Attribution

- The mahjong state/inference runtime used by this package (libriichi.so
  and parts of ai.py / compact.py) derives from the open-source Mortal
  project: https://github.com/Equim-chan/Mortal
  (commit 0cff2b52982be5b1163aa9a62fb01f03ce91e0d2), distributed under
  AGPL-3.0-or-later. See LICENSE.
- The model weights are distributed under Apache-2.0. See
  LICENSE-MODEL.txt. The bundled model is a locally fine-tuned derivative,
  distributed as a 4-bit quantized pack; its decisions agreed with the
  FP32 derivative on 98.8% of 2,304 non-forced positions in self-play.
- Base model: Mortal S42, distributed at
  https://huggingface.co/haor/Mortal-S42 (ONNX conversion:
  https://huggingface.co/xzhe121/Mortal-S42-onnx, pinned revision
  2f7ad48bc29dadd9805b1897f3c5b97b76c0d41a). Changes from the base
  model: a self-distilled derivative was produced locally (advantage
  head), then quantized to 4-bit for this package.
