Menü
Combine candidate submissions, post-process, verify, and submit to Kaggle.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Basierend auf der SOC-Berufsklassifikation
Use this skill for any task in the cayley IHES Picture Cube project — training, solving, evaluation, ensembling, submission. Covers the full ML+search pipeline.
Strat-5 acceptance gate for a megaminx V model. Runs 51-pid stratified eval at beam 65k, compares to m05 baseline (50/51 / mean 89.4), reports pass/fail.
Run the top-N long-pid sym-ensemble rescue + merge cycle on the current best megaminx submission.
One-shot status across GCP solves, Kaggle kernels, and local runs for the megaminx competition.
Verify a megaminx submission CSV, submit to Kaggle, and update HANDOFF / SUBMISSION_JOURNEY / to_do_shortlist with the new score.
Pull the latest cayleypy-tpu-beam-smoke kernel output, aggregate per-rank JSONs, merge with current best, verify, and prep for submission.
| name | source-command-submit |
| description | Combine candidate submissions, post-process, verify, and submit to Kaggle. |
Use this skill when the user asks to run the migrated source command submit.
Bundles the 4-command submission pipeline:
combine_submissions.py — take min-length per puzzle across all candidate CSVs + Kociemba fallback.post_process_submission.py — pair-cancel + BFS-d5 window shortening.verify_submission sanity check on the final CSV.kaggle competitions submit./submit <description> — uses all submissions/*.csv as candidates EXCEPT the output of previous /submit runs (named sub_*.csv)./submit --only a.csv b.csv ... "description" — specify candidates explicitly.Do the following, in order:
Pick candidate CSVs. Unless the user passed --only, default to submissions/fast_b8k_mitm6.csv submissions/s1_b4k.csv submissions/s2_b4k.csv submissions/s3_b4k.csv submissions/e3_khoruzhii_b65k.csv submissions/e5_khoruzhii_b65k.csv (these are the landmark runs that have always helped in ensembles — verify they exist before using).
Generate a timestamped output path submissions/sub_$(date +%Y%m%d_%H%M%S).csv — this is the raw combine output before post-processing.
Run the combine step:
.venv/Scripts/python.exe scripts/combine_submissions.py \
--candidates <CSVs> \
--fallback data/kociemba_fallback.csv \
--out <raw_out>
Run the post-process step (appending _pp to the output name):
.venv/Scripts/python.exe scripts/post_process_submission.py \
--in <raw_out> --out <pp_out> \
--bfs-table data/bfs_table_d5.pkl
Report the combined total move count and post-processed total.
Ask the user to confirm submission (the -m description is whatever was passed to /submit). Run:
export KAGGLE_API_TOKEN=$KAGGLE_API_TOKEN
.venv/Scripts/kaggle.exe competitions submit -c cayleypy-ihes-cube -f <pp_out> -m "<description>"
After submission, immediately check status with:
.venv/Scripts/kaggle.exe competitions submissions cayleypy-ihes-cube | head -3
SubmissionStatus.COMPLETE with score 999999 or similar on Kaggle (silent failure).data/kociemba_fallback.csv as the fallback floor; never use sample_fallback.csv (sample-quality, 500K+ moves).