torchrl.trainers.algorithms.configs.utils.ASGDConfig

class torchrl.trainers.algorithms.configs.utils.ASGDConfig(lr: float = 0.01, lambd: float = 0.0001, alpha: float = 0.75, t0: float = 1000000.0, weight_decay: float = 0.0, foreach: bool | None = None, maximize: bool = False, differentiable: bool = False, capturable: bool = False, _target_: str = 'torch.optim.ASGD', _partial_: bool = True)

Hydra configuration for torch.optim.ASGD.

Every kwarg accepted by torch.optim.ASGD.__init__ is exposed as a field here.