Calibrate the DEFAULT cost constant — it dominates real closures #23

Closed
opened 2026-07-08 21:38:37 +03:00 by issuer-agent · 1 comment

In the emmett test, 10,584 of 18,729 drvs hit the DEFAULT (1.0 min) bucket = 55% of total modeled work (10,584 of 19,269 min). The headline estimate is mostly one guessed constant. Ship a calibration: mine a corpus with the #12 miner and fit DEFAULT (and TRIVIAL/FIXED_OUTPUT) from measured medians; or at minimum document the sensitivity and add a --default-min knob.

In the emmett test, 10,584 of 18,729 drvs hit the DEFAULT (1.0 min) bucket = 55% of total modeled work (10,584 of 19,269 min). The headline estimate is mostly one guessed constant. Ship a calibration: mine a corpus with the #12 miner and fit DEFAULT (and TRIVIAL/FIXED_OUTPUT) from measured medians; or at minimum document the sensitivity and add a --default-min knob.
Owner

added cost(..., default_min=), costmodel.fit_defaults(), a --default-min FLOAT knob threaded through estimate(), and a calibrate <attr> --history FILE subcommand printing measured per-bucket medians vs current constants.

added `cost(..., default_min=)`, `costmodel.fit_defaults()`, a `--default-min FLOAT` knob threaded through estimate(), and a `calibrate <attr> --history FILE` subcommand printing measured per-bucket medians vs current constants.
oleks closed this issue 2026-07-08 22:45:23 +03:00
Sign in to join this conversation.
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: oleks/nix-estimator#23