KlaskLab/klasktron-il-benchmark

KlaskTron是一个用于模仿学习的Klask数据集，包含三个层次的数据：真实人类游戏数据、经过轴镜像增强的人类数据和大规模合成推演数据。数据集提供了Parquet和Pickle两种格式的数据加载方式，详细描述了数据列的含义和观察布局。此外，还提到了数据集的许可证、引用格式以及数据集的用途和特点。

KlaskTron is a Klask imitation-learning dataset with three tiers: real human play, axis-mirror-augmented human data, and large-scale synthetic rollouts. The dataset provides two formats for data loading: Parquet and Pickle, with detailed descriptions of the data columns and observation layout. Additionally, it mentions the datasets license, citation format, and the datasets purpose and features.