Aasdfip/rollouts_po2_train

该数据集包含多个字段，如步骤(step)、epsilon(eps)、目标(target)、上一次选择(last_choice)、当前选择(curr_choice)、地图概率(mpp)、图像(image)、机器人位置(agent_pos)、机器人朝向(agent_heading)、里程表(odomenter)、距离目标(distance_to_goal)、目标位置(goal_pos)、拒绝选项(rejected)、前沿(frontiers)、前沿ID(frontier_ids)、目标返回(target_return)、图像序列(images)、选中选项(chosen)、提示(prompt)等。数据集分为训练集(train)，共有39065个样本，大小为4453654121.64字节。此外，还提供了默认配置(default)，指定了训练数据的文件路径。

The dataset contains multiple fields such as step, epsilon (eps), target, last choice (last_choice), current choice (curr_choice), map probability (mpp), image, robot position (agent_pos), robot orientation (agent_heading), odometer, distance to goal (distance_to_goal), goal position (goal_pos), rejected options (rejected), frontiers, frontier IDs (frontier_ids), target return (target_return), image sequence (images), chosen options (chosen), and prompt. The dataset is split into a training set (train) with a total of 39,065 samples and a size of 4,453,654,121.64 bytes. Additionally, a default configuration is provided, specifying the file path for the training data.