torch.Tensor.crow_indices#

Tensor.crow_indices()→IntTensor#

Returns the tensor containing the compressed row indices of theselftensor whenself is a sparse CSR tensor of layoutsparse_csr.Thecrow_indices tensor is strictly of shape (self.size(0) + 1)and of typeint32 orint64. When using MKL routines such as sparsematrix multiplication, it is necessary to useint32 indexing in orderto avoid downcasting and potentially losing information.

Example:

>>>csr=torch.eye(5,5).to_sparse_csr()>>>csr.crow_indices()tensor([0, 1, 2, 3, 4, 5], dtype=torch.int32)