6. Trying out MAST

1. Go to $HOME/MAST/examples (or $WORK/MAST/examples or a similar folder, if you moved the $HOME/MAST folder from its default location.)

2. Select one of the examples. The fastest one is simple_optimization.inp

3. Copy that file:

   cp simple_optimization.inp test.inp
   

4. Modify the test.inp file with the correct mast_exec, mast_ppn, mast_queue, mast_walltime, and other settings described in Input File

5. Try to parse the input file, entering the following command as one line:

   nice -n 19 mast -i test.inp
   

   • The nice -n 19 keeps this command low priority, since it is being run on the headnode (but it is not too intensive).
   • The -i signals to MAST that it is processing an input file.

6. Your $MAST_SCRATCH directory should now have a recipe directory in it.

   • The recipe directory will have a name corresponding to the elements and the input file, and ending with a timestamp of YYYYMMDD”T”hhmmss.
   • The recipe directory will contain several subfolders, which are ingredient directories.

7. Go to that recipe directory.

   • To see the input options:

     • cat input.inp (should be identical to test.inp since no looping was used)

       • Note that you can use other viewing commands, not just cat, but be careful not to edit any of these files.

     • cat archive_input_options.txt (should show Al instead of element X1)

   • To see information about the ingredient relationships MAST detected from the recipe template:

     • cat archive_recipe_plan.txt
     • Look at the $personal_recipe section in the input.inp file

   • To see ingredient statuses at a glance:

     • cat status.txt

8. Run mast once: nice -n 19 mast

9. You should see a “mastmon” job appear on the queue specified in $MAST_CONTROL/mastmon_submit.sh

10. MAST should have detected that the first ingredient was ready to run, so when that process disappears, run mast again: nice -n 19 mast

11. Now you should see perfect_opt1 appear on the queue.

12. status.txt in the recipe directory in $MAST_SCRATCH should show that perfect_opt1 has a status of “Proceed to Queue”, or “P”.

13. When the queued perfect_opt1 job starts running, you should be able to see output files inside $MAST_SCRATCH/<recipe directory>/perfect_opt1

14. If you forgot some step above, or you encounter some errors, remove the recipe folder from $MAST_SCRATCH and start again from the beginning of this section.

15. The $MAST_CONTROL folder gives you error messages and other information. See Running MAST for tips.

6.1. Unit testing

Unit tests are available through the MAST tar.gz file. See Programming for MAST. (Unit tests are not installed by default using pip.)

To check the validity of the MAST source code, navigate to <MAST installation directory>/MAST/test and run the unit tests with:

nosetests --exe

Some tests may have been designated to be skipped. Errors should be reported to the MAST development team as an issue on the github site (see Programming for MAST).