from .. import _namespace as _ns
[docs]
def asarray(obj, /, *, dtype=None, device=None, copy=None):
return getattr(_ns, "asarray")(obj, dtype=dtype, device=device, copy=copy)
[docs]
def zeros(shape, *, dtype=None, device=None):
"""Create a sparse zero matrix (Array API compliant).
Parameters
----------
shape : tuple of int
Matrix shape (nrows, ncols).
dtype : dtype, optional
Data type (default: float64).
device : str, optional
Device (default: "cpu").
Returns
-------
CSR
Sparse zero matrix using Rust backend.
Examples
--------
>>> import lacuna.array_api as xp
>>> A = xp.zeros((10, 20))
>>> A.nnz
0
"""
import numpy as np
from ...sparse import CSR
if dtype is None:
dtype = np.float64
if device is not None and device != "cpu":
raise ValueError("Only 'cpu' device is currently supported")
return CSR.zeros(shape, dtype=dtype)
[docs]
def ones(shape, *, dtype=None, device=None):
return getattr(_ns, "ones")(shape, dtype=dtype, device=device)
[docs]
def full(shape, fill_value, *, dtype=None, device=None):
return getattr(_ns, "full")(shape, fill_value, dtype=dtype, device=device)
[docs]
def empty(shape, *, dtype=None, device=None):
return getattr(_ns, "empty")(shape, dtype=dtype, device=device)
[docs]
def zeros_like(x, *, dtype=None, device=None):
return getattr(_ns, "zeros_like")(x, dtype=dtype, device=device)
[docs]
def ones_like(x, *, dtype=None, device=None):
return getattr(_ns, "ones_like")(x, dtype=dtype, device=device)
[docs]
def full_like(x, fill_value, *, dtype=None, device=None):
return getattr(_ns, "full_like")(x, fill_value, dtype=dtype, device=device)
[docs]
def empty_like(x, *, dtype=None, device=None):
return getattr(_ns, "empty_like")(x, dtype=dtype, device=device)
[docs]
def arange(start, stop=None, step=1, *, dtype=None, device=None):
return getattr(_ns, "arange")(start, stop, step, dtype=dtype, device=device)
[docs]
def linspace(start, stop, num, *, dtype=None, device=None):
return getattr(_ns, "linspace")(start, stop, num, dtype=dtype, device=device)
[docs]
def meshgrid(*arrays, indexing="xy"):
return getattr(_ns, "meshgrid")(*arrays, indexing=indexing)
[docs]
def eye(n_rows, n_cols=None, k=0, *, dtype=None, device=None):
"""Create an identity matrix (Array API compliant).
Parameters
----------
n_rows : int
Number of rows (for square matrix, this is the size).
n_cols : int, optional
Number of columns (if None, creates square matrix).
k : int, optional
Diagonal offset (default: 0 for main diagonal).
dtype : dtype, optional
Data type (default: float64).
device : str, optional
Device (default: "cpu").
Returns
-------
CSR
Identity matrix using Rust backend.
Notes
-----
Currently only supports k=0 (main diagonal) and square matrices.
Examples
--------
>>> import lacuna.array_api as xp
>>> I = xp.eye(5)
>>> I.sum()
5.0
"""
import numpy as np
from ...sparse import CSR
if dtype is None:
dtype = np.float64
if device is not None and device != "cpu":
raise ValueError("Only 'cpu' device is currently supported")
if n_cols is not None and n_cols != n_rows:
raise NotImplementedError("Non-square identity matrices not yet supported")
if k != 0:
raise NotImplementedError("Off-diagonal identity (k != 0) not yet supported")
return CSR.eye(n_rows, dtype=dtype)
[docs]
def tril(x, k=0):
return getattr(_ns, "tril")(x, k)
[docs]
def triu(x, k=0):
return getattr(_ns, "triu")(x, k)
[docs]
def from_dlpack(dlpack_obj):
return getattr(_ns, "from_dlpack")(dlpack_obj)
