From 5a2d7e5221ca088a9213be005fb69f36bc742bf1 Mon Sep 17 00:00:00 2001 From: Filip Leonarski Date: Tue, 17 Nov 2026 09:51:06 +0100 Subject: [PATCH] CI battery: the placed model not refined (--extra=--no-refine) Refinement runs after every decision the battery scores, so leaving it out changes no verdict; the refinement arm covers it separately. A large model's refinement took most of the validation job on the runner's card (4nwv about 50 of its minutes). Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01UqXYdJz9sA2BQBEgKkuVuL --- .gitea/workflows/build_and_test.yml | 4 +++- tools/battery/README.md | 4 +++- 2 files changed, 6 insertions(+), 2 deletions(-) diff --git a/.gitea/workflows/build_and_test.yml b/.gitea/workflows/build_and_test.yml index e428b70b9..cfa1566bb 100644 --- a/.gitea/workflows/build_and_test.yml +++ b/.gitea/workflows/build_and_test.yml @@ -680,9 +680,11 @@ jobs: set -o pipefail # ends with one line per set (rugnux / reference, a tick or cross on the space group); # exits non-zero when a ci-tier set failed. The wrong-model null is the next step. + # --no-refine: the placed model is not refined - refinement runs after every decision this battery + # scores, and a large model's refinement took most of the job (4nwv ~50 min on the runner's card). JFJOCH_BATTERY_SITE=/tmp/site.json /tmp/bvenv/bin/python tools/battery/battery.py run \ --rugnux build/rugnux/rugnux --build-dir build --label "$LABEL" --no-nullmodel \ - $SUBSET 2>&1 | tee battery.log + --extra=--no-refine $SUBSET 2>&1 | tee battery.log - name: Wrong-model null if: ${{ !cancelled() && steps.build.outcome == 'success' && contains(env.SUBSET, '--tier ci') }} shell: bash diff --git a/tools/battery/README.md b/tools/battery/README.md index 4985b64dc..27f4dec17 100644 --- a/tools/battery/README.md +++ b/tools/battery/README.md @@ -317,7 +317,9 @@ a ci-tier set in the run failed - a wrong lattice or space group, a timeout or a when a pair of the wrong-model null (below) failed, and names those sets. This holds in any run that includes them (`--tier ci`, a full run, `--only` naming one), so a red CI job means the same thing whichever run it started. Nothing else fails a run: other sets' failures and every metric are reported, not judged. CI runs the two as separate steps - -`run --tier ci --no-nullmodel`, then `nullmodel` - so the job page shows which of them failed. +`run --tier ci --no-nullmodel --extra=--no-refine`, then `nullmodel` - so the job page shows which of them failed. +The placed model is not refined there: nothing the battery scores depends on it, and a large model's +refinement took most of the job on the runner's card. ### The wrong-model null (with every ci-tier run)