Where is distracting_wrapper? about agent-centric-representations HOT 3 CLOSED

Hi, you can find these two in the vd4rl repo here. We omitted these to provide the simplest possible implementation. Hope this helps!

from agent-centric-representations.

I am wondering how did you implement the distraction for Hard-Exo(c) (ie. Figure 1 of the paper). So, does it mean that you trained the agent using Hard-Exo(c) dataset but evaluate it on the original observation with distraction provided in vd4rl repo?

from agent-centric-representations.

Sorry about the delay in replying! Yes, all evaluations were on the standard set of distractions provided in the vd4rl repo. We created our own dataset for Hard-Exo (c), where one of the 9 agents corresponds to the controllable agent, and the rest are random agents, followed by downsampling this combined 3 x 3 rendering.

from agent-centric-representations.