google-deepmind / acme

A library of reinforcement learning components and agents

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Implementing HER with Replay Buffer/Reverb

lhendre opened this issue · comments

Hi,
Im doing some experiments with a project utilizing acme. We wanted to do some additional experiments utilizing HER(Hindsight Experience Replay). I have been working on including that, both with our own infrastructure and trying to see if we can use Reverb to do this but I have been running into issues. I wanted to know if you are aware of examples that have implemented HER with acme?

As a quick addition, for the "with our own infrastructure", we are creating our own replay buffer and utilizing it in our own environment loop, and theres some small modifications in other files to support it but thats where the bulk of the work is done. For reverb, I was looking at modifying the dqn agent