Add mapping function

runtime/onert/api/python/package/experimental/train/losses/registry.py

@@ -4,11 +4,11 @@
 
 class LossRegistry:
     """
-    Registry for creating losses by name.
+    Registry for creating and mapping losses by name or instance.
     """
     _losses = {
         "categorical_crossentropy": CategoricalCrossentropy,
-        "mean_squred_error": MeanSquaredError
+        "mean_squared_error": MeanSquaredError
     }
 
     @staticmethod
@@ -23,3 +23,27 @@ def create_loss(name):
         if name not in LossRegistry._losses:
             raise ValueError(f"Unknown Loss: {name}. Custom loss is not supported yet")
         return LossRegistry._losses[name]()
+
+    @staticmethod
+    def map_loss_function_to_enum(loss_instance):
+        """
+        Maps a LossFunction instance to the appropriate enum value.
+        Args:
+            loss_instance (BaseLoss): An instance of a loss function.
+        Returns:
+            loss_type: Corresponding enum value for the loss function.
+        Raises:
+            TypeError: If the loss_instance is not a recognized LossFunction type.
+        """
+        # Loss to Enum mapping
+        loss_to_enum = {
+            CategoricalCrossentropy: "CATEGORICAL_CROSSENTROPY",
+            MeanSquaredError: "MEAN_SQUARED_ERROR",
+        }
+        for loss_class, enum_value in loss_to_enum.items():
+            if isinstance(loss_instance, loss_class):
+                return enum_value
+        raise TypeError(
+            f"Unsupported loss function type: {type(loss_instance).__name__}. "
+            f"Supported types are: {list(loss_to_enum.keys())}."
+        )