3. Getting Started¶
In this section we will briefly describe how RepEx can be invoked, how input and resource configuration files should be used.
3.1. Invoking RepEx¶
To run RepEx users need to use a command line tool corresponding to MD package
kernel they intend to use. For example, if user wants to use Amber as MD kernel,
she would use repex-amber
command line tool. In addition to specifying an
appropriate command line tool, user need to specify a resource configuration file
and REMD simulation input file. The resulting invocation of RepEx should be:
repex-amber --input='tuu_remd_ace_ala_nme.json' --rconfig='stampede.json'
where:
--input=
- specifies the REMD simulation input file
--rconfig=
- specifies resource configuration file
Both REMD simulation input file and resource configuration file must conform to JSON format.
3.1.1. Resource configuration file¶
In resource configuration file must be provided the following parameters:
resource
- this is the name of the target machine. Currently supported machines are:
local.localhost
- your local system
xsede.stampede
- Stampede supercomputer at TACC
xsede.supermic
- SuperMIC supercomputer at LSU
xsede.comet
- Comet supercomputer at SDSC
xsede.gordon
- Gordon supercomputer at SDSC
epsrc.archer
- Archer supercomputer at EPCC
ncsa.bw_orte
- Blue Waters supercomputer at NCSA
username
- your username on the target machine
project
- your allocation on specified machine
cores
- number of cores you would like to allocate
runtime
- for how long you would like to allocate cores on target machine (in minutes).
In addition are provided the following optional parameters:
queue
- specifies which queue to use for job submission. Values are machine specific.cleanup
- specifies if files on remote machine must be deleted. Possible values are:True
orFalse
Example resource configuration file for Stampede supercomputer might look like this:
{
"target": {
"resource" : "xsede.stampede",
"username" : "octocat",
"project" : "TG-XYZ123456",
"queue" : "development",
"runtime" : "30",
"cleanup" : "False",
"cores" : "16"
}
}
3.1.2. REMD input file for Amber kernel¶
For use with Amber kernel, in REMD simulation input file must be provided the following parameters:
re_pattern
- this parameter specifies Replica Exchange Pattern to use, options are:S
- synchronous andA
- asynchronousexchange
- this parameter specifies type of REMD simulation, for 1D simulation options are:T-REMD
,S-REMD
andUS-REMD
number_of_cycles
- number of cycles for a given simulationnumber_of_replicas
- number of replicas to useinput_folder
- path to folder which contains simulation input filesinput_file_basename
- base name of generated input/output filesamber_input
- name of input file templateamber_parameters
- name of parameters fileamber_coordinates
- name of coordinates filereplica_mpi
- specifies ifsander
orsander.MPI
is used for MD-step. Options are:True
orFalse
replica_cores
- number of cores to use for MD-step for each replica, ifreplica_mpi
isFalse
this parameters must be equal to1
steps_per_cycle
- number of simulation time-stepsdownload_mdinfo
- specifies if Amber .mdinfo files must be downloaded. Options are:True
orFalse
. If this parameter is ommited, value defaults to “True”download_mdout
- specifies if Amber .mdout files must be downloaded. Options are:True
orFalse
. If this parameter is ommited, value defaults to “True”
Optional parameters are specific to each simulation type. Example REMD simulation input file for T-REMD simulation might look like this:
{
"remd.input": {
"re_pattern": "S",
"exchange": "T-REMD",
"number_of_cycles": "4",
"number_of_replicas": "16",
"input_folder": "t_remd_inputs",
"input_file_basename": "ace_ala_nme_remd",
"amber_input": "ace_ala_nme.mdin",
"amber_parameters": "ace_ala_nme.parm7",
"amber_coordinates": "ace_ala_nme.inpcrd",
"replica_mpi": "False",
"replica_cores": "1",
"min_temperature": "300",
"max_temperature": "600",
"steps_per_cycle": "1000",
"download_mdinfo": "True",
"download_mdout" : "True",
}
}