• Login to a machine at Edinburgh, student.compute for instance.
• Inside the header file for the script, header.emtrain.test, there are variables RUN_EMACC and RUN_EMTRAIN. Set them both to "./migrate -f"
• There is a similar variable, RUN_DISTRIBUTED, for the decoding script header, header.vit.test - change that too.

You'll probably need to change the path to Perl in the first line of migrate. There were other changes required to make it work at Edinburgh, migrate_edin is in CVS for reference. There's more documentation on migrate.

• • To run as a nice user, use the -nice_user option to migrate_edin

There's an alternative way of doing this, using Peter Bell's Python API to Condor. First, check the API out from CVS (the module is named condor_python_API). Then things are similar to using migrate:

• Inside the header file for the script, header.emtrain.test, set both RUN_EMACC and RUN_EMTRAIN to "condor_python_API/examples/parallelRunCommands.py ./ws06_edin_template.submit < ". I've attached ws06_edin_template.submit to this wiki page since it seems like the least confusing place to put it. You'll need to edit two lines there as described in the comments.
• Do the same to RUN_DISTRIBUTED in decoding script header header.vit.test

If you want the very latest version of the API you could apply for access to a CVS repository at Edinburgh called gmtk_tools (but there probably won't be any changes in the near future, I've been told).

-- ParthaLal - 10 Jul 2006

Attachment	Action	Size	Date	Who	Comment
ws06_edin_template.submit	manage	1.5 K	11 Jul 2006 - 13:52	ParthaLal