
The examples which include recurrent layers in config.yaml do not train. They do one step of training then... get stuck not training.
Hi @TashaSkyUp , which platform are you using? Also are you referring to the Hallway not training? We've tested the Hallway ourselves on Windows, and it is training with the default setup.
Thanks. Actually I think the error was because I had changed the yaml file
@TashaSkyUp , can you tell me what you had changed in the yaml file? I seem to be having this same problem. Training works for a few iterations then the brain just spits all zeros (while not moving, action space value of "0" in my project should make a particular movement). Maybe my problem is unrelated, but could you share what you had done wrong and how you fixed it?
I think it was that the time_horizon was greater than buffer_size.. when its set that way at step buffer_size the errors begin..
Ooooh.... Yes, that makes sense... that's exactly what my issue was. Thank you.
This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.
Most helpful comment
I think it was that the time_horizon was greater than buffer_size.. when its set that way at step buffer_size the errors begin..