🔍

LeakLens

The pre-flight checklist for machine learning datasets

📁 {{ meta.dataset_name }}

v{{ meta.version }}

⏱ {{ meta.runtime_seconds }}s

Train {{ meta.n_train_rows }}

{% if meta.n_test_rows %}

Test {{ meta.n_test_rows }}

{% endif %} {% if meta.target %}

Target {{ meta.target }}

{% endif %}

{% set risk_class = "high-risk" if risk_label == "HIGH RISK" else ("medium-risk" if risk_label == "MEDIUM RISK" else "low-risk") %}

❌

Critical Issues

Needs immediate attention

⚠

Warnings

Should review

✔

Passed Checks

No issues found

{{ checks_run | length }}/{{ checks_run | length }}

📊

Checks Executed

{{ passed_count }} passed · {{ checks_run|length - passed_count }} failed

📋 Checks Summary

{% for c in checks_status %}

{{ issue_icons.get(c.name, '')|safe }} {{ c.name }} {% if c.status == "passed" %}✔ Passed{% else %}{{ c.count }} found{% endif %}

{% endfor %}

Issues by Severity

{% set total_all = critical_count + warning_count + passed_count %}

{% if total_all > 0 %}

{% endif %}

Critical ({{ critical_count }}) Warning ({{ warning_count }}) Passed ({{ passed_count }})

Train / Test Overlap

{% set max_rows = [overlap.train_rows, overlap.test_rows, 1] | max %}

Train{{ overlap.train_rows }}

Test{{ overlap.test_rows }}

Overlap{{ overlap.overlap_pct }}%

⚠ Top Issues ({{ issues | length }})

{% if issues %} {% for issue in issues %}

{{ issue.severity.value }} {{ issue_icons.get(issue.title, '')|safe }} {{ issue.title }} {% if issue.column %}{{ issue.column }}{% endif %} {% if issue.details.get('psi') is not none %}PSI = {{ "%.3f"|format(issue.details['psi']) }} {% elif issue.details.get('ks_stat') is not none %}KS = {{ "%.3f"|format(issue.details['ks_stat']) }} {% elif issue.details.get('correlation') is not none %}corr = {{ "%.3f"|format(issue.details['correlation']) }} {% endif %} ▸

Explanation

{% if issue.details %}

Metrics

{% for k, v in issue.details.items() %} {% if k not in ("test",) %}

{{ "%.4f"|format(v) if v is number and (v is not integer or v|float != v) else v }}

{% endif %} {% endfor %}

{% endif %} {% set thr = none %} {% if issue.details.get('correlation') is not none %}{% set thr = thresholds.target_corr_threshold %} {% elif issue.details.get('cramers_v') is not none %}{% set thr = thresholds.cramers_v_threshold %} {% elif issue.details.get('mapped_fraction') is not none %}{% set thr = thresholds.near_identical_mapping_threshold %} {% elif issue.details.get('psi') is not none %}{% set thr = (thresholds.psi_critical if issue.severity.value == 'critical' else thresholds.psi_warning) %} {% elif issue.details.get('p_value') is not none %}{% set thr = thresholds.ks_alpha %} {% elif issue.details.get('overlap_pct') is not none %}{% set thr = (thresholds.contamination_critical_pct if issue.severity.value == 'critical' else thresholds.contamination_warning_pct) %} {% elif issue.details.get('top_value_freq') is not none %}{% set thr = thresholds.constant_feature_threshold %} {% endif %} {% if thr is not none %}

Threshold Used

This check flags values past {{ "%.3f"|format(thr) }} — fully configurable via Config().

{% endif %} {% if why_it_matters.get(issue.title) %}

Why it matters

{% endif %}

Recommendation

{{ issue.column and ("Review '" + issue.column + "' before training — " + issue.message.split('—')[0].strip().lower() + ".") or issue.message }}

{% endfor %} {% else %}

✔ No issues found across {{ checks_run | length }} checks.

{% endif %} {% if drift_ranking %} {% set has_real_drift = drift_ranking | selectattr("severity", "ne", "stable") | list | length > 0 %}

📈 Top Drifted Features

{% if not has_real_drift %}

✔ All checked features are stable — no significant drift detected.

{% endif %} {% for d in drift_ranking %}

{{ d.column }} {{ d.metric_label }} = {{ "%.3f"|format(d.value) }}

{% endfor %}

{% if numeric_drift_figures %}

📉 Numeric Distribution — {{ numeric_drift_figures[0].column }}

KS: {{ numeric_drift_figures[0].ks_stat }} p-value: {{ numeric_drift_figures[0].p_value }} {{ numeric_drift_figures[0].status }}

{% endif %}

{% endif %} {% if numeric_drift_figures | length > 1 %}

More Drifted Numeric Columns

{% for fig in numeric_drift_figures[1:] %}

KS: {{ fig.ks_stat }} {{ fig.status }}

{% endfor %}

{% endif %} {% if high_cardinality_cards %}

📁 High-Cardinality Columns (not plotted)

{% for hc in high_cardinality_cards %}

{{ hc.column }} 🔴 High Cardinality

Unique values{{ hc.unique_values }}

Unseen in test{{ hc.unseen_count }}

{% endfor %}

{% endif %}

📊 Quick Data Quality Strip

Missing Values (train)

{{ data_quality.missing_train_pct }}%

Duplicate Rows

{{ data_quality.duplicate_count }} ({{ data_quality.duplicate_pct }}%)

Constant Columns

High Cardinality

Dtype Mismatch

{% set seen_rec_keys = [] %} {% if recommendations %}

💡 Recommendations

{% for issue in issues %} {% set key = issue.title ~ '|' ~ (issue.column or '') %} {% if key not in seen_rec_keys and loop.index0 < 6 and fix_steps.get(issue.title) %} {% set _ = seen_rec_keys.append(key) %}

{{ issue_icons.get(issue.title, '')|safe }} {{ issue.title }}{% if issue.column %} — {{ issue.column }}{% endif %}

{% if why_it_matters.get(issue.title) %}

• {{ why_it_matters[issue.title] }}

{% endif %}

Suggested fix

✔ {{ step }}

{% endif %} {% endfor %}

{% endif %} {% set passed = checks_status | selectattr("status", "equalto", "passed") | list %} {% if passed %}

✅ Passed Checks ({{ passed | length }}) ▸

{% for p in passed %}

✔{{ p.name }}

{% endfor %}

{% endif %}