package.experimental.train.losses.registry.LossRegistry Class Reference

Static Public Member Functions
LossFunction	create_loss (str name)
loss_type	map_loss_function_to_enum (LossFunction loss_instance)

Static Protected Attributes
dict	_losses

Detailed Description

Registry for creating and mapping losses by name or instance.

Definition at line 8 of file registry.py.

Member Function Documentation

create_loss()

LossFunction package.experimental.train.losses.registry.LossRegistry.create_loss ( str  name )

static

Create a loss instance by name.
Args:
    name (str): Name of the loss.
Returns:
    BaseLoss: Loss instance.

Definition at line 18 of file registry.py.

    def create_loss(name: str) -> LossFunction:
        """
        Create a loss instance by name.
        Args:
            name (str): Name of the loss.
        Returns:
            BaseLoss: Loss instance.
        """
        if name not in LossRegistry._losses:
            raise ValueError(f"Unknown Loss: {name}. Custom loss is not supported yet")
        return LossRegistry._losses[name]()
 

map_loss_function_to_enum()

loss_type package.experimental.train.losses.registry.LossRegistry.map_loss_function_to_enum ( LossFunction  loss_instance )

static

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.

Definition at line 31 of file registry.py.

    def map_loss_function_to_enum(loss_instance: LossFunction) -> loss_type:
        """
        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: Dict[Type[LossFunction], loss_type] = {
            CategoricalCrossentropy: loss_type.CATEGORICAL_CROSSENTROPY,
            MeanSquaredError: loss_type.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())}.")

Field Documentation

_losses

dict package.experimental.train.losses.registry.LossRegistry._losses

staticprotected

Initial value:

=  {
        "categorical_crossentropy": CategoricalCrossentropy,
        "mean_squared_error": MeanSquaredError
    }

Definition at line 12 of file registry.py.

---

The documentation for this class was generated from the following file:

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