l5kit.planning.vectorized.open_loop_model module

class l5kit.planning.vectorized.open_loop_model.VectorizedModel(history_num_frames_ego: int, history_num_frames_agents: int, num_targets: int, weights_scaling: List[float], criterion: torch.nn.modules.module.Module, global_head_dropout: float, disable_other_agents: bool, disable_map: bool, disable_lane_boundaries: bool)

Bases: torch.nn.modules.module.Module

Vectorized planning model.

embed_polyline(features: torch.Tensor, mask: torch.Tensor) → Tuple[torch.Tensor, torch.Tensor]

Embeds the inputs, generates the positional embedding and calls the local subgraph.

Parameters

  • features – input features

  • mask – availability mask

Tensor features

[batch_size, num_elements, max_num_points, max_num_features]

Tensor mask

[batch_size, num_elements, max_num_points]

:return tuple of local subgraphout output, (in-)availability mask

forward(data_batch: Dict[str, torch.Tensor]) → Dict[str, torch.Tensor]

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

model_call(agents_polys: torch.Tensor, static_polys: torch.Tensor, agents_avail: torch.Tensor, static_avail: torch.Tensor, type_embedding: torch.Tensor, lane_bdry_len: int) → Tuple[torch.Tensor, Optional[torch.Tensor]]

Encapsulates calling the global_head (TODO?) and preparing needed data.

Parameters

  • agents_polys – dynamic elements - i.e. vectors corresponding to agents

  • static_polys – static elements - i.e. vectors corresponding to map elements

  • agents_avail – availability of agents

  • static_avail – availability of map elements

  • type_embedding –

  • lane_bdry_len –

training: bool