Quick Start Guide

This quick start guide gives an overview of the development workflow using this editor.

Installation

Follow the same installation procedure as other beat packages to setup a conda environment ready for development. Activate the conda environment (e.g. named “beat”).

$ conda activate beat
(beat) $

Note

In these instructions (beat) $ corresponds to your own shell prompt, which can be different. The actual command(s) to type comes after the $ marker.

All further instructions assume you have the environment where BEAT packages are installed properly activated.

Work flow

The BEAT editor allows you to implement and test your solution locally on either a reduced or different dataset. Once you have everything working as expected, push the resulting code to the platform and execute it there, for validation.

The following steps describe the procedure to start from an existing experiment and build on that.

1. Get (or pull) the reference experience you would like to base your work on:

   (beat) $ beat pull <experiment-author>/<experiments>/<toolchain-author>/<toolchain-name>/<toolchain-version>/<experiment>
   

   This will pull the experiment as well as all its dependencies from the platform so that it can be edited (and run) locally.

   Note

   Values within angle brackets (< and >) will depend on the asset name. Replace these values by appropriate actual values.

2. Create your own fork of the experiment:

   (beat) $ beat experiment fork <experiment-author>/<experiments>/<toolchain-author>/<toolchain-name>/<toolchain-version>/<experiment> <your-beat-user-id>/<experiments>/<toolchain-author>/<toolchain-name>/<toolchain-version>/<experiment-name>
   

3. Create your own fork from an algorithm you want to modify:

   (beat) $ beat algorithm fork <author>/<algorithm>/<version> <your-beat-user-id>/<algorithm>/<version>
   

   This will allow you to use your own modified algorithm for your fork.

   Note

   You will have to update your experiment fork in order to use your own algorithm(s).

   Note

   You do not need to fork an algorithm, you can use a different one as long as its inputs and outputs match the one you want to replace.

4. Edit the code:

   (beat) $ beat algorithm edit <author>/<algorithm>/<version>
   

   This will open your default configured editor to edit the code of the algorithm passed in parameter.

5. To edit any other asset, just start the editor, optionally pointing to the prefix of interest:

   (beat) $ beat --prefix <path/to/prefix> editor start
   

   This will start the BEAT editor so that you will have an easy access to all locally available assets. You’ll be able to edit them visually. The editor also allows to start your favorite editor for code or documentation modification.

6. Once the editing is done, you can run an experiment locally like this:

   $ beat exp run <experiment-author>/experiments/<toolchain-author>/<toolchain-name>/<toolchain-version>/<experiment-name>
   

7. Once the experiment runs successfully on your machine, you can upload it to the online platform and run it there.

   $ beat exp push <experiment-author>/experiments/<toolchain-author>/<toolchain-name>/<toolchain-version>/<experiment-name>
   $ beat exp start <experiment-author>/experiments/<toolchain-author>/<toolchain-name>/<toolchain-version>/<experiment-name>
   

   You can monitor the execution of your experiment at any moment with:

   $ beat exp monitor <experiment-author>/experiments/<toolchain-author>/<toolchain-name>/<toolchain-version>/<experiment-name>