fix: Fixing pytest discovery issues

ivy/functional/frontends/__init__.py

@@ -11,6 +11,7 @@
     "sklearn": "1.3.0",
     "xgboost": "1.7.6",
     "torchvision": "0.15.2.",
+    "mindspore": "2.0.0",
 }
 
 

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

@@ -1,7 +1,7 @@
 import ivy
 from ivy.func_wrapper import with_unsupported_dtypes
 from ivy.functional.frontends.jax.array import Array
-import ivy.functional.frontends.jax.numpy as jnp_frontend
+import ivy.functional.frontends.jax.numpy as jnp
 from ivy.functional.frontends.jax.func_wrapper import (
     to_ivy_arrays_and_back,
     outputs_to_frontend_arrays,
@@ -270,7 +270,7 @@ def setdiff1d(ar1, ar2, assume_unique=False, *, size=None, fill_value=None):
             if ivy.is_bool_dtype(ar1.dtype)
             else ivy.to_scalar(ivy.min(ar1))
         )
-        ar1 = jnp_frontend.unique(ar1, size=size and ar1.size, fill_value=val).ivy_array
+        ar1 = jnp.unique(ar1, size=size and ar1.size, fill_value=val).ivy_array
     mask = in1d(ar1, ar2, invert=True).ivy_array
     if size is None:
         return ar1[mask]
@@ -281,7 +281,7 @@ def setdiff1d(ar1, ar2, assume_unique=False, *, size=None, fill_value=None):
             mask = ivy.where(ivy.arange(ar1.size) < n_unique, mask, False)
         return ivy.where(
             ivy.arange(size) < mask.sum(dtype=ivy.int64),
-            ar1[jnp_frontend.where(mask, size=size)[0].ivy_array],
+            ar1[jnp.where(mask, size=size)[0].ivy_array],
             fill_value,
         )
 

ivy/functional/frontends/tensorflow/keras/activations.py

@@ -80,6 +80,14 @@ def get(identifier):
         raise ValueError(f"Could not interpret function identifier: {identifier}")
 
 
+# helper
+# note: defined to avoid AST call extraction of
+# 'tf_frontend.keras.activations.__dict__.items()
+# or 'tf_frontend.keras.activations.__dict__.values()'
+def get_tf_keras_activations():
+    return tf_frontend.keras.activations.__dict__.items()
+
+
 @to_ivy_arrays_and_back
 def hard_sigmoid(x):
     dtype_in = x.dtype
@@ -132,13 +140,15 @@ def serialize(activation, use_legacy_format=False, custom_objects=None):
                 if custom_func == activation:
                     return name
 
+        tf_keras_frontend_activations = get_tf_keras_activations()
+
         # Check if the function is in the ACTIVATION_FUNCTIONS list
         if activation.__name__ in ACTIVATION_FUNCTIONS:
             return activation.__name__
 
         # Check if the function is in the TensorFlow frontend activations
-        elif activation in tf_frontend.keras.activations.__dict__.values():
-            for name, tf_func in tf_frontend.keras.activations.__dict__.items():
+        elif activation in [fn for name, fn in tf_keras_frontend_activations]:
+            for name, tf_func in get_tf_keras_activations():
                 if tf_func == activation:
                     return name
 

ivy/functional/ivy/data_type.py

@@ -153,7 +153,7 @@ def _nested_get(f, base_set, merge_fn, get_fn, wrapper=set):
             continue
         is_frontend_fn = "frontend" in fn.__module__
         is_backend_fn = "backend" in fn.__module__ and not is_frontend_fn
-        is_einops_fn = "einops" in fn.__name__
+        is_einops_fn = hasattr(fn, "__name__") and "einops" in fn.__name__
         if is_backend_fn:
             f_supported = get_fn(fn, False)
             if hasattr(fn, "partial_mixed_handler"):
@@ -197,6 +197,10 @@ def _nested_get(f, base_set, merge_fn, get_fn, wrapper=set):
                     if "(" in key:
                         key = key.split("(")[0]
                     frontend_module = ".".join(key.split(".")[:-1])
+                    if (
+                        frontend_module == ""
+                    ):  # single edge case: fn='frontend_outputs_to_ivy_arrays'
+                        continue
                     frontend_fl = {key: frontend_fn}
                     res += list(
                         _get_functions_from_string(

ivy/functional/ivy/device.py

@@ -1169,7 +1169,7 @@ def _get_devices(fn: Callable, complement: bool = True) -> Tuple:
 
     is_backend_fn = "backend" in fn.__module__
     is_frontend_fn = "frontend" in fn.__module__
-    is_einops_fn = "einops" in fn.__name__
+    is_einops_fn = hasattr(fn, "__name__") and "einops" in fn.__name__
     if not is_backend_fn and not is_frontend_fn and not is_einops_fn:
         if complement:
             supported = set(all_devices).difference(supported)

ivy/functional/ivy/general.py

@@ -3989,7 +3989,7 @@ def _get_devices_and_dtypes(fn, recurse=True, complement=True):
 
     is_frontend_fn = "frontend" in fn.__module__
     is_backend_fn = "backend" in fn.__module__ and not is_frontend_fn
-    is_einops_fn = "einops" in fn.__name__
+    is_einops_fn = hasattr(fn, "__name__") and "einops" in fn.__name__
     if not is_backend_fn and not is_frontend_fn and not is_einops_fn:
         if complement:
             all_comb = _all_dnd_combinations()

ivy_tests/test_ivy/test_frontends/test_jax/test_numpy/test_indexing.py

@@ -12,7 +12,7 @@
     _helper_r_,
     _helper_c_,
 )
-import ivy.functional.frontends.jax.numpy as jnp_frontend
+import ivy.functional.frontends.jax.numpy as jnp
 
 
 # --- Helpers --- #
@@ -291,7 +291,7 @@ def test_jax_mask_indices(
 def test_jax_numpy_c_(inputs, backend_fw):
     ret_gt = c_.__getitem__(tuple(inputs))
     with BackendHandler.update_backend(backend_fw):
-        ret = jnp_frontend.c_.__getitem__(tuple(inputs))
+        ret = jnp.c_.__getitem__(tuple(inputs))
     assert np.allclose(ret.ivy_array, ret_gt)
 
 
@@ -331,7 +331,7 @@ def test_jax_numpy_r_(inputs, backend_fw):
     inputs, *_ = inputs
     ret_gt = r_.__getitem__(tuple(inputs))
     with BackendHandler.update_backend(backend_fw):
-        ret = jnp_frontend.r_.__getitem__(tuple(inputs))
+        ret = jnp.r_.__getitem__(tuple(inputs))
     assert np.allclose(ret.ivy_array, ret_gt)