get_path_to_best_clika_state_result

get_path_to_best_clika_state_result(

• output_dir: str
• key_name: Optional[str] =  None
• summary_json_group: Optional[Literal['training', 'evaluation']] =  None
• comparator_fn: Optional[Callable[[dict, dict], int]] =  None
• find_lowest: Optional[bool] =  None
• return_summary: bool =  False
• ignore_warnings: bool =  False

) ‑> Union[str, None, Tuple[Optional[str], Optional[dict]]]

Finds the best epoch results (.pompom file) in an output directory.

Parameters:

• output_dir (str) - path to output directory of the clika-compression training.
• key_name (str) - the name of the loss or metric to search for in the summary files.
• summary_json_group (str) - either 'training' or 'evaluation' - the group of the summary json file.
• comparator_fn (Callable[[dict, dict], bool]) - a function that compares two values - the first is the previous best and the second is the current file to compare. returns an int, -1 means that the first input (the best epoch so far) was preferable than second input (current epoch). 0 means the two objects are equal. 1 means that the second input (current epoch) was preferable than the first input (the best epoch so far).
• return_summary (bool) - if True will return the summary file as dict in addition to the path to the summary file.
• find_lowest (bool) - if True will return the lowest value (for losses), if False will return the highest value (usually for metrics).
• ignore_warnings (bool) - if True will ignore warnings.

Returns:

• str: path to the best epoch .pompom file.
• dict: the summary file as dict (only returned if return_summary is True).