mwhittaker / deeprl_project

Deep RL Final Project
1 stars 1 forks source link

try ppo on many CPUs at once with MPI #30

Open vlad17 opened 6 years ago

mwhittaker commented 6 years ago

Merge #22

I just realized that the baselines MPI stuff is if you have multiple GPUs, not for multiple instances.

So we really don't want to log every instance as we do here, it should really only log if rank == 0 and other wise the log should be disabled, as we had before.

My mistake.

SamvitJ commented 6 years ago

Deliverable: time to reward -15

SamvitJ commented 6 years ago

MPI configurations/commands: https://github.com/openai/baselines/tree/master/baselines/ppo1 Our file is called src/run_atari_ppo.py instead of run_atari.py

vlad17 commented 6 years ago

Regarding #22 / the logging part, check out the branch mpi-cpu, I already did that part.

This is also blocked on #16

vlad17 commented 6 years ago

See #33 . Note that my other prs should be merged first.