Closed accuracy-maker closed 10 months ago
I'd advise you to reduce your code as much as possible and monitor your agent's internal values to find out what's causing the problem. I don't think it's coming from panda-gym. I think it's coming from your agent. The logs say that std is nan, so it must be that log_std is diverging.
I implement the A2C algorithm to train the Panda Gym
PandaReachDense-v3
env, but I gotnan
when I trained the model. I didn't use the SB3 because I wanted to implement from scratch. Below is my code:There is the error: The whole notebook link is here: https://github.com/accuracy-maker/robotics-tutorial/blob/main/Robotics_A2C.ipynb