question

[yinshuangshuang621671]

May I ask a question: Why is the with torch.no_grad(): used in ppo.select_action ??

[GersiD]

If you don't use that then torch will accumulate gradients during the action selection process which can throw off the optimization step.