Found bug in pytorch reinforcement tutorials

I can reproduce the issue on Colab, while the tutorial runs on my local machine.
Could you please open an issue here?