Hi @Deathcalibur,
I’m using Learning Agents in UE 5.5, I’ve increased the Max Episode step num to 4096 potential need it higher because I’m trying to get my agents to navigate an environment for a while, but it seems to cap at 3333, is there a way to remove this limit?
Thanks,
James
Assuming you don’t have any completions which are firing, the only other thing is you need to set the max recorded steps higher perhaps? I usually set it to (agent num * step num)
Oh awesome, thank you for answering 
I have another question, is there a way to change the trainer settings in blueprints, I want to be able adjust the entrophy weight as time passes.
Currently no, this info is cached on the python side so editing it in the blueprint would not have the intended effect.
I see, thank you for your answer!
@Deathcalibur I’m having a strange issue when I’m running Inference, where the action values are really low and not reflecting the values of what I’m seeing when I’m training.
Run Inference range is really small
In training - range is normal -1 - 1
Do you know what might be wrong? I’ve never seen this before. When I print the values of the actions in the interactor they look normal in training but on inference they become small
Could it just be the action noise?
@Deathcalibur Oh I fixed it, I made a mistake of training as standalone, I thought it was action noise at first and played with that.
Was doing standalone as was giving me more speed but didn’t realise it wasn’t updating the policy.
Thank you for checking in with me
