This page is about the standardization of model training for ASR within CSTR.

Next Meeting


Generate Python libraries to do various common things to allow training modules to be written in a transparent way. Provide mechanism to plug and unplug modules from your training pipeline. Provide a mechanism to archive both modules themselves and also results of running a particular module (or set of modules) on a particular data set, with provenance, and enough information to recreate.


The base location for the repository is

  • request an account
  • structure of repository

Keeping track of data and modules

Completed modules should be checked into the subversion repository at..

Users should work on their own modules / data paths in their own space, but when a set of results is arrived at that are worth archiving, these should be entered into a particular directory structure with some specific rules for declaring how things were run.

-- JonathanKilgour - 13 Oct 2011

Topic revision: r3 - 04 Nov 2011 - 15:46:56 - JonathanKilgour
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback
This Wiki uses Cookies