Have we found the baselines to be reproducible?

[e-ndorfin]

Running the baseline ippo_ff_mpe.py gives me pretty terrible returns (running this with simple_spread means reward is just the distance between agents and landmarks) as I'm converging on -60 reward. Are there baseline metrics recorded to compare to?

[amacrutherford]

Hi, if you have a look at Figure 16 of our Appendix on Arxiv baselines for this env are reported.