The hwrt toolkit makes use of a configuration file. This file has to be in the home folder of the user an it has to be called .hwrtrc.

The configuration file is in YAML format. The possible values are:

  • root: This is a required configuration entry. Its value must be a path. hwrt will look for all configuration files in this path.
  • nntoolkit: The name of the executable in your path that does neural network training
  • preprocessing: A path to a Python script that contains your preprocessing classes. Have a look at the official preprocessing classes to see how they should be structured.
  • features: Just like preprocessing, this has to be a path to a Python script.
  • environment: Either 'production' or 'development'.

There are 3 configurations that are probably only interesting for me:

  • dbconfig: Only important if you want to access a MySQL db to get the data
  • dropbox_app_key and dropbox_app_secret: Only important if you want to upload data to DropBox


The following is an example ~/.hwrtrc configuration file:

root: /home/moose/GitHub/hwr-experiments
nntoolkit: nntoolkitfancyname
preprocessing: /home/moose/hwrt-config/
features: /home/moose/hwrt-config/
dbconfig: /home/moose/hwrt-config/db.config.yml
dropbox_app_key: 'INSERT_APP_KEY'
dropbox_app_secret: 'INSERT_APP_SECRET'
environment: 'production'

Table Of Contents

Previous topic


Next topic

Handwritten Data

This Page