RSVP for your your local TensorFlow Everywhere event today!

tfx.dsl.component.experimental.function_parser.parse_typehint_component_function

Parses the given component executor function.

tfx.dsl.component.experimental.function_parser.parse_typehint_component_function(
    func: types.FunctionType
) -> Tuple[Dict[Text, Type[artifact.Artifact]], Dict[Text, Type[artifact.
    Artifact]], Dict[Text, Type[Union[int, float, Text, bytes]]], Dict[Text,
    Any], Dict[Text, ArgFormats], Set[Text]]

This method parses a typehinted-annotated Python function that is intended to be used as a component and returns the information needed about the interface (inputs / outputs / returned output values) about that components, as well as a list of argument names and formats for determining the parameters that should be passed when calling func(*args).

func A component executor function to be parsed.

inputs A dictionary mapping each input name to its artifact type (as a subclass of tfx.types.Artifact).
outputs A dictionary mapping each output name to its artifact type (as a subclass of tfx.types.Artifact).
parameters A dictionary mapping each parameter name to its primitive type (one of int, float, Text and bytes).
arg_formats Dictionary representing the input arguments of the given component executor function. Each entry's key is the argument's string name; each entry's value is the format of the argument to be passed into the function (given by a value of the ArgFormats enum).
arg_defaults Dictionary mapping names of optional arguments to default values.
returned_outputs A set of output names that are declared as ValueArtifact returned outputs.