Runner

class xaicompare.runner.XAICompareRunner(model, x_test, y_test: Sequence | None = None, raw_text: Sequence | None = None, class_names: Sequence[str] | None = None, run_dir: str = 'runs/_latest', config: Dict[str, Any] | None = None, save_model: bool = True, *, model_type: str = 'sklearn', xai_methods: List[str] | None = None, top_k_local: int = 15)

Bases: object

Orchestrates a full ‘publish run’:

• Registry autodiscovery

• Model wrapping via registry

• Predictions + top-k probabilities

• XAI global + local explanations for one or more methods

• Text index

• Artifact persistence to a run directory

Usage:

runner = XAICompareRunner(
    model=my_model,
    x_test=x_test,
    y_test=y_test,
    raw_text=texts,
    class_names=class_names,
    run_dir="runs/_latest",
    config={"rows_limit_global": 200, "rows_limit_local": 200},
    save_model=True,
    model_type="sklearn",
    xai_methods=["shap_tree"],
    top_k_local=15,
)
runner.run()

_build_text_index()

_compute_predictions()

_ensure_registry()

static _pbar(iterable=None, total=None, desc=None)

_prepare_run_dir()

_run_xai()

_save_meta()

_save_model_if_requested()

static _to_yaml(d)

_wrap_model()

_write_common_artifacts()

_write_config_if_present()

run()

Main entry point :return: