latent.stats

Statistical analysis for evaluation workflows.

Submodules are loaded lazily via latent._lazy — importing one helper (e.g. log_to_mlflow) only loads latent.stats.mlflow, not the entire 28-module statistics ecosystem. Saves cold-start and lets vendor consumers ship a much smaller closure.

The StatisticalReport.optimization: OptimizationReport forward-ref is finalized inside latent.stats.models (the consumer just imports the class as usual; the rebuild happens at models.py load time, not here).

Quick start: >>> from latent.stats import analyze >>> report = analyze(scores={"f1": scores})