xiaoou 4e6ad234da feat: Neural-only training, experience replay, per-round eval ...

3个训练效率提升:
1. Neural filter: V训练只取player=2数据, 排除ISMCTS状态干扰
   旧格式(无player列)自动fallback全量数据
2. 经验回放: 保留最近3轮数据混入当前轮训练, 扩大分布覆盖
   存储到 paodekuai-trainer/data/replay_buffer.pkl
3. 每轮eval: 训练后跑200局--eval测胜率, 下降>10%自动回滚
   回滚到 model_prev.pt, 防止坏模型污染下轮数据

2026-07-13 11:26:40 +08:00

..

__pycache__

feat: 20k局/轮无限训练 + scan_wins支持存档目录

2026-07-13 08:43:39 +08:00

scan_wins.py

feat: 20k局/轮无限训练 + scan_wins支持存档目录

2026-07-13 08:43:39 +08:00

training_orchestrator.py

feat: Neural-only training, experience replay, per-round eval

2026-07-13 11:26:40 +08:00