reference/generated/torchrl.trainers.algorithms.configs.hooks.TdMpc2OptimizationStepperConfig

Run in Google Colab

Colab

Download Notebook

Notebook

View on GitHub

GitHub

torchrl.trainers.algorithms.configs.hooks.TdMpc2OptimizationStepperConfig

class torchrl.trainers.algorithms.configs.hooks.TdMpc2OptimizationStepperConfig(loss_module: Any = None, optimizer_model: Any = None, optimizer_actor: Any = None, target_tau: float = 0.01, zero_grad_set_to_none: bool = True, _target_: str = 'torchrl.trainers.algorithms.tdmpc2.TdMpc2OptimizationStepper')

Configuration for the TD-MPC2 optimization stepper.

The stepper owns the model and actor optimizers and performs the staged model, policy, and target-Q updates.

Examples

>>> from hydra.utils import instantiate
>>> from torchrl.trainers.algorithms.configs import (
...     TdMpc2OptimizationStepperConfig,
... )
>>> configured_stepper = instantiate(  
...     TdMpc2OptimizationStepperConfig(
...         loss_module=loss_module,
...         optimizer_model=optimizer_model,
...         optimizer_actor=optimizer_actor,
...     )
... )

See also TdMpc2OptimizationStepper.