kernelfoundry.algorithm.utils.load_from_db

Functions

load_best_kernel(problem_name, language[, ...])	Load best kernel generated so far from database
load_database(filter_by[, task_origin, ...])	Read database, filter with filter_by dict, and return as DataFrame

kernelfoundry.algorithm.utils.load_from_db.load_database(filter_by: dict, task_origin: str | None = None, max_improvement: float | None = None, min_improvement: float | None = None, job_options=None, limit: int | None = None)

Read database, filter with filter_by dict, and return as DataFrame

kernelfoundry.algorithm.utils.load_from_db.load_best_kernel(problem_name: str, language: str, sortby: str = 'improve_over_native', job_name: str | None = None) → Tuple[str, str]

Load best kernel generated so far from database

Parameters:

• problem_name (str) – the name of the task

• language (str) – output language

• sortby (str, optional) – which field to use to determine best kernel. Defaults to “improve_over_native”.

Returns:

code of best kernel, prior eval log

Return type:

Tuple[str, str]