search

shariqiqbal2810 / MAAC

Code for "Actor-Attention-Critic for Multi-Agent Reinforcement Learning" ICML 2019
MIT License
645 stars 169 forks source link

multi-agent particle environments #7

Closed SHYang1210 closed 5 years ago

SHYang1210 commented 5 years ago

when i run your multi-agent particle environments,the error: Traceback (most recent call last): File "/home/cherry/multiagent-particle-envs-master/bin/interactive.py", line 26, in env.render() File "/home/cherry/anaconda3/envs/shyang/lib/python3.6/site-packages/gym/core.py", line 108, in render raise NotImplementedError NotImplementedError

shariqiqbal2810 commented 5 years ago

Do your package versions match the ones listed in the README?

SHYang1210 commented 5 years ago

My pytorch is 0.4.0 and python is 3.6.1. The rest of the requirements are installed through PIP. I also feel that there is a problem with the version. Is it possible that my gym is the latest version installed directly by PIP?

shariqiqbal2810 commented 5 years ago

Yeah if you run pip install gym it will install the most recent version by default. Try pip install gym==0.9.4

shariqiqbal2810 commented 5 years ago

I've never tested simple_crypto.py, so I am not sure what the problem is there. It's possible that I broke something in that environment when making the modifications that I did.

What parts of main.py do you not understand? You need to pass in the name of the environment and a name for your saved model and all other parameters are optional. For example, this would run the multi_speaker_listener environment and save the model with the name "test" for 25000 episodes (with all other parameters at their defaults).

python main.py multi_speaker_listener test --n_episodes 25000

SHYang1210 commented 5 years ago

I used the command you gave me to appear in the console of pycharm: Episodes number-number of number form. Models folder will have a multi_speaker_listener folder.

shariqiqbal2810 commented 5 years ago

It sounds like you have it running properly!

shariqiqbal2810 commented 5 years ago

Yeah it doesn't render the environment during training since that would slow things down significantly. After training, you can load the saved model (I provide a method on the model class to load from a saved file) and render the environment to see how it's doing.

shariqiqbal2810 commented 5 years ago

I am going to go ahead and close this issue, since issues are generally created for reporting bugs in the code. My advice to you would be to read the code carefully, and it should answer most of your questions.