lilkm/pick_cube_octo_qc_fql_reembed

该数据集是一个机器人学习相关的数据集，包含了30个episodes，519帧图像，用于单一任务的训练。数据集中的文件以Parquet格式存储，并包括相应的视频文件。数据集特征包括行动、奖励、是否完成、惩罚、前端图像、手腕图像、状态、时间戳、帧索引、集索引、索引、任务索引和行动嵌入等。

This dataset is a robotics learning dataset containing 30 episodes and 519 frames of images for training a single task. The files in the dataset are stored in Parquet format and include corresponding video files. The dataset features include action, reward, done, penalty, front image, wrist image, state, timestamp, frame index, episode index, index, task index, and action embedding.