I noticed that the use_motion_hand_actions parameter is configured differently in the pnp_ground and pnp_table scenes.
Could you please explain the reason behind this difference? Is there any limitation or design consideration that prevents enabling use_motion_hand_actions=True when training the pnp_table scene?
I am interested in training pnp_table with infered hand actions from the released checkpoint and would like to know whether this setting is supported.
I noticed that the use_motion_hand_actions parameter is configured differently in the pnp_ground and pnp_table scenes.
Could you please explain the reason behind this difference? Is there any limitation or design consideration that prevents enabling use_motion_hand_actions=True when training the pnp_table scene?
I am interested in training pnp_table with infered hand actions from the released checkpoint and would like to know whether this setting is supported.