torch.sparse_csr

torch.sparse_csr is a module in PyTorch that provides support for compressed sparse row (CSR) matrices. CSR matrices are a popular format for storing sparse matrices in a compressed form, where only the non-zero elements are stored.

The torch.sparse_csr module provides functions to create, manipulate, and perform operations on CSR matrices. Some of the key functions in this module include:

- torch.sparse_csr_matrix: Creates a new CSR matrix from a dense or sparse input tensor.
- torch.sparse_csr_tensor: Creates a new CSR tensor from a COO (coordinate) tensor.
- torch.sparse_csr_matmul: Performs a matrix multiplication between two CSR matrices.
- torch.sparse_csr_add: Adds two CSR matrices.
- torch.sparse_csr_sub: Subtracts two CSR matrices.
- torch.sparse_csr_transpose: Transposes a CSR matrix.

Overall, the torch.sparse_csr module provides a powerful and efficient way to work with sparse matrices in PyTorch.