If you register a class which is not of type
Recipe which converts the normal model config to train and predict graph.
Enum to categorize and place custom components correctly in the graph.
Describes a graph component which was registered with the decorator.
This decorator can be used to register classes with the recipe.
component_types- Describes the types of a component which are then used to place the component in the graph.
Trueif the component requires training.
model_from- Can be used if this component requires a pre-loaded model such as
The registered class.
Converts the default config to graphs (see interface for full docstring).
Determine configuration from auto-filled configuration file.
Keys that are provided and have a value in the file are kept. Keys that are not provided are configured automatically.
Note that this needs to be called explicitly; ie. we cannot
auto-configure automatically from importers because importers are not
allowed to access code outside of
config_file_path- The path to the configuration file.
config- Configuration in dictionary format.
training_type- Optional training type to auto-configure. By default both core and NLU will be auto-configured.
Complete a config by adding automatic configuration for the specified keys.
config- The provided configuration.
keys_to_configure- Keys to be configured automatically (e.g.
The resulting configuration including both the provided and the automatically configured keys.