Dear authors,

Thanks for the brilliant research.

I have something I'd like you to tell me. Just write an issue for the first time.

I'm in trouble because I can't train Centralized.
Firstly, what is the difference in implementation(algorithm) between Centralized and Single?
Both network takes N agents actions and observations as input. Single network's output is one but Centralized network's output is N. Single type is the same as GAIL in the paper. Is my understanding correct?

What is the difference in performance between Centralized and Decentralized?
I want to tell me why the results of simple_tag and simple_push differ between D and C.

I want to use MAGAIL for our research.
Thanks.

Move irl folder to gail. Probably need to refactor some names.

This could pave the way for Lantao's stuff.

Move the MACK code from sandbox to RL folder, it seems easier to understand.

Hi, I got this error when I typed "python -m sandbox.mack.run_simple" in the multiagent-Gail folder.

Traceback (most recent call last):
File "C:\Users\Seonghyeon.conda\envs\cv\lib\runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "C:\Users\Seonghyeon.conda\envs\cv\lib\runpy.py", line 85, in _run_code
exec(code, run_globals)
File "D:\github_test\multiagent-gail\multiagent-gail\sandbox\mack\run_simple.py", line 63, in
main()
File "C:\Users\Seonghyeon.conda\envs\cv\lib\site-packages\click\core.py", line 764, in call
return self.main(*args, **kwargs)
File "C:\Users\Seonghyeon.conda\envs\cv\lib\site-packages\click\core.py", line 717, in main
rv = self.invoke(ctx)
File "C:\Users\Seonghyeon.conda\envs\cv\lib\site-packages\click\core.py", line 956, in invoke
return ctx.invoke(self.callback, **ctx.params)
File "C:\Users\Seonghyeon.conda\envs\cv\lib\site-packages\click\core.py", line 555, in invoke
return callback(*args, **kwargs)
File "D:\github_test\multiagent-gail\multiagent-gail\sandbox\mack\run_simple.py", line 58, in main
train(logdir + '/exps/mack/' + env_id + '/l-{}-b-{}/seed-{}'.format(lr, batch_size, seed),
TypeError: must be str, not NoneType

Dear authors,

Hi. Thank you for sharing your codes.

Recently, I've been interested in MAGAIL and tried to reproduce your results.
Firstly, I tried to train expert policy as recommended in README.md by python -m sandbox.mack.run_simple, but I failed. I thought there is a problem on action dimension, so I made all actions as multi-hot vectors and modified relevant terms. After training with MACK, however, it seems like agents cannot recover the appropriate policies similar to MADDPG.

So I wonder whether it is possible to share the weight files of expert so that readers can simply generate expert trajectories.

Thanks.

On running the code "python -m sandbox.mack.run_simple" from the terminal in a conda environment, I am getting the error, No module named 'make_env'.
The same error comes when I am running the code "python -m irl.mack.run_mack_gail [discrete]".