Exécutez n'importe quel Skill dans Manus
en un clic

Exécutez n'importe quel Skill dans Manus en un clic

lemma-model

Étoiles4

Forks1

Mis à jour15 juillet 2026 à 00:47

Select and evaluate a production-worthy model after an honest baseline and validation design exist; use for tuning, calibration, thresholding, and final evaluation.

Installation

Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.

Exécuter dans Manus

Source

tkpratardan

tkpratardan/lemma

Ouvrir le dépôt GitHub Voir les dépôts du créateur

Téléchargement

Exécuter dans Manus

Métiers associésSOC

Basé sur la classification professionnelle SOC

Scientifiques des donnéesProfessions informatiques et mathématiques·SOC 15-2051

Explorateur de fichiers

2 fichiers

SKILL.md

readonly

name	lemma-model
description	Select and evaluate a production-worthy model after an honest baseline and validation design exist; use for tuning, calibration, thresholding, and final evaluation.

Final modeling

Deliver

Provide a comparison against the baseline, the selected model and threshold, stability and subgroup evidence, a locked final evaluation, and operational limitations.

Check

Keep one validation design and decision metric across candidates.
Compare a small motivated model set; keep preprocessing and tuning within training data and set fixed seeds.
Diagnose train-validation gaps, fold or temporal stability, calibration, subgroup behavior, latency, and interpretability.
Challenge the most fragile feature or assumption with an ablation, shifted window, or leakage audit.
Lock the pipeline and threshold before touching the final test set once.

Do not tune against the test set, change metrics mid-search, compare different splits, omit the baseline, or hide subgroup regressions behind an aggregate gain.

For tuning and error-analysis detail, read references/deep-guide.md.

Plus depuis ce dépôt

même dépôt

lemma-baseline

tkpratardan/lemma

Establish a dumb baseline and an honest validation harness before any real model, so every later number means something.

2026-07-154

lemma-causal

tkpratardan/lemma

Rigor for causal questions and A/B tests (the effect of acting on X): confounding, post-treatment bias, valid control groups.

2026-07-154

lemma-describe

tkpratardan/lemma

Rigor for descriptive and diagnostic analytics (what happened and why): denominators, grain, and confounded slices, not model leakage.

2026-07-154

lemma-eda

tkpratardan/lemma

EDA kickoff for a fresh dataset: fixed opening scaffold (goal, imports, load, sanity), then chapters derived from the data; scan leakage, land a baseline.

2026-07-154

lemma-inference

tkpratardan/lemma

Rigor for statistical inference (is the difference real): hypothesis tests, power, multiple comparisons, effect size over p-value.

2026-07-154

lemma-leakage

tkpratardan/lemma

Audit a dataset or pipeline for the five leakages that inflate a metric: target, preprocessing, temporal, group, and sampling.

2026-07-154

name	lemma-model
description	Select and evaluate a production-worthy model after an honest baseline and validation design exist; use for tuning, calibration, thresholding, and final evaluation.

Final modeling

Deliver

Provide a comparison against the baseline, the selected model and threshold, stability and subgroup evidence, a locked final evaluation, and operational limitations.

Check

Keep one validation design and decision metric across candidates.
Compare a small motivated model set; keep preprocessing and tuning within training data and set fixed seeds.
Diagnose train-validation gaps, fold or temporal stability, calibration, subgroup behavior, latency, and interpretability.
Challenge the most fragile feature or assumption with an ablation, shifted window, or leakage audit.
Lock the pipeline and threshold before touching the final test set once.

Do not tune against the test set, change metrics mid-search, compare different splits, omit the baseline, or hide subgroup regressions behind an aggregate gain.

For tuning and error-analysis detail, read references/deep-guide.md.