We include a non-exhaustive number of examples for Jax-based systems, showing common use-cases for Mava. We also have a quickstart notebook that can be used to quickly create and train your first multi-agent system.

We also have various Jax-based examples which make use of the callback design paradigm. Our pre-implemented Jax-based systems are continually expanding so please check back often to see new implemented systems.

• IPPO: an IPPO system running on the discrete action space simple_spread MPE environment.
  • Feedforward
    • decentralised
    • decentralised record agents (recording agents acting in the environment).
    • decentralised single process (running in single process mode).
    • decentralised restore checkpoint (continuing training by restoring from an existing checkpoint).
    • decentralised evaluation intervals (perform evaluation at custom intervals for custom durations)

• IPPO: an IPPO system running on the discrete action space flatland environment.
  • Feedforward
    • decentralised

• IPPO: an IPPO system running on the discrete action space Cooperative pong MPE environment.
  • Feedforward
    • decentralised
    • decentralised (recording agents acting in the environment).

• IPPO: an IPPO system running on the Simple Spread environment.
  • Feedforward
    • decentralised

• IPPO: an IPPO system running on the discrete action space 3m SMAC environment.
  • Feedforward
    • decentralised
    • decentralised evaluation intervals (perform evaluation at custom intervals for custom durations)