36bda8432b2ed23f623c8c3031b2cba148bec93b,rllib/policy/torch_policy.py,TorchPolicy,__init__,#,45

Before Change

        self.model = model.to(self.device)
        // Combine view_requirements for Model and Policy.
        self.training_view_requirements = dict(
            **{
                SampleBatch.ACTIONS: ViewRequirement(
                    space=self.action_space, shift=0),
                SampleBatch.REWARDS: ViewRequirement(shift=0),
                SampleBatch.DONES: ViewRequirement(shift=0),
            },
            **self.model.inference_view_requirements)

        self.exploration = self._create_exploration()

After Change

            self.device = torch.device("cpu")
        self.model = model.to(self.device)
        // Combine view_requirements for Model and Policy.
        self.view_requirements.update(self.model.inference_view_requirements)

        self.exploration = self._create_exploration()
        self.unwrapped_model = model  // used to support DistributedDataParallel
        self._loss = loss

Instances