How can I use 'pd_ee_delta_pose' controller mode in interactive_panda.py

[Lebingelee]

Hello! I want to use pd_ee_delta_pose controller in order to collect demos. But in the example folders: mani_skill/examples/teleoperation/interactive_panda.py indicates that only supporting "pd_joint_pos", "pd_joint_pos_vel"
I tried to modify the env setting in following:

env = gym.make(
        args.env_id,
        obs_mode=args.obs_mode,
        control_mode=`pd_ee_delta_pose` # "pd_joint_pos" as original one,
        render_mode="rgb_array",
        reward_mode="none",
        enable_shadow=True,
        viewer_camera_configs=dict(shader_pack=args.viewer_shader)
    )

However, consoles output the following infomation:

File "xxx/xxx/envs/Maniskill_env/lib/python3.10/site-packages/mani_skill/agents/controllers/base_controller.py", line 325, in set_action
    assert action.shape == (
AssertionError: Received action of shape torch.Size([15]) but expected shape (1, 7)

[StoneT2000]

To collect pd ee delta pose data you can either

1. Use the original code and then replay to pd_ee_delta_pose actions with the replay tool
2. Rewrite the interactive panda code to allow pd ee delta pose control, and instead of using the motion planner to move the robot to target poses, you write your own code to generate pd ee delta pose actions give a target ee pose and the current ee pose.