Function config_for_class

axlearn/common/config.py:1144–1179

Returns an instance of ClassConfigBase, which is an object factory for `cls`. In other words, instantiating the config produces an instance of `cls`, where the configured attributes will be provided as arguments to `__init__`. Example: ``` class MyClass: def

(cls: type[T])

def config_for_class(cls: type[T]) -> Union[Any, ClassConfigBase[T]]:
    """Returns an instance of ClassConfigBase, which is an object factory for `cls`.

    In other words, instantiating the config produces an instance of `cls`, where the configured
    attributes will be provided as arguments to `__init__`.

    Example:
        ```
        class MyClass:
            def __init__(self, a: int, b: Optional[int] = None):
                self.a = a
                self.b = b

            def values(self):
                return (self.a, self.b)

        cfg = config_for_class(MyClass).set(a=2)

        # Should produce unique instances.
        assert cfg.instantiate() is not cfg.instantiate()

        # Should produce the correct values.
        assert cfg.instantiate().values() == cfg.instantiate().values()
        assert cfg.instantiate().values() == (2, None)
        assert cfg.set(b=3).instantiate().values() == (2, 3)
        ```

    Args:
        cls: The class to configure.

    Returns:
        A Config that when instantiated, invokes `cls.__init__` based on any config fields that have
        been set.
    """
    config_cls = _config_class_for_class(cls)
    return config_cls(klass=cls)


def maybe_set_config(cfg: _ConfigBase, **kwargs) -> _ConfigBase: