Hi,
I have been using ml-agents since 0.1, it’s very cool !
The question has maybe been already asked but I couldn’t find any road-map or something similar, I was wondering if there was any plan to allow training with the NN unity lib: Barracuda (e.g. requiring only Unity and not Python) ? I think RL is very under-estimated by game developers for example to automate animations.
I would like to do real time training during the game somehow with just a few layers allowing to learn with small amount of data (or even working around some off policy strategies such as imitation learning), it’s not feasible to have a Python running aside my Android app or something like that …