Skip to content

MLflow & Training Evidence

Proof of experiment tracking, model training, and registry operations.

Work in progress

This page is a placeholder. MLflow screenshots and run links will be added here.

Experiment Runs

Screenshot placeholder — MLflow Experiments UI showing training runs with metrics

Key Runs

Run Name Log-Loss Accuracy Status
xgb_baseline_v1 0.98 51.2% Archived
xgb_tuned_v1 0.96 52.8% Champion
lgbm_v1 0.97 52.1% Challenger

Model Registry

Screenshot placeholder — MLflow Model Registry showing soccer-predictor with versions and aliases

Version Stage Log-Loss Promoted
v1 Archived 0.98 2025-01-10
v2 Champion 0.96 2025-03-15
v3 Challenger In evaluation

Artifact Lineage

All artifacts tracked via DVC + logged to MLflow:

  • data/models/champion.pkl — DVC-tracked
  • MLflow artifact URI: mlflow-artifacts:/soccer-predictor/2/artifacts/model

Reproduce

dvc repro train_model
mlflow ui --backend-store-uri ./mlruns

See Training Pipeline (DVC).