End of Episode

Development Programming & Scripting
, ,
1

Hi, I’m trying to increase the complexity of the environment at the end when all the agents reset after every 5 minutes or agent reward > 1,2,3, then 2, etc.

I found a hacky way using a branch so it will only run once on Agent Reset Episode but I was wondering if there’s a function or if I missed something.

image
image838×257 41.4 KB

And what’s the best way to get the average sum reward to use that as a value I can increase the difficulty against?

Thank you, 5.4 has been amazing with the observation workflow :slight_smile:

Thanks,
James

2

Every listener has a “OnAgentsReset” callback where you can inject functionality. It might be helpful. You can also handle in in your manager object maybe.

We haven’t really added any sort of curriculum learning API to help here, but the way we designed LA it seems like this should be doable and we don’t prevent you from doing it :man_shrugging:

Hopefully that answered your question if I understood what you are asking for.

Thanks,
Brendan