一键在 Manus 中运行任何 Skill

kaggle-submit-check

星标1

分支0

更新时间2026年7月9日 17:49

Validate a submission.csv against sample_submission before spending a scarce daily submission slot — column names/order, id coverage, prob ranges/sums, NaNs, and canonical class order. Use right before every leaderboard submission.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

Erlemar

Erlemar/kaggle-tabular-multiclass-s6e7

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Skill: pre-submission format check

Submissions per day are scarce. Never burn one on a formatting bug. Validate first.

Run it

python templates/submit.py --check           # validate only
python templates/submit.py --check --submit -m "blend_v3: cv 0.4123"   # validate then submit (if Kaggle CLI set up)

What must pass (all of these)

Columns match sample_submission.csv exactly — same names, same order. For probability submissions the class-column order is the canonical order; a silent permutation destroys logloss.
Ids — every required test id present, exactly once; no missing, no extras.
Values — no NaN/inf; probabilities in [0,1]; each row's probabilities sum to ≈ 1 (for prob submissions). For argmax/label submissions, every value is a valid class label.
Post-processing applied — the metric's required transform (clip / label-map / snap) is done.
Row count = sample_submission row count.

Then submit (kernel vs CSV)

CSV competition: kaggle competitions submit -c <slug> -f submission.csv -m "<msg>", then check kaggle competitions submissions -c <slug>.
Kernel-only competition: the CSV upload does not score. Push the notebook and use the UI Submit to Competition action — it re-runs on the hidden test and spends the slot. (See the playbook's "Kernel-only" section.)

After submitting

Record CV → LB in EXPERIMENTS_LOG.md. Update the CV↔LB gap. If CV and LB disagree, the CV is suspect — investigate the split/leak before trusting more experiments.
Select finals by CV, not public LB. A safe pair: best-CV blend + a lower-variance hedge.

同仓库更多 Skills

同仓库

kaggle-blend

Erlemar/kaggle-tabular-multiclass-s6e7

Combine all model OOF predictions into the best validated blend/stack, apply the frozen recipe to test predictions, and write submission.csv. Use after the model agents finish, or any time the OOF pool changes.

2026-07-091

kaggle-cv-setup

Erlemar/kaggle-tabular-multiclass-s6e7

Create or validate the shared cross-validation split (artifacts/folds.parquet) that every model must use, and check that it tracks the leaderboard. Use before any model trains, or whenever CV and LB diverge.

2026-07-091

kaggle-new-competition

Erlemar/kaggle-tabular-multiclass-s6e7

Scaffold a new tabular-multiclass competition in this kit — download/place data, detect id/target/classes/metric, fill src/config.yaml, pick the fold scheme, and freeze the shared CV split. Use once at the very start, before EDA.

2026-07-091

name	kaggle-submit-check
description	Validate a submission.csv against sample_submission before spending a scarce daily submission slot — column names/order, id coverage, prob ranges/sums, NaNs, and canonical class order. Use right before every leaderboard submission.

Skill: pre-submission format check

Submissions per day are scarce. Never burn one on a formatting bug. Validate first.

Run it

python templates/submit.py --check           # validate only
python templates/submit.py --check --submit -m "blend_v3: cv 0.4123"   # validate then submit (if Kaggle CLI set up)

What must pass (all of these)

Columns match sample_submission.csv exactly — same names, same order. For probability submissions the class-column order is the canonical order; a silent permutation destroys logloss.
Ids — every required test id present, exactly once; no missing, no extras.
Values — no NaN/inf; probabilities in [0,1]; each row's probabilities sum to ≈ 1 (for prob submissions). For argmax/label submissions, every value is a valid class label.
Post-processing applied — the metric's required transform (clip / label-map / snap) is done.
Row count = sample_submission row count.

Then submit (kernel vs CSV)

CSV competition: kaggle competitions submit -c <slug> -f submission.csv -m "<msg>", then check kaggle competitions submissions -c <slug>.
Kernel-only competition: the CSV upload does not score. Push the notebook and use the UI Submit to Competition action — it re-runs on the hidden test and spends the slot. (See the playbook's "Kernel-only" section.)

After submitting

Record CV → LB in EXPERIMENTS_LOG.md. Update the CV↔LB gap. If CV and LB disagree, the CV is suspect — investigate the split/leak before trusting more experiments.
Select finals by CV, not public LB. A safe pair: best-CV blend + a lower-variance hedge.