ModelGate release evidence card

Decision: READY FOR HUMAN APPROVAL

Commit: d20327bb

Pipeline: https://gitlab.com/taku629/modelgate/-/pipelines/2994106830

Model SHA-256: 0287630482da1011b178845e912683709e7e2563a1809f691a876c9205485484

Model quality vs approved baseline

MetricCandidateBaseline
accuracy0.9860.986
f10.98890.9889
roc_auc0.99770.9977
p95_latency_ms0.1311.0

Evaluation gate

CheckResultDetail
accuracy regressionpassdrop +0.00 pt (limit 1.0 pt)
f1 regressionpassdrop +0.00 pt (limit 1.0 pt)
accuracy floorpassabsolute minimum
p95 latency (ms)passsingle-row predict

Drift: warn

worst compactness PSI 0.1905, worst concave points PSI 0.1903, fractal dimension error PSI 0.1552

Security scanners

• Scanner reports are attached to the pipeline's Security tab.

Dependency licenses

PackageVersionLicenseAllowed
fastapi0.115.6MIT Licenseyes
uvicorn0.32.1BSD-3-Clauseyes
scikit-learn1.5.2BSD Licenseyes
numpy2.1.3BSD Licenseyes
joblib1.4.2BSD 3-Clauseyes
pydantic2.10.3MITyes

Production deploy stays when: manual: a human approves after reading this card.