scipy 1.15.3__cp313-cp313-win_amd64.whl → 1.16.0rc2__cp313-cp313-win_amd64.whl

scipy/_lib/array_api_compat/torch/_info.py

@@ -34,7 +34,7 @@ class __array_namespace_info__:
 
     Examples
     --------
-    >>> info = np.__array_namespace_info__()
+    >>> info = xp.__array_namespace_info__()
     >>> info.default_dtypes()
     {'real floating': numpy.float64,
      'complex floating': numpy.complex128,
@@ -76,17 +76,17 @@ class __array_namespace_info__:
 
         Examples
         --------
-        >>> info = np.__array_namespace_info__()
+        >>> info = xp.__array_namespace_info__()
         >>> info.capabilities()
         {'boolean indexing': True,
-         'data-dependent shapes': True}
+         'data-dependent shapes': True,
+         'max dimensions': 64}
 
         """
         return {
             "boolean indexing": True,
             "data-dependent shapes": True,
-            # 'max rank' will be part of the 2024.12 standard
-            # "max rank": 64,
+            "max dimensions": 64,
         }
 
     def default_device(self):
@@ -102,15 +102,24 @@ class __array_namespace_info__:
 
         Returns
         -------
-        device : str
+        device : Device
             The default device used for new PyTorch arrays.
 
         Examples
         --------
-        >>> info = np.__array_namespace_info__()
+        >>> info = xp.__array_namespace_info__()
         >>> info.default_device()
-        'cpu'
+        device(type='cpu')
 
+        Notes
+        -----
+        This method returns the static default device when PyTorch is initialized.
+        However, the *current* device used by creation functions (``empty`` etc.)
+        can be changed at runtime.
+
+        See Also
+        --------
+        https://github.com/data-apis/array-api/issues/835
         """
         return torch.device("cpu")
 
@@ -120,9 +129,9 @@ class __array_namespace_info__:
 
         Parameters
         ----------
-        device : str, optional
-            The device to get the default data types for. For PyTorch, only
-            ``'cpu'`` is allowed.
+        device : Device, optional
+            The device to get the default data types for.
+            Unused for PyTorch, as all devices use the same default dtypes.
 
         Returns
         -------
@@ -139,7 +148,7 @@ class __array_namespace_info__:
 
         Examples
         --------
-        >>> info = np.__array_namespace_info__()
+        >>> info = xp.__array_namespace_info__()
         >>> info.default_dtypes()
         {'real floating': torch.float32,
          'complex floating': torch.complex64,
@@ -250,8 +259,9 @@ class __array_namespace_info__:
 
         Parameters
         ----------
-        device : str, optional
+        device : Device, optional
             The device to get the data types for.
+            Unused for PyTorch, as all devices use the same dtypes.
         kind : str or tuple of str, optional
             The kind of data types to return. If ``None``, all data types are
             returned. If a string, only data types of that kind are returned.
@@ -287,7 +297,7 @@ class __array_namespace_info__:
 
         Examples
         --------
-        >>> info = np.__array_namespace_info__()
+        >>> info = xp.__array_namespace_info__()
         >>> info.dtypes(kind='signed integer')
         {'int8': numpy.int8,
          'int16': numpy.int16,
@@ -310,7 +320,7 @@ class __array_namespace_info__:
 
         Returns
         -------
-        devices : list of str
+        devices : list[Device]
             The devices supported by PyTorch.
 
         See Also
@@ -322,7 +332,7 @@ class __array_namespace_info__:
 
         Examples
         --------
-        >>> info = np.__array_namespace_info__()
+        >>> info = xp.__array_namespace_info__()
         >>> info.devices()
         [device(type='cpu'), device(type='mps', index=0), device(type='meta')]
 
@@ -333,6 +343,7 @@ class __array_namespace_info__:
         # device:
         try:
             torch.device('notadevice')
+            raise AssertionError("unreachable")  # pragma: nocover
         except RuntimeError as e:
             # The error message is something like:
             # "Expected one of cpu, cuda, ipu, xpu, mkldnn, opengl, opencl, ideep, hip, ve, fpga, ort, xla, lazy, vulkan, mps, meta, hpu, mtia, privateuseone device type at start of device string: notadevice"

scipy/_lib/array_api_compat/torch/_typing.py

@@ -0,0 +1,3 @@
+__all__ = ["Array", "Device", "DType"]
+
+from torch import device as Device, dtype as DType, Tensor as Array

scipy/_lib/array_api_compat/torch/fft.py

@@ -1,76 +1,75 @@
 from __future__ import annotations
 
-from typing import TYPE_CHECKING
-if TYPE_CHECKING:
-    import torch
-    array = torch.Tensor
-    from typing import Union, Sequence, Literal
+from typing import Union, Sequence, Literal
 
-from torch.fft import * # noqa: F403
+import torch
 import torch.fft
+from torch.fft import * # noqa: F403
+
+from ._typing import Array
 
 # Several torch fft functions do not map axes to dim
 
 def fftn(
-    x: array,
+    x: Array,
     /,
     *,
     s: Sequence[int] = None,
     axes: Sequence[int] = None,
     norm: Literal["backward", "ortho", "forward"] = "backward",
     **kwargs,
-) -> array:
+) -> Array:
     return torch.fft.fftn(x, s=s, dim=axes, norm=norm, **kwargs)
 
 def ifftn(
-    x: array,
+    x: Array,
     /,
     *,
     s: Sequence[int] = None,
     axes: Sequence[int] = None,
     norm: Literal["backward", "ortho", "forward"] = "backward",
     **kwargs,
-) -> array:
+) -> Array:
     return torch.fft.ifftn(x, s=s, dim=axes, norm=norm, **kwargs)
 
 def rfftn(
-    x: array,
+    x: Array,
     /,
     *,
     s: Sequence[int] = None,
     axes: Sequence[int] = None,
     norm: Literal["backward", "ortho", "forward"] = "backward",
     **kwargs,
-) -> array:
+) -> Array:
     return torch.fft.rfftn(x, s=s, dim=axes, norm=norm, **kwargs)
 
 def irfftn(
-    x: array,
+    x: Array,
     /,
     *,
     s: Sequence[int] = None,
     axes: Sequence[int] = None,
     norm: Literal["backward", "ortho", "forward"] = "backward",
     **kwargs,
-) -> array:
+) -> Array:
     return torch.fft.irfftn(x, s=s, dim=axes, norm=norm, **kwargs)
 
 def fftshift(
-    x: array,
+    x: Array,
     /,
     *,
     axes: Union[int, Sequence[int]] = None,
     **kwargs,
-) -> array:
+) -> Array:
     return torch.fft.fftshift(x, dim=axes, **kwargs)
 
 def ifftshift(
-    x: array,
+    x: Array,
     /,
     *,
     axes: Union[int, Sequence[int]] = None,
     **kwargs,
-) -> array:
+) -> Array:
     return torch.fft.ifftshift(x, dim=axes, **kwargs)
 
 

scipy/_lib/array_api_compat/torch/linalg.py

@@ -1,14 +1,7 @@
 from __future__ import annotations
 
-from typing import TYPE_CHECKING
-if TYPE_CHECKING:
-    import torch
-    array = torch.Tensor
-    from torch import dtype as Dtype
-    from typing import Optional, Union, Tuple, Literal
-    inf = float('inf')
-
-from ._aliases import _fix_promotion, sum
+import torch
+from typing import Optional, Union, Tuple
 
 from torch.linalg import * # noqa: F403
 
@@ -19,15 +12,18 @@ linalg_all = [i for i in dir(torch_linalg) if not i.startswith('_')]
 
 # outer is implemented in torch but aren't in the linalg namespace
 from torch import outer
+from ._aliases import _fix_promotion, sum
 # These functions are in both the main and linalg namespaces
 from ._aliases import matmul, matrix_transpose, tensordot
+from ._typing import Array, DType
+from ..common._typing import JustInt, JustFloat
 
 # Note: torch.linalg.cross does not default to axis=-1 (it defaults to the
 # first axis with size 3), see https://github.com/pytorch/pytorch/issues/58743
 
 # torch.cross also does not support broadcasting when it would add new
 # dimensions https://github.com/pytorch/pytorch/issues/39656
-def cross(x1: array, x2: array, /, *, axis: int = -1) -> array:
+def cross(x1: Array, x2: Array, /, *, axis: int = -1) -> Array:
     x1, x2 = _fix_promotion(x1, x2, only_scalar=False)
     if not (-min(x1.ndim, x2.ndim) <= axis < max(x1.ndim, x2.ndim)):
         raise ValueError(f"axis {axis} out of bounds for cross product of arrays with shapes {x1.shape} and {x2.shape}")
@@ -36,7 +32,7 @@ def cross(x1: array, x2: array, /, *, axis: int = -1) -> array:
     x1, x2 = torch.broadcast_tensors(x1, x2)
     return torch_linalg.cross(x1, x2, dim=axis)
 
-def vecdot(x1: array, x2: array, /, *, axis: int = -1, **kwargs) -> array:
+def vecdot(x1: Array, x2: Array, /, *, axis: int = -1, **kwargs) -> Array:
     from ._aliases import isdtype
 
     x1, x2 = _fix_promotion(x1, x2, only_scalar=False)
@@ -58,7 +54,7 @@ def vecdot(x1: array, x2: array, /, *, axis: int = -1, **kwargs) -> array:
         return res[..., 0, 0]
     return torch.linalg.vecdot(x1, x2, dim=axis, **kwargs)
 
-def solve(x1: array, x2: array, /, **kwargs) -> array:
+def solve(x1: Array, x2: Array, /, **kwargs) -> Array:
     x1, x2 = _fix_promotion(x1, x2, only_scalar=False)
     # Torch tries to emulate NumPy 1 solve behavior by using batched 1-D solve
     # whenever
@@ -79,19 +75,20 @@ def solve(x1: array, x2: array, /, **kwargs) -> array:
     return torch.linalg.solve(x1, x2, **kwargs)
 
 # torch.trace doesn't support the offset argument and doesn't support stacking
-def trace(x: array, /, *, offset: int = 0, dtype: Optional[Dtype] = None) -> array:
+def trace(x: Array, /, *, offset: int = 0, dtype: Optional[DType] = None) -> Array:
     # Use our wrapped sum to make sure it does upcasting correctly
     return sum(torch.diagonal(x, offset=offset, dim1=-2, dim2=-1), axis=-1, dtype=dtype)
 
 def vector_norm(
-    x: array,
+    x: Array,
     /,
     *,
     axis: Optional[Union[int, Tuple[int, ...]]] = None,
     keepdims: bool = False,
-    ord: Union[int, float, Literal[inf, -inf]] = 2,
+    # JustFloat stands for inf | -inf, which are not valid for Literal
+    ord: JustInt | JustFloat = 2,
     **kwargs,
-) -> array:
+) -> Array:
     # torch.vector_norm incorrectly treats axis=() the same as axis=None
     if axis == ():
         out = kwargs.get('out')
@@ -119,3 +116,6 @@ __all__ = linalg_all + ['outer', 'matmul', 'matrix_transpose', 'tensordot',
 _all_ignore = ['torch_linalg', 'sum']
 
 del linalg_all
+
+def __dir__() -> list[str]:
+    return __all__

scipy/_lib/array_api_extra/__init__.py

@@ -1,15 +1,38 @@
-from __future__ import annotations
+"""Extra array functions built on top of the array API standard."""
 
-from ._funcs import atleast_nd, cov, create_diagonal, expand_dims, kron, sinc
+from ._delegation import isclose, pad
+from ._lib._at import at
+from ._lib._funcs import (
+    apply_where,
+    atleast_nd,
+    broadcast_shapes,
+    cov,
+    create_diagonal,
+    expand_dims,
+    kron,
+    nunique,
+    setdiff1d,
+    sinc,
+)
+from ._lib._lazy import lazy_apply
 
-__version__ = "0.2.0"
+__version__ = "0.8.0.dev0"
 
+# pylint: disable=duplicate-code
 __all__ = [
     "__version__",
+    "apply_where",
+    "at",
     "atleast_nd",
+    "broadcast_shapes",
     "cov",
     "create_diagonal",
     "expand_dims",
+    "isclose",
     "kron",
+    "lazy_apply",
+    "nunique",
+    "pad",
+    "setdiff1d",
     "sinc",
 ]

scipy/_lib/array_api_extra/_delegation.py

@@ -0,0 +1,171 @@
+"""Delegation to existing implementations for Public API Functions."""
+
+from collections.abc import Sequence
+from types import ModuleType
+from typing import Literal
+
+from ._lib import _funcs
+from ._lib._utils._compat import (
+    array_namespace,
+    is_cupy_namespace,
+    is_dask_namespace,
+    is_jax_namespace,
+    is_numpy_namespace,
+    is_pydata_sparse_namespace,
+    is_torch_namespace,
+)
+from ._lib._utils._helpers import asarrays
+from ._lib._utils._typing import Array
+
+__all__ = ["isclose", "pad"]
+
+
+def isclose(
+    a: Array | complex,
+    b: Array | complex,
+    *,
+    rtol: float = 1e-05,
+    atol: float = 1e-08,
+    equal_nan: bool = False,
+    xp: ModuleType | None = None,
+) -> Array:
+    """
+    Return a boolean array where two arrays are element-wise equal within a tolerance.
+
+    The tolerance values are positive, typically very small numbers. The relative
+    difference ``(rtol * abs(b))`` and the absolute difference `atol` are added together
+    to compare against the absolute difference between `a` and `b`.
+
+    NaNs are treated as equal if they are in the same place and if ``equal_nan=True``.
+    Infs are treated as equal if they are in the same place and of the same sign in both
+    arrays.
+
+    Parameters
+    ----------
+    a, b : Array | int | float | complex | bool
+        Input objects to compare. At least one must be an array.
+    rtol : array_like, optional
+        The relative tolerance parameter (see Notes).
+    atol : array_like, optional
+        The absolute tolerance parameter (see Notes).
+    equal_nan : bool, optional
+        Whether to compare NaN's as equal. If True, NaN's in `a` will be considered
+        equal to NaN's in `b` in the output array.
+    xp : array_namespace, optional
+        The standard-compatible namespace for `a` and `b`. Default: infer.
+
+    Returns
+    -------
+    Array
+        A boolean array of shape broadcasted from `a` and `b`, containing ``True`` where
+        `a` is close to `b`, and ``False`` otherwise.
+
+    Warnings
+    --------
+    The default `atol` is not appropriate for comparing numbers with magnitudes much
+    smaller than one (see notes).
+
+    See Also
+    --------
+    math.isclose : Similar function in stdlib for Python scalars.
+
+    Notes
+    -----
+    For finite values, `isclose` uses the following equation to test whether two
+    floating point values are equivalent::
+
+        absolute(a - b) <= (atol + rtol * absolute(b))
+
+    Unlike the built-in `math.isclose`,
+    the above equation is not symmetric in `a` and `b`,
+    so that ``isclose(a, b)`` might be different from ``isclose(b, a)`` in some rare
+    cases.
+
+    The default value of `atol` is not appropriate when the reference value `b` has
+    magnitude smaller than one. For example, it is unlikely that ``a = 1e-9`` and
+    ``b = 2e-9`` should be considered "close", yet ``isclose(1e-9, 2e-9)`` is ``True``
+    with default settings. Be sure to select `atol` for the use case at hand, especially
+    for defining the threshold below which a non-zero value in `a` will be considered
+    "close" to a very small or zero value in `b`.
+
+    The comparison of `a` and `b` uses standard broadcasting, which means that `a` and
+    `b` need not have the same shape in order for ``isclose(a, b)`` to evaluate to
+    ``True``.
+
+    `isclose` is not defined for non-numeric data types.
+    ``bool`` is considered a numeric data-type for this purpose.
+    """
+    xp = array_namespace(a, b) if xp is None else xp
+
+    if (
+        is_numpy_namespace(xp)
+        or is_cupy_namespace(xp)
+        or is_dask_namespace(xp)
+        or is_jax_namespace(xp)
+    ):
+        return xp.isclose(a, b, rtol=rtol, atol=atol, equal_nan=equal_nan)
+
+    if is_torch_namespace(xp):
+        a, b = asarrays(a, b, xp=xp)  # Array API 2024.12 support
+        return xp.isclose(a, b, rtol=rtol, atol=atol, equal_nan=equal_nan)
+
+    return _funcs.isclose(a, b, rtol=rtol, atol=atol, equal_nan=equal_nan, xp=xp)
+
+
+def pad(
+    x: Array,
+    pad_width: int | tuple[int, int] | Sequence[tuple[int, int]],
+    mode: Literal["constant"] = "constant",
+    *,
+    constant_values: complex = 0,
+    xp: ModuleType | None = None,
+) -> Array:
+    """
+    Pad the input array.
+
+    Parameters
+    ----------
+    x : array
+        Input array.
+    pad_width : int or tuple of ints or sequence of pairs of ints
+        Pad the input array with this many elements from each side.
+        If a sequence of tuples, ``[(before_0, after_0), ... (before_N, after_N)]``,
+        each pair applies to the corresponding axis of ``x``.
+        A single tuple, ``(before, after)``, is equivalent to a list of ``x.ndim``
+        copies of this tuple.
+    mode : str, optional
+        Only "constant" mode is currently supported, which pads with
+        the value passed to `constant_values`.
+    constant_values : python scalar, optional
+        Use this value to pad the input. Default is zero.
+    xp : array_namespace, optional
+        The standard-compatible namespace for `x`. Default: infer.
+
+    Returns
+    -------
+    array
+        The input array,
+        padded with ``pad_width`` elements equal to ``constant_values``.
+    """
+    xp = array_namespace(x) if xp is None else xp
+
+    if mode != "constant":
+        msg = "Only `'constant'` mode is currently supported"
+        raise NotImplementedError(msg)
+
+    if (
+        is_numpy_namespace(xp)
+        or is_cupy_namespace(xp)
+        or is_jax_namespace(xp)
+        or is_pydata_sparse_namespace(xp)
+    ):
+        return xp.pad(x, pad_width, mode, constant_values=constant_values)
+
+    # https://github.com/pytorch/pytorch/blob/cf76c05b4dc629ac989d1fb8e789d4fac04a095a/torch/_numpy/_funcs_impl.py#L2045-L2056
+    if is_torch_namespace(xp):
+        pad_width = xp.asarray(pad_width)
+        pad_width = xp.broadcast_to(pad_width, (x.ndim, 2))
+        pad_width = xp.flip(pad_width, axis=(0,)).flatten()
+        return xp.nn.functional.pad(x, tuple(pad_width), value=constant_values)  # type: ignore[arg-type]  # pyright: ignore[reportArgumentType]
+
+    return _funcs.pad(x, pad_width, constant_values=constant_values, xp=xp)

scipy/_lib/array_api_extra/_lib/__init__.py

@@ -0,0 +1 @@
+"""Internals of array-api-extra."""