Add files using upload-large-folder tool

.gitattributes

@@ -348,3 +348,9 @@ tuning-competition-baseline/.venv/lib/python3.11/site-packages/nvidia/cudnn/lib/
 .venv/lib/python3.11/site-packages/torchvision/_C.so filter=lfs diff=lfs merge=lfs -text
 .venv/lib/python3.11/site-packages/torchvision/transforms/__pycache__/transforms.cpython-311.pyc filter=lfs diff=lfs merge=lfs -text
 .venv/lib/python3.11/site-packages/torchvision/transforms/v2/functional/__pycache__/_geometry.cpython-311.pyc filter=lfs diff=lfs merge=lfs -text
+.venv/lib/python3.11/site-packages/safetensors/_safetensors_rust.abi3.so filter=lfs diff=lfs merge=lfs -text
+.venv/lib/python3.11/site-packages/psutil/_psutil_linux.abi3.so filter=lfs diff=lfs merge=lfs -text
+.venv/lib/python3.11/site-packages/psutil/tests/__pycache__/test_linux.cpython-311.pyc filter=lfs diff=lfs merge=lfs -text
+.venv/lib/python3.11/site-packages/psutil/tests/__pycache__/__init__.cpython-311.pyc filter=lfs diff=lfs merge=lfs -text
+.venv/lib/python3.11/site-packages/psutil/__pycache__/_pslinux.cpython-311.pyc filter=lfs diff=lfs merge=lfs -text
+.venv/lib/python3.11/site-packages/psutil/tests/__pycache__/test_process.cpython-311.pyc filter=lfs diff=lfs merge=lfs -text

.venv/lib/python3.11/site-packages/numpy/__config__.py

@@ -0,0 +1,162 @@
+# This file is generated by numpy's build process
+# It contains system_info results at the time of building this package.
+from enum import Enum
+from numpy.core._multiarray_umath import (
+    __cpu_features__,
+    __cpu_baseline__,
+    __cpu_dispatch__,
+)
+
+__all__ = ["show"]
+_built_with_meson = True
+
+
+class DisplayModes(Enum):
+    stdout = "stdout"
+    dicts = "dicts"
+
+
+def _cleanup(d):
+    """
+    Removes empty values in a `dict` recursively
+    This ensures we remove values that Meson could not provide to CONFIG
+    """
+    if isinstance(d, dict):
+        return {k: _cleanup(v) for k, v in d.items() if v and _cleanup(v)}
+    else:
+        return d
+
+
+CONFIG = _cleanup(
+    {
+        "Compilers": {
+            "c": {
+                "name": "gcc",
+                "linker": r"ld.bfd",
+                "version": "10.2.1",
+                "commands": r"cc",
+                "args": r"-fno-strict-aliasing",
+                "linker args": r"-Wl,--strip-debug, -fno-strict-aliasing",
+            },
+            "cython": {
+                "name": "cython",
+                "linker": r"cython",
+                "version": "3.0.8",
+                "commands": r"cython",
+                "args": r"",
+                "linker args": r"",
+            },
+            "c++": {
+                "name": "gcc",
+                "linker": r"ld.bfd",
+                "version": "10.2.1",
+                "commands": r"c++",
+                "args": r"",
+                "linker args": r"-Wl,--strip-debug",
+            },
+        },
+        "Machine Information": {
+            "host": {
+                "cpu": "x86_64",
+                "family": "x86_64",
+                "endian": "little",
+                "system": "linux",
+            },
+            "build": {
+                "cpu": "x86_64",
+                "family": "x86_64",
+                "endian": "little",
+                "system": "linux",
+            },
+            "cross-compiled": bool("False".lower().replace("false", "")),
+        },
+        "Build Dependencies": {
+            "blas": {
+                "name": "openblas64",
+                "found": bool("True".lower().replace("false", "")),
+                "version": "0.3.23.dev",
+                "detection method": "pkgconfig",
+                "include directory": r"/usr/local/include",
+                "lib directory": r"/usr/local/lib",
+                "openblas configuration": r"USE_64BITINT=1 DYNAMIC_ARCH=1 DYNAMIC_OLDER= NO_CBLAS= NO_LAPACK= NO_LAPACKE= NO_AFFINITY=1 USE_OPENMP= HASWELL MAX_THREADS=2",
+                "pc file directory": r"/usr/local/lib/pkgconfig",
+            },
+            "lapack": {
+                "name": "dep140213194937296",
+                "found": bool("True".lower().replace("false", "")),
+                "version": "1.26.4",
+                "detection method": "internal",
+                "include directory": r"unknown",
+                "lib directory": r"unknown",
+                "openblas configuration": r"unknown",
+                "pc file directory": r"unknown",
+            },
+        },
+        "Python Information": {
+            "path": r"/opt/python/cp311-cp311/bin/python",
+            "version": "3.11",
+        },
+        "SIMD Extensions": {
+            "baseline": __cpu_baseline__,
+            "found": [
+                feature for feature in __cpu_dispatch__ if __cpu_features__[feature]
+            ],
+            "not found": [
+                feature for feature in __cpu_dispatch__ if not __cpu_features__[feature]
+            ],
+        },
+    }
+)
+
+
+def _check_pyyaml():
+    import yaml
+
+    return yaml
+
+
+def show(mode=DisplayModes.stdout.value):
+    """
+    Show libraries and system information on which NumPy was built
+    and is being used
+
+    Parameters
+    ----------
+    mode : {`'stdout'`, `'dicts'`}, optional.
+        Indicates how to display the config information.
+        `'stdout'` prints to console, `'dicts'` returns a dictionary
+        of the configuration.
+
+    Returns
+    -------
+    out : {`dict`, `None`}
+        If mode is `'dicts'`, a dict is returned, else None
+
+    See Also
+    --------
+    get_include : Returns the directory containing NumPy C
+                  header files.
+
+    Notes
+    -----
+    1. The `'stdout'` mode will give more readable
+       output if ``pyyaml`` is installed
+
+    """
+    if mode == DisplayModes.stdout.value:
+        try:  # Non-standard library, check import
+            yaml = _check_pyyaml()
+
+            print(yaml.dump(CONFIG))
+        except ModuleNotFoundError:
+            import warnings
+            import json
+
+            warnings.warn("Install `pyyaml` for better output", stacklevel=1)
+            print(json.dumps(CONFIG, indent=2))
+    elif mode == DisplayModes.dicts.value:
+        return CONFIG
+    else:
+        raise AttributeError(
+            f"Invalid `mode`, use one of: {', '.join([e.value for e in DisplayModes])}"
+        )

.venv/lib/python3.11/site-packages/numpy/__init__.pxd

@@ -0,0 +1,1015 @@
+# NumPy static imports for Cython < 3.0
+#
+# If any of the PyArray_* functions are called, import_array must be
+# called first.
+#
+# Author: Dag Sverre Seljebotn
+#
+
+DEF _buffer_format_string_len = 255
+
+cimport cpython.buffer as pybuf
+from cpython.ref cimport Py_INCREF
+from cpython.mem cimport PyObject_Malloc, PyObject_Free
+from cpython.object cimport PyObject, PyTypeObject
+from cpython.buffer cimport PyObject_GetBuffer
+from cpython.type cimport type
+cimport libc.stdio as stdio
+
+cdef extern from "Python.h":
+    ctypedef int Py_intptr_t
+    bint PyObject_TypeCheck(object obj, PyTypeObject* type)
+
+cdef extern from "numpy/arrayobject.h":
+    ctypedef Py_intptr_t npy_intp
+    ctypedef size_t npy_uintp
+
+    cdef enum NPY_TYPES:
+        NPY_BOOL
+        NPY_BYTE
+        NPY_UBYTE
+        NPY_SHORT
+        NPY_USHORT
+        NPY_INT
+        NPY_UINT
+        NPY_LONG
+        NPY_ULONG
+        NPY_LONGLONG
+        NPY_ULONGLONG
+        NPY_FLOAT
+        NPY_DOUBLE
+        NPY_LONGDOUBLE
+        NPY_CFLOAT
+        NPY_CDOUBLE
+        NPY_CLONGDOUBLE
+        NPY_OBJECT
+        NPY_STRING
+        NPY_UNICODE
+        NPY_VOID
+        NPY_DATETIME
+        NPY_TIMEDELTA
+        NPY_NTYPES
+        NPY_NOTYPE
+
+        NPY_INT8
+        NPY_INT16
+        NPY_INT32
+        NPY_INT64
+        NPY_INT128
+        NPY_INT256
+        NPY_UINT8
+        NPY_UINT16
+        NPY_UINT32
+        NPY_UINT64
+        NPY_UINT128
+        NPY_UINT256
+        NPY_FLOAT16
+        NPY_FLOAT32
+        NPY_FLOAT64
+        NPY_FLOAT80
+        NPY_FLOAT96
+        NPY_FLOAT128
+        NPY_FLOAT256
+        NPY_COMPLEX32
+        NPY_COMPLEX64
+        NPY_COMPLEX128
+        NPY_COMPLEX160
+        NPY_COMPLEX192
+        NPY_COMPLEX256
+        NPY_COMPLEX512
+
+        NPY_INTP
+
+    ctypedef enum NPY_ORDER:
+        NPY_ANYORDER
+        NPY_CORDER
+        NPY_FORTRANORDER
+        NPY_KEEPORDER
+
+    ctypedef enum NPY_CASTING:
+        NPY_NO_CASTING
+        NPY_EQUIV_CASTING
+        NPY_SAFE_CASTING
+        NPY_SAME_KIND_CASTING
+        NPY_UNSAFE_CASTING
+
+    ctypedef enum NPY_CLIPMODE:
+        NPY_CLIP
+        NPY_WRAP
+        NPY_RAISE
+
+    ctypedef enum NPY_SCALARKIND:
+        NPY_NOSCALAR,
+        NPY_BOOL_SCALAR,
+        NPY_INTPOS_SCALAR,
+        NPY_INTNEG_SCALAR,
+        NPY_FLOAT_SCALAR,
+        NPY_COMPLEX_SCALAR,
+        NPY_OBJECT_SCALAR
+
+    ctypedef enum NPY_SORTKIND:
+        NPY_QUICKSORT
+        NPY_HEAPSORT
+        NPY_MERGESORT
+
+    ctypedef enum NPY_SEARCHSIDE:
+        NPY_SEARCHLEFT
+        NPY_SEARCHRIGHT
+
+    enum:
+        # DEPRECATED since NumPy 1.7 ! Do not use in new code!
+        NPY_C_CONTIGUOUS
+        NPY_F_CONTIGUOUS
+        NPY_CONTIGUOUS
+        NPY_FORTRAN
+        NPY_OWNDATA
+        NPY_FORCECAST
+        NPY_ENSURECOPY
+        NPY_ENSUREARRAY
+        NPY_ELEMENTSTRIDES
+        NPY_ALIGNED
+        NPY_NOTSWAPPED
+        NPY_WRITEABLE
+        NPY_ARR_HAS_DESCR
+
+        NPY_BEHAVED
+        NPY_BEHAVED_NS
+        NPY_CARRAY
+        NPY_CARRAY_RO
+        NPY_FARRAY
+        NPY_FARRAY_RO
+        NPY_DEFAULT
+
+        NPY_IN_ARRAY
+        NPY_OUT_ARRAY
+        NPY_INOUT_ARRAY
+        NPY_IN_FARRAY
+        NPY_OUT_FARRAY
+        NPY_INOUT_FARRAY
+
+        NPY_UPDATE_ALL
+
+    enum:
+        # Added in NumPy 1.7 to replace the deprecated enums above.
+        NPY_ARRAY_C_CONTIGUOUS
+        NPY_ARRAY_F_CONTIGUOUS
+        NPY_ARRAY_OWNDATA
+        NPY_ARRAY_FORCECAST
+        NPY_ARRAY_ENSURECOPY
+        NPY_ARRAY_ENSUREARRAY
+        NPY_ARRAY_ELEMENTSTRIDES
+        NPY_ARRAY_ALIGNED
+        NPY_ARRAY_NOTSWAPPED
+        NPY_ARRAY_WRITEABLE
+        NPY_ARRAY_WRITEBACKIFCOPY
+
+        NPY_ARRAY_BEHAVED
+        NPY_ARRAY_BEHAVED_NS
+        NPY_ARRAY_CARRAY
+        NPY_ARRAY_CARRAY_RO
+        NPY_ARRAY_FARRAY
+        NPY_ARRAY_FARRAY_RO
+        NPY_ARRAY_DEFAULT
+
+        NPY_ARRAY_IN_ARRAY
+        NPY_ARRAY_OUT_ARRAY
+        NPY_ARRAY_INOUT_ARRAY
+        NPY_ARRAY_IN_FARRAY
+        NPY_ARRAY_OUT_FARRAY
+        NPY_ARRAY_INOUT_FARRAY
+
+        NPY_ARRAY_UPDATE_ALL
+
+    cdef enum:
+        NPY_MAXDIMS
+
+    npy_intp NPY_MAX_ELSIZE
+
+    ctypedef void (*PyArray_VectorUnaryFunc)(void *, void *, npy_intp, void *,  void *)
+
+    ctypedef struct PyArray_ArrayDescr:
+        # shape is a tuple, but Cython doesn't support "tuple shape"
+        # inside a non-PyObject declaration, so we have to declare it
+        # as just a PyObject*.
+        PyObject* shape
+
+    ctypedef struct PyArray_Descr:
+        pass
+
+    ctypedef class numpy.dtype [object PyArray_Descr, check_size ignore]:
+        # Use PyDataType_* macros when possible, however there are no macros
+        # for accessing some of the fields, so some are defined.
+        cdef PyTypeObject* typeobj
+        cdef char kind
+        cdef char type
+        # Numpy sometimes mutates this without warning (e.g. it'll
+        # sometimes change "|" to "<" in shared dtype objects on
+        # little-endian machines). If this matters to you, use
+        # PyArray_IsNativeByteOrder(dtype.byteorder) instead of
+        # directly accessing this field.
+        cdef char byteorder
+        cdef char flags
+        cdef int type_num
+        cdef int itemsize "elsize"
+        cdef int alignment
+        cdef object fields
+        cdef tuple names
+        # Use PyDataType_HASSUBARRAY to test whether this field is
+        # valid (the pointer can be NULL). Most users should access
+        # this field via the inline helper method PyDataType_SHAPE.
+        cdef PyArray_ArrayDescr* subarray
+
+    ctypedef class numpy.flatiter [object PyArrayIterObject, check_size ignore]:
+        # Use through macros
+        pass
+
+    ctypedef class numpy.broadcast [object PyArrayMultiIterObject, check_size ignore]:
+        cdef int numiter
+        cdef npy_intp size, index
+        cdef int nd
+        cdef npy_intp *dimensions
+        cdef void **iters
+
+    ctypedef struct PyArrayObject:
+        # For use in situations where ndarray can't replace PyArrayObject*,
+        # like PyArrayObject**.
+        pass
+
+    ctypedef class numpy.ndarray [object PyArrayObject, check_size ignore]:
+        cdef __cythonbufferdefaults__ = {"mode": "strided"}
+
+        cdef:
+            # Only taking a few of the most commonly used and stable fields.
+            # One should use PyArray_* macros instead to access the C fields.
+            char *data
+            int ndim "nd"
+            npy_intp *shape "dimensions"
+            npy_intp *strides
+            dtype descr  # deprecated since NumPy 1.7 !
+            PyObject* base #  NOT PUBLIC, DO NOT USE !
+
+
+
+    ctypedef unsigned char      npy_bool
+
+    ctypedef signed char      npy_byte
+    ctypedef signed short     npy_short
+    ctypedef signed int       npy_int
+    ctypedef signed long      npy_long
+    ctypedef signed long long npy_longlong
+
+    ctypedef unsigned char      npy_ubyte
+    ctypedef unsigned short     npy_ushort
+    ctypedef unsigned int       npy_uint
+    ctypedef unsigned long      npy_ulong
+    ctypedef unsigned long long npy_ulonglong
+
+    ctypedef float        npy_float
+    ctypedef double       npy_double
+    ctypedef long double  npy_longdouble
+
+    ctypedef signed char        npy_int8
+    ctypedef signed short       npy_int16
+    ctypedef signed int         npy_int32
+    ctypedef signed long long   npy_int64
+    ctypedef signed long long   npy_int96
+    ctypedef signed long long   npy_int128
+
+    ctypedef unsigned char      npy_uint8
+    ctypedef unsigned short     npy_uint16
+    ctypedef unsigned int       npy_uint32
+    ctypedef unsigned long long npy_uint64
+    ctypedef unsigned long long npy_uint96
+    ctypedef unsigned long long npy_uint128
+
+    ctypedef float        npy_float32
+    ctypedef double       npy_float64
+    ctypedef long double  npy_float80
+    ctypedef long double  npy_float96
+    ctypedef long double  npy_float128
+
+    ctypedef struct npy_cfloat:
+        float real
+        float imag
+
+    ctypedef struct npy_cdouble:
+        double real
+        double imag
+
+    ctypedef struct npy_clongdouble:
+        long double real
+        long double imag
+
+    ctypedef struct npy_complex64:
+        float real
+        float imag
+
+    ctypedef struct npy_complex128:
+        double real
+        double imag
+
+    ctypedef struct npy_complex160:
+        long double real
+        long double imag
+
+    ctypedef struct npy_complex192:
+        long double real
+        long double imag
+
+    ctypedef struct npy_complex256:
+        long double real
+        long double imag
+
+    ctypedef struct PyArray_Dims:
+        npy_intp *ptr
+        int len
+
+    int _import_array() except -1
+    # A second definition so _import_array isn't marked as used when we use it here.
+    # Do not use - subject to change any time.
+    int __pyx_import_array "_import_array"() except -1
+
+    #
+    # Macros from ndarrayobject.h
+    #
+    bint PyArray_CHKFLAGS(ndarray m, int flags) nogil
+    bint PyArray_IS_C_CONTIGUOUS(ndarray arr) nogil
+    bint PyArray_IS_F_CONTIGUOUS(ndarray arr) nogil
+    bint PyArray_ISCONTIGUOUS(ndarray m) nogil
+    bint PyArray_ISWRITEABLE(ndarray m) nogil
+    bint PyArray_ISALIGNED(ndarray m) nogil
+
+    int PyArray_NDIM(ndarray) nogil
+    bint PyArray_ISONESEGMENT(ndarray) nogil
+    bint PyArray_ISFORTRAN(ndarray) nogil
+    int PyArray_FORTRANIF(ndarray) nogil
+
+    void* PyArray_DATA(ndarray) nogil
+    char* PyArray_BYTES(ndarray) nogil
+
+    npy_intp* PyArray_DIMS(ndarray) nogil
+    npy_intp* PyArray_STRIDES(ndarray) nogil
+    npy_intp PyArray_DIM(ndarray, size_t) nogil
+    npy_intp PyArray_STRIDE(ndarray, size_t) nogil
+
+    PyObject *PyArray_BASE(ndarray) nogil  # returns borrowed reference!
+    PyArray_Descr *PyArray_DESCR(ndarray) nogil  # returns borrowed reference to dtype!
+    int PyArray_FLAGS(ndarray) nogil
+    npy_intp PyArray_ITEMSIZE(ndarray) nogil
+    int PyArray_TYPE(ndarray arr) nogil
+
+    object PyArray_GETITEM(ndarray arr, void *itemptr)
+    int PyArray_SETITEM(ndarray arr, void *itemptr, object obj) except -1
+
+    bint PyTypeNum_ISBOOL(int) nogil
+    bint PyTypeNum_ISUNSIGNED(int) nogil
+    bint PyTypeNum_ISSIGNED(int) nogil
+    bint PyTypeNum_ISINTEGER(int) nogil
+    bint PyTypeNum_ISFLOAT(int) nogil
+    bint PyTypeNum_ISNUMBER(int) nogil
+    bint PyTypeNum_ISSTRING(int) nogil
+    bint PyTypeNum_ISCOMPLEX(int) nogil
+    bint PyTypeNum_ISPYTHON(int) nogil
+    bint PyTypeNum_ISFLEXIBLE(int) nogil
+    bint PyTypeNum_ISUSERDEF(int) nogil
+    bint PyTypeNum_ISEXTENDED(int) nogil
+    bint PyTypeNum_ISOBJECT(int) nogil
+
+    bint PyDataType_ISBOOL(dtype) nogil
+    bint PyDataType_ISUNSIGNED(dtype) nogil
+    bint PyDataType_ISSIGNED(dtype) nogil
+    bint PyDataType_ISINTEGER(dtype) nogil
+    bint PyDataType_ISFLOAT(dtype) nogil
+    bint PyDataType_ISNUMBER(dtype) nogil
+    bint PyDataType_ISSTRING(dtype) nogil
+    bint PyDataType_ISCOMPLEX(dtype) nogil
+    bint PyDataType_ISPYTHON(dtype) nogil
+    bint PyDataType_ISFLEXIBLE(dtype) nogil
+    bint PyDataType_ISUSERDEF(dtype) nogil
+    bint PyDataType_ISEXTENDED(dtype) nogil
+    bint PyDataType_ISOBJECT(dtype) nogil
+    bint PyDataType_HASFIELDS(dtype) nogil
+    bint PyDataType_HASSUBARRAY(dtype) nogil
+
+    bint PyArray_ISBOOL(ndarray) nogil
+    bint PyArray_ISUNSIGNED(ndarray) nogil
+    bint PyArray_ISSIGNED(ndarray) nogil
+    bint PyArray_ISINTEGER(ndarray) nogil
+    bint PyArray_ISFLOAT(ndarray) nogil
+    bint PyArray_ISNUMBER(ndarray) nogil
+    bint PyArray_ISSTRING(ndarray) nogil
+    bint PyArray_ISCOMPLEX(ndarray) nogil
+    bint PyArray_ISPYTHON(ndarray) nogil
+    bint PyArray_ISFLEXIBLE(ndarray) nogil
+    bint PyArray_ISUSERDEF(ndarray) nogil
+    bint PyArray_ISEXTENDED(ndarray) nogil
+    bint PyArray_ISOBJECT(ndarray) nogil
+    bint PyArray_HASFIELDS(ndarray) nogil
+
+    bint PyArray_ISVARIABLE(ndarray) nogil
+
+    bint PyArray_SAFEALIGNEDCOPY(ndarray) nogil
+    bint PyArray_ISNBO(char) nogil              # works on ndarray.byteorder
+    bint PyArray_IsNativeByteOrder(char) nogil # works on ndarray.byteorder
+    bint PyArray_ISNOTSWAPPED(ndarray) nogil
+    bint PyArray_ISBYTESWAPPED(ndarray) nogil
+
+    bint PyArray_FLAGSWAP(ndarray, int) nogil
+
+    bint PyArray_ISCARRAY(ndarray) nogil
+    bint PyArray_ISCARRAY_RO(ndarray) nogil
+    bint PyArray_ISFARRAY(ndarray) nogil
+    bint PyArray_ISFARRAY_RO(ndarray) nogil
+    bint PyArray_ISBEHAVED(ndarray) nogil
+    bint PyArray_ISBEHAVED_RO(ndarray) nogil
+
+
+    bint PyDataType_ISNOTSWAPPED(dtype) nogil
+    bint PyDataType_ISBYTESWAPPED(dtype) nogil
+
+    bint PyArray_DescrCheck(object)
+
+    bint PyArray_Check(object)
+    bint PyArray_CheckExact(object)
+
+    # Cannot be supported due to out arg:
+    # bint PyArray_HasArrayInterfaceType(object, dtype, object, object&)
+    # bint PyArray_HasArrayInterface(op, out)
+
+
+    bint PyArray_IsZeroDim(object)
+    # Cannot be supported due to ## ## in macro:
+    # bint PyArray_IsScalar(object, verbatim work)
+    bint PyArray_CheckScalar(object)
+    bint PyArray_IsPythonNumber(object)
+    bint PyArray_IsPythonScalar(object)
+    bint PyArray_IsAnyScalar(object)
+    bint PyArray_CheckAnyScalar(object)
+
+    ndarray PyArray_GETCONTIGUOUS(ndarray)
+    bint PyArray_SAMESHAPE(ndarray, ndarray) nogil
+    npy_intp PyArray_SIZE(ndarray) nogil
+    npy_intp PyArray_NBYTES(ndarray) nogil
+
+    object PyArray_FROM_O(object)
+    object PyArray_FROM_OF(object m, int flags)
+    object PyArray_FROM_OT(object m, int type)
+    object PyArray_FROM_OTF(object m, int type, int flags)
+    object PyArray_FROMANY(object m, int type, int min, int max, int flags)
+    object PyArray_ZEROS(int nd, npy_intp* dims, int type, int fortran)
+    object PyArray_EMPTY(int nd, npy_intp* dims, int type, int fortran)
+    void PyArray_FILLWBYTE(object, int val)
+    npy_intp PyArray_REFCOUNT(object)
+    object PyArray_ContiguousFromAny(op, int, int min_depth, int max_depth)
+    unsigned char PyArray_EquivArrTypes(ndarray a1, ndarray a2)
+    bint PyArray_EquivByteorders(int b1, int b2) nogil
+    object PyArray_SimpleNew(int nd, npy_intp* dims, int typenum)
+    object PyArray_SimpleNewFromData(int nd, npy_intp* dims, int typenum, void* data)
+    #object PyArray_SimpleNewFromDescr(int nd, npy_intp* dims, dtype descr)
+    object PyArray_ToScalar(void* data, ndarray arr)
+
+    void* PyArray_GETPTR1(ndarray m, npy_intp i) nogil
+    void* PyArray_GETPTR2(ndarray m, npy_intp i, npy_intp j) nogil
+    void* PyArray_GETPTR3(ndarray m, npy_intp i, npy_intp j, npy_intp k) nogil
+    void* PyArray_GETPTR4(ndarray m, npy_intp i, npy_intp j, npy_intp k, npy_intp l) nogil
+
+    # Cannot be supported due to out arg
+    # void PyArray_DESCR_REPLACE(descr)
+
+
+    object PyArray_Copy(ndarray)
+    object PyArray_FromObject(object op, int type, int min_depth, int max_depth)
+    object PyArray_ContiguousFromObject(object op, int type, int min_depth, int max_depth)
+    object PyArray_CopyFromObject(object op, int type, int min_depth, int max_depth)
+
+    object PyArray_Cast(ndarray mp, int type_num)
+    object PyArray_Take(ndarray ap, object items, int axis)
+    object PyArray_Put(ndarray ap, object items, object values)
+
+    void PyArray_ITER_RESET(flatiter it) nogil
+    void PyArray_ITER_NEXT(flatiter it) nogil
+    void PyArray_ITER_GOTO(flatiter it, npy_intp* destination) nogil
+    void PyArray_ITER_GOTO1D(flatiter it, npy_intp ind) nogil
+    void* PyArray_ITER_DATA(flatiter it) nogil
+    bint PyArray_ITER_NOTDONE(flatiter it) nogil
+
+    void PyArray_MultiIter_RESET(broadcast multi) nogil
+    void PyArray_MultiIter_NEXT(broadcast multi) nogil
+    void PyArray_MultiIter_GOTO(broadcast multi, npy_intp dest) nogil
+    void PyArray_MultiIter_GOTO1D(broadcast multi, npy_intp ind) nogil
+    void* PyArray_MultiIter_DATA(broadcast multi, npy_intp i) nogil
+    void PyArray_MultiIter_NEXTi(broadcast multi, npy_intp i) nogil
+    bint PyArray_MultiIter_NOTDONE(broadcast multi) nogil
+
+    # Functions from __multiarray_api.h
+
+    # Functions taking dtype and returning object/ndarray are disabled
+    # for now as they steal dtype references. I'm conservative and disable
+    # more than is probably needed until it can be checked further.
+    int PyArray_SetNumericOps (object) except -1
+    object PyArray_GetNumericOps ()
+    int PyArray_INCREF (ndarray) except *  # uses PyArray_Item_INCREF...
+    int PyArray_XDECREF (ndarray) except *  # uses PyArray_Item_DECREF...
+    void PyArray_SetStringFunction (object, int)
+    dtype PyArray_DescrFromType (int)
+    object PyArray_TypeObjectFromType (int)
+    char * PyArray_Zero (ndarray)
+    char * PyArray_One (ndarray)
+    #object PyArray_CastToType (ndarray, dtype, int)
+    int PyArray_CastTo (ndarray, ndarray) except -1
+    int PyArray_CastAnyTo (ndarray, ndarray) except -1
+    int PyArray_CanCastSafely (int, int)  # writes errors
+    npy_bool PyArray_CanCastTo (dtype, dtype)  # writes errors
+    int PyArray_ObjectType (object, int) except 0
+    dtype PyArray_DescrFromObject (object, dtype)
+    #ndarray* PyArray_ConvertToCommonType (object, int *)
+    dtype PyArray_DescrFromScalar (object)
+    dtype PyArray_DescrFromTypeObject (object)
+    npy_intp PyArray_Size (object)
+    #object PyArray_Scalar (void *, dtype, object)
+    #object PyArray_FromScalar (object, dtype)
+    void PyArray_ScalarAsCtype (object, void *)
+    #int PyArray_CastScalarToCtype (object, void *, dtype)
+    #int PyArray_CastScalarDirect (object, dtype, void *, int)
+    object PyArray_ScalarFromObject (object)
+    #PyArray_VectorUnaryFunc * PyArray_GetCastFunc (dtype, int)
+    object PyArray_FromDims (int, int *, int)
+    #object PyArray_FromDimsAndDataAndDescr (int, int *, dtype, char *)
+    #object PyArray_FromAny (object, dtype, int, int, int, object)
+    object PyArray_EnsureArray (object)
+    object PyArray_EnsureAnyArray (object)
+    #object PyArray_FromFile (stdio.FILE *, dtype, npy_intp, char *)
+    #object PyArray_FromString (char *, npy_intp, dtype, npy_intp, char *)
+    #object PyArray_FromBuffer (object, dtype, npy_intp, npy_intp)
+    #object PyArray_FromIter (object, dtype, npy_intp)
+    object PyArray_Return (ndarray)
+    #object PyArray_GetField (ndarray, dtype, int)
+    #int PyArray_SetField (ndarray, dtype, int, object) except -1
+    object PyArray_Byteswap (ndarray, npy_bool)
+    object PyArray_Resize (ndarray, PyArray_Dims *, int, NPY_ORDER)
+    int PyArray_MoveInto (ndarray, ndarray) except -1
+    int PyArray_CopyInto (ndarray, ndarray) except -1
+    int PyArray_CopyAnyInto (ndarray, ndarray) except -1
+    int PyArray_CopyObject (ndarray, object) except -1
+    object PyArray_NewCopy (ndarray, NPY_ORDER)
+    object PyArray_ToList (ndarray)
+    object PyArray_ToString (ndarray, NPY_ORDER)
+    int PyArray_ToFile (ndarray, stdio.FILE *, char *, char *) except -1
+    int PyArray_Dump (object, object, int) except -1
+    object PyArray_Dumps (object, int)
+    int PyArray_ValidType (int)  # Cannot error
+    void PyArray_UpdateFlags (ndarray, int)
+    object PyArray_New (type, int, npy_intp *, int, npy_intp *, void *, int, int, object)
+    #object PyArray_NewFromDescr (type, dtype, int, npy_intp *, npy_intp *, void *, int, object)
+    #dtype PyArray_DescrNew (dtype)
+    dtype PyArray_DescrNewFromType (int)
+    double PyArray_GetPriority (object, double)  # clears errors as of 1.25
+    object PyArray_IterNew (object)
+    object PyArray_MultiIterNew (int, ...)
+
+    int PyArray_PyIntAsInt (object) except? -1
+    npy_intp PyArray_PyIntAsIntp (object)
+    int PyArray_Broadcast (broadcast) except -1
+    void PyArray_FillObjectArray (ndarray, object) except *
+    int PyArray_FillWithScalar (ndarray, object) except -1
+    npy_bool PyArray_CheckStrides (int, int, npy_intp, npy_intp, npy_intp *, npy_intp *)
+    dtype PyArray_DescrNewByteorder (dtype, char)
+    object PyArray_IterAllButAxis (object, int *)
+    #object PyArray_CheckFromAny (object, dtype, int, int, int, object)
+    #object PyArray_FromArray (ndarray, dtype, int)
+    object PyArray_FromInterface (object)
+    object PyArray_FromStructInterface (object)
+    #object PyArray_FromArrayAttr (object, dtype, object)
+    #NPY_SCALARKIND PyArray_ScalarKind (int, ndarray*)
+    int PyArray_CanCoerceScalar (int, int, NPY_SCALARKIND)
+    object PyArray_NewFlagsObject (object)
+    npy_bool PyArray_CanCastScalar (type, type)
+    #int PyArray_CompareUCS4 (npy_ucs4 *, npy_ucs4 *, register size_t)
+    int PyArray_RemoveSmallest (broadcast) except -1
+    int PyArray_ElementStrides (object)
+    void PyArray_Item_INCREF (char *, dtype) except *
+    void PyArray_Item_XDECREF (char *, dtype) except *
+    object PyArray_FieldNames (object)
+    object PyArray_Transpose (ndarray, PyArray_Dims *)
+    object PyArray_TakeFrom (ndarray, object, int, ndarray, NPY_CLIPMODE)
+    object PyArray_PutTo (ndarray, object, object, NPY_CLIPMODE)
+    object PyArray_PutMask (ndarray, object, object)
+    object PyArray_Repeat (ndarray, object, int)
+    object PyArray_Choose (ndarray, object, ndarray, NPY_CLIPMODE)
+    int PyArray_Sort (ndarray, int, NPY_SORTKIND) except -1
+    object PyArray_ArgSort (ndarray, int, NPY_SORTKIND)
+    object PyArray_SearchSorted (ndarray, object, NPY_SEARCHSIDE, PyObject *)
+    object PyArray_ArgMax (ndarray, int, ndarray)
+    object PyArray_ArgMin (ndarray, int, ndarray)
+    object PyArray_Reshape (ndarray, object)
+    object PyArray_Newshape (ndarray, PyArray_Dims *, NPY_ORDER)
+    object PyArray_Squeeze (ndarray)
+    #object PyArray_View (ndarray, dtype, type)
+    object PyArray_SwapAxes (ndarray, int, int)
+    object PyArray_Max (ndarray, int, ndarray)
+    object PyArray_Min (ndarray, int, ndarray)
+    object PyArray_Ptp (ndarray, int, ndarray)
+    object PyArray_Mean (ndarray, int, int, ndarray)
+    object PyArray_Trace (ndarray, int, int, int, int, ndarray)
+    object PyArray_Diagonal (ndarray, int, int, int)
+    object PyArray_Clip (ndarray, object, object, ndarray)
+    object PyArray_Conjugate (ndarray, ndarray)
+    object PyArray_Nonzero (ndarray)
+    object PyArray_Std (ndarray, int, int, ndarray, int)
+    object PyArray_Sum (ndarray, int, int, ndarray)
+    object PyArray_CumSum (ndarray, int, int, ndarray)
+    object PyArray_Prod (ndarray, int, int, ndarray)
+    object PyArray_CumProd (ndarray, int, int, ndarray)
+    object PyArray_All (ndarray, int, ndarray)
+    object PyArray_Any (ndarray, int, ndarray)
+    object PyArray_Compress (ndarray, object, int, ndarray)
+    object PyArray_Flatten (ndarray, NPY_ORDER)
+    object PyArray_Ravel (ndarray, NPY_ORDER)
+    npy_intp PyArray_MultiplyList (npy_intp *, int)
+    int PyArray_MultiplyIntList (int *, int)
+    void * PyArray_GetPtr (ndarray, npy_intp*)
+    int PyArray_CompareLists (npy_intp *, npy_intp *, int)
+    #int PyArray_AsCArray (object*, void *, npy_intp *, int, dtype)
+    #int PyArray_As1D (object*, char **, int *, int)
+    #int PyArray_As2D (object*, char ***, int *, int *, int)
+    int PyArray_Free (object, void *)
+    #int PyArray_Converter (object, object*)
+    int PyArray_IntpFromSequence (object, npy_intp *, int) except -1
+    object PyArray_Concatenate (object, int)
+    object PyArray_InnerProduct (object, object)
+    object PyArray_MatrixProduct (object, object)
+    object PyArray_CopyAndTranspose (object)
+    object PyArray_Correlate (object, object, int)
+    int PyArray_TypestrConvert (int, int)
+    #int PyArray_DescrConverter (object, dtype*) except 0
+    #int PyArray_DescrConverter2 (object, dtype*) except 0
+    int PyArray_IntpConverter (object, PyArray_Dims *) except 0
+    #int PyArray_BufferConverter (object, chunk) except 0
+    int PyArray_AxisConverter (object, int *) except 0
+    int PyArray_BoolConverter (object, npy_bool *) except 0
+    int PyArray_ByteorderConverter (object, char *) except 0
+    int PyArray_OrderConverter (object, NPY_ORDER *) except 0
+    unsigned char PyArray_EquivTypes (dtype, dtype)  # clears errors
+    #object PyArray_Zeros (int, npy_intp *, dtype, int)
+    #object PyArray_Empty (int, npy_intp *, dtype, int)
+    object PyArray_Where (object, object, object)
+    object PyArray_Arange (double, double, double, int)
+    #object PyArray_ArangeObj (object, object, object, dtype)
+    int PyArray_SortkindConverter (object, NPY_SORTKIND *) except 0
+    object PyArray_LexSort (object, int)
+    object PyArray_Round (ndarray, int, ndarray)
+    unsigned char PyArray_EquivTypenums (int, int)
+    int PyArray_RegisterDataType (dtype) except -1
+    int PyArray_RegisterCastFunc (dtype, int, PyArray_VectorUnaryFunc *) except -1
+    int PyArray_RegisterCanCast (dtype, int, NPY_SCALARKIND) except -1
+    #void PyArray_InitArrFuncs (PyArray_ArrFuncs *)
+    object PyArray_IntTupleFromIntp (int, npy_intp *)
+    int PyArray_TypeNumFromName (char *)
+    int PyArray_ClipmodeConverter (object, NPY_CLIPMODE *) except 0
+    #int PyArray_OutputConverter (object, ndarray*) except 0
+    object PyArray_BroadcastToShape (object, npy_intp *, int)
+    void _PyArray_SigintHandler (int)
+    void* _PyArray_GetSigintBuf ()
+    #int PyArray_DescrAlignConverter (object, dtype*) except 0
+    #int PyArray_DescrAlignConverter2 (object, dtype*) except 0
+    int PyArray_SearchsideConverter (object, void *) except 0
+    object PyArray_CheckAxis (ndarray, int *, int)
+    npy_intp PyArray_OverflowMultiplyList (npy_intp *, int)
+    int PyArray_CompareString (char *, char *, size_t)
+    int PyArray_SetBaseObject(ndarray, base) except -1 # NOTE: steals a reference to base! Use "set_array_base()" instead.
+
+
+# Typedefs that matches the runtime dtype objects in
+# the numpy module.
+
+# The ones that are commented out needs an IFDEF function
+# in Cython to enable them only on the right systems.
+
+ctypedef npy_int8       int8_t
+ctypedef npy_int16      int16_t
+ctypedef npy_int32      int32_t
+ctypedef npy_int64      int64_t
+#ctypedef npy_int96      int96_t
+#ctypedef npy_int128     int128_t
+
+ctypedef npy_uint8      uint8_t
+ctypedef npy_uint16     uint16_t
+ctypedef npy_uint32     uint32_t
+ctypedef npy_uint64     uint64_t
+#ctypedef npy_uint96     uint96_t
+#ctypedef npy_uint128    uint128_t
+
+ctypedef npy_float32    float32_t
+ctypedef npy_float64    float64_t
+#ctypedef npy_float80    float80_t
+#ctypedef npy_float128   float128_t
+
+ctypedef float complex  complex64_t
+ctypedef double complex complex128_t
+
+# The int types are mapped a bit surprising --
+# numpy.int corresponds to 'l' and numpy.long to 'q'
+ctypedef npy_long       int_t
+ctypedef npy_longlong   longlong_t
+
+ctypedef npy_ulong      uint_t
+ctypedef npy_ulonglong  ulonglong_t
+
+ctypedef npy_intp       intp_t
+ctypedef npy_uintp      uintp_t
+
+ctypedef npy_double     float_t
+ctypedef npy_double     double_t
+ctypedef npy_longdouble longdouble_t
+
+ctypedef npy_cfloat      cfloat_t
+ctypedef npy_cdouble     cdouble_t
+ctypedef npy_clongdouble clongdouble_t
+
+ctypedef npy_cdouble     complex_t
+
+cdef inline object PyArray_MultiIterNew1(a):
+    return PyArray_MultiIterNew(1, <void*>a)
+
+cdef inline object PyArray_MultiIterNew2(a, b):
+    return PyArray_MultiIterNew(2, <void*>a, <void*>b)
+
+cdef inline object PyArray_MultiIterNew3(a, b, c):
+    return PyArray_MultiIterNew(3, <void*>a, <void*>b, <void*> c)
+
+cdef inline object PyArray_MultiIterNew4(a, b, c, d):
+    return PyArray_MultiIterNew(4, <void*>a, <void*>b, <void*>c, <void*> d)
+
+cdef inline object PyArray_MultiIterNew5(a, b, c, d, e):
+    return PyArray_MultiIterNew(5, <void*>a, <void*>b, <void*>c, <void*> d, <void*> e)
+
+cdef inline tuple PyDataType_SHAPE(dtype d):
+    if PyDataType_HASSUBARRAY(d):
+        return <tuple>d.subarray.shape
+    else:
+        return ()
+
+
+cdef extern from "numpy/ndarrayobject.h":
+    PyTypeObject PyTimedeltaArrType_Type
+    PyTypeObject PyDatetimeArrType_Type
+    ctypedef int64_t npy_timedelta
+    ctypedef int64_t npy_datetime
+
+cdef extern from "numpy/ndarraytypes.h":
+    ctypedef struct PyArray_DatetimeMetaData:
+        NPY_DATETIMEUNIT base
+        int64_t num
+
+cdef extern from "numpy/arrayscalars.h":
+
+    # abstract types
+    ctypedef class numpy.generic [object PyObject]:
+        pass
+    ctypedef class numpy.number [object PyObject]:
+        pass
+    ctypedef class numpy.integer [object PyObject]:
+        pass
+    ctypedef class numpy.signedinteger [object PyObject]:
+        pass
+    ctypedef class numpy.unsignedinteger [object PyObject]:
+        pass
+    ctypedef class numpy.inexact [object PyObject]:
+        pass
+    ctypedef class numpy.floating [object PyObject]:
+        pass
+    ctypedef class numpy.complexfloating [object PyObject]:
+        pass
+    ctypedef class numpy.flexible [object PyObject]:
+        pass
+    ctypedef class numpy.character [object PyObject]:
+        pass
+
+    ctypedef struct PyDatetimeScalarObject:
+        # PyObject_HEAD
+        npy_datetime obval
+        PyArray_DatetimeMetaData obmeta
+
+    ctypedef struct PyTimedeltaScalarObject:
+        # PyObject_HEAD
+        npy_timedelta obval
+        PyArray_DatetimeMetaData obmeta
+
+    ctypedef enum NPY_DATETIMEUNIT:
+        NPY_FR_Y
+        NPY_FR_M
+        NPY_FR_W
+        NPY_FR_D
+        NPY_FR_B
+        NPY_FR_h
+        NPY_FR_m
+        NPY_FR_s
+        NPY_FR_ms
+        NPY_FR_us
+        NPY_FR_ns
+        NPY_FR_ps
+        NPY_FR_fs
+        NPY_FR_as
+        NPY_FR_GENERIC
+
+
+#
+# ufunc API
+#
+
+cdef extern from "numpy/ufuncobject.h":
+
+    ctypedef void (*PyUFuncGenericFunction) (char **, npy_intp *, npy_intp *, void *)
+
+    ctypedef class numpy.ufunc [object PyUFuncObject, check_size ignore]:
+        cdef:
+            int nin, nout, nargs
+            int identity
+            PyUFuncGenericFunction *functions
+            void **data
+            int ntypes
+            int check_return
+            char *name
+            char *types
+            char *doc
+            void *ptr
+            PyObject *obj
+            PyObject *userloops
+
+    cdef enum:
+        PyUFunc_Zero
+        PyUFunc_One
+        PyUFunc_None
+        UFUNC_ERR_IGNORE
+        UFUNC_ERR_WARN
+        UFUNC_ERR_RAISE
+        UFUNC_ERR_CALL
+        UFUNC_ERR_PRINT
+        UFUNC_ERR_LOG
+        UFUNC_MASK_DIVIDEBYZERO
+        UFUNC_MASK_OVERFLOW
+        UFUNC_MASK_UNDERFLOW
+        UFUNC_MASK_INVALID
+        UFUNC_SHIFT_DIVIDEBYZERO
+        UFUNC_SHIFT_OVERFLOW
+        UFUNC_SHIFT_UNDERFLOW
+        UFUNC_SHIFT_INVALID
+        UFUNC_FPE_DIVIDEBYZERO
+        UFUNC_FPE_OVERFLOW
+        UFUNC_FPE_UNDERFLOW
+        UFUNC_FPE_INVALID
+        UFUNC_ERR_DEFAULT
+        UFUNC_ERR_DEFAULT2
+
+    object PyUFunc_FromFuncAndData(PyUFuncGenericFunction *,
+          void **, char *, int, int, int, int, char *, char *, int)
+    int PyUFunc_RegisterLoopForType(ufunc, int,
+                                    PyUFuncGenericFunction, int *, void *) except -1
+    void PyUFunc_f_f_As_d_d \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_d_d \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_f_f \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_g_g \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_F_F_As_D_D \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_F_F \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_D_D \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_G_G \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_O_O \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_ff_f_As_dd_d \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_ff_f \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_dd_d \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_gg_g \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_FF_F_As_DD_D \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_DD_D \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_FF_F \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_GG_G \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_OO_O \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_O_O_method \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_OO_O_method \
+         (char **, npy_intp *, npy_intp *, void *)
+    void PyUFunc_On_Om \
+         (char **, npy_intp *, npy_intp *, void *)
+    int PyUFunc_GetPyValues \
+        (char *, int *, int *, PyObject **)
+    int PyUFunc_checkfperr \
+           (int, PyObject *, int *)
+    void PyUFunc_clearfperr()
+    int PyUFunc_getfperr()
+    int PyUFunc_handlefperr \
+        (int, PyObject *, int, int *) except -1
+    int PyUFunc_ReplaceLoopBySignature \
+        (ufunc, PyUFuncGenericFunction, int *, PyUFuncGenericFunction *)
+    object PyUFunc_FromFuncAndDataAndSignature \
+             (PyUFuncGenericFunction *, void **, char *, int, int, int,
+              int, char *, char *, int, char *)
+
+    int _import_umath() except -1
+
+cdef inline void set_array_base(ndarray arr, object base):
+    Py_INCREF(base) # important to do this before stealing the reference below!
+    PyArray_SetBaseObject(arr, base)
+
+cdef inline object get_array_base(ndarray arr):
+    base = PyArray_BASE(arr)
+    if base is NULL:
+        return None
+    return <object>base
+
+# Versions of the import_* functions which are more suitable for
+# Cython code.
+cdef inline int import_array() except -1:
+    try:
+        __pyx_import_array()
+    except Exception:
+        raise ImportError("numpy.core.multiarray failed to import")
+
+cdef inline int import_umath() except -1:
+    try:
+        _import_umath()
+    except Exception:
+        raise ImportError("numpy.core.umath failed to import")
+
+cdef inline int import_ufunc() except -1:
+    try:
+        _import_umath()
+    except Exception:
+        raise ImportError("numpy.core.umath failed to import")
+
+cdef extern from *:
+    # Leave a marker that the NumPy declarations came from this file
+    # See https://github.com/cython/cython/issues/3573
+    """
+    /* NumPy API declarations from "numpy/__init__.pxd" */
+    """
+
+
+cdef inline bint is_timedelta64_object(object obj):
+    """
+    Cython equivalent of `isinstance(obj, np.timedelta64)`
+
+    Parameters
+    ----------
+    obj : object
+
+    Returns
+    -------
+    bool
+    """
+    return PyObject_TypeCheck(obj, &PyTimedeltaArrType_Type)
+
+
+cdef inline bint is_datetime64_object(object obj):
+    """
+    Cython equivalent of `isinstance(obj, np.datetime64)`
+
+    Parameters
+    ----------
+    obj : object
+
+    Returns
+    -------
+    bool
+    """
+    return PyObject_TypeCheck(obj, &PyDatetimeArrType_Type)
+
+
+cdef inline npy_datetime get_datetime64_value(object obj) nogil:
+    """
+    returns the int64 value underlying scalar numpy datetime64 object
+
+    Note that to interpret this as a datetime, the corresponding unit is
+    also needed.  That can be found using `get_datetime64_unit`.
+    """
+    return (<PyDatetimeScalarObject*>obj).obval
+
+
+cdef inline npy_timedelta get_timedelta64_value(object obj) nogil:
+    """
+    returns the int64 value underlying scalar numpy timedelta64 object
+    """
+    return (<PyTimedeltaScalarObject*>obj).obval
+
+
+cdef inline NPY_DATETIMEUNIT get_datetime64_unit(object obj) nogil:
+    """
+    returns the unit part of the dtype for a numpy datetime64 object.
+    """
+    return <NPY_DATETIMEUNIT>(<PyDatetimeScalarObject*>obj).obmeta.base

.venv/lib/python3.11/site-packages/numpy/__init__.py

@@ -0,0 +1,461 @@
+"""
+NumPy
+=====
+
+Provides
+  1. An array object of arbitrary homogeneous items
+  2. Fast mathematical operations over arrays
+  3. Linear Algebra, Fourier Transforms, Random Number Generation
+
+How to use the documentation
+----------------------------
+Documentation is available in two forms: docstrings provided
+with the code, and a loose standing reference guide, available from
+`the NumPy homepage <https://numpy.org>`_.
+
+We recommend exploring the docstrings using
+`IPython <https://ipython.org>`_, an advanced Python shell with
+TAB-completion and introspection capabilities.  See below for further
+instructions.
+
+The docstring examples assume that `numpy` has been imported as ``np``::
+
+  >>> import numpy as np
+
+Code snippets are indicated by three greater-than signs::
+
+  >>> x = 42
+  >>> x = x + 1
+
+Use the built-in ``help`` function to view a function's docstring::
+
+  >>> help(np.sort)
+  ... # doctest: +SKIP
+
+For some objects, ``np.info(obj)`` may provide additional help.  This is
+particularly true if you see the line "Help on ufunc object:" at the top
+of the help() page.  Ufuncs are implemented in C, not Python, for speed.
+The native Python help() does not know how to view their help, but our
+np.info() function does.
+
+To search for documents containing a keyword, do::
+
+  >>> np.lookfor('keyword')
+  ... # doctest: +SKIP
+
+General-purpose documents like a glossary and help on the basic concepts
+of numpy are available under the ``doc`` sub-module::
+
+  >>> from numpy import doc
+  >>> help(doc)
+  ... # doctest: +SKIP
+
+Available subpackages
+---------------------
+lib
+    Basic functions used by several sub-packages.
+random
+    Core Random Tools
+linalg
+    Core Linear Algebra Tools
+fft
+    Core FFT routines
+polynomial
+    Polynomial tools
+testing
+    NumPy testing tools
+distutils
+    Enhancements to distutils with support for
+    Fortran compilers support and more  (for Python <= 3.11).
+
+Utilities
+---------
+test
+    Run numpy unittests
+show_config
+    Show numpy build configuration
+matlib
+    Make everything matrices.
+__version__
+    NumPy version string
+
+Viewing documentation using IPython
+-----------------------------------
+
+Start IPython and import `numpy` usually under the alias ``np``: `import
+numpy as np`.  Then, directly past or use the ``%cpaste`` magic to paste
+examples into the shell.  To see which functions are available in `numpy`,
+type ``np.<TAB>`` (where ``<TAB>`` refers to the TAB key), or use
+``np.*cos*?<ENTER>`` (where ``<ENTER>`` refers to the ENTER key) to narrow
+down the list.  To view the docstring for a function, use
+``np.cos?<ENTER>`` (to view the docstring) and ``np.cos??<ENTER>`` (to view
+the source code).
+
+Copies vs. in-place operation
+-----------------------------
+Most of the functions in `numpy` return a copy of the array argument
+(e.g., `np.sort`).  In-place versions of these functions are often
+available as array methods, i.e. ``x = np.array([1,2,3]); x.sort()``.
+Exceptions to this rule are documented.
+
+"""
+import sys
+import warnings
+
+from ._globals import _NoValue, _CopyMode
+# These exceptions were moved in 1.25 and are hidden from __dir__()
+from .exceptions import (
+    ComplexWarning, ModuleDeprecationWarning, VisibleDeprecationWarning,
+    TooHardError, AxisError)
+
+
+# If a version with git hash was stored, use that instead
+from . import version
+from .version import __version__
+
+# We first need to detect if we're being called as part of the numpy setup
+# procedure itself in a reliable manner.
+try:
+    __NUMPY_SETUP__
+except NameError:
+    __NUMPY_SETUP__ = False
+
+if __NUMPY_SETUP__:
+    sys.stderr.write('Running from numpy source directory.\n')
+else:
+    # Allow distributors to run custom init code before importing numpy.core
+    from . import _distributor_init
+
+    try:
+        from numpy.__config__ import show as show_config
+    except ImportError as e:
+        msg = """Error importing numpy: you should not try to import numpy from
+        its source directory; please exit the numpy source tree, and relaunch
+        your python interpreter from there."""
+        raise ImportError(msg) from e
+
+    __all__ = [
+        'exceptions', 'ModuleDeprecationWarning', 'VisibleDeprecationWarning',
+        'ComplexWarning', 'TooHardError', 'AxisError']
+
+    # mapping of {name: (value, deprecation_msg)}
+    __deprecated_attrs__ = {}
+
+    from . import core
+    from .core import *
+    from . import compat
+    from . import exceptions
+    from . import dtypes
+    from . import lib
+    # NOTE: to be revisited following future namespace cleanup.
+    # See gh-14454 and gh-15672 for discussion.
+    from .lib import *
+
+    from . import linalg
+    from . import fft
+    from . import polynomial
+    from . import random
+    from . import ctypeslib
+    from . import ma
+    from . import matrixlib as _mat
+    from .matrixlib import *
+
+    # Deprecations introduced in NumPy 1.20.0, 2020-06-06
+    import builtins as _builtins
+
+    _msg = (
+        "module 'numpy' has no attribute '{n}'.\n"
+        "`np.{n}` was a deprecated alias for the builtin `{n}`. "
+        "To avoid this error in existing code, use `{n}` by itself. "
+        "Doing this will not modify any behavior and is safe. {extended_msg}\n"
+        "The aliases was originally deprecated in NumPy 1.20; for more "
+        "details and guidance see the original release note at:\n"
+        "    https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations")
+
+    _specific_msg = (
+        "If you specifically wanted the numpy scalar type, use `np.{}` here.")
+
+    _int_extended_msg = (
+        "When replacing `np.{}`, you may wish to use e.g. `np.int64` "
+        "or `np.int32` to specify the precision. If you wish to review "
+        "your current use, check the release note link for "
+        "additional information.")
+
+    _type_info = [
+        ("object", ""),  # The NumPy scalar only exists by name.
+        ("bool", _specific_msg.format("bool_")),
+        ("float", _specific_msg.format("float64")),
+        ("complex", _specific_msg.format("complex128")),
+        ("str", _specific_msg.format("str_")),
+        ("int", _int_extended_msg.format("int"))]
+
+    __former_attrs__ = {
+         n: _msg.format(n=n, extended_msg=extended_msg)
+         for n, extended_msg in _type_info
+     }
+
+    # Future warning introduced in NumPy 1.24.0, 2022-11-17
+    _msg = (
+        "`np.{n}` is a deprecated alias for `{an}`.  (Deprecated NumPy 1.24)")
+
+    # Some of these are awkward (since `np.str` may be preferable in the long
+    # term), but overall the names ending in 0 seem undesirable
+    _type_info = [
+        ("bool8", bool_, "np.bool_"),
+        ("int0", intp, "np.intp"),
+        ("uint0", uintp, "np.uintp"),
+        ("str0", str_, "np.str_"),
+        ("bytes0", bytes_, "np.bytes_"),
+        ("void0", void, "np.void"),
+        ("object0", object_,
+            "`np.object0` is a deprecated alias for `np.object_`. "
+            "`object` can be used instead.  (Deprecated NumPy 1.24)")]
+
+    # Some of these could be defined right away, but most were aliases to
+    # the Python objects and only removed in NumPy 1.24.  Defining them should
+    # probably wait for NumPy 1.26 or 2.0.
+    # When defined, these should possibly not be added to `__all__` to avoid
+    # import with `from numpy import *`.
+    __future_scalars__ = {"bool", "long", "ulong", "str", "bytes", "object"}
+
+    __deprecated_attrs__.update({
+        n: (alias, _msg.format(n=n, an=an)) for n, alias, an in _type_info})
+
+    import math
+
+    __deprecated_attrs__['math'] = (math,
+        "`np.math` is a deprecated alias for the standard library `math` "
+        "module (Deprecated Numpy 1.25). Replace usages of `np.math` with "
+        "`math`")
+
+    del math, _msg, _type_info
+
+    from .core import abs
+    # now that numpy modules are imported, can initialize limits
+    core.getlimits._register_known_types()
+
+    __all__.extend(['__version__', 'show_config'])
+    __all__.extend(core.__all__)
+    __all__.extend(_mat.__all__)
+    __all__.extend(lib.__all__)
+    __all__.extend(['linalg', 'fft', 'random', 'ctypeslib', 'ma'])
+
+    # Remove min and max from __all__ to avoid `from numpy import *` override
+    # the builtins min/max. Temporary fix for 1.25.x/1.26.x, see gh-24229.
+    __all__.remove('min')
+    __all__.remove('max')
+    __all__.remove('round')
+
+    # Remove one of the two occurrences of `issubdtype`, which is exposed as
+    # both `numpy.core.issubdtype` and `numpy.lib.issubdtype`.
+    __all__.remove('issubdtype')
+
+    # These are exported by np.core, but are replaced by the builtins below
+    # remove them to ensure that we don't end up with `np.long == np.int_`,
+    # which would be a breaking change.
+    del long, unicode
+    __all__.remove('long')
+    __all__.remove('unicode')
+
+    # Remove things that are in the numpy.lib but not in the numpy namespace
+    # Note that there is a test (numpy/tests/test_public_api.py:test_numpy_namespace)
+    # that prevents adding more things to the main namespace by accident.
+    # The list below will grow until the `from .lib import *` fixme above is
+    # taken care of
+    __all__.remove('Arrayterator')
+    del Arrayterator
+
+    # These names were removed in NumPy 1.20.  For at least one release,
+    # attempts to access these names in the numpy namespace will trigger
+    # a warning, and calling the function will raise an exception.
+    _financial_names = ['fv', 'ipmt', 'irr', 'mirr', 'nper', 'npv', 'pmt',
+                        'ppmt', 'pv', 'rate']
+    __expired_functions__ = {
+        name: (f'In accordance with NEP 32, the function {name} was removed '
+               'from NumPy version 1.20.  A replacement for this function '
+               'is available in the numpy_financial library: '
+               'https://pypi.org/project/numpy-financial')
+        for name in _financial_names}
+
+    # Filter out Cython harmless warnings
+    warnings.filterwarnings("ignore", message="numpy.dtype size changed")
+    warnings.filterwarnings("ignore", message="numpy.ufunc size changed")
+    warnings.filterwarnings("ignore", message="numpy.ndarray size changed")
+
+    # oldnumeric and numarray were removed in 1.9. In case some packages import
+    # but do not use them, we define them here for backward compatibility.
+    oldnumeric = 'removed'
+    numarray = 'removed'
+
+    def __getattr__(attr):
+        # Warn for expired attributes, and return a dummy function
+        # that always raises an exception.
+        import warnings
+        import math
+        try:
+            msg = __expired_functions__[attr]
+        except KeyError:
+            pass
+        else:
+            warnings.warn(msg, DeprecationWarning, stacklevel=2)
+
+            def _expired(*args, **kwds):
+                raise RuntimeError(msg)
+
+            return _expired
+
+        # Emit warnings for deprecated attributes
+        try:
+            val, msg = __deprecated_attrs__[attr]
+        except KeyError:
+            pass
+        else:
+            warnings.warn(msg, DeprecationWarning, stacklevel=2)
+            return val
+
+        if attr in __future_scalars__:
+            # And future warnings for those that will change, but also give
+            # the AttributeError
+            warnings.warn(
+                f"In the future `np.{attr}` will be defined as the "
+                "corresponding NumPy scalar.", FutureWarning, stacklevel=2)
+
+        if attr in __former_attrs__:
+            raise AttributeError(__former_attrs__[attr])
+
+        if attr == 'testing':
+            import numpy.testing as testing
+            return testing
+        elif attr == 'Tester':
+            "Removed in NumPy 1.25.0"
+            raise RuntimeError("Tester was removed in NumPy 1.25.")
+
+        raise AttributeError("module {!r} has no attribute "
+                             "{!r}".format(__name__, attr))
+
+    def __dir__():
+        public_symbols = globals().keys() | {'testing'}
+        public_symbols -= {
+            "core", "matrixlib",
+            # These were moved in 1.25 and may be deprecated eventually:
+            "ModuleDeprecationWarning", "VisibleDeprecationWarning",
+            "ComplexWarning", "TooHardError", "AxisError"
+        }
+        return list(public_symbols)
+
+    # Pytest testing
+    from numpy._pytesttester import PytestTester
+    test = PytestTester(__name__)
+    del PytestTester
+
+    def _sanity_check():
+        """
+        Quick sanity checks for common bugs caused by environment.
+        There are some cases e.g. with wrong BLAS ABI that cause wrong
+        results under specific runtime conditions that are not necessarily
+        achieved during test suite runs, and it is useful to catch those early.
+
+        See https://github.com/numpy/numpy/issues/8577 and other
+        similar bug reports.
+
+        """
+        try:
+            x = ones(2, dtype=float32)
+            if not abs(x.dot(x) - float32(2.0)) < 1e-5:
+                raise AssertionError()
+        except AssertionError:
+            msg = ("The current Numpy installation ({!r}) fails to "
+                   "pass simple sanity checks. This can be caused for example "
+                   "by incorrect BLAS library being linked in, or by mixing "
+                   "package managers (pip, conda, apt, ...). Search closed "
+                   "numpy issues for similar problems.")
+            raise RuntimeError(msg.format(__file__)) from None
+
+    _sanity_check()
+    del _sanity_check
+
+    def _mac_os_check():
+        """
+        Quick Sanity check for Mac OS look for accelerate build bugs.
+        Testing numpy polyfit calls init_dgelsd(LAPACK)
+        """
+        try:
+            c = array([3., 2., 1.])
+            x = linspace(0, 2, 5)
+            y = polyval(c, x)
+            _ = polyfit(x, y, 2, cov=True)
+        except ValueError:
+            pass
+
+    if sys.platform == "darwin":
+        from . import exceptions
+        with warnings.catch_warnings(record=True) as w:
+            _mac_os_check()
+            # Throw runtime error, if the test failed Check for warning and error_message
+            if len(w) > 0:
+                for _wn in w:
+                    if _wn.category is exceptions.RankWarning:
+                        # Ignore other warnings, they may not be relevant (see gh-25433).
+                        error_message = f"{_wn.category.__name__}: {str(_wn.message)}"
+                        msg = (
+                            "Polyfit sanity test emitted a warning, most likely due "
+                            "to using a buggy Accelerate backend."
+                            "\nIf you compiled yourself, more information is available at:"
+                            "\nhttps://numpy.org/devdocs/building/index.html"
+                            "\nOtherwise report this to the vendor "
+                            "that provided NumPy.\n\n{}\n".format(error_message))
+                        raise RuntimeError(msg)
+                del _wn
+            del w
+    del _mac_os_check
+
+    # We usually use madvise hugepages support, but on some old kernels it
+    # is slow and thus better avoided.
+    # Specifically kernel version 4.6 had a bug fix which probably fixed this:
+    # https://github.com/torvalds/linux/commit/7cf91a98e607c2f935dbcc177d70011e95b8faff
+    import os
+    use_hugepage = os.environ.get("NUMPY_MADVISE_HUGEPAGE", None)
+    if sys.platform == "linux" and use_hugepage is None:
+        # If there is an issue with parsing the kernel version,
+        # set use_hugepages to 0. Usage of LooseVersion will handle
+        # the kernel version parsing better, but avoided since it
+        # will increase the import time. See: #16679 for related discussion.
+        try:
+            use_hugepage = 1
+            kernel_version = os.uname().release.split(".")[:2]
+            kernel_version = tuple(int(v) for v in kernel_version)
+            if kernel_version < (4, 6):
+                use_hugepage = 0
+        except ValueError:
+            use_hugepages = 0
+    elif use_hugepage is None:
+        # This is not Linux, so it should not matter, just enable anyway
+        use_hugepage = 1
+    else:
+        use_hugepage = int(use_hugepage)
+
+    # Note that this will currently only make a difference on Linux
+    core.multiarray._set_madvise_hugepage(use_hugepage)
+    del use_hugepage
+
+    # Give a warning if NumPy is reloaded or imported on a sub-interpreter
+    # We do this from python, since the C-module may not be reloaded and
+    # it is tidier organized.
+    core.multiarray._multiarray_umath._reload_guard()
+
+    # default to "weak" promotion for "NumPy 2".
+    core._set_promotion_state(
+        os.environ.get("NPY_PROMOTION_STATE",
+                       "weak" if _using_numpy2_behavior() else "legacy"))
+
+    # Tell PyInstaller where to find hook-numpy.py
+    def _pyinstaller_hooks_dir():
+        from pathlib import Path
+        return [str(Path(__file__).with_name("_pyinstaller").resolve())]
+
+    # Remove symbols imported for internal use
+    del os
+
+
+# Remove symbols imported for internal use
+del sys, warnings

.venv/lib/python3.11/site-packages/numpy/_globals.py

@@ -0,0 +1,95 @@
+"""
+Module defining global singleton classes.
+
+This module raises a RuntimeError if an attempt to reload it is made. In that
+way the identities of the classes defined here are fixed and will remain so
+even if numpy itself is reloaded. In particular, a function like the following
+will still work correctly after numpy is reloaded::
+
+    def foo(arg=np._NoValue):
+        if arg is np._NoValue:
+            ...
+
+That was not the case when the singleton classes were defined in the numpy
+``__init__.py`` file. See gh-7844 for a discussion of the reload problem that
+motivated this module.
+
+"""
+import enum
+
+from ._utils import set_module as _set_module
+
+__all__ = ['_NoValue', '_CopyMode']
+
+
+# Disallow reloading this module so as to preserve the identities of the
+# classes defined here.
+if '_is_loaded' in globals():
+    raise RuntimeError('Reloading numpy._globals is not allowed')
+_is_loaded = True
+
+
+class _NoValueType:
+    """Special keyword value.
+
+    The instance of this class may be used as the default value assigned to a
+    keyword if no other obvious default (e.g., `None`) is suitable,
+
+    Common reasons for using this keyword are:
+
+    - A new keyword is added to a function, and that function forwards its
+      inputs to another function or method which can be defined outside of
+      NumPy. For example, ``np.std(x)`` calls ``x.std``, so when a ``keepdims``
+      keyword was added that could only be forwarded if the user explicitly
+      specified ``keepdims``; downstream array libraries may not have added
+      the same keyword, so adding ``x.std(..., keepdims=keepdims)``
+      unconditionally could have broken previously working code.
+    - A keyword is being deprecated, and a deprecation warning must only be
+      emitted when the keyword is used.
+
+    """
+    __instance = None
+    def __new__(cls):
+        # ensure that only one instance exists
+        if not cls.__instance:
+            cls.__instance = super().__new__(cls)
+        return cls.__instance
+
+    def __repr__(self):
+        return "<no value>"
+
+
+_NoValue = _NoValueType()
+
+
+@_set_module("numpy")
+class _CopyMode(enum.Enum):
+    """
+    An enumeration for the copy modes supported
+    by numpy.copy() and numpy.array(). The following three modes are supported,
+
+    - ALWAYS: This means that a deep copy of the input
+              array will always be taken.
+    - IF_NEEDED: This means that a deep copy of the input
+                 array will be taken only if necessary.
+    - NEVER: This means that the deep copy will never be taken.
+             If a copy cannot be avoided then a `ValueError` will be
+             raised.
+
+    Note that the buffer-protocol could in theory do copies.  NumPy currently
+    assumes an object exporting the buffer protocol will never do this.
+    """
+
+    ALWAYS = True
+    IF_NEEDED = False
+    NEVER = 2
+
+    def __bool__(self):
+        # For backwards compatibility
+        if self == _CopyMode.ALWAYS:
+            return True
+
+        if self == _CopyMode.IF_NEEDED:
+            return False
+
+        raise ValueError(f"{self} is neither True nor False.")

.venv/lib/python3.11/site-packages/numpy/array_api/__init__.py

@@ -0,0 +1,387 @@
+"""
+A NumPy sub-namespace that conforms to the Python array API standard.
+
+This submodule accompanies NEP 47, which proposes its inclusion in NumPy. It
+is still considered experimental, and will issue a warning when imported.
+
+This is a proof-of-concept namespace that wraps the corresponding NumPy
+functions to give a conforming implementation of the Python array API standard
+(https://data-apis.github.io/array-api/latest/). The standard is currently in
+an RFC phase and comments on it are both welcome and encouraged. Comments
+should be made either at https://github.com/data-apis/array-api or at
+https://github.com/data-apis/consortium-feedback/discussions.
+
+NumPy already follows the proposed spec for the most part, so this module
+serves mostly as a thin wrapper around it. However, NumPy also implements a
+lot of behavior that is not included in the spec, so this serves as a
+restricted subset of the API. Only those functions that are part of the spec
+are included in this namespace, and all functions are given with the exact
+signature given in the spec, including the use of position-only arguments, and
+omitting any extra keyword arguments implemented by NumPy but not part of the
+spec. The behavior of some functions is also modified from the NumPy behavior
+to conform to the standard. Note that the underlying array object itself is
+wrapped in a wrapper Array() class, but is otherwise unchanged. This submodule
+is implemented in pure Python with no C extensions.
+
+The array API spec is designed as a "minimal API subset" and explicitly allows
+libraries to include behaviors not specified by it. But users of this module
+that intend to write portable code should be aware that only those behaviors
+that are listed in the spec are guaranteed to be implemented across libraries.
+Consequently, the NumPy implementation was chosen to be both conforming and
+minimal, so that users can use this implementation of the array API namespace
+and be sure that behaviors that it defines will be available in conforming
+namespaces from other libraries.
+
+A few notes about the current state of this submodule:
+
+- There is a test suite that tests modules against the array API standard at
+  https://github.com/data-apis/array-api-tests. The test suite is still a work
+  in progress, but the existing tests pass on this module, with a few
+  exceptions:
+
+  - DLPack support (see https://github.com/data-apis/array-api/pull/106) is
+    not included here, as it requires a full implementation in NumPy proper
+    first.
+
+  The test suite is not yet complete, and even the tests that exist are not
+  guaranteed to give a comprehensive coverage of the spec. Therefore, when
+  reviewing and using this submodule, you should refer to the standard
+  documents themselves. There are some tests in numpy.array_api.tests, but
+  they primarily focus on things that are not tested by the official array API
+  test suite.
+
+- There is a custom array object, numpy.array_api.Array, which is returned by
+  all functions in this module. All functions in the array API namespace
+  implicitly assume that they will only receive this object as input. The only
+  way to create instances of this object is to use one of the array creation
+  functions. It does not have a public constructor on the object itself. The
+  object is a small wrapper class around numpy.ndarray. The main purpose of it
+  is to restrict the namespace of the array object to only those dtypes and
+  only those methods that are required by the spec, as well as to limit/change
+  certain behavior that differs in the spec. In particular:
+
+  - The array API namespace does not have scalar objects, only 0-D arrays.
+    Operations on Array that would create a scalar in NumPy create a 0-D
+    array.
+
+  - Indexing: Only a subset of indices supported by NumPy are required by the
+    spec. The Array object restricts indexing to only allow those types of
+    indices that are required by the spec. See the docstring of the
+    numpy.array_api.Array._validate_indices helper function for more
+    information.
+
+  - Type promotion: Some type promotion rules are different in the spec. In
+    particular, the spec does not have any value-based casting. The spec also
+    does not require cross-kind casting, like integer -> floating-point. Only
+    those promotions that are explicitly required by the array API
+    specification are allowed in this module. See NEP 47 for more info.
+
+  - Functions do not automatically call asarray() on their input, and will not
+    work if the input type is not Array. The exception is array creation
+    functions, and Python operators on the Array object, which accept Python
+    scalars of the same type as the array dtype.
+
+- All functions include type annotations, corresponding to those given in the
+  spec (see _typing.py for definitions of some custom types). These do not
+  currently fully pass mypy due to some limitations in mypy.
+
+- Dtype objects are just the NumPy dtype objects, e.g., float64 =
+  np.dtype('float64'). The spec does not require any behavior on these dtype
+  objects other than that they be accessible by name and be comparable by
+  equality, but it was considered too much extra complexity to create custom
+  objects to represent dtypes.
+
+- All places where the implementations in this submodule are known to deviate
+  from their corresponding functions in NumPy are marked with "# Note:"
+  comments.
+
+Still TODO in this module are:
+
+- DLPack support for numpy.ndarray is still in progress. See
+  https://github.com/numpy/numpy/pull/19083.
+
+- The copy=False keyword argument to asarray() is not yet implemented. This
+  requires support in numpy.asarray() first.
+
+- Some functions are not yet fully tested in the array API test suite, and may
+  require updates that are not yet known until the tests are written.
+
+- The spec is still in an RFC phase and may still have minor updates, which
+  will need to be reflected here.
+
+- Complex number support in array API spec is planned but not yet finalized,
+  as are the fft extension and certain linear algebra functions such as eig
+  that require complex dtypes.
+
+"""
+
+import warnings
+
+warnings.warn(
+    "The numpy.array_api submodule is still experimental. See NEP 47.", stacklevel=2
+)
+
+__array_api_version__ = "2022.12"
+
+__all__ = ["__array_api_version__"]
+
+from ._constants import e, inf, nan, pi, newaxis
+
+__all__ += ["e", "inf", "nan", "pi", "newaxis"]
+
+from ._creation_functions import (
+    asarray,
+    arange,
+    empty,
+    empty_like,
+    eye,
+    from_dlpack,
+    full,
+    full_like,
+    linspace,
+    meshgrid,
+    ones,
+    ones_like,
+    tril,
+    triu,
+    zeros,
+    zeros_like,
+)
+
+__all__ += [
+    "asarray",
+    "arange",
+    "empty",
+    "empty_like",
+    "eye",
+    "from_dlpack",
+    "full",
+    "full_like",
+    "linspace",
+    "meshgrid",
+    "ones",
+    "ones_like",
+    "tril",
+    "triu",
+    "zeros",
+    "zeros_like",
+]
+
+from ._data_type_functions import (
+    astype,
+    broadcast_arrays,
+    broadcast_to,
+    can_cast,
+    finfo,
+    isdtype,
+    iinfo,
+    result_type,
+)
+
+__all__ += [
+    "astype",
+    "broadcast_arrays",
+    "broadcast_to",
+    "can_cast",
+    "finfo",
+    "iinfo",
+    "result_type",
+]
+
+from ._dtypes import (
+    int8,
+    int16,
+    int32,
+    int64,
+    uint8,
+    uint16,
+    uint32,
+    uint64,
+    float32,
+    float64,
+    complex64,
+    complex128,
+    bool,
+)
+
+__all__ += [
+    "int8",
+    "int16",
+    "int32",
+    "int64",
+    "uint8",
+    "uint16",
+    "uint32",
+    "uint64",
+    "float32",
+    "float64",
+    "bool",
+]
+
+from ._elementwise_functions import (
+    abs,
+    acos,
+    acosh,
+    add,
+    asin,
+    asinh,
+    atan,
+    atan2,
+    atanh,
+    bitwise_and,
+    bitwise_left_shift,
+    bitwise_invert,
+    bitwise_or,
+    bitwise_right_shift,
+    bitwise_xor,
+    ceil,
+    conj,
+    cos,
+    cosh,
+    divide,
+    equal,
+    exp,
+    expm1,
+    floor,
+    floor_divide,
+    greater,
+    greater_equal,
+    imag,
+    isfinite,
+    isinf,
+    isnan,
+    less,
+    less_equal,
+    log,
+    log1p,
+    log2,
+    log10,
+    logaddexp,
+    logical_and,
+    logical_not,
+    logical_or,
+    logical_xor,
+    multiply,
+    negative,
+    not_equal,
+    positive,
+    pow,
+    real,
+    remainder,
+    round,
+    sign,
+    sin,
+    sinh,
+    square,
+    sqrt,
+    subtract,
+    tan,
+    tanh,
+    trunc,
+)
+
+__all__ += [
+    "abs",
+    "acos",
+    "acosh",
+    "add",
+    "asin",
+    "asinh",
+    "atan",
+    "atan2",
+    "atanh",
+    "bitwise_and",
+    "bitwise_left_shift",
+    "bitwise_invert",
+    "bitwise_or",
+    "bitwise_right_shift",
+    "bitwise_xor",
+    "ceil",
+    "cos",
+    "cosh",
+    "divide",
+    "equal",
+    "exp",
+    "expm1",
+    "floor",
+    "floor_divide",
+    "greater",
+    "greater_equal",
+    "isfinite",
+    "isinf",
+    "isnan",
+    "less",
+    "less_equal",
+    "log",
+    "log1p",
+    "log2",
+    "log10",
+    "logaddexp",
+    "logical_and",
+    "logical_not",
+    "logical_or",
+    "logical_xor",
+    "multiply",
+    "negative",
+    "not_equal",
+    "positive",
+    "pow",
+    "remainder",
+    "round",
+    "sign",
+    "sin",
+    "sinh",
+    "square",
+    "sqrt",
+    "subtract",
+    "tan",
+    "tanh",
+    "trunc",
+]
+
+from ._indexing_functions import take
+
+__all__ += ["take"]
+
+# linalg is an extension in the array API spec, which is a sub-namespace. Only
+# a subset of functions in it are imported into the top-level namespace.
+from . import linalg
+
+__all__ += ["linalg"]
+
+from .linalg import matmul, tensordot, matrix_transpose, vecdot
+
+__all__ += ["matmul", "tensordot", "matrix_transpose", "vecdot"]
+
+from ._manipulation_functions import (
+    concat,
+    expand_dims,
+    flip,
+    permute_dims,
+    reshape,
+    roll,
+    squeeze,
+    stack,
+)
+
+__all__ += ["concat", "expand_dims", "flip", "permute_dims", "reshape", "roll", "squeeze", "stack"]
+
+from ._searching_functions import argmax, argmin, nonzero, where
+
+__all__ += ["argmax", "argmin", "nonzero", "where"]
+
+from ._set_functions import unique_all, unique_counts, unique_inverse, unique_values
+
+__all__ += ["unique_all", "unique_counts", "unique_inverse", "unique_values"]
+
+from ._sorting_functions import argsort, sort
+
+__all__ += ["argsort", "sort"]
+
+from ._statistical_functions import max, mean, min, prod, std, sum, var
+
+__all__ += ["max", "mean", "min", "prod", "std", "sum", "var"]
+
+from ._utility_functions import all, any
+
+__all__ += ["all", "any"]

.venv/lib/python3.11/site-packages/numpy/array_api/_array_object.py

@@ -0,0 +1,1133 @@
+"""
+Wrapper class around the ndarray object for the array API standard.
+
+The array API standard defines some behaviors differently than ndarray, in
+particular, type promotion rules are different (the standard has no
+value-based casting). The standard also specifies a more limited subset of
+array methods and functionalities than are implemented on ndarray. Since the
+goal of the array_api namespace is to be a minimal implementation of the array
+API standard, we need to define a separate wrapper class for the array_api
+namespace.
+
+The standard compliant class is only a wrapper class. It is *not* a subclass
+of ndarray.
+"""
+
+from __future__ import annotations
+
+import operator
+from enum import IntEnum
+from ._creation_functions import asarray
+from ._dtypes import (
+    _all_dtypes,
+    _boolean_dtypes,
+    _integer_dtypes,
+    _integer_or_boolean_dtypes,
+    _floating_dtypes,
+    _complex_floating_dtypes,
+    _numeric_dtypes,
+    _result_type,
+    _dtype_categories,
+)
+
+from typing import TYPE_CHECKING, Optional, Tuple, Union, Any, SupportsIndex
+import types
+
+if TYPE_CHECKING:
+    from ._typing import Any, PyCapsule, Device, Dtype
+    import numpy.typing as npt
+
+import numpy as np
+
+from numpy import array_api
+
+
+class Array:
+    """
+    n-d array object for the array API namespace.
+
+    See the docstring of :py:obj:`np.ndarray <numpy.ndarray>` for more
+    information.
+
+    This is a wrapper around numpy.ndarray that restricts the usage to only
+    those things that are required by the array API namespace. Note,
+    attributes on this object that start with a single underscore are not part
+    of the API specification and should only be used internally. This object
+    should not be constructed directly. Rather, use one of the creation
+    functions, such as asarray().
+
+    """
+    _array: np.ndarray[Any, Any]
+
+    # Use a custom constructor instead of __init__, as manually initializing
+    # this class is not supported API.
+    @classmethod
+    def _new(cls, x, /):
+        """
+        This is a private method for initializing the array API Array
+        object.
+
+        Functions outside of the array_api submodule should not use this
+        method. Use one of the creation functions instead, such as
+        ``asarray``.
+
+        """
+        obj = super().__new__(cls)
+        # Note: The spec does not have array scalars, only 0-D arrays.
+        if isinstance(x, np.generic):
+            # Convert the array scalar to a 0-D array
+            x = np.asarray(x)
+        if x.dtype not in _all_dtypes:
+            raise TypeError(
+                f"The array_api namespace does not support the dtype '{x.dtype}'"
+            )
+        obj._array = x
+        return obj
+
+    # Prevent Array() from working
+    def __new__(cls, *args, **kwargs):
+        raise TypeError(
+            "The array_api Array object should not be instantiated directly. Use an array creation function, such as asarray(), instead."
+        )
+
+    # These functions are not required by the spec, but are implemented for
+    # the sake of usability.
+
+    def __str__(self: Array, /) -> str:
+        """
+        Performs the operation __str__.
+        """
+        return self._array.__str__().replace("array", "Array")
+
+    def __repr__(self: Array, /) -> str:
+        """
+        Performs the operation __repr__.
+        """
+        suffix = f", dtype={self.dtype.name})"
+        if 0 in self.shape:
+            prefix = "empty("
+            mid = str(self.shape)
+        else:
+            prefix = "Array("
+            mid = np.array2string(self._array, separator=', ', prefix=prefix, suffix=suffix)
+        return prefix + mid + suffix
+
+    # This function is not required by the spec, but we implement it here for
+    # convenience so that np.asarray(np.array_api.Array) will work.
+    def __array__(self, dtype: None | np.dtype[Any] = None) -> npt.NDArray[Any]:
+        """
+        Warning: this method is NOT part of the array API spec. Implementers
+        of other libraries need not include it, and users should not assume it
+        will be present in other implementations.
+
+        """
+        return np.asarray(self._array, dtype=dtype)
+
+    # These are various helper functions to make the array behavior match the
+    # spec in places where it either deviates from or is more strict than
+    # NumPy behavior
+
+    def _check_allowed_dtypes(self, other: bool | int | float | Array, dtype_category: str, op: str) -> Array:
+        """
+        Helper function for operators to only allow specific input dtypes
+
+        Use like
+
+            other = self._check_allowed_dtypes(other, 'numeric', '__add__')
+            if other is NotImplemented:
+                return other
+        """
+
+        if self.dtype not in _dtype_categories[dtype_category]:
+            raise TypeError(f"Only {dtype_category} dtypes are allowed in {op}")
+        if isinstance(other, (int, complex, float, bool)):
+            other = self._promote_scalar(other)
+        elif isinstance(other, Array):
+            if other.dtype not in _dtype_categories[dtype_category]:
+                raise TypeError(f"Only {dtype_category} dtypes are allowed in {op}")
+        else:
+            return NotImplemented
+
+        # This will raise TypeError for type combinations that are not allowed
+        # to promote in the spec (even if the NumPy array operator would
+        # promote them).
+        res_dtype = _result_type(self.dtype, other.dtype)
+        if op.startswith("__i"):
+            # Note: NumPy will allow in-place operators in some cases where
+            # the type promoted operator does not match the left-hand side
+            # operand. For example,
+
+            # >>> a = np.array(1, dtype=np.int8)
+            # >>> a += np.array(1, dtype=np.int16)
+
+            # The spec explicitly disallows this.
+            if res_dtype != self.dtype:
+                raise TypeError(
+                    f"Cannot perform {op} with dtypes {self.dtype} and {other.dtype}"
+                )
+
+        return other
+
+    # Helper function to match the type promotion rules in the spec
+    def _promote_scalar(self, scalar):
+        """
+        Returns a promoted version of a Python scalar appropriate for use with
+        operations on self.
+
+        This may raise an OverflowError in cases where the scalar is an
+        integer that is too large to fit in a NumPy integer dtype, or
+        TypeError when the scalar type is incompatible with the dtype of self.
+        """
+        # Note: Only Python scalar types that match the array dtype are
+        # allowed.
+        if isinstance(scalar, bool):
+            if self.dtype not in _boolean_dtypes:
+                raise TypeError(
+                    "Python bool scalars can only be promoted with bool arrays"
+                )
+        elif isinstance(scalar, int):
+            if self.dtype in _boolean_dtypes:
+                raise TypeError(
+                    "Python int scalars cannot be promoted with bool arrays"
+                )
+            if self.dtype in _integer_dtypes:
+                info = np.iinfo(self.dtype)
+                if not (info.min <= scalar <= info.max):
+                    raise OverflowError(
+                        "Python int scalars must be within the bounds of the dtype for integer arrays"
+                    )
+            # int + array(floating) is allowed
+        elif isinstance(scalar, float):
+            if self.dtype not in _floating_dtypes:
+                raise TypeError(
+                    "Python float scalars can only be promoted with floating-point arrays."
+                )
+        elif isinstance(scalar, complex):
+            if self.dtype not in _complex_floating_dtypes:
+                raise TypeError(
+                    "Python complex scalars can only be promoted with complex floating-point arrays."
+                )
+        else:
+            raise TypeError("'scalar' must be a Python scalar")
+
+        # Note: scalars are unconditionally cast to the same dtype as the
+        # array.
+
+        # Note: the spec only specifies integer-dtype/int promotion
+        # behavior for integers within the bounds of the integer dtype.
+        # Outside of those bounds we use the default NumPy behavior (either
+        # cast or raise OverflowError).
+        return Array._new(np.array(scalar, self.dtype))
+
+    @staticmethod
+    def _normalize_two_args(x1, x2) -> Tuple[Array, Array]:
+        """
+        Normalize inputs to two arg functions to fix type promotion rules
+
+        NumPy deviates from the spec type promotion rules in cases where one
+        argument is 0-dimensional and the other is not. For example:
+
+        >>> import numpy as np
+        >>> a = np.array([1.0], dtype=np.float32)
+        >>> b = np.array(1.0, dtype=np.float64)
+        >>> np.add(a, b) # The spec says this should be float64
+        array([2.], dtype=float32)
+
+        To fix this, we add a dimension to the 0-dimension array before passing it
+        through. This works because a dimension would be added anyway from
+        broadcasting, so the resulting shape is the same, but this prevents NumPy
+        from not promoting the dtype.
+        """
+        # Another option would be to use signature=(x1.dtype, x2.dtype, None),
+        # but that only works for ufuncs, so we would have to call the ufuncs
+        # directly in the operator methods. One should also note that this
+        # sort of trick wouldn't work for functions like searchsorted, which
+        # don't do normal broadcasting, but there aren't any functions like
+        # that in the array API namespace.
+        if x1.ndim == 0 and x2.ndim != 0:
+            # The _array[None] workaround was chosen because it is relatively
+            # performant. broadcast_to(x1._array, x2.shape) is much slower. We
+            # could also manually type promote x2, but that is more complicated
+            # and about the same performance as this.
+            x1 = Array._new(x1._array[None])
+        elif x2.ndim == 0 and x1.ndim != 0:
+            x2 = Array._new(x2._array[None])
+        return (x1, x2)
+
+    # Note: A large fraction of allowed indices are disallowed here (see the
+    # docstring below)
+    def _validate_index(self, key):
+        """
+        Validate an index according to the array API.
+
+        The array API specification only requires a subset of indices that are
+        supported by NumPy. This function will reject any index that is
+        allowed by NumPy but not required by the array API specification. We
+        always raise ``IndexError`` on such indices (the spec does not require
+        any specific behavior on them, but this makes the NumPy array API
+        namespace a minimal implementation of the spec). See
+        https://data-apis.org/array-api/latest/API_specification/indexing.html
+        for the full list of required indexing behavior
+
+        This function raises IndexError if the index ``key`` is invalid. It
+        only raises ``IndexError`` on indices that are not already rejected by
+        NumPy, as NumPy will already raise the appropriate error on such
+        indices. ``shape`` may be None, in which case, only cases that are
+        independent of the array shape are checked.
+
+        The following cases are allowed by NumPy, but not specified by the array
+        API specification:
+
+        - Indices to not include an implicit ellipsis at the end. That is,
+          every axis of an array must be explicitly indexed or an ellipsis
+          included. This behaviour is sometimes referred to as flat indexing.
+
+        - The start and stop of a slice may not be out of bounds. In
+          particular, for a slice ``i:j:k`` on an axis of size ``n``, only the
+          following are allowed:
+
+          - ``i`` or ``j`` omitted (``None``).
+          - ``-n <= i <= max(0, n - 1)``.
+          - For ``k > 0`` or ``k`` omitted (``None``), ``-n <= j <= n``.
+          - For ``k < 0``, ``-n - 1 <= j <= max(0, n - 1)``.
+
+        - Boolean array indices are not allowed as part of a larger tuple
+          index.
+
+        - Integer array indices are not allowed (with the exception of 0-D
+          arrays, which are treated the same as scalars).
+
+        Additionally, it should be noted that indices that would return a
+        scalar in NumPy will return a 0-D array. Array scalars are not allowed
+        in the specification, only 0-D arrays. This is done in the
+        ``Array._new`` constructor, not this function.
+
+        """
+        _key = key if isinstance(key, tuple) else (key,)
+        for i in _key:
+            if isinstance(i, bool) or not (
+                isinstance(i, SupportsIndex)  # i.e. ints
+                or isinstance(i, slice)
+                or i == Ellipsis
+                or i is None
+                or isinstance(i, Array)
+                or isinstance(i, np.ndarray)
+            ):
+                raise IndexError(
+                    f"Single-axes index {i} has {type(i)=}, but only "
+                    "integers, slices (:), ellipsis (...), newaxis (None), "
+                    "zero-dimensional integer arrays and boolean arrays "
+                    "are specified in the Array API."
+                )
+
+        nonexpanding_key = []
+        single_axes = []
+        n_ellipsis = 0
+        key_has_mask = False
+        for i in _key:
+            if i is not None:
+                nonexpanding_key.append(i)
+                if isinstance(i, Array) or isinstance(i, np.ndarray):
+                    if i.dtype in _boolean_dtypes:
+                        key_has_mask = True
+                    single_axes.append(i)
+                else:
+                    # i must not be an array here, to avoid elementwise equals
+                    if i == Ellipsis:
+                        n_ellipsis += 1
+                    else:
+                        single_axes.append(i)
+
+        n_single_axes = len(single_axes)
+        if n_ellipsis > 1:
+            return  # handled by ndarray
+        elif n_ellipsis == 0:
+            # Note boolean masks must be the sole index, which we check for
+            # later on.
+            if not key_has_mask and n_single_axes < self.ndim:
+                raise IndexError(
+                    f"{self.ndim=}, but the multi-axes index only specifies "
+                    f"{n_single_axes} dimensions. If this was intentional, "
+                    "add a trailing ellipsis (...) which expands into as many "
+                    "slices (:) as necessary - this is what np.ndarray arrays "
+                    "implicitly do, but such flat indexing behaviour is not "
+                    "specified in the Array API."
+                )
+
+        if n_ellipsis == 0:
+            indexed_shape = self.shape
+        else:
+            ellipsis_start = None
+            for pos, i in enumerate(nonexpanding_key):
+                if not (isinstance(i, Array) or isinstance(i, np.ndarray)):
+                    if i == Ellipsis:
+                        ellipsis_start = pos
+                        break
+            assert ellipsis_start is not None  # sanity check
+            ellipsis_end = self.ndim - (n_single_axes - ellipsis_start)
+            indexed_shape = (
+                self.shape[:ellipsis_start] + self.shape[ellipsis_end:]
+            )
+        for i, side in zip(single_axes, indexed_shape):
+            if isinstance(i, slice):
+                if side == 0:
+                    f_range = "0 (or None)"
+                else:
+                    f_range = f"between -{side} and {side - 1} (or None)"
+                if i.start is not None:
+                    try:
+                        start = operator.index(i.start)
+                    except TypeError:
+                        pass  # handled by ndarray
+                    else:
+                        if not (-side <= start <= side):
+                            raise IndexError(
+                                f"Slice {i} contains {start=}, but should be "
+                                f"{f_range} for an axis of size {side} "
+                                "(out-of-bounds starts are not specified in "
+                                "the Array API)"
+                            )
+                if i.stop is not None:
+                    try:
+                        stop = operator.index(i.stop)
+                    except TypeError:
+                        pass  # handled by ndarray
+                    else:
+                        if not (-side <= stop <= side):
+                            raise IndexError(
+                                f"Slice {i} contains {stop=}, but should be "
+                                f"{f_range} for an axis of size {side} "
+                                "(out-of-bounds stops are not specified in "
+                                "the Array API)"
+                            )
+            elif isinstance(i, Array):
+                if i.dtype in _boolean_dtypes and len(_key) != 1:
+                    assert isinstance(key, tuple)  # sanity check
+                    raise IndexError(
+                        f"Single-axes index {i} is a boolean array and "
+                        f"{len(key)=}, but masking is only specified in the "
+                        "Array API when the array is the sole index."
+                    )
+                elif i.dtype in _integer_dtypes and i.ndim != 0:
+                    raise IndexError(
+                        f"Single-axes index {i} is a non-zero-dimensional "
+                        "integer array, but advanced integer indexing is not "
+                        "specified in the Array API."
+                    )
+            elif isinstance(i, tuple):
+                raise IndexError(
+                    f"Single-axes index {i} is a tuple, but nested tuple "
+                    "indices are not specified in the Array API."
+                )
+
+    # Everything below this line is required by the spec.
+
+    def __abs__(self: Array, /) -> Array:
+        """
+        Performs the operation __abs__.
+        """
+        if self.dtype not in _numeric_dtypes:
+            raise TypeError("Only numeric dtypes are allowed in __abs__")
+        res = self._array.__abs__()
+        return self.__class__._new(res)
+
+    def __add__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __add__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__add__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__add__(other._array)
+        return self.__class__._new(res)
+
+    def __and__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __and__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__and__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__and__(other._array)
+        return self.__class__._new(res)
+
+    def __array_namespace__(
+        self: Array, /, *, api_version: Optional[str] = None
+    ) -> types.ModuleType:
+        if api_version is not None and not api_version.startswith("2021."):
+            raise ValueError(f"Unrecognized array API version: {api_version!r}")
+        return array_api
+
+    def __bool__(self: Array, /) -> bool:
+        """
+        Performs the operation __bool__.
+        """
+        # Note: This is an error here.
+        if self._array.ndim != 0:
+            raise TypeError("bool is only allowed on arrays with 0 dimensions")
+        res = self._array.__bool__()
+        return res
+
+    def __complex__(self: Array, /) -> complex:
+        """
+        Performs the operation __complex__.
+        """
+        # Note: This is an error here.
+        if self._array.ndim != 0:
+            raise TypeError("complex is only allowed on arrays with 0 dimensions")
+        res = self._array.__complex__()
+        return res
+
+    def __dlpack__(self: Array, /, *, stream: None = None) -> PyCapsule:
+        """
+        Performs the operation __dlpack__.
+        """
+        return self._array.__dlpack__(stream=stream)
+
+    def __dlpack_device__(self: Array, /) -> Tuple[IntEnum, int]:
+        """
+        Performs the operation __dlpack_device__.
+        """
+        # Note: device support is required for this
+        return self._array.__dlpack_device__()
+
+    def __eq__(self: Array, other: Union[int, float, bool, Array], /) -> Array:
+        """
+        Performs the operation __eq__.
+        """
+        # Even though "all" dtypes are allowed, we still require them to be
+        # promotable with each other.
+        other = self._check_allowed_dtypes(other, "all", "__eq__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__eq__(other._array)
+        return self.__class__._new(res)
+
+    def __float__(self: Array, /) -> float:
+        """
+        Performs the operation __float__.
+        """
+        # Note: This is an error here.
+        if self._array.ndim != 0:
+            raise TypeError("float is only allowed on arrays with 0 dimensions")
+        if self.dtype in _complex_floating_dtypes:
+            raise TypeError("float is not allowed on complex floating-point arrays")
+        res = self._array.__float__()
+        return res
+
+    def __floordiv__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __floordiv__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__floordiv__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__floordiv__(other._array)
+        return self.__class__._new(res)
+
+    def __ge__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __ge__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__ge__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__ge__(other._array)
+        return self.__class__._new(res)
+
+    def __getitem__(
+        self: Array,
+        key: Union[
+            int,
+            slice,
+            ellipsis,
+            Tuple[Union[int, slice, ellipsis, None], ...],
+            Array,
+        ],
+        /,
+    ) -> Array:
+        """
+        Performs the operation __getitem__.
+        """
+        # Note: Only indices required by the spec are allowed. See the
+        # docstring of _validate_index
+        self._validate_index(key)
+        if isinstance(key, Array):
+            # Indexing self._array with array_api arrays can be erroneous
+            key = key._array
+        res = self._array.__getitem__(key)
+        return self._new(res)
+
+    def __gt__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __gt__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__gt__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__gt__(other._array)
+        return self.__class__._new(res)
+
+    def __int__(self: Array, /) -> int:
+        """
+        Performs the operation __int__.
+        """
+        # Note: This is an error here.
+        if self._array.ndim != 0:
+            raise TypeError("int is only allowed on arrays with 0 dimensions")
+        if self.dtype in _complex_floating_dtypes:
+            raise TypeError("int is not allowed on complex floating-point arrays")
+        res = self._array.__int__()
+        return res
+
+    def __index__(self: Array, /) -> int:
+        """
+        Performs the operation __index__.
+        """
+        res = self._array.__index__()
+        return res
+
+    def __invert__(self: Array, /) -> Array:
+        """
+        Performs the operation __invert__.
+        """
+        if self.dtype not in _integer_or_boolean_dtypes:
+            raise TypeError("Only integer or boolean dtypes are allowed in __invert__")
+        res = self._array.__invert__()
+        return self.__class__._new(res)
+
+    def __le__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __le__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__le__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__le__(other._array)
+        return self.__class__._new(res)
+
+    def __lshift__(self: Array, other: Union[int, Array], /) -> Array:
+        """
+        Performs the operation __lshift__.
+        """
+        other = self._check_allowed_dtypes(other, "integer", "__lshift__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__lshift__(other._array)
+        return self.__class__._new(res)
+
+    def __lt__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __lt__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__lt__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__lt__(other._array)
+        return self.__class__._new(res)
+
+    def __matmul__(self: Array, other: Array, /) -> Array:
+        """
+        Performs the operation __matmul__.
+        """
+        # matmul is not defined for scalars, but without this, we may get
+        # the wrong error message from asarray.
+        other = self._check_allowed_dtypes(other, "numeric", "__matmul__")
+        if other is NotImplemented:
+            return other
+        res = self._array.__matmul__(other._array)
+        return self.__class__._new(res)
+
+    def __mod__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __mod__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__mod__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__mod__(other._array)
+        return self.__class__._new(res)
+
+    def __mul__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __mul__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__mul__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__mul__(other._array)
+        return self.__class__._new(res)
+
+    def __ne__(self: Array, other: Union[int, float, bool, Array], /) -> Array:
+        """
+        Performs the operation __ne__.
+        """
+        other = self._check_allowed_dtypes(other, "all", "__ne__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__ne__(other._array)
+        return self.__class__._new(res)
+
+    def __neg__(self: Array, /) -> Array:
+        """
+        Performs the operation __neg__.
+        """
+        if self.dtype not in _numeric_dtypes:
+            raise TypeError("Only numeric dtypes are allowed in __neg__")
+        res = self._array.__neg__()
+        return self.__class__._new(res)
+
+    def __or__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __or__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__or__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__or__(other._array)
+        return self.__class__._new(res)
+
+    def __pos__(self: Array, /) -> Array:
+        """
+        Performs the operation __pos__.
+        """
+        if self.dtype not in _numeric_dtypes:
+            raise TypeError("Only numeric dtypes are allowed in __pos__")
+        res = self._array.__pos__()
+        return self.__class__._new(res)
+
+    def __pow__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __pow__.
+        """
+        from ._elementwise_functions import pow
+
+        other = self._check_allowed_dtypes(other, "numeric", "__pow__")
+        if other is NotImplemented:
+            return other
+        # Note: NumPy's __pow__ does not follow type promotion rules for 0-d
+        # arrays, so we use pow() here instead.
+        return pow(self, other)
+
+    def __rshift__(self: Array, other: Union[int, Array], /) -> Array:
+        """
+        Performs the operation __rshift__.
+        """
+        other = self._check_allowed_dtypes(other, "integer", "__rshift__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rshift__(other._array)
+        return self.__class__._new(res)
+
+    def __setitem__(
+        self,
+        key: Union[
+            int, slice, ellipsis, Tuple[Union[int, slice, ellipsis], ...], Array
+        ],
+        value: Union[int, float, bool, Array],
+        /,
+    ) -> None:
+        """
+        Performs the operation __setitem__.
+        """
+        # Note: Only indices required by the spec are allowed. See the
+        # docstring of _validate_index
+        self._validate_index(key)
+        if isinstance(key, Array):
+            # Indexing self._array with array_api arrays can be erroneous
+            key = key._array
+        self._array.__setitem__(key, asarray(value)._array)
+
+    def __sub__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __sub__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__sub__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__sub__(other._array)
+        return self.__class__._new(res)
+
+    # PEP 484 requires int to be a subtype of float, but __truediv__ should
+    # not accept int.
+    def __truediv__(self: Array, other: Union[float, Array], /) -> Array:
+        """
+        Performs the operation __truediv__.
+        """
+        other = self._check_allowed_dtypes(other, "floating-point", "__truediv__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__truediv__(other._array)
+        return self.__class__._new(res)
+
+    def __xor__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __xor__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__xor__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__xor__(other._array)
+        return self.__class__._new(res)
+
+    def __iadd__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __iadd__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__iadd__")
+        if other is NotImplemented:
+            return other
+        self._array.__iadd__(other._array)
+        return self
+
+    def __radd__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __radd__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__radd__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__radd__(other._array)
+        return self.__class__._new(res)
+
+    def __iand__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __iand__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__iand__")
+        if other is NotImplemented:
+            return other
+        self._array.__iand__(other._array)
+        return self
+
+    def __rand__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __rand__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__rand__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rand__(other._array)
+        return self.__class__._new(res)
+
+    def __ifloordiv__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __ifloordiv__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__ifloordiv__")
+        if other is NotImplemented:
+            return other
+        self._array.__ifloordiv__(other._array)
+        return self
+
+    def __rfloordiv__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __rfloordiv__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__rfloordiv__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rfloordiv__(other._array)
+        return self.__class__._new(res)
+
+    def __ilshift__(self: Array, other: Union[int, Array], /) -> Array:
+        """
+        Performs the operation __ilshift__.
+        """
+        other = self._check_allowed_dtypes(other, "integer", "__ilshift__")
+        if other is NotImplemented:
+            return other
+        self._array.__ilshift__(other._array)
+        return self
+
+    def __rlshift__(self: Array, other: Union[int, Array], /) -> Array:
+        """
+        Performs the operation __rlshift__.
+        """
+        other = self._check_allowed_dtypes(other, "integer", "__rlshift__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rlshift__(other._array)
+        return self.__class__._new(res)
+
+    def __imatmul__(self: Array, other: Array, /) -> Array:
+        """
+        Performs the operation __imatmul__.
+        """
+        # matmul is not defined for scalars, but without this, we may get
+        # the wrong error message from asarray.
+        other = self._check_allowed_dtypes(other, "numeric", "__imatmul__")
+        if other is NotImplemented:
+            return other
+        res = self._array.__imatmul__(other._array)
+        return self.__class__._new(res)
+
+    def __rmatmul__(self: Array, other: Array, /) -> Array:
+        """
+        Performs the operation __rmatmul__.
+        """
+        # matmul is not defined for scalars, but without this, we may get
+        # the wrong error message from asarray.
+        other = self._check_allowed_dtypes(other, "numeric", "__rmatmul__")
+        if other is NotImplemented:
+            return other
+        res = self._array.__rmatmul__(other._array)
+        return self.__class__._new(res)
+
+    def __imod__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __imod__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__imod__")
+        if other is NotImplemented:
+            return other
+        self._array.__imod__(other._array)
+        return self
+
+    def __rmod__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __rmod__.
+        """
+        other = self._check_allowed_dtypes(other, "real numeric", "__rmod__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rmod__(other._array)
+        return self.__class__._new(res)
+
+    def __imul__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __imul__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__imul__")
+        if other is NotImplemented:
+            return other
+        self._array.__imul__(other._array)
+        return self
+
+    def __rmul__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __rmul__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__rmul__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rmul__(other._array)
+        return self.__class__._new(res)
+
+    def __ior__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __ior__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__ior__")
+        if other is NotImplemented:
+            return other
+        self._array.__ior__(other._array)
+        return self
+
+    def __ror__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __ror__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__ror__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__ror__(other._array)
+        return self.__class__._new(res)
+
+    def __ipow__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __ipow__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__ipow__")
+        if other is NotImplemented:
+            return other
+        self._array.__ipow__(other._array)
+        return self
+
+    def __rpow__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __rpow__.
+        """
+        from ._elementwise_functions import pow
+
+        other = self._check_allowed_dtypes(other, "numeric", "__rpow__")
+        if other is NotImplemented:
+            return other
+        # Note: NumPy's __pow__ does not follow the spec type promotion rules
+        # for 0-d arrays, so we use pow() here instead.
+        return pow(other, self)
+
+    def __irshift__(self: Array, other: Union[int, Array], /) -> Array:
+        """
+        Performs the operation __irshift__.
+        """
+        other = self._check_allowed_dtypes(other, "integer", "__irshift__")
+        if other is NotImplemented:
+            return other
+        self._array.__irshift__(other._array)
+        return self
+
+    def __rrshift__(self: Array, other: Union[int, Array], /) -> Array:
+        """
+        Performs the operation __rrshift__.
+        """
+        other = self._check_allowed_dtypes(other, "integer", "__rrshift__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rrshift__(other._array)
+        return self.__class__._new(res)
+
+    def __isub__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __isub__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__isub__")
+        if other is NotImplemented:
+            return other
+        self._array.__isub__(other._array)
+        return self
+
+    def __rsub__(self: Array, other: Union[int, float, Array], /) -> Array:
+        """
+        Performs the operation __rsub__.
+        """
+        other = self._check_allowed_dtypes(other, "numeric", "__rsub__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rsub__(other._array)
+        return self.__class__._new(res)
+
+    def __itruediv__(self: Array, other: Union[float, Array], /) -> Array:
+        """
+        Performs the operation __itruediv__.
+        """
+        other = self._check_allowed_dtypes(other, "floating-point", "__itruediv__")
+        if other is NotImplemented:
+            return other
+        self._array.__itruediv__(other._array)
+        return self
+
+    def __rtruediv__(self: Array, other: Union[float, Array], /) -> Array:
+        """
+        Performs the operation __rtruediv__.
+        """
+        other = self._check_allowed_dtypes(other, "floating-point", "__rtruediv__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rtruediv__(other._array)
+        return self.__class__._new(res)
+
+    def __ixor__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __ixor__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__ixor__")
+        if other is NotImplemented:
+            return other
+        self._array.__ixor__(other._array)
+        return self
+
+    def __rxor__(self: Array, other: Union[int, bool, Array], /) -> Array:
+        """
+        Performs the operation __rxor__.
+        """
+        other = self._check_allowed_dtypes(other, "integer or boolean", "__rxor__")
+        if other is NotImplemented:
+            return other
+        self, other = self._normalize_two_args(self, other)
+        res = self._array.__rxor__(other._array)
+        return self.__class__._new(res)
+
+    def to_device(self: Array, device: Device, /, stream: None = None) -> Array:
+        if stream is not None:
+            raise ValueError("The stream argument to to_device() is not supported")
+        if device == 'cpu':
+            return self
+        raise ValueError(f"Unsupported device {device!r}")
+
+    @property
+    def dtype(self) -> Dtype:
+        """
+        Array API compatible wrapper for :py:meth:`np.ndarray.dtype <numpy.ndarray.dtype>`.
+
+        See its docstring for more information.
+        """
+        return self._array.dtype
+
+    @property
+    def device(self) -> Device:
+        return "cpu"
+
+    # Note: mT is new in array API spec (see matrix_transpose)
+    @property
+    def mT(self) -> Array:
+        from .linalg import matrix_transpose
+        return matrix_transpose(self)
+
+    @property
+    def ndim(self) -> int:
+        """
+        Array API compatible wrapper for :py:meth:`np.ndarray.ndim <numpy.ndarray.ndim>`.
+
+        See its docstring for more information.
+        """
+        return self._array.ndim
+
+    @property
+    def shape(self) -> Tuple[int, ...]:
+        """
+        Array API compatible wrapper for :py:meth:`np.ndarray.shape <numpy.ndarray.shape>`.
+
+        See its docstring for more information.
+        """
+        return self._array.shape
+
+    @property
+    def size(self) -> int:
+        """
+        Array API compatible wrapper for :py:meth:`np.ndarray.size <numpy.ndarray.size>`.
+
+        See its docstring for more information.
+        """
+        return self._array.size
+
+    @property
+    def T(self) -> Array:
+        """
+        Array API compatible wrapper for :py:meth:`np.ndarray.T <numpy.ndarray.T>`.
+
+        See its docstring for more information.
+        """
+        # Note: T only works on 2-dimensional arrays. See the corresponding
+        # note in the specification:
+        # https://data-apis.org/array-api/latest/API_specification/array_object.html#t
+        if self.ndim != 2:
+            raise ValueError("x.T requires x to have 2 dimensions. Use x.mT to transpose stacks of matrices and permute_dims() to permute dimensions.")
+        return self.__class__._new(self._array.T)

.venv/lib/python3.11/site-packages/numpy/array_api/_constants.py

@@ -0,0 +1,7 @@
+import numpy as np
+
+e = np.e
+inf = np.inf
+nan = np.nan
+pi = np.pi
+newaxis = np.newaxis

.venv/lib/python3.11/site-packages/numpy/array_api/_creation_functions.py

@@ -0,0 +1,351 @@
+from __future__ import annotations
+
+
+from typing import TYPE_CHECKING, List, Optional, Tuple, Union
+
+if TYPE_CHECKING:
+    from ._typing import (
+        Array,
+        Device,
+        Dtype,
+        NestedSequence,
+        SupportsBufferProtocol,
+    )
+    from collections.abc import Sequence
+from ._dtypes import _all_dtypes
+
+import numpy as np
+
+
+def _check_valid_dtype(dtype):
+    # Note: Only spelling dtypes as the dtype objects is supported.
+
+    # We use this instead of "dtype in _all_dtypes" because the dtype objects
+    # define equality with the sorts of things we want to disallow.
+    for d in (None,) + _all_dtypes:
+        if dtype is d:
+            return
+    raise ValueError("dtype must be one of the supported dtypes")
+
+
+def asarray(
+    obj: Union[
+        Array,
+        bool,
+        int,
+        float,
+        NestedSequence[bool | int | float],
+        SupportsBufferProtocol,
+    ],
+    /,
+    *,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+    copy: Optional[Union[bool, np._CopyMode]] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.asarray <numpy.asarray>`.
+
+    See its docstring for more information.
+    """
+    # _array_object imports in this file are inside the functions to avoid
+    # circular imports
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    if copy in (False, np._CopyMode.IF_NEEDED):
+        # Note: copy=False is not yet implemented in np.asarray
+        raise NotImplementedError("copy=False is not yet implemented")
+    if isinstance(obj, Array):
+        if dtype is not None and obj.dtype != dtype:
+            copy = True
+        if copy in (True, np._CopyMode.ALWAYS):
+            return Array._new(np.array(obj._array, copy=True, dtype=dtype))
+        return obj
+    if dtype is None and isinstance(obj, int) and (obj > 2 ** 64 or obj < -(2 ** 63)):
+        # Give a better error message in this case. NumPy would convert this
+        # to an object array. TODO: This won't handle large integers in lists.
+        raise OverflowError("Integer out of bounds for array dtypes")
+    res = np.asarray(obj, dtype=dtype)
+    return Array._new(res)
+
+
+def arange(
+    start: Union[int, float],
+    /,
+    stop: Optional[Union[int, float]] = None,
+    step: Union[int, float] = 1,
+    *,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.arange <numpy.arange>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.arange(start, stop=stop, step=step, dtype=dtype))
+
+
+def empty(
+    shape: Union[int, Tuple[int, ...]],
+    *,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.empty <numpy.empty>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.empty(shape, dtype=dtype))
+
+
+def empty_like(
+    x: Array, /, *, dtype: Optional[Dtype] = None, device: Optional[Device] = None
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.empty_like <numpy.empty_like>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.empty_like(x._array, dtype=dtype))
+
+
+def eye(
+    n_rows: int,
+    n_cols: Optional[int] = None,
+    /,
+    *,
+    k: int = 0,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.eye <numpy.eye>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.eye(n_rows, M=n_cols, k=k, dtype=dtype))
+
+
+def from_dlpack(x: object, /) -> Array:
+    from ._array_object import Array
+
+    return Array._new(np.from_dlpack(x))
+
+
+def full(
+    shape: Union[int, Tuple[int, ...]],
+    fill_value: Union[int, float],
+    *,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.full <numpy.full>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    if isinstance(fill_value, Array) and fill_value.ndim == 0:
+        fill_value = fill_value._array
+    res = np.full(shape, fill_value, dtype=dtype)
+    if res.dtype not in _all_dtypes:
+        # This will happen if the fill value is not something that NumPy
+        # coerces to one of the acceptable dtypes.
+        raise TypeError("Invalid input to full")
+    return Array._new(res)
+
+
+def full_like(
+    x: Array,
+    /,
+    fill_value: Union[int, float],
+    *,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.full_like <numpy.full_like>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    res = np.full_like(x._array, fill_value, dtype=dtype)
+    if res.dtype not in _all_dtypes:
+        # This will happen if the fill value is not something that NumPy
+        # coerces to one of the acceptable dtypes.
+        raise TypeError("Invalid input to full_like")
+    return Array._new(res)
+
+
+def linspace(
+    start: Union[int, float],
+    stop: Union[int, float],
+    /,
+    num: int,
+    *,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+    endpoint: bool = True,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linspace <numpy.linspace>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.linspace(start, stop, num, dtype=dtype, endpoint=endpoint))
+
+
+def meshgrid(*arrays: Array, indexing: str = "xy") -> List[Array]:
+    """
+    Array API compatible wrapper for :py:func:`np.meshgrid <numpy.meshgrid>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    # Note: unlike np.meshgrid, only inputs with all the same dtype are
+    # allowed
+
+    if len({a.dtype for a in arrays}) > 1:
+        raise ValueError("meshgrid inputs must all have the same dtype")
+
+    return [
+        Array._new(array)
+        for array in np.meshgrid(*[a._array for a in arrays], indexing=indexing)
+    ]
+
+
+def ones(
+    shape: Union[int, Tuple[int, ...]],
+    *,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.ones <numpy.ones>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.ones(shape, dtype=dtype))
+
+
+def ones_like(
+    x: Array, /, *, dtype: Optional[Dtype] = None, device: Optional[Device] = None
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.ones_like <numpy.ones_like>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.ones_like(x._array, dtype=dtype))
+
+
+def tril(x: Array, /, *, k: int = 0) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.tril <numpy.tril>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    if x.ndim < 2:
+        # Note: Unlike np.tril, x must be at least 2-D
+        raise ValueError("x must be at least 2-dimensional for tril")
+    return Array._new(np.tril(x._array, k=k))
+
+
+def triu(x: Array, /, *, k: int = 0) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.triu <numpy.triu>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    if x.ndim < 2:
+        # Note: Unlike np.triu, x must be at least 2-D
+        raise ValueError("x must be at least 2-dimensional for triu")
+    return Array._new(np.triu(x._array, k=k))
+
+
+def zeros(
+    shape: Union[int, Tuple[int, ...]],
+    *,
+    dtype: Optional[Dtype] = None,
+    device: Optional[Device] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.zeros <numpy.zeros>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.zeros(shape, dtype=dtype))
+
+
+def zeros_like(
+    x: Array, /, *, dtype: Optional[Dtype] = None, device: Optional[Device] = None
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.zeros_like <numpy.zeros_like>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    _check_valid_dtype(dtype)
+    if device not in ["cpu", None]:
+        raise ValueError(f"Unsupported device {device!r}")
+    return Array._new(np.zeros_like(x._array, dtype=dtype))

.venv/lib/python3.11/site-packages/numpy/array_api/_data_type_functions.py

@@ -0,0 +1,197 @@
+from __future__ import annotations
+
+from ._array_object import Array
+from ._dtypes import (
+    _all_dtypes,
+    _boolean_dtypes,
+    _signed_integer_dtypes,
+    _unsigned_integer_dtypes,
+    _integer_dtypes,
+    _real_floating_dtypes,
+    _complex_floating_dtypes,
+    _numeric_dtypes,
+    _result_type,
+)
+
+from dataclasses import dataclass
+from typing import TYPE_CHECKING, List, Tuple, Union
+
+if TYPE_CHECKING:
+    from ._typing import Dtype
+    from collections.abc import Sequence
+
+import numpy as np
+
+
+# Note: astype is a function, not an array method as in NumPy.
+def astype(x: Array, dtype: Dtype, /, *, copy: bool = True) -> Array:
+    if not copy and dtype == x.dtype:
+        return x
+    return Array._new(x._array.astype(dtype=dtype, copy=copy))
+
+
+def broadcast_arrays(*arrays: Array) -> List[Array]:
+    """
+    Array API compatible wrapper for :py:func:`np.broadcast_arrays <numpy.broadcast_arrays>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    return [
+        Array._new(array) for array in np.broadcast_arrays(*[a._array for a in arrays])
+    ]
+
+
+def broadcast_to(x: Array, /, shape: Tuple[int, ...]) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.broadcast_to <numpy.broadcast_to>`.
+
+    See its docstring for more information.
+    """
+    from ._array_object import Array
+
+    return Array._new(np.broadcast_to(x._array, shape))
+
+
+def can_cast(from_: Union[Dtype, Array], to: Dtype, /) -> bool:
+    """
+    Array API compatible wrapper for :py:func:`np.can_cast <numpy.can_cast>`.
+
+    See its docstring for more information.
+    """
+    if isinstance(from_, Array):
+        from_ = from_.dtype
+    elif from_ not in _all_dtypes:
+        raise TypeError(f"{from_=}, but should be an array_api array or dtype")
+    if to not in _all_dtypes:
+        raise TypeError(f"{to=}, but should be a dtype")
+    # Note: We avoid np.can_cast() as it has discrepancies with the array API,
+    # since NumPy allows cross-kind casting (e.g., NumPy allows bool -> int8).
+    # See https://github.com/numpy/numpy/issues/20870
+    try:
+        # We promote `from_` and `to` together. We then check if the promoted
+        # dtype is `to`, which indicates if `from_` can (up)cast to `to`.
+        dtype = _result_type(from_, to)
+        return to == dtype
+    except TypeError:
+        # _result_type() raises if the dtypes don't promote together
+        return False
+
+
+# These are internal objects for the return types of finfo and iinfo, since
+# the NumPy versions contain extra data that isn't part of the spec.
+@dataclass
+class finfo_object:
+    bits: int
+    # Note: The types of the float data here are float, whereas in NumPy they
+    # are scalars of the corresponding float dtype.
+    eps: float
+    max: float
+    min: float
+    smallest_normal: float
+    dtype: Dtype
+
+
+@dataclass
+class iinfo_object:
+    bits: int
+    max: int
+    min: int
+    dtype: Dtype
+
+
+def finfo(type: Union[Dtype, Array], /) -> finfo_object:
+    """
+    Array API compatible wrapper for :py:func:`np.finfo <numpy.finfo>`.
+
+    See its docstring for more information.
+    """
+    fi = np.finfo(type)
+    # Note: The types of the float data here are float, whereas in NumPy they
+    # are scalars of the corresponding float dtype.
+    return finfo_object(
+        fi.bits,
+        float(fi.eps),
+        float(fi.max),
+        float(fi.min),
+        float(fi.smallest_normal),
+        fi.dtype,
+    )
+
+
+def iinfo(type: Union[Dtype, Array], /) -> iinfo_object:
+    """
+    Array API compatible wrapper for :py:func:`np.iinfo <numpy.iinfo>`.
+
+    See its docstring for more information.
+    """
+    ii = np.iinfo(type)
+    return iinfo_object(ii.bits, ii.max, ii.min, ii.dtype)
+
+
+# Note: isdtype is a new function from the 2022.12 array API specification.
+def isdtype(
+    dtype: Dtype, kind: Union[Dtype, str, Tuple[Union[Dtype, str], ...]]
+) -> bool:
+    """
+    Returns a boolean indicating whether a provided dtype is of a specified data type ``kind``.
+
+    See
+    https://data-apis.org/array-api/latest/API_specification/generated/array_api.isdtype.html
+    for more details
+    """
+    if isinstance(kind, tuple):
+        # Disallow nested tuples
+        if any(isinstance(k, tuple) for k in kind):
+            raise TypeError("'kind' must be a dtype, str, or tuple of dtypes and strs")
+        return any(isdtype(dtype, k) for k in kind)
+    elif isinstance(kind, str):
+        if kind == 'bool':
+            return dtype in _boolean_dtypes
+        elif kind == 'signed integer':
+            return dtype in _signed_integer_dtypes
+        elif kind == 'unsigned integer':
+            return dtype in _unsigned_integer_dtypes
+        elif kind == 'integral':
+            return dtype in _integer_dtypes
+        elif kind == 'real floating':
+            return dtype in _real_floating_dtypes
+        elif kind == 'complex floating':
+            return dtype in _complex_floating_dtypes
+        elif kind == 'numeric':
+            return dtype in _numeric_dtypes
+        else:
+            raise ValueError(f"Unrecognized data type kind: {kind!r}")
+    elif kind in _all_dtypes:
+        return dtype == kind
+    else:
+        raise TypeError(f"'kind' must be a dtype, str, or tuple of dtypes and strs, not {type(kind).__name__}")
+
+def result_type(*arrays_and_dtypes: Union[Array, Dtype]) -> Dtype:
+    """
+    Array API compatible wrapper for :py:func:`np.result_type <numpy.result_type>`.
+
+    See its docstring for more information.
+    """
+    # Note: we use a custom implementation that gives only the type promotions
+    # required by the spec rather than using np.result_type. NumPy implements
+    # too many extra type promotions like int64 + uint64 -> float64, and does
+    # value-based casting on scalar arrays.
+    A = []
+    for a in arrays_and_dtypes:
+        if isinstance(a, Array):
+            a = a.dtype
+        elif isinstance(a, np.ndarray) or a not in _all_dtypes:
+            raise TypeError("result_type() inputs must be array_api arrays or dtypes")
+        A.append(a)
+
+    if len(A) == 0:
+        raise ValueError("at least one array or dtype is required")
+    elif len(A) == 1:
+        return A[0]
+    else:
+        t = A[0]
+        for t2 in A[1:]:
+            t = _result_type(t, t2)
+        return t

.venv/lib/python3.11/site-packages/numpy/array_api/_dtypes.py

@@ -0,0 +1,180 @@
+import numpy as np
+
+# Note: we use dtype objects instead of dtype classes. The spec does not
+# require any behavior on dtypes other than equality.
+int8 = np.dtype("int8")
+int16 = np.dtype("int16")
+int32 = np.dtype("int32")
+int64 = np.dtype("int64")
+uint8 = np.dtype("uint8")
+uint16 = np.dtype("uint16")
+uint32 = np.dtype("uint32")
+uint64 = np.dtype("uint64")
+float32 = np.dtype("float32")
+float64 = np.dtype("float64")
+complex64 = np.dtype("complex64")
+complex128 = np.dtype("complex128")
+# Note: This name is changed
+bool = np.dtype("bool")
+
+_all_dtypes = (
+    int8,
+    int16,
+    int32,
+    int64,
+    uint8,
+    uint16,
+    uint32,
+    uint64,
+    float32,
+    float64,
+    complex64,
+    complex128,
+    bool,
+)
+_boolean_dtypes = (bool,)
+_real_floating_dtypes = (float32, float64)
+_floating_dtypes = (float32, float64, complex64, complex128)
+_complex_floating_dtypes = (complex64, complex128)
+_integer_dtypes = (int8, int16, int32, int64, uint8, uint16, uint32, uint64)
+_signed_integer_dtypes = (int8, int16, int32, int64)
+_unsigned_integer_dtypes = (uint8, uint16, uint32, uint64)
+_integer_or_boolean_dtypes = (
+    bool,
+    int8,
+    int16,
+    int32,
+    int64,
+    uint8,
+    uint16,
+    uint32,
+    uint64,
+)
+_real_numeric_dtypes = (
+    float32,
+    float64,
+    int8,
+    int16,
+    int32,
+    int64,
+    uint8,
+    uint16,
+    uint32,
+    uint64,
+)
+_numeric_dtypes = (
+    float32,
+    float64,
+    complex64,
+    complex128,
+    int8,
+    int16,
+    int32,
+    int64,
+    uint8,
+    uint16,
+    uint32,
+    uint64,
+)
+
+_dtype_categories = {
+    "all": _all_dtypes,
+    "real numeric": _real_numeric_dtypes,
+    "numeric": _numeric_dtypes,
+    "integer": _integer_dtypes,
+    "integer or boolean": _integer_or_boolean_dtypes,
+    "boolean": _boolean_dtypes,
+    "real floating-point": _floating_dtypes,
+    "complex floating-point": _complex_floating_dtypes,
+    "floating-point": _floating_dtypes,
+}
+
+
+# Note: the spec defines a restricted type promotion table compared to NumPy.
+# In particular, cross-kind promotions like integer + float or boolean +
+# integer are not allowed, even for functions that accept both kinds.
+# Additionally, NumPy promotes signed integer + uint64 to float64, but this
+# promotion is not allowed here. To be clear, Python scalar int objects are
+# allowed to promote to floating-point dtypes, but only in array operators
+# (see Array._promote_scalar) method in _array_object.py.
+_promotion_table = {
+    (int8, int8): int8,
+    (int8, int16): int16,
+    (int8, int32): int32,
+    (int8, int64): int64,
+    (int16, int8): int16,
+    (int16, int16): int16,
+    (int16, int32): int32,
+    (int16, int64): int64,
+    (int32, int8): int32,
+    (int32, int16): int32,
+    (int32, int32): int32,
+    (int32, int64): int64,
+    (int64, int8): int64,
+    (int64, int16): int64,
+    (int64, int32): int64,
+    (int64, int64): int64,
+    (uint8, uint8): uint8,
+    (uint8, uint16): uint16,
+    (uint8, uint32): uint32,
+    (uint8, uint64): uint64,
+    (uint16, uint8): uint16,
+    (uint16, uint16): uint16,
+    (uint16, uint32): uint32,
+    (uint16, uint64): uint64,
+    (uint32, uint8): uint32,
+    (uint32, uint16): uint32,
+    (uint32, uint32): uint32,
+    (uint32, uint64): uint64,
+    (uint64, uint8): uint64,
+    (uint64, uint16): uint64,
+    (uint64, uint32): uint64,
+    (uint64, uint64): uint64,
+    (int8, uint8): int16,
+    (int8, uint16): int32,
+    (int8, uint32): int64,
+    (int16, uint8): int16,
+    (int16, uint16): int32,
+    (int16, uint32): int64,
+    (int32, uint8): int32,
+    (int32, uint16): int32,
+    (int32, uint32): int64,
+    (int64, uint8): int64,
+    (int64, uint16): int64,
+    (int64, uint32): int64,
+    (uint8, int8): int16,
+    (uint16, int8): int32,
+    (uint32, int8): int64,
+    (uint8, int16): int16,
+    (uint16, int16): int32,
+    (uint32, int16): int64,
+    (uint8, int32): int32,
+    (uint16, int32): int32,
+    (uint32, int32): int64,
+    (uint8, int64): int64,
+    (uint16, int64): int64,
+    (uint32, int64): int64,
+    (float32, float32): float32,
+    (float32, float64): float64,
+    (float64, float32): float64,
+    (float64, float64): float64,
+    (complex64, complex64): complex64,
+    (complex64, complex128): complex128,
+    (complex128, complex64): complex128,
+    (complex128, complex128): complex128,
+    (float32, complex64): complex64,
+    (float32, complex128): complex128,
+    (float64, complex64): complex128,
+    (float64, complex128): complex128,
+    (complex64, float32): complex64,
+    (complex64, float64): complex128,
+    (complex128, float32): complex128,
+    (complex128, float64): complex128,
+    (bool, bool): bool,
+}
+
+
+def _result_type(type1, type2):
+    if (type1, type2) in _promotion_table:
+        return _promotion_table[type1, type2]
+    raise TypeError(f"{type1} and {type2} cannot be type promoted together")

.venv/lib/python3.11/site-packages/numpy/array_api/_elementwise_functions.py

@@ -0,0 +1,765 @@
+from __future__ import annotations
+
+from ._dtypes import (
+    _boolean_dtypes,
+    _floating_dtypes,
+    _real_floating_dtypes,
+    _complex_floating_dtypes,
+    _integer_dtypes,
+    _integer_or_boolean_dtypes,
+    _real_numeric_dtypes,
+    _numeric_dtypes,
+    _result_type,
+)
+from ._array_object import Array
+
+import numpy as np
+
+
+def abs(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.abs <numpy.abs>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in abs")
+    return Array._new(np.abs(x._array))
+
+
+# Note: the function name is different here
+def acos(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.arccos <numpy.arccos>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in acos")
+    return Array._new(np.arccos(x._array))
+
+
+# Note: the function name is different here
+def acosh(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.arccosh <numpy.arccosh>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in acosh")
+    return Array._new(np.arccosh(x._array))
+
+
+def add(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.add <numpy.add>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in add")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.add(x1._array, x2._array))
+
+
+# Note: the function name is different here
+def asin(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.arcsin <numpy.arcsin>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in asin")
+    return Array._new(np.arcsin(x._array))
+
+
+# Note: the function name is different here
+def asinh(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.arcsinh <numpy.arcsinh>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in asinh")
+    return Array._new(np.arcsinh(x._array))
+
+
+# Note: the function name is different here
+def atan(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.arctan <numpy.arctan>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in atan")
+    return Array._new(np.arctan(x._array))
+
+
+# Note: the function name is different here
+def atan2(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.arctan2 <numpy.arctan2>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _real_floating_dtypes or x2.dtype not in _real_floating_dtypes:
+        raise TypeError("Only real floating-point dtypes are allowed in atan2")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.arctan2(x1._array, x2._array))
+
+
+# Note: the function name is different here
+def atanh(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.arctanh <numpy.arctanh>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in atanh")
+    return Array._new(np.arctanh(x._array))
+
+
+def bitwise_and(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.bitwise_and <numpy.bitwise_and>`.
+
+    See its docstring for more information.
+    """
+    if (
+        x1.dtype not in _integer_or_boolean_dtypes
+        or x2.dtype not in _integer_or_boolean_dtypes
+    ):
+        raise TypeError("Only integer or boolean dtypes are allowed in bitwise_and")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.bitwise_and(x1._array, x2._array))
+
+
+# Note: the function name is different here
+def bitwise_left_shift(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.left_shift <numpy.left_shift>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _integer_dtypes or x2.dtype not in _integer_dtypes:
+        raise TypeError("Only integer dtypes are allowed in bitwise_left_shift")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    # Note: bitwise_left_shift is only defined for x2 nonnegative.
+    if np.any(x2._array < 0):
+        raise ValueError("bitwise_left_shift(x1, x2) is only defined for x2 >= 0")
+    return Array._new(np.left_shift(x1._array, x2._array))
+
+
+# Note: the function name is different here
+def bitwise_invert(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.invert <numpy.invert>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _integer_or_boolean_dtypes:
+        raise TypeError("Only integer or boolean dtypes are allowed in bitwise_invert")
+    return Array._new(np.invert(x._array))
+
+
+def bitwise_or(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.bitwise_or <numpy.bitwise_or>`.
+
+    See its docstring for more information.
+    """
+    if (
+        x1.dtype not in _integer_or_boolean_dtypes
+        or x2.dtype not in _integer_or_boolean_dtypes
+    ):
+        raise TypeError("Only integer or boolean dtypes are allowed in bitwise_or")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.bitwise_or(x1._array, x2._array))
+
+
+# Note: the function name is different here
+def bitwise_right_shift(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.right_shift <numpy.right_shift>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _integer_dtypes or x2.dtype not in _integer_dtypes:
+        raise TypeError("Only integer dtypes are allowed in bitwise_right_shift")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    # Note: bitwise_right_shift is only defined for x2 nonnegative.
+    if np.any(x2._array < 0):
+        raise ValueError("bitwise_right_shift(x1, x2) is only defined for x2 >= 0")
+    return Array._new(np.right_shift(x1._array, x2._array))
+
+
+def bitwise_xor(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.bitwise_xor <numpy.bitwise_xor>`.
+
+    See its docstring for more information.
+    """
+    if (
+        x1.dtype not in _integer_or_boolean_dtypes
+        or x2.dtype not in _integer_or_boolean_dtypes
+    ):
+        raise TypeError("Only integer or boolean dtypes are allowed in bitwise_xor")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.bitwise_xor(x1._array, x2._array))
+
+
+def ceil(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.ceil <numpy.ceil>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in ceil")
+    if x.dtype in _integer_dtypes:
+        # Note: The return dtype of ceil is the same as the input
+        return x
+    return Array._new(np.ceil(x._array))
+
+
+def conj(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.conj <numpy.conj>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _complex_floating_dtypes:
+        raise TypeError("Only complex floating-point dtypes are allowed in conj")
+    return Array._new(np.conj(x))
+
+
+def cos(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.cos <numpy.cos>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in cos")
+    return Array._new(np.cos(x._array))
+
+
+def cosh(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.cosh <numpy.cosh>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in cosh")
+    return Array._new(np.cosh(x._array))
+
+
+def divide(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.divide <numpy.divide>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _floating_dtypes or x2.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in divide")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.divide(x1._array, x2._array))
+
+
+def equal(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.equal <numpy.equal>`.
+
+    See its docstring for more information.
+    """
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.equal(x1._array, x2._array))
+
+
+def exp(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.exp <numpy.exp>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in exp")
+    return Array._new(np.exp(x._array))
+
+
+def expm1(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.expm1 <numpy.expm1>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in expm1")
+    return Array._new(np.expm1(x._array))
+
+
+def floor(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.floor <numpy.floor>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in floor")
+    if x.dtype in _integer_dtypes:
+        # Note: The return dtype of floor is the same as the input
+        return x
+    return Array._new(np.floor(x._array))
+
+
+def floor_divide(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.floor_divide <numpy.floor_divide>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _real_numeric_dtypes or x2.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in floor_divide")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.floor_divide(x1._array, x2._array))
+
+
+def greater(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.greater <numpy.greater>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _real_numeric_dtypes or x2.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in greater")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.greater(x1._array, x2._array))
+
+
+def greater_equal(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.greater_equal <numpy.greater_equal>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _real_numeric_dtypes or x2.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in greater_equal")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.greater_equal(x1._array, x2._array))
+
+
+def imag(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.imag <numpy.imag>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _complex_floating_dtypes:
+        raise TypeError("Only complex floating-point dtypes are allowed in imag")
+    return Array._new(np.imag(x))
+
+
+def isfinite(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.isfinite <numpy.isfinite>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in isfinite")
+    return Array._new(np.isfinite(x._array))
+
+
+def isinf(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.isinf <numpy.isinf>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in isinf")
+    return Array._new(np.isinf(x._array))
+
+
+def isnan(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.isnan <numpy.isnan>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in isnan")
+    return Array._new(np.isnan(x._array))
+
+
+def less(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.less <numpy.less>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _real_numeric_dtypes or x2.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in less")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.less(x1._array, x2._array))
+
+
+def less_equal(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.less_equal <numpy.less_equal>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _real_numeric_dtypes or x2.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in less_equal")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.less_equal(x1._array, x2._array))
+
+
+def log(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.log <numpy.log>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in log")
+    return Array._new(np.log(x._array))
+
+
+def log1p(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.log1p <numpy.log1p>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in log1p")
+    return Array._new(np.log1p(x._array))
+
+
+def log2(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.log2 <numpy.log2>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in log2")
+    return Array._new(np.log2(x._array))
+
+
+def log10(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.log10 <numpy.log10>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in log10")
+    return Array._new(np.log10(x._array))
+
+
+def logaddexp(x1: Array, x2: Array) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.logaddexp <numpy.logaddexp>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _real_floating_dtypes or x2.dtype not in _real_floating_dtypes:
+        raise TypeError("Only real floating-point dtypes are allowed in logaddexp")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.logaddexp(x1._array, x2._array))
+
+
+def logical_and(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.logical_and <numpy.logical_and>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _boolean_dtypes or x2.dtype not in _boolean_dtypes:
+        raise TypeError("Only boolean dtypes are allowed in logical_and")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.logical_and(x1._array, x2._array))
+
+
+def logical_not(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.logical_not <numpy.logical_not>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _boolean_dtypes:
+        raise TypeError("Only boolean dtypes are allowed in logical_not")
+    return Array._new(np.logical_not(x._array))
+
+
+def logical_or(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.logical_or <numpy.logical_or>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _boolean_dtypes or x2.dtype not in _boolean_dtypes:
+        raise TypeError("Only boolean dtypes are allowed in logical_or")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.logical_or(x1._array, x2._array))
+
+
+def logical_xor(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.logical_xor <numpy.logical_xor>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _boolean_dtypes or x2.dtype not in _boolean_dtypes:
+        raise TypeError("Only boolean dtypes are allowed in logical_xor")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.logical_xor(x1._array, x2._array))
+
+
+def multiply(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.multiply <numpy.multiply>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in multiply")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.multiply(x1._array, x2._array))
+
+
+def negative(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.negative <numpy.negative>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in negative")
+    return Array._new(np.negative(x._array))
+
+
+def not_equal(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.not_equal <numpy.not_equal>`.
+
+    See its docstring for more information.
+    """
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.not_equal(x1._array, x2._array))
+
+
+def positive(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.positive <numpy.positive>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in positive")
+    return Array._new(np.positive(x._array))
+
+
+# Note: the function name is different here
+def pow(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.power <numpy.power>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in pow")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.power(x1._array, x2._array))
+
+
+def real(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.real <numpy.real>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _complex_floating_dtypes:
+        raise TypeError("Only complex floating-point dtypes are allowed in real")
+    return Array._new(np.real(x))
+
+
+def remainder(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.remainder <numpy.remainder>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _real_numeric_dtypes or x2.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in remainder")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.remainder(x1._array, x2._array))
+
+
+def round(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.round <numpy.round>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in round")
+    return Array._new(np.round(x._array))
+
+
+def sign(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.sign <numpy.sign>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in sign")
+    return Array._new(np.sign(x._array))
+
+
+def sin(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.sin <numpy.sin>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in sin")
+    return Array._new(np.sin(x._array))
+
+
+def sinh(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.sinh <numpy.sinh>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in sinh")
+    return Array._new(np.sinh(x._array))
+
+
+def square(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.square <numpy.square>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in square")
+    return Array._new(np.square(x._array))
+
+
+def sqrt(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.sqrt <numpy.sqrt>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in sqrt")
+    return Array._new(np.sqrt(x._array))
+
+
+def subtract(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.subtract <numpy.subtract>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in subtract")
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.subtract(x1._array, x2._array))
+
+
+def tan(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.tan <numpy.tan>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in tan")
+    return Array._new(np.tan(x._array))
+
+
+def tanh(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.tanh <numpy.tanh>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _floating_dtypes:
+        raise TypeError("Only floating-point dtypes are allowed in tanh")
+    return Array._new(np.tanh(x._array))
+
+
+def trunc(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.trunc <numpy.trunc>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in trunc")
+    if x.dtype in _integer_dtypes:
+        # Note: The return dtype of trunc is the same as the input
+        return x
+    return Array._new(np.trunc(x._array))

.venv/lib/python3.11/site-packages/numpy/array_api/_indexing_functions.py

@@ -0,0 +1,20 @@
+from __future__ import annotations
+
+from ._array_object import Array
+from ._dtypes import _integer_dtypes
+
+import numpy as np
+
+def take(x: Array, indices: Array, /, *, axis: Optional[int] = None) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.take <numpy.take>`.
+
+    See its docstring for more information.
+    """
+    if axis is None and x.ndim != 1:
+        raise ValueError("axis must be specified when ndim > 1")
+    if indices.dtype not in _integer_dtypes:
+        raise TypeError("Only integer dtypes are allowed in indexing")
+    if indices.ndim != 1:
+        raise ValueError("Only 1-dim indices array is supported")
+    return Array._new(np.take(x._array, indices._array, axis=axis))

.venv/lib/python3.11/site-packages/numpy/array_api/_manipulation_functions.py

@@ -0,0 +1,112 @@
+from __future__ import annotations
+
+from ._array_object import Array
+from ._data_type_functions import result_type
+
+from typing import List, Optional, Tuple, Union
+
+import numpy as np
+
+# Note: the function name is different here
+def concat(
+    arrays: Union[Tuple[Array, ...], List[Array]], /, *, axis: Optional[int] = 0
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.concatenate <numpy.concatenate>`.
+
+    See its docstring for more information.
+    """
+    # Note: Casting rules here are different from the np.concatenate default
+    # (no for scalars with axis=None, no cross-kind casting)
+    dtype = result_type(*arrays)
+    arrays = tuple(a._array for a in arrays)
+    return Array._new(np.concatenate(arrays, axis=axis, dtype=dtype))
+
+
+def expand_dims(x: Array, /, *, axis: int) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.expand_dims <numpy.expand_dims>`.
+
+    See its docstring for more information.
+    """
+    return Array._new(np.expand_dims(x._array, axis))
+
+
+def flip(x: Array, /, *, axis: Optional[Union[int, Tuple[int, ...]]] = None) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.flip <numpy.flip>`.
+
+    See its docstring for more information.
+    """
+    return Array._new(np.flip(x._array, axis=axis))
+
+
+# Note: The function name is different here (see also matrix_transpose).
+# Unlike transpose(), the axes argument is required.
+def permute_dims(x: Array, /, axes: Tuple[int, ...]) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.transpose <numpy.transpose>`.
+
+    See its docstring for more information.
+    """
+    return Array._new(np.transpose(x._array, axes))
+
+
+# Note: the optional argument is called 'shape', not 'newshape'
+def reshape(x: Array, 
+            /, 
+            shape: Tuple[int, ...],
+            *,
+            copy: Optional[Bool] = None) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.reshape <numpy.reshape>`.
+
+    See its docstring for more information.
+    """
+
+    data = x._array
+    if copy:
+        data = np.copy(data)
+
+    reshaped = np.reshape(data, shape)
+
+    if copy is False and not np.shares_memory(data, reshaped):
+        raise AttributeError("Incompatible shape for in-place modification.")
+
+    return Array._new(reshaped)
+
+
+def roll(
+    x: Array,
+    /,
+    shift: Union[int, Tuple[int, ...]],
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.roll <numpy.roll>`.
+
+    See its docstring for more information.
+    """
+    return Array._new(np.roll(x._array, shift, axis=axis))
+
+
+def squeeze(x: Array, /, axis: Union[int, Tuple[int, ...]]) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.squeeze <numpy.squeeze>`.
+
+    See its docstring for more information.
+    """
+    return Array._new(np.squeeze(x._array, axis=axis))
+
+
+def stack(arrays: Union[Tuple[Array, ...], List[Array]], /, *, axis: int = 0) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.stack <numpy.stack>`.
+
+    See its docstring for more information.
+    """
+    # Call result type here just to raise on disallowed type combinations
+    result_type(*arrays)
+    arrays = tuple(a._array for a in arrays)
+    return Array._new(np.stack(arrays, axis=axis))

.venv/lib/python3.11/site-packages/numpy/array_api/_searching_functions.py

@@ -0,0 +1,51 @@
+from __future__ import annotations
+
+from ._array_object import Array
+from ._dtypes import _result_type, _real_numeric_dtypes
+
+from typing import Optional, Tuple
+
+import numpy as np
+
+
+def argmax(x: Array, /, *, axis: Optional[int] = None, keepdims: bool = False) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.argmax <numpy.argmax>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in argmax")
+    return Array._new(np.asarray(np.argmax(x._array, axis=axis, keepdims=keepdims)))
+
+
+def argmin(x: Array, /, *, axis: Optional[int] = None, keepdims: bool = False) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.argmin <numpy.argmin>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in argmin")
+    return Array._new(np.asarray(np.argmin(x._array, axis=axis, keepdims=keepdims)))
+
+
+def nonzero(x: Array, /) -> Tuple[Array, ...]:
+    """
+    Array API compatible wrapper for :py:func:`np.nonzero <numpy.nonzero>`.
+
+    See its docstring for more information.
+    """
+    return tuple(Array._new(i) for i in np.nonzero(x._array))
+
+
+def where(condition: Array, x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.where <numpy.where>`.
+
+    See its docstring for more information.
+    """
+    # Call result type here just to raise on disallowed type combinations
+    _result_type(x1.dtype, x2.dtype)
+    x1, x2 = Array._normalize_two_args(x1, x2)
+    return Array._new(np.where(condition._array, x1._array, x2._array))

.venv/lib/python3.11/site-packages/numpy/array_api/_set_functions.py

@@ -0,0 +1,106 @@
+from __future__ import annotations
+
+from ._array_object import Array
+
+from typing import NamedTuple
+
+import numpy as np
+
+# Note: np.unique() is split into four functions in the array API:
+# unique_all, unique_counts, unique_inverse, and unique_values (this is done
+# to remove polymorphic return types).
+
+# Note: The various unique() functions are supposed to return multiple NaNs.
+# This does not match the NumPy behavior, however, this is currently left as a
+# TODO in this implementation as this behavior may be reverted in np.unique().
+# See https://github.com/numpy/numpy/issues/20326.
+
+# Note: The functions here return a namedtuple (np.unique() returns a normal
+# tuple).
+
+class UniqueAllResult(NamedTuple):
+    values: Array
+    indices: Array
+    inverse_indices: Array
+    counts: Array
+
+
+class UniqueCountsResult(NamedTuple):
+    values: Array
+    counts: Array
+
+
+class UniqueInverseResult(NamedTuple):
+    values: Array
+    inverse_indices: Array
+
+
+def unique_all(x: Array, /) -> UniqueAllResult:
+    """
+    Array API compatible wrapper for :py:func:`np.unique <numpy.unique>`.
+
+    See its docstring for more information.
+    """
+    values, indices, inverse_indices, counts = np.unique(
+        x._array,
+        return_counts=True,
+        return_index=True,
+        return_inverse=True,
+        equal_nan=False,
+    )
+    # np.unique() flattens inverse indices, but they need to share x's shape
+    # See https://github.com/numpy/numpy/issues/20638
+    inverse_indices = inverse_indices.reshape(x.shape)
+    return UniqueAllResult(
+        Array._new(values),
+        Array._new(indices),
+        Array._new(inverse_indices),
+        Array._new(counts),
+    )
+
+
+def unique_counts(x: Array, /) -> UniqueCountsResult:
+    res = np.unique(
+        x._array,
+        return_counts=True,
+        return_index=False,
+        return_inverse=False,
+        equal_nan=False,
+    )
+
+    return UniqueCountsResult(*[Array._new(i) for i in res])
+
+
+def unique_inverse(x: Array, /) -> UniqueInverseResult:
+    """
+    Array API compatible wrapper for :py:func:`np.unique <numpy.unique>`.
+
+    See its docstring for more information.
+    """
+    values, inverse_indices = np.unique(
+        x._array,
+        return_counts=False,
+        return_index=False,
+        return_inverse=True,
+        equal_nan=False,
+    )
+    # np.unique() flattens inverse indices, but they need to share x's shape
+    # See https://github.com/numpy/numpy/issues/20638
+    inverse_indices = inverse_indices.reshape(x.shape)
+    return UniqueInverseResult(Array._new(values), Array._new(inverse_indices))
+
+
+def unique_values(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.unique <numpy.unique>`.
+
+    See its docstring for more information.
+    """
+    res = np.unique(
+        x._array,
+        return_counts=False,
+        return_index=False,
+        return_inverse=False,
+        equal_nan=False,
+    )
+    return Array._new(res)

.venv/lib/python3.11/site-packages/numpy/array_api/_sorting_functions.py

@@ -0,0 +1,54 @@
+from __future__ import annotations
+
+from ._array_object import Array
+from ._dtypes import _real_numeric_dtypes
+
+import numpy as np
+
+
+# Note: the descending keyword argument is new in this function
+def argsort(
+    x: Array, /, *, axis: int = -1, descending: bool = False, stable: bool = True
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.argsort <numpy.argsort>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in argsort")
+    # Note: this keyword argument is different, and the default is different.
+    kind = "stable" if stable else "quicksort"
+    if not descending:
+        res = np.argsort(x._array, axis=axis, kind=kind)
+    else:
+        # As NumPy has no native descending sort, we imitate it here. Note that
+        # simply flipping the results of np.argsort(x._array, ...) would not
+        # respect the relative order like it would in native descending sorts.
+        res = np.flip(
+            np.argsort(np.flip(x._array, axis=axis), axis=axis, kind=kind),
+            axis=axis,
+        )
+        # Rely on flip()/argsort() to validate axis
+        normalised_axis = axis if axis >= 0 else x.ndim + axis
+        max_i = x.shape[normalised_axis] - 1
+        res = max_i - res
+    return Array._new(res)
+
+# Note: the descending keyword argument is new in this function
+def sort(
+    x: Array, /, *, axis: int = -1, descending: bool = False, stable: bool = True
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.sort <numpy.sort>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in sort")
+    # Note: this keyword argument is different, and the default is different.
+    kind = "stable" if stable else "quicksort"
+    res = np.sort(x._array, axis=axis, kind=kind)
+    if descending:
+        res = np.flip(res, axis=axis)
+    return Array._new(res)

.venv/lib/python3.11/site-packages/numpy/array_api/_statistical_functions.py

@@ -0,0 +1,122 @@
+from __future__ import annotations
+
+from ._dtypes import (
+    _real_floating_dtypes,
+    _real_numeric_dtypes,
+    _numeric_dtypes,
+)
+from ._array_object import Array
+from ._dtypes import float32, float64, complex64, complex128
+
+from typing import TYPE_CHECKING, Optional, Tuple, Union
+
+if TYPE_CHECKING:
+    from ._typing import Dtype
+
+import numpy as np
+
+
+def max(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    keepdims: bool = False,
+) -> Array:
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in max")
+    return Array._new(np.max(x._array, axis=axis, keepdims=keepdims))
+
+
+def mean(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    keepdims: bool = False,
+) -> Array:
+    if x.dtype not in _real_floating_dtypes:
+        raise TypeError("Only real floating-point dtypes are allowed in mean")
+    return Array._new(np.mean(x._array, axis=axis, keepdims=keepdims))
+
+
+def min(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    keepdims: bool = False,
+) -> Array:
+    if x.dtype not in _real_numeric_dtypes:
+        raise TypeError("Only real numeric dtypes are allowed in min")
+    return Array._new(np.min(x._array, axis=axis, keepdims=keepdims))
+
+
+def prod(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    dtype: Optional[Dtype] = None,
+    keepdims: bool = False,
+) -> Array:
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in prod")
+    # Note: sum() and prod() always upcast for dtype=None. `np.prod` does that
+    # for integers, but not for float32 or complex64, so we need to
+    # special-case it here
+    if dtype is None:
+        if x.dtype == float32:
+            dtype = float64
+        elif x.dtype == complex64:
+            dtype = complex128
+    return Array._new(np.prod(x._array, dtype=dtype, axis=axis, keepdims=keepdims))
+
+
+def std(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    correction: Union[int, float] = 0.0,
+    keepdims: bool = False,
+) -> Array:
+    # Note: the keyword argument correction is different here
+    if x.dtype not in _real_floating_dtypes:
+        raise TypeError("Only real floating-point dtypes are allowed in std")
+    return Array._new(np.std(x._array, axis=axis, ddof=correction, keepdims=keepdims))
+
+
+def sum(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    dtype: Optional[Dtype] = None,
+    keepdims: bool = False,
+) -> Array:
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError("Only numeric dtypes are allowed in sum")
+    # Note: sum() and prod() always upcast for dtype=None. `np.sum` does that
+    # for integers, but not for float32 or complex64, so we need to
+    # special-case it here
+    if dtype is None:
+        if x.dtype == float32:
+            dtype = float64
+        elif x.dtype == complex64:
+            dtype = complex128
+    return Array._new(np.sum(x._array, axis=axis, dtype=dtype, keepdims=keepdims))
+
+
+def var(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    correction: Union[int, float] = 0.0,
+    keepdims: bool = False,
+) -> Array:
+    # Note: the keyword argument correction is different here
+    if x.dtype not in _real_floating_dtypes:
+        raise TypeError("Only real floating-point dtypes are allowed in var")
+    return Array._new(np.var(x._array, axis=axis, ddof=correction, keepdims=keepdims))

.venv/lib/python3.11/site-packages/numpy/array_api/_typing.py

@@ -0,0 +1,76 @@
+"""
+This file defines the types for type annotations.
+
+These names aren't part of the module namespace, but they are used in the
+annotations in the function signatures. The functions in the module are only
+valid for inputs that match the given type annotations.
+"""
+
+from __future__ import annotations
+
+__all__ = [
+    "Array",
+    "Device",
+    "Dtype",
+    "SupportsDLPack",
+    "SupportsBufferProtocol",
+    "PyCapsule",
+]
+
+import sys
+
+from typing import (
+    Any,
+    Literal,
+    Sequence,
+    Type,
+    Union,
+    TypeVar,
+    Protocol,
+)
+
+from ._array_object import Array
+from numpy import (
+    dtype,
+    int8,
+    int16,
+    int32,
+    int64,
+    uint8,
+    uint16,
+    uint32,
+    uint64,
+    float32,
+    float64,
+)
+
+_T_co = TypeVar("_T_co", covariant=True)
+
+class NestedSequence(Protocol[_T_co]):
+    def __getitem__(self, key: int, /) -> _T_co | NestedSequence[_T_co]: ...
+    def __len__(self, /) -> int: ...
+
+Device = Literal["cpu"]
+
+Dtype = dtype[Union[
+    int8,
+    int16,
+    int32,
+    int64,
+    uint8,
+    uint16,
+    uint32,
+    uint64,
+    float32,
+    float64,
+]]
+
+if sys.version_info >= (3, 12):
+    from collections.abc import Buffer as SupportsBufferProtocol
+else:
+    SupportsBufferProtocol = Any
+
+PyCapsule = Any
+
+class SupportsDLPack(Protocol):
+    def __dlpack__(self, /, *, stream: None = ...) -> PyCapsule: ...

.venv/lib/python3.11/site-packages/numpy/array_api/_utility_functions.py

@@ -0,0 +1,37 @@
+from __future__ import annotations
+
+from ._array_object import Array
+
+from typing import Optional, Tuple, Union
+
+import numpy as np
+
+
+def all(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    keepdims: bool = False,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.all <numpy.all>`.
+
+    See its docstring for more information.
+    """
+    return Array._new(np.asarray(np.all(x._array, axis=axis, keepdims=keepdims)))
+
+
+def any(
+    x: Array,
+    /,
+    *,
+    axis: Optional[Union[int, Tuple[int, ...]]] = None,
+    keepdims: bool = False,
+) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.any <numpy.any>`.
+
+    See its docstring for more information.
+    """
+    return Array._new(np.asarray(np.any(x._array, axis=axis, keepdims=keepdims)))

.venv/lib/python3.11/site-packages/numpy/array_api/linalg.py

@@ -0,0 +1,466 @@
+from __future__ import annotations
+
+from ._dtypes import (
+    _floating_dtypes,
+    _numeric_dtypes,
+    float32,
+    float64,
+    complex64,
+    complex128
+)
+from ._manipulation_functions import reshape
+from ._elementwise_functions import conj
+from ._array_object import Array
+
+from ..core.numeric import normalize_axis_tuple
+
+from typing import TYPE_CHECKING
+if TYPE_CHECKING:
+    from ._typing import Literal, Optional, Sequence, Tuple, Union, Dtype
+
+from typing import NamedTuple
+
+import numpy.linalg
+import numpy as np
+
+class EighResult(NamedTuple):
+    eigenvalues: Array
+    eigenvectors: Array
+
+class QRResult(NamedTuple):
+    Q: Array
+    R: Array
+
+class SlogdetResult(NamedTuple):
+    sign: Array
+    logabsdet: Array
+
+class SVDResult(NamedTuple):
+    U: Array
+    S: Array
+    Vh: Array
+
+# Note: the inclusion of the upper keyword is different from
+# np.linalg.cholesky, which does not have it.
+def cholesky(x: Array, /, *, upper: bool = False) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.cholesky <numpy.linalg.cholesky>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.cholesky.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in cholesky')
+    L = np.linalg.cholesky(x._array)
+    if upper:
+        U = Array._new(L).mT
+        if U.dtype in [complex64, complex128]:
+            U = conj(U)
+        return U
+    return Array._new(L)
+
+# Note: cross is the numpy top-level namespace, not np.linalg
+def cross(x1: Array, x2: Array, /, *, axis: int = -1) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.cross <numpy.cross>`.
+
+    See its docstring for more information.
+    """
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError('Only numeric dtypes are allowed in cross')
+    # Note: this is different from np.cross(), which broadcasts
+    if x1.shape != x2.shape:
+        raise ValueError('x1 and x2 must have the same shape')
+    if x1.ndim == 0:
+        raise ValueError('cross() requires arrays of dimension at least 1')
+    # Note: this is different from np.cross(), which allows dimension 2
+    if x1.shape[axis] != 3:
+        raise ValueError('cross() dimension must equal 3')
+    return Array._new(np.cross(x1._array, x2._array, axis=axis))
+
+def det(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.det <numpy.linalg.det>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.det.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in det')
+    return Array._new(np.linalg.det(x._array))
+
+# Note: diagonal is the numpy top-level namespace, not np.linalg
+def diagonal(x: Array, /, *, offset: int = 0) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.diagonal <numpy.diagonal>`.
+
+    See its docstring for more information.
+    """
+    # Note: diagonal always operates on the last two axes, whereas np.diagonal
+    # operates on the first two axes by default
+    return Array._new(np.diagonal(x._array, offset=offset, axis1=-2, axis2=-1))
+
+
+def eigh(x: Array, /) -> EighResult:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.eigh <numpy.linalg.eigh>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.eigh.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in eigh')
+
+    # Note: the return type here is a namedtuple, which is different from
+    # np.eigh, which only returns a tuple.
+    return EighResult(*map(Array._new, np.linalg.eigh(x._array)))
+
+
+def eigvalsh(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.eigvalsh <numpy.linalg.eigvalsh>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.eigvalsh.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in eigvalsh')
+
+    return Array._new(np.linalg.eigvalsh(x._array))
+
+def inv(x: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.inv <numpy.linalg.inv>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.inv.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in inv')
+
+    return Array._new(np.linalg.inv(x._array))
+
+
+# Note: matmul is the numpy top-level namespace but not in np.linalg
+def matmul(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.matmul <numpy.matmul>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to numeric dtypes only is different from
+    # np.matmul.
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError('Only numeric dtypes are allowed in matmul')
+
+    return Array._new(np.matmul(x1._array, x2._array))
+
+
+# Note: the name here is different from norm(). The array API norm is split
+# into matrix_norm and vector_norm().
+
+# The type for ord should be Optional[Union[int, float, Literal[np.inf,
+# -np.inf, 'fro', 'nuc']]], but Literal does not support floating-point
+# literals.
+def matrix_norm(x: Array, /, *, keepdims: bool = False, ord: Optional[Union[int, float, Literal['fro', 'nuc']]] = 'fro') -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.norm <numpy.linalg.norm>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.norm.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in matrix_norm')
+
+    return Array._new(np.linalg.norm(x._array, axis=(-2, -1), keepdims=keepdims, ord=ord))
+
+
+def matrix_power(x: Array, n: int, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.matrix_power <numpy.matrix_power>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.matrix_power.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed for the first argument of matrix_power')
+
+    # np.matrix_power already checks if n is an integer
+    return Array._new(np.linalg.matrix_power(x._array, n))
+
+# Note: the keyword argument name rtol is different from np.linalg.matrix_rank
+def matrix_rank(x: Array, /, *, rtol: Optional[Union[float, Array]] = None) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.matrix_rank <numpy.matrix_rank>`.
+
+    See its docstring for more information.
+    """
+    # Note: this is different from np.linalg.matrix_rank, which supports 1
+    # dimensional arrays.
+    if x.ndim < 2:
+        raise np.linalg.LinAlgError("1-dimensional array given. Array must be at least two-dimensional")
+    S = np.linalg.svd(x._array, compute_uv=False)
+    if rtol is None:
+        tol = S.max(axis=-1, keepdims=True) * max(x.shape[-2:]) * np.finfo(S.dtype).eps
+    else:
+        if isinstance(rtol, Array):
+            rtol = rtol._array
+        # Note: this is different from np.linalg.matrix_rank, which does not multiply
+        # the tolerance by the largest singular value.
+        tol = S.max(axis=-1, keepdims=True)*np.asarray(rtol)[..., np.newaxis]
+    return Array._new(np.count_nonzero(S > tol, axis=-1))
+
+
+# Note: this function is new in the array API spec. Unlike transpose, it only
+# transposes the last two axes.
+def matrix_transpose(x: Array, /) -> Array:
+    if x.ndim < 2:
+        raise ValueError("x must be at least 2-dimensional for matrix_transpose")
+    return Array._new(np.swapaxes(x._array, -1, -2))
+
+# Note: outer is the numpy top-level namespace, not np.linalg
+def outer(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.outer <numpy.outer>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to numeric dtypes only is different from
+    # np.outer.
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError('Only numeric dtypes are allowed in outer')
+
+    # Note: the restriction to only 1-dim arrays is different from np.outer
+    if x1.ndim != 1 or x2.ndim != 1:
+        raise ValueError('The input arrays to outer must be 1-dimensional')
+
+    return Array._new(np.outer(x1._array, x2._array))
+
+# Note: the keyword argument name rtol is different from np.linalg.pinv
+def pinv(x: Array, /, *, rtol: Optional[Union[float, Array]] = None) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.pinv <numpy.linalg.pinv>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.pinv.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in pinv')
+
+    # Note: this is different from np.linalg.pinv, which does not multiply the
+    # default tolerance by max(M, N).
+    if rtol is None:
+        rtol = max(x.shape[-2:]) * np.finfo(x.dtype).eps
+    return Array._new(np.linalg.pinv(x._array, rcond=rtol))
+
+def qr(x: Array, /, *, mode: Literal['reduced', 'complete'] = 'reduced') -> QRResult:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.qr <numpy.linalg.qr>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.qr.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in qr')
+
+    # Note: the return type here is a namedtuple, which is different from
+    # np.linalg.qr, which only returns a tuple.
+    return QRResult(*map(Array._new, np.linalg.qr(x._array, mode=mode)))
+
+def slogdet(x: Array, /) -> SlogdetResult:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.slogdet <numpy.linalg.slogdet>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.slogdet.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in slogdet')
+
+    # Note: the return type here is a namedtuple, which is different from
+    # np.linalg.slogdet, which only returns a tuple.
+    return SlogdetResult(*map(Array._new, np.linalg.slogdet(x._array)))
+
+# Note: unlike np.linalg.solve, the array API solve() only accepts x2 as a
+# vector when it is exactly 1-dimensional. All other cases treat x2 as a stack
+# of matrices. The np.linalg.solve behavior of allowing stacks of both
+# matrices and vectors is ambiguous c.f.
+# https://github.com/numpy/numpy/issues/15349 and
+# https://github.com/data-apis/array-api/issues/285.
+
+# To workaround this, the below is the code from np.linalg.solve except
+# only calling solve1 in the exactly 1D case.
+def _solve(a, b):
+    from ..linalg.linalg import (_makearray, _assert_stacked_2d,
+                                 _assert_stacked_square, _commonType,
+                                 isComplexType, get_linalg_error_extobj,
+                                 _raise_linalgerror_singular)
+    from ..linalg import _umath_linalg
+
+    a, _ = _makearray(a)
+    _assert_stacked_2d(a)
+    _assert_stacked_square(a)
+    b, wrap = _makearray(b)
+    t, result_t = _commonType(a, b)
+
+    # This part is different from np.linalg.solve
+    if b.ndim == 1:
+        gufunc = _umath_linalg.solve1
+    else:
+        gufunc = _umath_linalg.solve
+
+    # This does nothing currently but is left in because it will be relevant
+    # when complex dtype support is added to the spec in 2022.
+    signature = 'DD->D' if isComplexType(t) else 'dd->d'
+    with np.errstate(call=_raise_linalgerror_singular, invalid='call',
+                     over='ignore', divide='ignore', under='ignore'):
+        r = gufunc(a, b, signature=signature)
+
+    return wrap(r.astype(result_t, copy=False))
+
+def solve(x1: Array, x2: Array, /) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.solve <numpy.linalg.solve>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.solve.
+    if x1.dtype not in _floating_dtypes or x2.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in solve')
+
+    return Array._new(_solve(x1._array, x2._array))
+
+def svd(x: Array, /, *, full_matrices: bool = True) -> SVDResult:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.svd <numpy.linalg.svd>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.svd.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in svd')
+
+    # Note: the return type here is a namedtuple, which is different from
+    # np.svd, which only returns a tuple.
+    return SVDResult(*map(Array._new, np.linalg.svd(x._array, full_matrices=full_matrices)))
+
+# Note: svdvals is not in NumPy (but it is in SciPy). It is equivalent to
+# np.linalg.svd(compute_uv=False).
+def svdvals(x: Array, /) -> Union[Array, Tuple[Array, ...]]:
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in svdvals')
+    return Array._new(np.linalg.svd(x._array, compute_uv=False))
+
+# Note: tensordot is the numpy top-level namespace but not in np.linalg
+
+# Note: axes must be a tuple, unlike np.tensordot where it can be an array or array-like.
+def tensordot(x1: Array, x2: Array, /, *, axes: Union[int, Tuple[Sequence[int], Sequence[int]]] = 2) -> Array:
+    # Note: the restriction to numeric dtypes only is different from
+    # np.tensordot.
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError('Only numeric dtypes are allowed in tensordot')
+
+    return Array._new(np.tensordot(x1._array, x2._array, axes=axes))
+
+# Note: trace is the numpy top-level namespace, not np.linalg
+def trace(x: Array, /, *, offset: int = 0, dtype: Optional[Dtype] = None) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.trace <numpy.trace>`.
+
+    See its docstring for more information.
+    """
+    if x.dtype not in _numeric_dtypes:
+        raise TypeError('Only numeric dtypes are allowed in trace')
+
+    # Note: trace() works the same as sum() and prod() (see
+    # _statistical_functions.py)
+    if dtype is None:
+        if x.dtype == float32:
+            dtype = float64
+        elif x.dtype == complex64:
+            dtype = complex128
+    # Note: trace always operates on the last two axes, whereas np.trace
+    # operates on the first two axes by default
+    return Array._new(np.asarray(np.trace(x._array, offset=offset, axis1=-2, axis2=-1, dtype=dtype)))
+
+# Note: vecdot is not in NumPy
+def vecdot(x1: Array, x2: Array, /, *, axis: int = -1) -> Array:
+    if x1.dtype not in _numeric_dtypes or x2.dtype not in _numeric_dtypes:
+        raise TypeError('Only numeric dtypes are allowed in vecdot')
+    ndim = max(x1.ndim, x2.ndim)
+    x1_shape = (1,)*(ndim - x1.ndim) + tuple(x1.shape)
+    x2_shape = (1,)*(ndim - x2.ndim) + tuple(x2.shape)
+    if x1_shape[axis] != x2_shape[axis]:
+        raise ValueError("x1 and x2 must have the same size along the given axis")
+
+    x1_, x2_ = np.broadcast_arrays(x1._array, x2._array)
+    x1_ = np.moveaxis(x1_, axis, -1)
+    x2_ = np.moveaxis(x2_, axis, -1)
+
+    res = x1_[..., None, :] @ x2_[..., None]
+    return Array._new(res[..., 0, 0])
+
+
+# Note: the name here is different from norm(). The array API norm is split
+# into matrix_norm and vector_norm().
+
+# The type for ord should be Optional[Union[int, float, Literal[np.inf,
+# -np.inf]]] but Literal does not support floating-point literals.
+def vector_norm(x: Array, /, *, axis: Optional[Union[int, Tuple[int, ...]]] = None, keepdims: bool = False, ord: Optional[Union[int, float]] = 2) -> Array:
+    """
+    Array API compatible wrapper for :py:func:`np.linalg.norm <numpy.linalg.norm>`.
+
+    See its docstring for more information.
+    """
+    # Note: the restriction to floating-point dtypes only is different from
+    # np.linalg.norm.
+    if x.dtype not in _floating_dtypes:
+        raise TypeError('Only floating-point dtypes are allowed in norm')
+
+    # np.linalg.norm tries to do a matrix norm whenever axis is a 2-tuple or
+    # when axis=None and the input is 2-D, so to force a vector norm, we make
+    # it so the input is 1-D (for axis=None), or reshape so that norm is done
+    # on a single dimension.
+    a = x._array
+    if axis is None:
+        # Note: np.linalg.norm() doesn't handle 0-D arrays
+        a = a.ravel()
+        _axis = 0
+    elif isinstance(axis, tuple):
+        # Note: The axis argument supports any number of axes, whereas
+        # np.linalg.norm() only supports a single axis for vector norm.
+        normalized_axis = normalize_axis_tuple(axis, x.ndim)
+        rest = tuple(i for i in range(a.ndim) if i not in normalized_axis)
+        newshape = axis + rest
+        a = np.transpose(a, newshape).reshape(
+            (np.prod([a.shape[i] for i in axis], dtype=int), *[a.shape[i] for i in rest]))
+        _axis = 0
+    else:
+        _axis = axis
+
+    res = Array._new(np.linalg.norm(a, axis=_axis, ord=ord))
+
+    if keepdims:
+        # We can't reuse np.linalg.norm(keepdims) because of the reshape hacks
+        # above to avoid matrix norm logic.
+        shape = list(x.shape)
+        _axis = normalize_axis_tuple(range(x.ndim) if axis is None else axis, x.ndim)
+        for i in _axis:
+            shape[i] = 1
+        res = reshape(res, tuple(shape))
+
+    return res
+
+__all__ = ['cholesky', 'cross', 'det', 'diagonal', 'eigh', 'eigvalsh', 'inv', 'matmul', 'matrix_norm', 'matrix_power', 'matrix_rank', 'matrix_transpose', 'outer', 'pinv', 'qr', 'slogdet', 'solve', 'svd', 'svdvals', 'tensordot', 'trace', 'vecdot', 'vector_norm']

.venv/lib/python3.11/site-packages/numpy/array_api/setup.py

@@ -0,0 +1,12 @@
+def configuration(parent_package="", top_path=None):
+    from numpy.distutils.misc_util import Configuration
+
+    config = Configuration("array_api", parent_package, top_path)
+    config.add_subpackage("tests")
+    return config
+
+
+if __name__ == "__main__":
+    from numpy.distutils.core import setup
+
+    setup(configuration=configuration)

.venv/lib/python3.11/site-packages/numpy/array_api/tests/__init__.py

@@ -0,0 +1,7 @@
+"""
+Tests for the array API namespace.
+
+Note, full compliance with the array API can be tested with the official array API test
+suite https://github.com/data-apis/array-api-tests. This test suite primarily
+focuses on those things that are not tested by the official test suite.
+"""