micahr234/play_ns_cartpole

该数据集包含强化学习环境中的交互数据，用于训练和评估智能体。特征包括环境名称（env_name）、环境索引（env_idx）、全局步数（global_step）、回合步数（episode_step）、动作（action）、奖励（reward）、变换后的奖励（xformed_reward）、完成标志（done）、观测值（observation）以及元数据如质量杆标志（metadata_masspole_flag）、质量杆值（metadata_masspole）、重力标志（metadata_gravity_flag）、重力值（metadata_gravity）和q星值（metadata_q_star）。数据分为训练集（1000万个示例）和评估集（20万个示例），总大小约为1.4GB。This dataset contains interaction data from reinforcement learning environments, used for training and evaluating agents. Features include environment name (env_name), environment index (env_idx), global step (global_step), episode step (episode_step), action (action), reward (reward), transformed reward (xformed_reward), done flag (done), observation (observation), and metadata such as masspole flag (metadata_masspole_flag), masspole value (metadata_masspole), gravity flag (metadata_gravity_flag), gravity value (metadata_gravity), and q-star value (metadata_q_star). The data is split into a training set (10 million examples) and an evaluation set (200,000 examples), with a total size of approximately 1.4GB.