torchrl.trainers.algorithms.configs.trainers.TdMpc2TrainerConfig#
- class torchrl.trainers.algorithms.configs.trainers.TdMpc2TrainerConfig(collector: Any, total_frames: int | None, loss_module: Any = None, world_model: Any = None, policy_prior: Any = None, q_ensemble: Any = None, planner: Any = None, optimizer_model: Any | None = None, optimizer_actor: Any | None = None, optimization_stepper: Any | None = None, replay_buffer: Any = None, batch_size: int | None = None, logger: Any = None, save_trainer_file: Any = None, optim_steps_per_batch: int = 1, frame_skip: int = 1, seed_pretrain_steps: int = 0, clip_grad_norm: bool = True, clip_norm: float | None = 20.0, progress_bar: bool = True, seed: int | None = None, save_trainer_interval: int = 10000, log_interval: int = 10000, num_epochs: int = 1, async_collection: bool = False, log_timings: bool = False, auto_log_optim_steps: bool = True, learner_backend: str = 'local', learner_backend_options: dict[str, Any] | None = None, learner_poll_interval: float = 0.05, lr: float = 0.0003, encoder_lr_scale: float = 0.3, actor_eps: float = 1e-05, weight_decay: float = 0.0, capturable: bool | None = None, target_tau: float = 0.01, zero_grad_set_to_none: bool = True, hooks: list[Any] | None = None, checkpoint: Any = None, checkpoint_rotation: Any = None, checkpoint_metadata: Any = None, _target_: str = 'torchrl.trainers.algorithms.configs.trainers._make_tdmpc2_trainer')[source]#
Hydra configuration for
TdMpc2Trainer.See also