bug: expected np.ndarray (got tuple)


`state = torch.from_numpy(state).float().unsqueeze(0)` report this bug.

The input `state` is: 
![image](https://github.com/sweetice/Deep-reinforcement-learning-with-pytorch/assets/29271051/e788c8b9-a7f0-4e1b-b1f8-7cad1beb4429)

The torch version is: 1.11.0+cu113.

Any suggestion is appreciated~~~