==========
Benchmarks
==========

In order to ensure consistency of the implementations for various algorithm
as the library grows, there are a few automated benchmarks that are ran with
every new release.

.. _Deep Reinforcement Learning that Matters: https://arxiv.org/abs/1709.06560

Following th guidance from `Deep Reinforcement Learning that Matters`_ every
implementation is tested along various axes of variability:

    - Network architecture

    - Reward scale

    - Random seeds and trials

    - Environments

.. todo:
    embed all the benchmark plots