Question about deploy on my own dataset and settings.

[Promethe-us]

Thanks for your great work! I am trying to deploy this on my own setting (an realaman arm + an inspire hand).
I meed a misunderstanding when prepare the datsdet:
-During infering, you use [t0, t0+1] as obs to predict [t0+1, t0+16] action, and infer all 16 actions. How should I prepare my dataset? obs in [t0~t0+16] and action in [t0+1, t0+17] ??

My dataset code:

    def _sample_to_data(self, sample): # np.ndarray
        data = {
            'obs': {
                'target_object_pc_bodycoord': sample['target_object_pc_bodycoord'][1:], # (16)
                'current_object_pc_bodycoord': sample['current_object_pc_bodycoord'][1:], # (16)
                'current_hand_pc_bodycoord': sample['current_hand_pc_bodycoord'][1:], # (16)
                'current_robot_table_pc_bodycoord': np.concatenate([sample['current_robot_pc_bodycoord'], sample['current_table_pc_bodycoord']], axis=1)[1:], # (16)
                'ont_vector': sample['ont_vector'][1:], # (16)
                'state_jointangle': sample['state_jointangle'][:-1], # (16)
            },
            'action': sample['state_jointangle'][1:],
            # 'delta_action': sample['state_jointangle'][1:]-sample['state_jointangle'][:-1]
        }
        return data