Shouldn't you reset z at the beginning of each episode? Otherwise, after the first steps it will also update the weights of the latest steps of the previous episode.
From your post:
https://miro.medium.com/max/1400/1*G3bXDd1QeeUphO0RyQOgVA.png
Thanks for this amazing repository! :)
Shouldn't you reset z at the beginning of each episode? Otherwise, after the first steps it will also update the weights of the latest steps of the previous episode.
From your post:
https://miro.medium.com/max/1400/1*G3bXDd1QeeUphO0RyQOgVA.png
Thanks for this amazing repository! :)