search

dkkim93 / meta-mapg

Source code for "A Policy Gradient Algorithm for Learning to Learn in Multiagent Reinforcement Learning" (ICML 2021)
MIT License
32 stars 4 forks source link

2-Agent HalfCheetah code #1

Closed SigmaBM closed 2 years ago

SigmaBM commented 2 years ago

Hi, when will you upload the 2-Agent HalfCheetah code?

I implemented this part myself, but got 100 average reward for agent i using Meta-MAPG, only half of that shown in your paper. Maybe I missed some details. So I'm looking forward to your source code. Thank you!

dkkim93 commented 2 years ago

Hello. Thank you for your interest in our paper. :) The performance difference can be resulted due to using a different meta-population. I will upload the source code along with the meta-population that we used in the paper by the end of this week or next week! Thank you.

dkkim93 commented 2 years ago

Hello. I have uploaded the 2-Agent HalfCheetah code to the main branch. Please refer to the updated README.md for further details on running the 2-Agent HalfCheetah experiment. Please feel free to re-open this issue if you may have any questions. Have a great day! :)