(self)
| 81 | to_torch(resets, dtype=torch.uint8, device=self.rl_device), []) |
| 82 | |
| 83 | def reset(self): |
| 84 | actions = 0.01 * (1 - 2 * np.random.rand(self.num_envs, self.num_actions)).astype('f') |
| 85 | |
| 86 | # step the simulator |
| 87 | obs, rewards, resets, extras = self.task.step(actions) |
| 88 | |
| 89 | return to_torch(np.clip(obs, -self.clip_obs, self.clip_obs), dtype=torch.float, device=self.rl_device) |
| 90 | |
| 91 | |
| 92 | # C++ GPU Class |