[Dataset Replay] Request for per-task simulation environment configurations and replay support
Thank you for releasing InternData-A1. Would it be possible to open-source the per-task simulation environment configurations, including the corresponding assets, robot/controller settings, camera parameters, task definitions, initial object states, and a mapping between dataset episodes and simulation environments? A replay script would be especially helpful for reproducing the released trajectories and obtaining additional information such as depth, point clouds, object poses, segmentation masks, and contact states. This would support research on 3D world models, controlled action perturbations, and counterfactual rollout generation. Could you also clarify whether the released episodes can currently be replayed with InternDataEngine and whether the required configurations or tools are planned for release?
Please use InternDataEngine and SIM1, which consitute the overall data generation process.
我现在想从发布的数据集 回溯每个任务对应的skills字段配置
比如https://huggingface.co/datasets/InternRobotics/InternData-A1/tree/main/sim_updated_lerobotv30/basic_tasks/lift2中flip_the_package这个任务用的是
期待您的回答