Polars Eval Metrics

Welcome to Polars Eval Metrics

Tests Documentation

A high-performance model evaluation framework built on Polars lazy evaluation.

Features

  • Fast: Leverages Polars lazy evaluation for optimal performance
  • Flexible: Support for custom metrics and expressions
  • Registry-based: Global registry system for errors, metrics, and summaries
  • Type-safe: Pydantic models with validation
  • Simple: Clean API with sensible defaults
  • Extensible: Easy to add new metrics and aggregation types