koobas.hobune.stream
fixed FPS/update calls for learning neural network
Hello,
I am trying to make environment for training neural network. I don't need real time between frames to be always the same, but I need to be able to specify how many frames will be rendered in simulation second. So if I set 200 fps/update calls it can be processed over more than one second. Also I need physics to be evaluated (I am using wheel colliders). I was thinking about using time scale, but since I can not predict how much time will next frame require, I don't think that would work. Maybe downscale time and set target frame rate, but that still does not guarantee anything.
Thank you for any suggestion :)
