Penguin-N/wofost-weather-regime-pool

该数据集是一个WOFOST-Gym天气池，用于Weather-Regime OOD基准测试。包含以下分割：train（非极端天气，每种作物1600个场景）、val（组合验证池，每种作物640个场景）、val_id（保留的非极端天气，每种作物128个场景）以及val_drought、val_wet、val_hot、val_cold（每种作物/天气状况128个场景）。作物类型包括chickpea（鹰嘴豆）和potato（马铃薯）。此池中没有test分割。捆绑的天气缓存以meteo_cache.tar.gz存储，并由AgriManager的weather_pool.ensure_pool()运行时助手自动提取。This dataset is a WOFOST-Gym weather pool intended for the Weather-Regime OOD benchmark test. It comprises the following data splits: train (non-extreme weather, 1600 scenarios per crop), val (combined validation pool, 640 scenarios per crop), val_id (held-out non-extreme weather, 128 scenarios per crop), along with val_drought, val_wet, val_hot, val_cold (128 scenarios per crop/weather condition). The supported crop types are chickpea and potato. No test split is included in this pool. The bundled weather cache is stored as meteo_cache.tar.gz, and will be automatically extracted by the AgriManager's weather_pool.ensure_pool() runtime assistant.