Feat (QuantTensor): QuantTensor x Tensor elementary ops dequantize to…

… Tensor (#668)

src/brevitas/quant_tensor/__init__.py

@@ -295,7 +295,7 @@ def cat(tensors, dim, out=None):
             else:
                 tensors = [qt.value if isinstance(qt, QuantTensor) else qt for qt in tensors]
                 output_value = torch.cat(tensors, dim=dim)
-                return QuantTensor(output_value)
+                return output_value
 
     # Reference: https://docs.python.org/3/reference/datamodel.html#emulating-numeric-types
 
@@ -366,9 +366,9 @@ def __add__(self, other):
                 signed=output_signed,
                 training=output_training)
         elif isinstance(other, QuantTensor):
-            output = QuantTensor(self.value + other.value)
+            output = self.value + other.value
         else:
-            output = QuantTensor(self.value + other)
+            output = self.value + other
         return output
 
     def __radd__(self, other):
@@ -396,9 +396,9 @@ def __mul__(self, other):
                 signed=output_signed,
                 training=output_training)
         elif isinstance(other, QuantTensor):
-            output = QuantTensor(self.value * other.value)
+            output = self.value * other.value
         else:
-            output = QuantTensor(self.value * other)
+            output = self.value * other
         return output
 
     def __sub__(self, other):
@@ -423,9 +423,9 @@ def __truediv__(self, other):
                 signed=output_signed,
                 training=output_training)
         elif isinstance(other, QuantTensor):
-            output = QuantTensor(self.value / other.value)
+            output = self.value / other.value
         else:
-            output = QuantTensor(self.value / other)
+            output = self.value / other
         return output
 
     def __abs__(self):