e4b6611cb73ef7658f028831be1aa6bd85ecbed0,src/garage/torch/policies/stochastic_policy.py,StochasticPolicy,get_action,#StochasticPolicy#Any#,14

Before Change

            observation = observation.unsqueeze(0)
            dist, info = self.forward(observation)
            return dist.sample().squeeze(0).cpu().numpy(), {
                k: v.squeeze(0).detach().cpu().numpy()
                for (k, v) in info.items()
            }

    def get_actions(self, observations):

After Change

                    * np.ndarray[float]: Standard deviation of logarithmic
                        values of the distribution.
        
        if not isinstance(observation, np.ndarray) and not isinstance(
                observation, torch.Tensor):
            observation = self._env_spec.observation_space.flatten(observation)
        with torch.no_grad():
            if not isinstance(observation, torch.Tensor):
                observation = torch.as_tensor(observation).float().to(
                    global_device())

Instances