All Tools
B
AgentsFreeOpen Source
BENCHMARL
Standardized benchmarking library for multi-agent RL
MIT
ABOUT
MARL papers often mix algorithm code with environment wiring, so results cannot be compared fairly. BenchMARL separates algorithms, tasks, and models on TorchRL with shared configs so researchers can swap one axis and keep the rest of the experiment identical.
INSTALL
pip install benchmarlINTEGRATION GUIDE
1. Compare MAPPO, QMIX, or IPPO on the same VMAS or PettingZoo task
2. Run a reproducible MARL sweep from BenchMARL configs instead of custom scripts
3. Benchmark a new multi-agent algorithm against TorchRL baselines
4. Standardize logging and seeds across MARL experiments for a paper
TAGS
pythonpytorchmarlmulti-agentbenchmarktorchrl