Driver handling in svdvals function in torch_frontend (#23718)

Co-authored-by: ivy-branch <ivy.branch@lets-unify.ai>
Co-authored-by: juliagsy <67888047+juliagsy@users.noreply.github.com>

ivy/functional/backends/jax/linear_algebra.py

@@ -371,7 +371,10 @@ def svd(
     {"0.4.19 and below": ("bfloat16", "float16", "complex")},
     backend_version,
 )
-def svdvals(x: JaxArray, /, *, out: Optional[JaxArray] = None) -> JaxArray:
+def svdvals(
+    x: JaxArray, /, *, driver: Optional[str] = None, out: Optional[JaxArray] = None
+) -> JaxArray:
+    # TODO: handling the driver argument
     return jnp.linalg.svd(x, compute_uv=False)
 
 

ivy/functional/backends/mxnet/linear_algebra.py

@@ -228,8 +228,10 @@ def svdvals(
     x: Union[(None, mx.ndarray.NDArray)],
     /,
     *,
+    driver: Optional[str] = None,
     out: Optional[Union[(None, mx.ndarray.NDArray)]] = None,
 ) -> Union[(None, mx.ndarray.NDArray)]:
+    # TODO: handling the driver argument
     raise IvyNotImplementedException()
 
 

ivy/functional/backends/numpy/linear_algebra.py

@@ -298,7 +298,10 @@ def svd(
 
 
 @with_unsupported_dtypes({"1.26.1 and below": ("float16",)}, backend_version)
-def svdvals(x: np.ndarray, /, *, out: Optional[np.ndarray] = None) -> np.ndarray:
+def svdvals(
+    x: np.ndarray, /, *, driver: Optional[str] = None, out: Optional[np.ndarray] = None
+) -> np.ndarray:
+    # TODO: handling the driver argument
     return np.linalg.svd(x, compute_uv=False)
 
 

ivy/functional/backends/paddle/linear_algebra.py

@@ -521,8 +521,13 @@ def svd(
     backend_version,
 )
 def svdvals(
-    x: paddle.Tensor, /, *, out: Optional[paddle.Tensor] = None
+    x: paddle.Tensor,
+    /,
+    *,
+    driver: Optional[str] = None,
+    out: Optional[paddle.Tensor] = None,
 ) -> paddle.Tensor:
+    # TODO:handling the driver argument
     return paddle_backend.svd(x)[1]
 
 

ivy/functional/backends/tensorflow/linear_algebra.py

@@ -564,8 +564,10 @@ def svdvals(
     x: Union[tf.Tensor, tf.Variable],
     /,
     *,
+    driver: Optional[str] = None,
     out: Optional[Union[tf.Tensor, tf.Variable]] = None,
 ) -> Union[tf.Tensor, tf.Variable]:
+    # TODO: handling the driver argument
     ret = tf.linalg.svd(x, compute_uv=False)
     return ret
 

ivy/functional/backends/torch/linear_algebra.py

@@ -414,9 +414,15 @@ def svd(
         return results(D)
 
 
-@with_unsupported_dtypes({"2.1.0 and below": ("float16", "bfloat16")}, backend_version)
-def svdvals(x: torch.Tensor, /, *, out: Optional[torch.Tensor] = None) -> torch.Tensor:
-    return torch.linalg.svdvals(x, out=out)
+@with_unsupported_dtypes({"2.0.1 and below": ("float16", "bfloat16")}, backend_version)
+def svdvals(
+    x: torch.Tensor,
+    /,
+    *,
+    driver: Optional[str] = None,
+    out: Optional[torch.Tensor] = None,
+) -> torch.Tensor:
+    return torch.linalg.svdvals(x, driver=driver, out=out)
 
 
 svdvals.support_native_out = True

ivy/functional/frontends/torch/linalg.py

@@ -314,8 +314,10 @@ def svd(A, /, *, full_matrices=True, driver=None, out=None):
     {"2.1.0 and below": ("float32", "float64", "complex32", "complex64")}, "torch"
 )
 def svdvals(A, *, driver=None, out=None):
-    # TODO: add handling for driver
-    return ivy.svdvals(A, out=out)
+    if driver in ["gesvd", "gesvdj", "gesvda", None]:
+        return ivy.svdvals(A, driver=driver, out=out)
+    else:
+        raise ValueError("Unsupported SVD driver")
 
 
 @to_ivy_arrays_and_back

ivy/functional/ivy/linear_algebra.py

@@ -2264,7 +2264,11 @@ def svd(
 @handle_array_function
 @handle_device
 def svdvals(
-    x: Union[ivy.Array, ivy.NativeArray], /, *, out: Optional[ivy.Array] = None
+    x: Union[ivy.Array, ivy.NativeArray],
+    /,
+    *,
+    driver: Optional[str] = None,
+    out: Optional[ivy.Array] = None,
 ) -> ivy.Array:
     """
     Return the singular values of a matrix (or a stack of matrices) ``x``.
@@ -2274,6 +2278,9 @@ def svdvals(
     x
         input array having shape ``(..., M, N)`` and whose innermost two dimensions form
         ``MxN`` matrices.
+    driver
+        optional output array,name of the cuSOLVER method to be used. This keyword argument only works on CUDA inputs.
+        Available options are: None, gesvd, gesvdj, and gesvda.Default: None.
     out
         optional output array, for writing the result to. It must have a shape that the
         inputs broadcast to.
@@ -2387,7 +2394,7 @@ def svdvals(
         b: ivy.array([23.16134834, 10.35037804, 4.31025076, 1.35769391])
     }
     """
-    return current_backend(x).svdvals(x, out=out)
+    return current_backend(x).svdvals(x, driver=driver, out=out)
 
 
 @handle_exceptions

ivy_tests/test_ivy/test_frontends/test_torch/test_linalg.py

@@ -1156,10 +1156,12 @@ def test_torch_svd(
 @handle_frontend_test(
     fn_tree="torch.linalg.svdvals",
     dtype_and_x=_get_dtype_and_matrix(batch=True),
+    driver=st.sampled_from([None, "gesvd", "gesvdj", "gesvda"]),
 )
 def test_torch_svdvals(
     *,
     dtype_and_x,
+    driver,
     on_device,
     fn_tree,
     frontend,
@@ -1174,6 +1176,7 @@ def test_torch_svdvals(
         test_flags=test_flags,
         fn_tree=fn_tree,
         on_device=on_device,
+        driver=driver,
         A=x[0],
     )