jax.experimental.sparse.BCSR

class jax.experimental.sparse.BCSR(args, *, shape, indices_sorted=False, unique_indices=False)

Experimental batched CSR matrix implemented in JAX.

Parameters:

• args (tuple[Array, Array, Array])

• shape (Shape)

• indices_sorted (bool)

• unique_indices (bool)

__init__(args, *, shape, indices_sorted=False, unique_indices=False)

Parameters:

• args (tuple[Array, Array, Array])

• shape (Sequence[int])

• indices_sorted (bool)

• unique_indices (bool)

Methods

__init__(args, *, shape[, indices_sorted, ...])
block_until_ready()
from_bcoo(arr)
from_scipy_sparse(mat, *[, index_dtype, ...])	Create a BCSR array from a scipy.sparse array.
fromdense(mat, *[, nse, index_dtype, ...])	Create a BCSR array from a (dense) Array.
sum(*args, **kwargs)
sum_duplicates([nse, remove_zeros])	Return a copy of the array with duplicate indices summed.
to_bcoo()
todense()	Create a dense version of the array.
transpose(*args, **kwargs)
tree_flatten()
tree_unflatten(aux_data, children)

Attributes

T
dtype
n_batch
n_dense
n_sparse
ndim
nse
size
data
indices
indptr
shape
indices_sorted
unique_indices