Our great sponsors
-
Mava
🦁 A research-friendly codebase for fast experimentation of multi-agent reinforcement learning in JAX
-
epymarl
An extension of the PyMARL codebase that includes additional algorithms and environment support
-
InfluxDB
Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.
I'm new to MARL and I'm looking for some open source implementations that I could use in a project. I have some previous experience in single agent RL, mainly with SB3 and gym, but just now started reading some MARL papers. I'm mainly looking for a good balance between performance, good documentation and ease of use. So far, I've taken look at Mava and RLlib. Mava seems like a very complete option, though I'm not at all familiar with the API and it maybe something simpler could also do the trick. As for the environment library, I was considering PettingZoo, since it has a very similar api to gym. Thought I might as well ask here first, as people can suggest other options for me to investigate or even give me some pros and cons they have learned from past experience.
If you're thinking of pettingzoo, you should try https://github.com/uoe-agents/epymarl. I use it actively for my work.
Recently dm-acme also added support for multi-agent environments. Acme: https://github.com/deepmind/acme
Related posts
- Starting wth Multi Agent Reinforcement Learning
- How much of a MuJoCo simulation or real life robot can you train on a 3090?
- Is there a way to get PPO controlled agents to move a little more gracefully?
- Worthwhile to convert custom env to be dm_env compatible?
- [D] Physics and Reinforcement Learning - Discussion of Deepmind's work