Mountain Car Continuous does not learn

[wilhem]

Hi guys,

I was trying Phil's code on my new server. The code he did is available here and here. Since I didn't change the code (the only change is the new gymnasium library, which requires this change here):

observation = env.reset()[0]

in the line: link.

The problem is: even if the code runs without any error, the score is accumulating to huge numbers (-100000) in the same episode. Since it doesn't trigger, it stucks in the first episode without going to the next.

Why it is happening?

Thanks