Difference between revisions of "CMT Command Line Tools"

From CSDMS
m (More description on how to run the cmt command)
m (Alert about the debug queue)
Line 76: Line 76:
 
* The <tt>-C \$TMPDIR</tt> argument tells cmt to use a temporary folder on the compute node to hold all of your input and output data.
 
* The <tt>-C \$TMPDIR</tt> argument tells cmt to use a temporary folder on the compute node to hold all of your input and output data.
 
* The <tt>--launch-args="-q debug"</tt> argument puts your job into the debug queue so that it will start quickly.
 
* The <tt>--launch-args="-q debug"</tt> argument puts your job into the debug queue so that it will start quickly.
 +
 +
{{Alert Box|message=If you think your simulation will take more than 2 hours to complete you must not submit it to the debug queue. For longer runs, leave out the <tt>--launch-args</tt> option completely.
 +
}}
  
 
== The CMT Resource File ==
 
== The CMT Resource File ==

Revision as of 08:35, 3 May 2013

The CMT Command Line Tools

Warning 32.png Before beginning, you will need to have:
  • Set up an account on the CSDMS cluster, and be able to connect to it through an ssh client.

Load the CMT environment

Once logged into the CSDMS High Performance Computing Cluster, beach.colorado.edu, you will need to set up your environment to use the CMT command line tools. This is most easily done using modules.

$ module purge
$ module load cmtcl

This clears any previously loaded modules and then adds the cmt command to your path. If you have done this successfully, the following should print out a brief help message that describes some of the options for the cmt command,

$ cmt --help

Running a CMT Configuration

To run an example we first must prepare a CMT resource file. The resource file will tell CMT what components to use and how to connect them. The most basic way to run CMT on the command line is,

$ cmt my_resource_file.rc

where my_resource_file.rc is described in the next section. Usually this file is prepared by the CMT graphical user interface, however, it is a fairly simple format and can be created by-hand without too much effort. It is much easier to start with a template though so I will provide links to some sample resource files.


Warning 32.png Please note that these files are provided for reference and at this time are not guaranteed to even work. The probability of working generally decreases as one moves down the list. I hope to have them all working as soon as possible but, until then, beware!

Some sample resource files that run the Child model:

Some sample resource files that run the Sedflux3D model:

A connect Child-Sedflux simulation:

Running an Example

On beach.colorado.edu, I put some example rc files in the directory,

 /scratch/cmt/sample_rc_files

These are the same ones referenced above. In addition there are some sample input files in the directory,

 /scratch/cmt/sample_input_files

There are some input files for both Sedflux and Child. Note that the sample RC files reference these input files. If you want to use your own input files you will have to create your own copy of them and change your RC file to point to your own copy.

To run an example simulation, first create a folder that will hold your input and output and copy the sample RC file into this folder.

$ mkdir cmt_examples
$ cd cmt_examples
$ cp /scratch/cmt/sample_rc_files/child_00.rc .

Now (from your cmt_examples folder) you can run CMT,

$ cmt -C \$TMPDIR --launch-args="-q debug" child_00.rc

Your CMT configuration described by the child_00.rc resource file is now running on one of the compute nodes of the cluster. When it competes the input and output will be placed in sub folder of your cmt_examples folder, whose name will be the job id that the cmt command returned.

A couple notes:

  • The -C \$TMPDIR argument tells cmt to use a temporary folder on the compute node to hold all of your input and output data.
  • The --launch-args="-q debug" argument puts your job into the debug queue so that it will start quickly.
Warning 32.png If you think your simulation will take more than 2 hours to complete you must not submit it to the debug queue. For longer runs, leave out the --launch-args option completely.

The CMT Resource File

The CMT resource file tells the CMT what to do through three main commands:

  1. instantiate. Create an instance of a component.
  2. connect. Connect two components.
  3. parameters. Set a model parameter

There are other too, but these are the main ones.

instantiate

Instantiate a component of type class_name and call it instance_name. Later on in the file we will use the instance name when referring to this component.

instantiate <class_name> <instance_name>

Examples of instance names are edu.csdms.models.Child and edu.csdms.models.Sedflux3D. Instance names can be pretty much any old string. So, an example instantiate command would be,

instantiate edu.csdms.models.Child Child

connect

Connect the component instances user_name and provider_name. The two components will be connected through their named ports. All most always the user and provider port names will be the same but the need not be.

connect <user_name> <user_port> <provider_name> <provider_port>

As an example, the Child component provides information about its subaerial delta through a provider port called SubaerialDelta. The Sedflux3D component is able to (optionally) use data provided by such a port. Thus, these two components can be connected through this port with the following command,

connect Sedflux3D SubaerialDelta Child SubaerialDelta

In this case, the Child component provides the Sedflux3D component with information about sediment erosion and deposition rates on land.

parameters

For the component <instance_name>, set the configuration parameter <param_key> to be <param_value>.

parameters <instance_name> Configure <param_key> <param_value>