Examples¶

Use these examples as a starting point. They are grounded in the code and are kept minimal so you can adapt them quickly.

Quickstart scripts (recommended)¶

Bandit OPE: examples/quickstart/bandit_ope.py
MDP OPE: examples/quickstart/mdp_ope.py

Run them locally:

python -m examples.quickstart.bandit_ope
python -m examples.quickstart.mdp_ope

10-line end-to-end example¶

from crl.benchmarks.bandit_synth import SyntheticBandit, SyntheticBanditConfig
from crl.ope import evaluate_ope

benchmark = SyntheticBandit(SyntheticBanditConfig(seed=0))
dataset = benchmark.sample(num_samples=500, seed=1)
report = evaluate_ope(dataset=dataset, policy=benchmark.target_policy)

print(report.summary_table())

Notebook gallery (source)¶

The repo includes notebooks that walk through diagnostics, sensitivity analysis, and estimator selection. Each notebook is paired with a .py script via Jupytext so you can diff and run them easily. See the source directory:

notebooks/

Recommended notebooks:

00_introduction.ipynb — estimand-first tour + report export
02_bandit_ope_walkthrough.ipynb — diagnostics-driven bandit OPE
03_mdp_ope_walkthrough.ipynb — trajectory estimators and horizon effects
10_bandit_ope_end_to_end.ipynb — full workflow with sensitivity analysis