Python API for Complex Storage and storage copy logic (pytorch#35771)

Summary:
Following up on this: pytorch#35851 cross dtype storage copy is not being used internally, so I have not included cross dtype copy for complex.
Pull Request resolved: pytorch#35771

Differential Revision: D21319650

Pulled By: anjali411

fbshipit-source-id: 07c72996ee598eba0cf401ad61534494d6f5b5b3

Author: anjali411

aten/src/ATen/DLConvertor.cpp

@@ -53,10 +53,13 @@ DLDataType getDLDataType(const Tensor& t) {
       break;
     case ScalarType::ComplexHalf:
       throw std::logic_error("ComplexHalf is not supported by dlpack");
+      break;
     case ScalarType::ComplexFloat:
       throw std::logic_error("ComplexFloat is not supported by dlpack");
+      break;
     case ScalarType::ComplexDouble:
       throw std::logic_error("ComplexDouble is not supported by dlpack");
+      break;
     case ScalarType::Undefined:
       throw std::logic_error("Undefined is not a valid ScalarType");
     case ScalarType::NumOptions:

aten/src/ATen/function_wrapper.py

@@ -240,6 +240,8 @@ def TypedDict(name, attrs, total=True):  # type: ignore
     ('Short', 'int16_t', 'Long', False),
     ('Half', 'Half', 'Double', True),
     ('BFloat16', 'BFloat16', 'BFloat16AccrealNotDefined', True),
+    ('ComplexFloat', 'ComplexFloat', 'ComplexDouble', False),
+    ('ComplexDouble', 'ComplexDouble', 'ComplexDouble', False),
 ]
 
 static_dispatch_backends = ['CPU', 'QuantizedCPU']