Not sure which subforum to post this in, but given that ECS/DOTS is making the final preparations for 1.0 , can we get an indication from the new ml-agents team of what’s being planned (if anything) for integration of mlagents/barracuda and dots?
I remember a while back this question was asked fairly regularly and the answer was that there was an experimental update being worked on, but I haven’t heard any noises about it in a while.
There is an experimental ml-agents for dots package. It is pretty early in development, a few features are missing and academy has to be implemented by the user, but it works and I was able to train ~1k agents on the same scene (I have little experience with dots, so it is probably far from its limit)
I actually tried using their project but I get a lot of errors due different versions.
They use an old version of Entities and more packages that don’t correspond with their versions and I use the newest.
I hope that DOTS 1.0 will get some updates about that.