feat: add torch.special.erfc in torch frontend and turn the implement…

…ation in pointwise ops into an alias for the special func

erfissues.py

@@ -0,0 +1,5 @@
+import torch
+print(torch.special.erfc(torch.tensor([0, -1., 10.])))
+from ivy.functional.frontends import torch as ivy_torch
+print(ivy_torch.special.erfc(ivy_torch.tensor([0, -1., 10.])))
+print(ivy_torch.erfc(ivy_torch.tensor([0, -1., 10.])))

ivy/functional/frontends/torch/__init__.py

@@ -261,6 +261,7 @@ def promote_types_of_torch_inputs(
 
 from . import nn
 from .nn.functional import softmax, relu, lstm
+from . import special
 from . import tensor
 from .tensor import *
 from . import blas_and_lapack_ops

ivy/functional/frontends/torch/pointwise_ops.py

@@ -10,6 +10,9 @@
 )
 
 
+erfc = torch_frontend.special.erfc
+
+
 @to_ivy_arrays_and_back
 def abs(input, *, out=None):
     return ivy.abs(input, out=out)
@@ -189,12 +192,6 @@ def erf(input, *, out=None):
     return ivy.erf(input, out=out)
 
 
-@with_unsupported_dtypes({"2.2 and below": ("float16", "complex")}, "torch")
-@to_ivy_arrays_and_back
-def erfc(input, *, out=None):
-    return 1.0 - ivy.erf(input, out=out)
-
-
 @with_unsupported_dtypes({"2.2 and below": ("float16",)}, "torch")
 @to_ivy_arrays_and_back
 def exp(input, *, out=None):

ivy/functional/frontends/torch/special/__init__.py

@@ -0,0 +1 @@
+from .special_funcs import *

ivy/functional/frontends/torch/special/special_funcs.py

@@ -0,0 +1,13 @@
+import ivy
+from ivy.func_wrapper import (
+    with_unsupported_dtypes,
+)
+from ivy.functional.frontends.torch.func_wrapper import (
+    to_ivy_arrays_and_back,
+)
+
+
+@with_unsupported_dtypes({"2.2 and below": ("float16", "complex")}, "torch")
+@to_ivy_arrays_and_back
+def erfc(input, *, out=None):
+    return 1.0 - ivy.erf(input, out=out)