Method reset

unihsi/env/tasks/vec_task.py:133–139

(self)

        return torch.clamp(self.task.obs_buf, -self.clip_obs, self.clip_obs).to(self.rl_device), self.task.rew_buf.to(self.rl_device), self.task.reset_buf.to(self.rl_device), self.task.extras

    def reset(self):
        actions = 0.01 * (1 - 2 * torch.rand([self.task.num_envs, self.task.num_actions], dtype=torch.float32, device=self.rl_device))

        # step the simulator
        self.task.step(actions)

        return torch.clamp(self.task.obs_buf, -self.clip_obs, self.clip_obs).to(self.rl_device)