daniel/Artifex
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Resume Guard overnight from base evaluation

Browse source
This commit is contained in:
Daniel Maddern 2026-08-17 13:14:13 +07:00
parent 38dfb059d7
commit 6ff055b9ce
1 changed files with 2 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
scripts/run_guard3b_overnight_spark.sh
View file

@ -57,7 +57,8 @@ fi
cp "$PROJECT/artifex_overnight/guard3b_v03_canary/pilot_metrics.json" "$RUN_ROOT/canary_metrics.json" cp "$PROJECT/artifex_overnight/guard3b_v03_canary/pilot_metrics.json" "$RUN_ROOT/canary_metrics.json"
# Base and the initial Challenger use the identical fixed harness/version. # Base and the initial Challenger use the identical fixed harness/version.
evaluate base # Reuse a completed base result when resuming after an infrastructure repair.
if [[ ! -s "$RUN_ROOT/base_evmbench.json" ]]; then evaluate base; else docker_run python /workspace/scripts/score_evmbench_strict.py "/workspace/artifex_overnight/guard3b_v03_overnight/base_evmbench.json" > "$RUN_ROOT/base_strict.json" 2>&1; fi
cp "$RUN_ROOT/base_strict.json" "$RUN_ROOT/current_strict.json" cp "$RUN_ROOT/base_strict.json" "$RUN_ROOT/current_strict.json"
CURRENT=guard3b_v03_canary CURRENT=guard3b_v03_canary
evaluate canary "$CURRENT" evaluate canary "$CURRENT"