afdrl

Scalable Asynchronous Federated Deep Reinforcement Learning Simulator

about

AFDRL simulates a large-scale deterministic federation of clients seeking to maximize reward in an ALE game.

building

CMake is required to build afdrl.

    $ mkdir build
    $ cd build
    $ cmake -DCMAKE_BUILD_TYPE=Release ..
    $ make
    $ afdrl/afdrl

usage

To run a simulation of 64 participants, execute the following:

  $ afdrl/afdrl -c 64 --gpu