run

This module implements a class to configure and run the IDyOM model.

For more concrete examples on how to configure and run the IDyOM model, please see the tutorial.

To run the IDyOM model with py2lispIDyOM, the IDyOMExperiment object is provided to configure the model. You will set all the dataset paths, model parameters here.

class py2lispIDyOM.run.IDyOMExperiment(test_dataset_path, pretrain_dataset_path=None, experiment_history_folder_path=None, experiment_logger_name=None, idyom_config=<factory>)[source]

A class to configure and run an IDyOM experiment

Parameters:

  • test_dataset_path (str) – the path to your test dataset (required)

  • pretrain_dataset_path (str) – the path to your pretrain dataset

  • experiment_history_folder_path (str) – the path to which you want to save all the result data/plots, defaults to None.

  • experiment_logger_name (str) – the name of the experiment logger for the current experiment, defaults to the current timestamp.

generate_lisp_script(write=True)[source]

Generate the LISP script for the IDyOM model configurations.

Parameters:

write (bool) – whether to write the file or not, defaults to True.

Returns:

the path to the lisp script file.

Return type:

str

run()[source]

Run the IDyOM model.

set_parameters(**kwargs)[source]

Set the IDyOM model parameters.

Parameters:

kwargs – see the API reference (the section below) for a complete list of valid parameters (keyword arguments).

Important notes

Parameters to configure the IDyOM model are almost the same as the those listed and described in the IDyOM parameters documentation, EXCEPT that users are not allowed to assign dataset-id and pretraining-ids, and output-path in py2lispIDyOM.

Instead, users need to supply the relevant dataset paths for the test dataset and pretrain dataset, and unique dataset ID will automatically then be assigned to those user-specified dataset respectively. The output file of the IDyOM model are saved to the corresponding experiment logger.

Valid parameters to configure the IDyOM model

Required parameters:

  • target_viewpoints: List[SingleViewpoint]

  • source_viewpoints: Union[Literal[‘:select’], List[Union[SingleViewpoint, Tuple[SingleViewpoint]]]]

Statistical modelling parameters:

  • models: Literal[‘:stm’, ‘:ltm’, ‘:ltm+’, ‘:both’, ‘:both+’]

  • ltmo: Literal[‘:ltmo’]

  • stmo: Literal[‘:stmo’]

  • ltmo_order_bound``stmo_order_bound: int

  • ltmo_mixtures, stmo_mixtures: bool

  • ltmo_update_exclusion, stmo_update_exclusion: bool

  • ltmo_escape, stmo_escape: Literal[‘:a’, ‘:b’, ‘:c’, ‘:d’, ‘:x’]

Training parameters:

  • k: Union[int, Literal[“:full”]], default is 10

  • resampling_indices: List[int]

Viewpoint selection parameters:

  • basis: Union[List[SingleViewpoint], Literal[‘:pitch-full’, ‘:pitch-short’, ‘:bioi’, ‘:onset’, ‘:auto’]]

  • dp: int

  • max_links: int

  • min_links: int

  • viewpoint_selection_output: str

Output parameters:

  • detail: Literal[1, 2, 3]

  • overwrite: bool

  • separator: str

Caching parameters:

  • use_resampling_set_cache: bool

  • use_ltms_cache: bool