ravel_indices

iskra.sparse.ravel_indices(indices: Tensor, shape: Size | tuple[int, ...]) → Tensor[SOURCE]

Converts a COO indices into a linear index corresponding to a strided layout.

Example

The sparse tensor ``` a = [[1.0, 0.0],

[0.0, 2.0]]

``` has COO indices a.indices() = [[0, 1], [0, 1]]. The linear index otutput is ravel_index(a.indices()) = [0, 3].

Parameters:

• indices (Tensor[dim, nnz]) – Indices of a sparse tensor in COO format.

• shape (torch.Size | tuple[int, ...]) – Shape of the tensor.

Returns:

​ (Tensor[nnz]) – Linear indices into a strided layout tensor of same shape.