Add implementation of dpnp.linalg.svdvals() (#2094)

* Add implementation of dpnp.linalg.svdvals

* Add TestSvdvals to test_linalg.py

* Add dpnp.linalg.svdvals to docs

* Update doc/reference/linalg.rst

---------

Co-authored-by: Anton <[email protected]>

Author: vlad-perevezentsev

doc/reference/linalg.rst

@@ -30,8 +30,10 @@ Decompositions
    :nosignatures:
 
    dpnp.linalg.cholesky
+   dpnp.linalg.outer
    dpnp.linalg.qr
    dpnp.linalg.svd
+   dpnp.linalg.svdvals
 
 Matrix eigenvalues
 ------------------

dpnp/linalg/dpnp_iface_linalg.py

@@ -83,6 +83,7 @@
     "qr",
     "solve",
     "svd",
+    "svdvals",
     "slogdet",
     "tensorinv",
     "tensorsolve",
@@ -1315,6 +1316,62 @@ def svd(a, full_matrices=True, compute_uv=True, hermitian=False):
     return dpnp_svd(a, full_matrices, compute_uv, hermitian)
 
 
+def svdvals(x, /):
+    """
+    Returns the singular values of a matrix (or a stack of matrices) `x`.
+
+    When `x` is a stack of matrices, the function will compute
+    the singular values for each matrix in the stack.
+
+    Calling ``dpnp.linalg.svdvals(x)`` to get singular values is the same as
+    ``dpnp.linalg.svd(x, compute_uv=False, hermitian=False)``.
+
+    For full documentation refer to :obj:`numpy.linalg.svdvals`.
+
+    Parameters
+    ----------
+    x : (..., M, N) {dpnp.ndarray, usm_ndarray}
+        Input array with ``x.ndim >= 2`` and whose last two dimensions
+        form matrices on which to perform singular value decomposition.
+
+    Returns
+    -------
+    out : (..., K) dpnp.ndarray
+        Vector(s) of singular values of length K, where K = min(M, N).
+
+
+    See Also
+    --------
+    :obj:`dpnp.linalg.svd` : Compute the singular value decomposition.
+
+
+    Examples
+    --------
+    >>> import dpnp as np
+    >>> a = np.array([[3, 0], [0, 4]])
+    >>> np.linalg.svdvals(a)
+    array([4., 3.])
+
+    This is equivalent to calling:
+
+    >>> np.linalg.svd(a, compute_uv=False, hermitian=False)
+    array([4., 3.])
+
+    Stack of matrices:
+
+    >>> b = np.array([[[6, 0], [0, 8]], [[9, 0], [0, 12]]])
+    >>> np.linalg.svdvals(b)
+    array([[ 8.,  6.],
+           [12.,  9.]])
+
+    """
+
+    dpnp.check_supported_arrays_type(x)
+    assert_stacked_2d(x)
+
+    return dpnp_svd(x, full_matrices=True, compute_uv=False, hermitian=False)
+
+
 def slogdet(a):
     """
     Compute the sign and (natural) logarithm of the determinant of an array.

tests/test_linalg.py

@@ -2993,6 +2993,50 @@ def test_svd_errors(self):
         assert_raises(inp.linalg.LinAlgError, inp.linalg.svd, a_dp_ndim_1)
 
 
+# numpy.linalg.svdvals() is available since numpy >= 2.0
+@testing.with_requires("numpy>=2.0")
+class TestSvdvals:
+    @pytest.mark.parametrize("dtype", get_all_dtypes(no_bool=True))
+    @pytest.mark.parametrize(
+        "shape",
+        [(3, 5), (4, 2), (2, 3, 3), (3, 5, 2)],
+        ids=["(3,5)", "(4,2)", "(2,3,3)", "(3,5,2)"],
+    )
+    def test_svdvals(self, dtype, shape):
+        a = numpy.arange(numpy.prod(shape), dtype=dtype).reshape(shape)
+        dp_a = inp.array(a)
+
+        expected = numpy.linalg.svdvals(a)
+        result = inp.linalg.svdvals(dp_a)
+
+        assert_dtype_allclose(result, expected)
+
+    @pytest.mark.parametrize(
+        "shape",
+        [(0, 0), (1, 0, 0), (0, 2, 2)],
+        ids=["(0,0)", "(1,0,0)", "(0,2,2)"],
+    )
+    def test_svdvals_empty(self, shape):
+        a = generate_random_numpy_array(shape, inp.default_float_type())
+        dp_a = inp.array(a)
+
+        expected = numpy.linalg.svdvals(a)
+        result = inp.linalg.svdvals(dp_a)
+
+        assert_dtype_allclose(result, expected)
+
+    def test_svdvals_errors(self):
+        a_dp = inp.array([[1, 2], [3, 4]], dtype="float32")
+
+        # unsupported type
+        a_np = inp.asnumpy(a_dp)
+        assert_raises(TypeError, inp.linalg.svdvals, a_np)
+
+        # a.ndim < 2
+        a_dp_ndim_1 = a_dp.flatten()
+        assert_raises(inp.linalg.LinAlgError, inp.linalg.svdvals, a_dp_ndim_1)
+
+
 class TestPinv:
     def get_tol(self, dtype):
         tol = 1e-06