Train dataset Difference with Diffusion-Planner

[charriry]

Thank you for your open-source contributions, but I have encountered a problem.

in Diffusion-Planner code

data = {
            "map_name": map_name,
            "token": token,
            "ego_current_state": ego_current_state,
            "ego_agent_future": ego_agent_future,
            "neighbor_agents_past": neighbor_agents_past,
            "neighbor_agents_future": neighbor_agents_future,
            "static_objects": static_objects
        }

but in this project,I noted :

data = NuPlanDataSample(
            batched=False,
            ego_past=ego_agent_past,
            ego_current=ego_current_state,
            ego_future=ego_agent_future,
            neighbor_past=neighbor_agents_past,
            neighbor_future=neighbor_agents_future,
            neighbor_future_observed=neighbor_future_observed,
            lanes=lanes,
            lanes_speedlimit=lanes_speed_limit,
            lanes_has_speedlimit=lanes_has_speed_limit,
            routes=route_lanes,
            routes_speedlimit=route_lanes_speed_limit,
            routes_has_speedlimit=route_lanes_has_speed_limit,
            map_objects=static_objects
        )

I noticed that the data in Diffusion-planner is missing a field ego_agent_past compared to Flow-Planner, which is preventing the training process from proceeding.Then in ./flow_planner/data/dataset/nuplan line 245 ,using ego_agent_past = torch.from_numpy(data['ego_agent_past']) ,
Does this mean that data from Diffusion cannot be used directly within Flow-Planner?