VeraIsHere/sokoban_processed

此文件夹包含从https://huggingface.co/datasets/Xiaofeng77/sokoban派生的Sokoban数据集的处理版本。数据被打包以便快速本地加载和视觉检查。文件包括`train.parquet`（3,982行）和`test.parquet`（1,602行）；对应的PNG文件位于`images/`目录中。列包括：`data_source`（原始Hugging Face数据集的来源分割/配置，如`sokoban_6x6_1horizon`）、`prompt`（用于引出Sokoban解决方案的聊天式多轮提示）、`ability`（创建样本时使用的启发式/求解器，如`bfs`）、`reward_model`（包含评估信息的小字典，如`ground_truth`动作索引和`style`）、`extra_info`（环境元数据，如网格大小、初始网格布局、箱子/玩家起始位置、最大步数、渲染模式）、`id`（唯一示例标识符）、`images`（包含单个base64编码的PNG字符串的数组，用于板渲染）和`image_path`（`images/`目录中保存的PNG的绝对路径）。

This folder contains a processed version of the Sokoban dataset derived from https://huggingface.co/datasets/Xiaofeng77/sokoban. The data is packaged for fast local loading and visual inspection. Files: `train.parquet` (3,982 rows) and `test.parquet` (1,602 rows); corresponding PNGs live in `images/`. Columns: `data_source` (Source split/config from the original Hugging Face dataset, e.g., `sokoban_6x6_1horizon`), `prompt` (Chat-style, multi-turn prompt used to elicit a Sokoban solution), `ability` (Heuristic/solver used when creating the sample, e.g., `bfs`), `reward_model` (Small dict with evaluation info such as `ground_truth` action indices and `style`), `extra_info` (Environment metadata such as grid size, initial grid layout, box/player start positions, max steps, render mode), `id` (Unique example identifier), `images` (Array with a single base64-encoded PNG string for the board rendering), and `image_path` (Absolute path to the saved PNG inside `images/`).