OptimizerHook

class torchrl.trainers.OptimizerHook(optimizer: optim.Optimizer, loss_components: Sequence[str] | None = None)

為一或多個損失元件新增最佳化器。

引數:

• optimizer (optim.Optimizer) – 應用於 loss_components 的最佳化器。

• loss_components (Sequence[str], optional) – 損失 TensorDict 中用於對相應值應用最佳化器的鍵。如果省略，則最佳化器將應用於所有名稱以 loss_ 開頭的元件。

示例

>>> optimizer_hook = OptimizerHook(optimizer, ["loss_actor"])
>>> trainer.register_op("optimizer", optimizer_hook)

register(trainer, name='optimizer') → None

Registers the hook in the trainer at a default location.

引數:

• trainer (Trainer) – the trainer where the hook must be registered.

• name (str) – the name of the hook.

注意

To register the hook at another location than the default, use register_op().