.

Parameters

• input (Tensor) – the input tensor.
• dim0 (int) – the first dimension to be transposed
• dim1 (int) – the second dimension to be transposed

Example:

x = torch.randn(2, 3) x tensor([[ 1.0028, -0.9893, 0.5809], [-0.1669, 0.7299, 0.4942]]) torch.transpose(x, 0, 1) tensor([[ 1.0028, -0.1669], [-0.9893, 0.7299], [ 0.5809, 0.4942]])

See also torch.t().