Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

https://github.com/enlite-ai/maze

Maze Applied Reinforcement Learning Framework
https://github.com/enlite-ai/maze

applied-machine-learning automation data-science decision-making deep-learning distributed documentation framework machine-learning monitoring optimization python reinforcement-learning simulation

Last synced: 2 months ago
JSON representation

Maze Applied Reinforcement Learning Framework

Lists

README 

        
![Banner](https://github.com/enlite-ai/maze/raw/main/docs/source/logos/main_logo.png)  



    Language grade: Python



![PyPI](https://img.shields.io/pypi/v/maze-rl)

![PyPI - Python Version](https://img.shields.io/pypi/pyversions/maze-rl)

[![Maze Docker Image](https://github.com/enlite-ai/maze/actions/workflows/github-ci.yml/badge.svg)](https://github.com/enlite-ai/maze/actions/workflows/github-ci.yml)

![Read the Docs](https://img.shields.io/readthedocs/maze-rl)

[![contributions welcome](https://img.shields.io/badge/contributions-welcome-brightgreen.svg?style=flat)](https://github.com/enlite-ai/maze/issues)



# Applied Reinforcement Learning with Python



MazeRL is an application oriented Deep Reinforcement Learning (RL) framework, addressing real-world decision problems.

Our vision is to cover the complete development life cycle of RL applications ranging from simulation 

engineering up to agent development, training and deployment.



*This is a preliminary, non-stable release of Maze. It is not yet complete and not all of our interfaces have settled

yet. Hence, there might be some breaking changes on our way towards the first stable release.*



## Spotlight Features



Below we list a few selected Maze features.



 - Design and visualize your policy and value networks with the 

   [Perception Module](https://maze-rl.readthedocs.io/en/latest/policy_and_value_networks/perception_overview.html). 

   It is based on PyTorch and provides a large variety of neural network building blocks and model styles. 

   Quickly compose powerful representation learners from building blocks such as: dense, 

   convolution, graph convolution and attention, recurrent architectures, action- and observation masking, 

   self-attention etc.

 - Create the conditions for efficient RL training without writing boiler plate code, e.g. by supporting 

   best practices like [pre-processing](https://maze-rl.readthedocs.io/en/latest/environment_customization/observation_pre_processing.html) and 

   [normalizing](https://maze-rl.readthedocs.io/en/latest/environment_customization/observation_normalization.html) your observations.

 - Maze supports advanced [environment structures](https://maze-rl.readthedocs.io/en/latest/concepts_and_structure/env_hierarchy.html) reflecting 

   the requirements of real-world industrial decision problems such as multi-step and multi-agent scenarios. 

   You can of course work with existing [Gym-compatible environments](https://maze-rl.readthedocs.io/en/latest/best_practices_and_tutorials/integrating_gym_environment.html).

 - Use the provided [Maze trainers](https://maze-rl.readthedocs.io/en/latest/trainers/maze_trainers.html) (A2C, PPO, Impala, SAC, Evolution Strategies), 

   which are supporting dictionary action and observation spaces as well as multi-step (auto-regressive policies) training. 

   Or stick to your favorite tools and trainers by [combining Maze with other RL frameworks](todo/best_practices_and_tutorials/maze_and_other_frameworks.html).

 - Out of the box support for advanced training workflows such as [imitation learning from teacher policies and 

   policy fine-tuning](https://maze-rl.readthedocs.io/en/latest/workflow/imitation_and_fine_tuning.html). 

 - Keep even complex application and experiment configuration manageable with the [Hydra Config System](https://maze-rl.readthedocs.io/en/latest/concepts_and_structure/hydra.html).

 

## Get Started



* Make sure [PyTorch](https://pytorch.org/get-started/locally/) is installed and then get the latest released version of Maze as follows:



      pip install -U maze-rl

      

  [Read more about other options](https://maze-rl.readthedocs.io/en/latest/getting_started/installation.html) like the installation of the latest 

  development version.  



  :zap:  We encourage you to start with **Python 3.7**, as many popular environments like Atari or Box2D can not easily 

  be installed in newer Python environments. Maze itself supports newer Python versions, but for Python 3.9 you might have to

  [install additional binary dependencies manually](https://maze-rl.readthedocs.io/en/latest/getting_started/installation.html)

* Alternatively you can work with Maze in a Docker container with pre-installed Jupyter lab: Run `docker run -p 8888:8888 enliteai/maze:playground` and open `localhost:8888` in your browser. This loads Jupyter 

* To see Maze in action, check out a [first example](https://maze-rl.readthedocs.io/en/latest/getting_started/first_example.html).

* [Try your own Gym env](https://maze-rl.readthedocs.io/en/latest/best_practices_and_tutorials/integrating_gym_environment.html)

  or visit our [Maze step-by-step tutorial](https://maze-rl.readthedocs.io/en/latest/getting_started/step_by_step_tutorial.html).



    

        Pip


        Installation

    

    

        First Example


        First Example

    

    

        Tutorial


        Step by Step Tutorial

    

    

        Documentation


        Documentation

    



* Clone this [project template repo](https://github.com/enlite-ai/maze-cartpole>) to start your own Maze project.



## Learn more about Maze



The [documentation](https://maze-rl.readthedocs.io/en/latest/index.html#documentation-overview) is the starting point to learn more about

  the underlying concepts, but most importantly also provides code snippets and minimum working examples to 

  get you started quickly.



* The *Workflow* section guides you through typical tasks in a RL project

  * [Training](https://maze-rl.readthedocs.io/en/latest/workflow/training.html)

  * [Rollouts](https://maze-rl.readthedocs.io/en/latest/workflow/rollouts.html)

  * [Collection and Visualizing Rollouts](https://maze-rl.readthedocs.io/en/latest/workflow/rollouts_trajectories_viewer.html)

  * [Imitation Learning and Fine-Tuning](https://maze-rl.readthedocs.io/en/latest/workflow/imitation_and_fine_tuning.html) 



* *Policy and Value Networks* introduces you to the 

  [Perception Module](https://maze-rl.readthedocs.io/en/latest/policy_and_value_networks/perception_overview.html), 

  how to [customize action spaces and the underlying action probability distributions](https://maze-rl.readthedocs.io/en/latest/policy_and_value_networks/distributions_and_action_heads.html) 

  and two styles of policy and value networks construction:

  

  * [Template models](https://maze-rl.readthedocs.io/en/latest/policy_and_value_networks/perception_template_models.html) 

    are composed directly from an environment's observation and action space,

    allowing you to train with suitable agent networks on a new environment within minutes.

    

  * [Custom models](https://maze-rl.readthedocs.io/en/latest/policy_and_value_networks/perception_custom_models.html) gives you the full 

    flexibility of application specific models, either with the provided Maze building blocks or 

    directly with PyTorch. 



* Learn more about core *concepts and structures* such as the 

  [Maze environment hierarchy](https://maze-rl.readthedocs.io/en/latest/concepts_and_structure/env_hierarchy.html), the 

  [Maze event system](https://maze-rl.readthedocs.io/en/latest/concepts_and_structure/event_system.html) providing a convenient way to collect 

  statistics and KPIs, enable flexible reward formulation and supporting offline analysis. 



* [Structured Environments and Action Masking](https://maze-rl.readthedocs.io/en/latest/best_practices_and_tutorials/struct_env_tutorial.html) 

  introduces you to a general concept, which can greatly improve

  the performance of the trained agents in practical RL problems.



## License



Maze is freely available for research and non-commercial use. A commercial license is available, if interested please 

contact us on our [company website](https://enlite.ai) or write us an [email](mailto:[email protected]).



We believe in Open Source principles and aim at transitioning Maze to a commercial Open Source project, 

releasing larger parts of the framework under a permissive license in the near future.