From b270eb58e22deb9f6a1c5342db41d531232ded0d Mon Sep 17 00:00:00 2001 From: yurenh Date: Mon, 31 Aug 2026 18:34:12 -0500 Subject: results flow: collect.py (node-side bundle) + plot_ladder.py (gap-vs-scale analysis) Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_01GkgLsACEF6CCP7EUfA5fZe --- README.md | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'README.md') diff --git a/README.md b/README.md index f7f803d..128eed0 100644 --- a/README.md +++ b/README.md @@ -43,3 +43,13 @@ query path; forward differences (n+1 instead of 2n queries) as a cheaper biased ## Citation Paper in preparation ("Backpropagation Without Jacobians"); see the main zobp research repo for the method, theory and part-1/part-3 results. License: Apache-2.0. + +## After the runs (results flow) +Checkpoints stay on the node (`runs/` is gitignored). Collect the small JSONL logs and push them back: +``` +python scripts/collect.py --runs runs --out results/h200node1 +git add results && git commit -m "ladder results" && git push +``` +Analysis (anywhere): `python scripts/plot_ladder.py --results results/h200node1` -> per-run table, loss +curves, and the gap-vs-scale figure (the paper's part-2 headline). If a specific checkpoint is needed for +the estimator audits, scp just that `runs//ckpt.pt`. -- cgit v1.2.3