ray.rllib.core.rl_module.rl_module.RLModule._forward#

RLModule._forward(batch: Dict[str, Any], **kwargs) → Dict[str, Any][source]#

Generic forward pass method, used in all phases of training and evaluation.

If you need a more nuanced distinction between forward passes in the different phases of training and evaluation, override the following methods instead:

For distinct action computation logic w/o exploration, override the self._forward_inference() method. For distinct action computation logic with exploration, override the self._forward_exploration() method. For distinct forward pass logic before loss computation, override the self._forward_train() method.

Parameters:

  • batch – The input batch.

  • **kwargs – Additional keyword arguments.

Returns:

The output of the forward pass.