The nvmath.bindings module is experimental and potentially subject to future changes.

create_block_sparse_tensor_descriptor

nvmath.bindings.cutensor.create_block_sparse_tensor_descriptor(

intptr_t handle,

uint32_t num_modes,

uint64_t num_non_zero_blocks,

num_sections_per_mode,

extent,

non_zero_coordinates,

stride,

int data_type,

) → intptr_t

See cutensorCreateBlockSparseTensorDescriptor.