facebookresearch/BenchMARL

View on GitHub

BenchMARL is a library for benchmarking Multi-Agent Reinforcement Learning (MARL). BenchMARL allows to quickly compare different MARL algorithms, tasks, and models while being systematically grounded in its two core tenets: reproducibility and standardization.

Stars: 661
Forks: 135
Open beginner issues: 0
Indexed issues: 30

Dominant language: Python
License: MIT
Last GitHub push: Feb 7, 2026
Latest indexed: Sep 15, 2026
Contributing guide: Contributing guide
Code of conduct: Code of conduct
Beginner labels: good first issue help wanted
PR merge metrics: No merged PRs in 30d

Open All

30 open issues indexed

Suggestion of integrating HARL algorithms Open

facebookresearch/BenchMARL#52 · 1 comment · 1 reaction · 0 assignees · Indexed Sep 15, 2026
Request for Example of AEC API Usage with Agent Masking in Petting Zoo Open

facebookresearch/BenchMARL#76 · 6 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
[DO NOT CLOSE] Library TODOs and call for contributions Open

good first issue help wanted

facebookresearch/BenchMARL#94 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
How to integrate algo `Multi-Agent Transformer` in BenchMARL Open

facebookresearch/BenchMARL#117 · 9 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
[FAQ] GNNs frequent questions Open

question

facebookresearch/BenchMARL#157 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Fails to export policy Open

facebookresearch/BenchMARL#188 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
How to add something to the buffer Open

facebookresearch/BenchMARL#193 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Prroblem when creating new tasks from a new environment Open

facebookresearch/BenchMARL#194 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
No reward in Meltingpot tasks Open

facebookresearch/BenchMARL#196 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Custom environment Open

facebookresearch/BenchMARL#203 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Multi-agent environment with early agents termination Open

facebookresearch/BenchMARL#204 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Adding Level Based Foraging to BenchMARL. Open

facebookresearch/BenchMARL#205 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Tensormismatch restore_experiment Open

facebookresearch/BenchMARL#206 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Training from checkpoints / executing trained policies Open

facebookresearch/BenchMARL#207 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Probabilistic actors during evaluation Open

facebookresearch/BenchMARL#208 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Key 'wandb_extra_kwargs' not in 'ExperimentConfig' Open

facebookresearch/BenchMARL#209 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Adding MaMUJOCO to benchmarl Open

facebookresearch/BenchMARL#210 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
GNN critic error during training Open

facebookresearch/BenchMARL#219 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Data Ordering Consistency in MASAC Algorithm - Critical for Custom Critic Implementation Open

facebookresearch/BenchMARL#222 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
unity mlagent integration Open

facebookresearch/BenchMARL#234 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Support for custom environments Open

facebookresearch/BenchMARL#238 · 4 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Model weights fail to update while MLPs update with group_map on MeltingPot/predator_prey__orchard Open

facebookresearch/BenchMARL#242 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Proposing to add HATRPO and HAPPO algorithms Open

facebookresearch/BenchMARL#243 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Slowness when using RNN Open

facebookresearch/BenchMARL#246 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Compatibility and hidden state handling of RNN with Off-Policy Algorithms Open

facebookresearch/BenchMARL#248 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
MAPPO fails to converge on VMAS Balance task with default configuration Open

facebookresearch/BenchMARL#249 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Support for Offline MARL Open

facebookresearch/BenchMARL#250 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
MAPPO does not learn expected behavior on VMAS dispersion and passage Open

facebookresearch/BenchMARL#251 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Environment transform weights not loaded at eval time Open

facebookresearch/BenchMARL#254 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Is per-agent advantage normalization intentionally left out? Open

facebookresearch/BenchMARL#255 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026

facebookresearch/BenchMARL

Get new issues in your inbox