12:32 < Trion> zoq: Should I also open a PR for the Evolution Agent files? and should we have a separate branch for Agent development?

13:04 < zoq> Trion: Yeah, sure open a PR for the Agent, and I would recommend to developed the agent in another branch in your own repo, makes the update process easier, but it doesn't matter for the PR.

13:05 < Trion> It looks quite complete :-) https://goo.gl/a3KdAe Generation 4! PewPew!

13:10 < zoq> cool, how many iterations?

13:19 < rcurtin> I couldn't get the video to load :(

13:20 < rcurtin> probably using a browser that doesn't work or my packages are out of date or something... sigh

13:35 < Trion> 5 parallel workers, they play until the episode ends, So 5x4 Episodes :P I think I should have a matric for iterations

13:37 < Trion> rcurtin: :P After I fix one last issue, you can run it on your own!

13:37 < zoq> Trion: That sounds good :)

13:38 < Trion> Best thing about OpenAI's ES algorithm is that it doesn't need backpropagation so you can go crazy even on crappy CPU!

13:40 < Trion> zoq: :( Creating an Environment* is not working

13:40 < zoq> Trion

13:40 < Trion> Environment* env = envs[i];

13:40 < zoq> ah

13:41 < zoq> do you have a minimal example?

13:43 < Trion> envs is a vector of Space-Invaders Environments, I want a reference to one of those. Earlier I made a new environment but 1 worker randomly died and when OpenMP started it again, it crashed the whole system

13:50 < zoq> that might have several reasons, maybe the PR give us some more hints?

13:52 < Trion> ok, PR incoming!

14:02 < Trion> zoq: https://github.com/zoq/gym_tcp_api/pull/3/files#diff-bd5226d202ba33440764a87ba77f5475R53

14:05 < zoq> thanks, I'll take a look later today