experiment_runner.utils

Pure helpers for the experiment runner.

Small, side-effect-free utilities shared by the runner entry point: archive wildcard expansion, metric-score aggregation, and the metrics.csv row and log-label formatting. Kept apart from experiment_runner.main so the orchestration module holds only the pipeline itself.

This is the runner’s own helper module and is unrelated to xai4tsc.utils, which serves the importable package.

Attributes

logger

Functions

expand_datasets(→ list[dict])

Expand archive wildcards and resolve per-member overrides.

Module Contents

experiment_runner.utils.logger

experiment_runner.utils.expand_datasets(datasets: list[dict]) → list[dict]

Expand archive wildcards and resolve per-member overrides.

Each entry is either a concrete dataset or an archive keyword (UCR / UEA) that expands to all its member datasets. An entry’s inline settings apply to every member it produces; an optional overrides map ({dataset_name: {setting: value}}) sets per-member exceptions that take precedence over the entry’s inline settings.

Parameters:

datasets (list[dict]) – Raw data_config["datasets"] entries.

Returns:

Concrete dataset entries with overrides resolved and stripped.

Return type:

list[dict]