uot-bench exposes its pipeline through console scripts installed with the
package. Every command is also available as python -m <module> — the two
forms are exactly equivalent.
| Console script | python -m equivalent |
What it does | Details |
|---|---|---|---|
uot-serialize |
python -m uot.problems.problem_serializer |
Generate and persist problems from a YAML generator config | serialize |
uot-benchmark |
python -m uot.experiments.synthetic.benchmark |
Run solvers over problems, write results CSV | benchmark |
uot-color-transfer |
python -m uot.experiments.real_data.color_transfer.color_transfer |
Color transfer experiment | color-transfer |
uot-color-transfer-viz |
python -m uot.experiments.real_data.color_transfer.visualization |
Visual dashboard for color transfer results | |
uot-mnist-distances |
python -m uot.experiments.real_data.mnist_classification.count_pairwise_distances |
Step 1 of MNIST: pairwise OT distances | mnist |
uot-mnist-classification |
python -m uot.experiments.real_data.mnist_classification.mnist_classification |
Step 2 of MNIST: KNN classification | mnist |
uot-inspect-store |
python -m uot.problems.inspect_store |
Visualize a serialized problem dataset |
Run any command with --help for the full flag listing.
configs/generators/example.yaml
│
▼
uot-serialize --config <generator.yaml> --export-dir datasets/synthetic
│
▼ (or skip serialize and use --generators-config for online generation)
│
uot-benchmark --config <runner.yaml> --dataset datasets/synthetic
--folds 3 --export results/my_run.csv
│
▼
results/my_run.csv ← pandas DataFrame, one row per (problem, solver, params, fold)
The two config files serve different purposes:
Generator config (configs/generators/) — defines what problems to create:
distribution family, dimensions, number of samples.
See Generating datasets.
Runner config (configs/runners/) — defines which solvers to run and
how to measure them: solver class, parameter grids, measurement function.
See Running benchmarks.
They are intentionally separate so you can reuse the same dataset across many runner configs, and run any runner against any dataset.
See SLURM for sbatch wrappers around the benchmark and color-transfer commands.