search

inoryy / reaver

Reaver: Modular Deep Reinforcement Learning Framework. Focused on StarCraft II. Supports Gym, Atari, and MuJoCo.
MIT License
554 stars 89 forks source link

Issue with running on google colab #22

Closed jackylee1 closed 5 years ago

jackylee1 commented 5 years ago

the last step,it shows: ERROR: ld.so: object '/usr/lib/x86_64-linux-gnu/libtcmalloc.so.4' from LD_PRELOAD cannot be preloaded (cannot open shared object file): ignored.

inoryy commented 5 years ago

That's ok, if you look at the original colab output you'll see those errors as well. Happens because google's tcmalloc is uninstalled (conflicts with SC2), but the preload config isn't updated. screenshot from 2019-01-27 13-21-52

jackylee1 commented 5 years ago

but i didn't see any output

inoryy commented 5 years ago

Ok, I'll see if I can replicate

inoryy commented 5 years ago

@jackylee1 ok I was able to replicate, going to re-run whole notebook just in case, but for a quickfix replace last line with:

!python -m reaver.run --env MoveToBeacon --agent a2c --n_envs 2 --log_freq 25 --n_updates 5000 2> stderr.log
inoryy commented 5 years ago

I've noticed that some of my latest changes slowed down learning process on 1-4 envs, so I'm going to lock colab to the initial 2.0 version. I've updated the notebook and re-ran it, works fine. If you want to use the latest reaver version then use the command above, but note that training on colab could take up to 2 hours.