easyagents 1.0.1

reinforcement learning for practitioners.

Reinforcement Learning for Practitioners (v1)

Status: under active development, breaking changes may occur

EasyAgents is a high level reinforcement learning api, written in Python and running on top of OpenAI gym using algorithms implemented in tf-Agents, OpenAI baselines and huskarl.

Use EasyAgents if

• you have implemented your own environment and want to experiment with it
• you want try out different libraries and algorithms, but don't want to learn the details of each implementation
• you are looking for an easy and simple way to get started with reinforcement learning

Try it on colab:

• Cartpole on colab introduction: training, plotting & switching algorithms. based on the classic reinforcement learning example balancing a stick on a cart.
• Berater on colab example of a custom environment & training. gym environment based on a routing problem.
• LineWorld on colab implement your own environment, workshop example [work in progress]

In collaboration with Oliver Zeigermann.

Scenario: simple (quick test, plot state)

from easyagents.agents import PpoAgent
from easyagents.callbacks import plot, duration

ppoAgent = PpoAgent('CartPole-v0')
ppoAgent.train([plot.State(), duration.Fast()])

Scenario: more detailed (custom network & training)

from easyagents.agents import PpoAgent
from easyagents.callbacks import plot, duration

ppoAgent = PpoAgent( 'Orso-v1', fc_layers=(500,500,500))
ppoAgent.train(learning_rate=0.0001, [plot.State()], num_iterations = 500, max_steps_per_episode = 50)

Guiding Principles

• easily train, evaluate & debug policies for (you own) gym environment over "designing new algorithms"
• simple & consistent over "flexible & powerful"
• inspired by keras:
  • same api across all algorithms
  • support different implementations of the same algorithm

Design ideas

• separate "public api" from concrete implementation using a frontend / backend architecture (inspired by scikit learn, matplotlib, keras)
• pluggable backends
• extensible through callbacks (inspired by keras). separate callback types for training, evaluation and monitoring
• pre-configurable, algorithm specific train & play loops

Installation

Install from pypi using pip:

pip install easyagents

Vocabulary

some terms explained

Don't use EasyAgents if

• you would like to leverage implementation specific advantages of an algorithm
• you want to do distributed or in parallel reinforcement learning

Note

• This repository is under active development and in an early stage. Thus any- and everything may (and probably should) change.
• If you have any difficulties in installing or using easyagents please let us know. We'll try to do our best to help you.
• Any ideas, help, suggestions, comments etc in python / open source development / reinforcement learning / whatever are more than welcome. Thanks a lot in advance.