Chapter 4 - 01 cartpole.py: ValueError: probabilities do not sum to 1 about deep-reinforcement-learning-hands-on-second-edition HOT 2 CLOSED

Interesting. I'm currently working on porting examples to torch 1.7, but this example worked for me.

Is the error consistent (so, you're getting this error every time you run the example)?
This error means some of values produced by NN are too large, which might happen when something really bad happened with the training process (i.e. optimisation has diverged), but I've never seen this in this example - it normally just converge very quickly.

So, if this is something happened once - just ignore this :)

from deep-reinforcement-learning-hands-on-second-edition.

Hi @bezineb5 ,

Please reopen if the issue still persists.

Thanks,

from deep-reinforcement-learning-hands-on-second-edition.