ScaRLib-experiments-startup

This repository is a startup template for ScaRLib, it shows how to organize a new project and how to manage dependencies.

Project structure

All the files needed to describe the experiment are in the src/main/scala/experiment folder. As described in ScaRLib, to run a learning the user must always define:

• The action space
• The reward function
• The state
• The neural network used to approximate the Q-function

Depending on if the user want to use the module alchemist-scafi or only the module scarlib-core some other components are needed:

• alchemist-scafi
  • The scafi logic
  • The alchemist specification
• scarlib-core
  • The environment

Finally, all these elements must be merged to create the learning system in the file <experiment-name>.scala

Preliminaries

ScaRLib relies on some third-party tools, namely:

• Gradle
• ScalaPy
• PyTorch and TensorBoard

ScalaPy

Depending on the underlying OS some further configuration may be required, a detailed guide can be found here.

Python tools

Requirements are in the file requirements.txt and can be installed using the command:

pip3 install -r requirements.txt

How to use

Once a new repository has been created from this template, the following steps need to be taken:

• The user must define his own custom components that describe the task on which he want to perform learning.

• The user must define a gradle task to run the learning and the evaluation. Some template tasks are provided in the file build.gradle.kts