Contextual bandit pipeline and off-policy evaluation tooling for adaptive game difficulty.
bandit/- Training, evaluation, and serving pipeline for contextual banditsope/- Off-policy evaluation utilities and experiments
Copy the example configs in bandit/config/ and fill in the real values:
pipeline_params.example.jsonredis_config.example.jsoncloudflare_kv_config.example.json
pip install -r bandit/requirements.txtSee bandit/README.md for detailed pipeline steps and ope/requirements_obp.txt for OPE dependencies.