Skip to content.

-- KarenLivescu - 15 Dec 2005

WS06 > ComputeResources > DirectoryStructure
-- SimonKing - 13 Jul 2006

Directory structure

The directory structure is the same at all sites (JHU, EDI, UIUC, UW) and these sites should be kept in sync as often as practical - see SyncFilesBetweenSites for instructions.


/export/ws06afsr/                - the top level directory at JHU
                 data            - static data (PLPs, waveforms, video, etc)
                 GMTK            - GMTK source code (do not execute binaries from here!)
                 laboratory      - stuff from the pre-workshop summer school
                 packages        - external packages such as sclite
                 projects        - the main workspace: see below
                 rsync.sh        - the script for synchronising sites
                 src             - the CVS root (do not go here!)
                 src_checked_out - a checked out version of src (see below)
                 users           - personal stuff you want to access at all sites

Paths at the different sites

The path to the top level is as follows:

  • JHU /export/ws06afsr/
  • EDI /group/cstr/projects/dbns/export/ws06afsr/
  • UIUC //cworkspace/ifp-32-1/hasegawa/jhu06export/ws06afsr/
  • UW ???/export/ws06afsr/

Project space

The project space is organised into projects and subprojects - here's a snapshot

analysis
audio_only/
          svitchboard
audiovisual/
          AVICAR
          cuave_baselines
          bin
observation_modeling/
                    generative
                    hybrid
                    tandem
pronunciation_modeling
SLT_paper

The idea is that a small subset of people are working within each (sub)project directory. They should agree amongst themselves when to run the synchronisation script.

Using the shared src code

I suggest that you do not run scripts/binaries from /export/ws06afsr/src_checked_out. Instead, within your (sub)project directory, you should check out your own copy of the source code. Only ever do this on the JHU system: the CVS files themselves are not kept in sync across the different sites.

By having your own copy of the src, you get to control when it is updated (e.g. not in the middle of an experiment run). You can check out whichever modules of the src you need, using the instruction at CvsDocumentation.