Add blinded ML debugging benchmark harness

2026-08-12 12:10:51 +08:00 · 2026-07-12 22:18:26 +08:00
parent 9774c4bb1d
commit a0fcfa291e
8 changed files with 452 additions and 3 deletions
@@ -3,3 +3,12 @@ default:

 audit:
  python3 scripts/audit.py --self-test .
+
+benchmark-validate:
+  python3 benchmark/validate.py
+
+benchmark-run run_id model workers="4":
+  python3 benchmark/run.py --run-id {{run_id}} --model {{model}} --workers {{workers}}
+
+benchmark-score run_id:
+  python3 benchmark/score.py --run-id {{run_id}}