Merge branch 'unifyai:master' into master

ivy/functional/backends/paddle/creation.py

@@ -387,7 +387,7 @@ def _slice_at_axis(sl, axis):
 
 
 @with_unsupported_device_and_dtypes(
-    {"2.4.2 and below": {"cpu": ("uint16", "bfloat16")}}, backend_version
+    {"2.4.2 and below": {"cpu": ("uint16", "bfloat16", "float16")}}, backend_version
 )
 def linspace(
     start: Union[paddle.Tensor, float],

ivy/functional/backends/paddle/elementwise.py

@@ -922,6 +922,10 @@ def exp2(
         return ivy.pow(2, x)
 
 
+@with_supported_dtypes(
+    {"2.4.2 and below": ("float64", "float32", "int64", "int64")},
+    backend_version,
+)
 def subtract(
     x1: Union[float, paddle.Tensor],
     x2: Union[float, paddle.Tensor],
@@ -931,8 +935,6 @@ def subtract(
     out: Optional[paddle.Tensor] = None,
 ) -> paddle.Tensor:
     x1, x2, ret_dtype = _elementwise_helper(x1, x2)
-    if x1.dtype in [paddle.int8, paddle.uint8, paddle.float16, paddle.bool]:
-        x1, x2 = x1.astype("float32"), x2.astype("float32")
     if alpha not in (1, None):
         x2 = paddle_backend.multiply(x2, alpha)
         x1, x2 = ivy.promote_types_of_inputs(x1, x2)

ivy/functional/backends/tensorflow/experimental/layers.py

@@ -549,6 +549,7 @@ def fft(
         )
     if norm != "backward" and norm != "ortho" and norm != "forward":
         raise ivy.utils.exceptions.IvyError(f"Unrecognized normalization mode {norm}")
+    x = tf.cast(x, tf.complex128)
     if x.shape[dim] != n:
         s = list(x.shape)
         if s[dim] > n:

ivy/functional/frontends/jax/numpy/creation.py

@@ -6,6 +6,7 @@
     to_ivy_arrays_and_back,
     outputs_to_frontend_arrays,
     handle_jax_dtype,
+    inputs_to_ivy_arrays
 )
 
 from ivy.func_wrapper import handle_out_argument
@@ -243,3 +244,9 @@ def compress(condition, a, *, axis=None, out=None):
         )
     arr = arr[: condition_arr.shape[0]]
     return ivy.moveaxis(arr[condition_arr], 0, axis)
+
+
+@inputs_to_ivy_arrays
+def iterable(y):
+    return hasattr(y, "__iter__") and y.ndim > 0
+

ivy/functional/frontends/numpy/fft/discrete_fourier_transform.py

@@ -75,7 +75,7 @@ def fftshift(x, axes=None):
     return roll
 
 
-@with_unsupported_dtypes({"1.9.0 and below": ("float16",)}, "torch")
+@with_unsupported_dtypes({"1.24.3 and below": ("float16",)}, "numpy")
 @to_ivy_arrays_and_back
 def rfft(a, n=None, axis=-1, norm=None):
     if norm is None:
@@ -84,18 +84,17 @@ def rfft(a, n=None, axis=-1, norm=None):
     return ivy.dft(a, axis=axis, inverse=False, onesided=True, dft_length=n, norm=norm)
 
 
+@with_unsupported_dtypes({"1.24.3 and below": ("float16",)}, "numpy")
 @to_ivy_arrays_and_back
-@with_unsupported_dtypes({"1.12.0 and below": ("float16",)}, "numpy")
 def ihfft(a, n=None, axis=-1, norm=None):
-    a = ivy.array(a, dtype=ivy.float64)
     if n is None:
         n = a.shape[axis]
     norm = _swap_direction(norm)
     output = ivy.conj(rfft(a, n, axis, norm=norm).ivy_array)
     return output
 
 
-@with_unsupported_dtypes({"2.4.2 and below": ("int",)}, "paddle")
+@with_unsupported_dtypes({"1.24.3 and below": ("int",)}, "numpy")
 @to_ivy_arrays_and_back
 def fftfreq(n, d=1.0):
     if not isinstance(

ivy/functional/frontends/torch/tensor.py

@@ -920,7 +920,7 @@ def is_complex(self):
 
     def addr(self, vec1, vec2, *, beta=1, alpha=1, out=None):
         return torch_frontend.addr(self, vec1, vec2, beta=beta, alpha=alpha, out=out)
-    
+
     def addr_(self, vec1, vec2, *, beta=1, alpha=1):
         self.ivy_array = self.addr(vec1, vec2, beta=beta, alpha=alpha).ivy_array
         return self

ivy/functional/ivy/data_type.py

@@ -1426,7 +1426,7 @@ def default_complex_dtype(
     input
         Number or array for inferring the complex dtype.
     complex_dtype
-        The float dtype to be returned.
+        The complex dtype to be returned.
     as_native
         Whether to return the complex dtype as native dtype.
 

ivy/functional/ivy/experimental/layers.py

@@ -1006,7 +1006,7 @@ def embedding(
 @handle_exceptions
 @handle_nestable
 @handle_out_argument
-@to_native_arrays_and_back
+@inputs_to_ivy_arrays
 def dft(
     x: Union[ivy.Array, ivy.NativeArray],
     /,
@@ -1070,9 +1070,9 @@ def dft(
         The signal_dim at the specified axis is equal to the dft_length.
     """
     if inverse:
-        res = ifft(x, axis, norm=norm, n=dft_length, out=out)
+        res = ivy.ifft(x, axis, norm=norm, n=dft_length, out=out)
     else:
-        res = fft(x, axis, norm=norm, n=dft_length, out=out)
+        res = ivy.fft(x, axis, norm=norm, n=dft_length, out=out)
 
     if onesided:
         slices = [slice(0, a) for a in res.shape]

ivy_tests/test_ivy/test_frontends/test_jax/test_jax_numpy_creation.py

@@ -923,3 +923,28 @@ def test_jax_numpy_compress(
         a=arr[0],
         axis=ax,
     )
+
+
+@handle_frontend_test(
+    fn_tree="jax.numpy.iterable",
+    dtype_and_x=helpers.dtype_and_values(
+        available_dtypes=helpers.get_dtypes("valid"),
+    ),
+)
+def test_jax_numpy_iterable(
+    dtype_and_x,
+    test_flags,
+    frontend,
+    fn_tree,
+    on_device,
+):
+    dtype, x = dtype_and_x
+    helpers.test_frontend_function(
+        input_dtypes=dtype,
+        frontend=frontend,
+        test_flags=test_flags,
+        fn_tree=fn_tree,
+        on_device=on_device,
+        y=x[0],
+    )
+

ivy_tests/test_ivy/test_frontends/test_numpy/test_fft/test_discrete_fourier_transform.py

@@ -1,6 +1,9 @@
+# global
+from hypothesis import strategies as st
+
+# local
 import ivy_tests.test_ivy.helpers as helpers
 from ivy_tests.test_ivy.helpers import handle_frontend_test
-from hypothesis import strategies as st
 
 
 @handle_frontend_test(

ivy_tests/test_ivy/test_frontends/test_numpy/test_mathematical_functions/test_arithmetic_operations.py

@@ -172,6 +172,8 @@ def test_numpy_divide(
         test_flags=test_flags,
         fn_tree=fn_tree,
         on_device=on_device,
+        atol=1e-03,
+        rtol=1e-03,
         x1=xs[0],
         x2=xs[1],
         out=None,

ivy_tests/test_ivy/test_frontends/test_numpy/test_mathematical_functions/test_trigonometric_functions.py

@@ -39,6 +39,7 @@ def test_numpy_cos(
         test_flags=test_flags,
         fn_tree=fn_tree,
         on_device=on_device,
+        atol=1e-03,
         x=x[0],
         out=None,
         where=where,

ivy_tests/test_ivy/test_frontends/test_torch/test_tensor.py

@@ -8823,4 +8823,4 @@ def test_torch_instance_addr_(
         frontend=frontend,
         atol_=1e-02,
         on_device=on_device,
-    )
+    )

requirements/requirements.txt

@@ -8,3 +8,4 @@ nvidia-ml-py<=11.495.46 # mod_name=pynvml
 diskcache
 google-auth # mod_name=google.auth
 requests
+pyvis