mateoguaman/language_table_blocktoblock_oracle_sim

该数据集是通过LeRobot v3.0转换自Google Research的Language Table项目中的language_table_blocktoblock_oracle_sim数据集。原始数据是从Google Cloud Storage流式传输，并转换为LeRobot v2.0格式，保留了所有字段（动作、状态、效应器目标、奖励、完成、指令），然后无损视频拼接为v3.0格式。数据集包含200,000个episodes，12,770,620帧，896个任务，帧率为10fps。数据以parquet文件格式存储，视频以mp4格式存储。数据集的结构包括元数据文件info.json，详细描述了数据集的版本、机器人类型、总集数、总帧数、总任务数、块大小、帧率、分割、数据路径、视频路径以及各种特征（如观察图像、状态、效应器目标翻译、动作、奖励、完成等）的详细信息。

This dataset was created using LeRobot v3.0 conversion of the language_table_blocktoblock_oracle_sim dataset from the Language Table project (Google Research). The original RLDS data was streamed from Google Cloud Storage, converted to LeRobot v2.0 format preserving all fields (action, state, effector target, reward, done, instruction), then restructured to v3.0 with lossless video concatenation. The dataset contains 200,000 episodes, 12,770,620 frames, 896 tasks, with a frame rate of 10fps. The data is stored in parquet files, and videos are stored in mp4 format. The dataset structure includes a metadata file info.json, which details the dataset version, robot type, total episodes, total frames, total tasks, chunk size, frame rate, splits, data path, video path, and various features (such as observation images, state, effector target translation, action, reward, done, etc.).