autointent.advisor.reduce_to_fit#

autointent.advisor.reduce_to_fit(config, stats, hardware, *, max_iters=20, refit_after=False)#

Iteratively drop the most expensive infeasible module until the search space fits.

Behavior:

  • If config is already feasible, returns (config, report) unchanged.

  • Otherwise picks the OVER-driving scoring-node module with the largest cost along whichever budget breached (VRAM > time > RAM) and removes it from the search_space, then re-runs preflight. Disk is deliberately not in that order: driver rows carry no per-module disk figure, so disk pressure reduces by the VRAM proxy — download size tracks model size.

  • Repeats until feasible, max_iters reached, or no droppable module remains — in the last two cases raises ReduceToFitError carrying the pruned config and final report.

Parameters:

  • config (dict[str, Any]) – an OptimizationConfig-shaped dict (same input as run_preflight()).

  • stats (autointent.advisor._report.DatasetStats) – dataset stats to score against.

  • hardware (Any) – detected hardware profile.

  • max_iters (int) – safety cap; a valid pipeline has ≤ ~10 scoring modules so hitting the default cap means the picker is stuck (raises).

  • refit_after (bool) – forwarded to run_preflight().

Returns:

(pruned_config, report) where report.is_feasible is True.

Raises:

ReduceToFitError – nothing fits after pruning.

Return type:

tuple[dict[str, Any], autointent.advisor._report.PreflightReport]

On this page