uot-bench

uot-bench is a Python toolkit for optimal transport solvers and benchmarking. It provides JAX-first implementations of common OT methods, utilities for generating problems and measures, and a configurable pipeline for running experiments at scale.


Where to start

Goal Where to go
Install the package Install
Run a solver in 5 lines of Python Quickstart
Understand the core concepts (Problem, Generator, Solver, Experiment) Concepts
Write a custom Problem, Generator, or Solver Library guide
Run experiments via YAML config files CLI & YAML
API reference Reference

Key abstractions

from uot import (
    Problem, Generator,           # base classes — subclass these
    TwoMarginalProblem,           # concrete problem type
    BaseSolver, SolverConfig,     # solver infrastructure
    Experiment, run_pipeline,     # experiment runner
)

See guide/concepts.md for a diagram of how these pieces fit together.