isip_run has two modes: 1) Server(Default) 2) Cluster.
isip_run -p param.sof [arguments] [files]
or
isip_run -mode cluster -cpus N -command_to_execute "some command" -walltime 01:00:00 -list list.sof 

isip_run -help;
isip_run -parameter_file params.sof -list identifiers.sof -identifier_list identifier_list.sof -machine_list machine_list.sof -accumulator_list accumulator_list.sof [-database_file machine_database.sof] [-verbose brief]
isip_run -mode cluster -cpus 63 -command_to_execute "isip_transform -param param.sof -verbosity all" -list list.sof -walltime 09:00:00

This utility is used for submitting jobs in parallel, we can submit jobs both to the servers as well as a cluster. It automatically splits a list into individual jobs and submits it to the cluster. In the cluster mode we can also run unix commands in parallel.

• -help: displays a help message
• -mode: choose the mode of operation: Server or Cluster (Default: Server)
  
  If mode is Server we can use the following options:
  
  
• -verbose: log verbosity level
  • NONE: output the number of files processed
  • BRIEF: output also names of the files processed
  • DETAILED: output also hypotheses, scores and frames number
  
  
• -split_mode: identifier split mode (NONE, SPEED, MEMORY)
• -parameter_file params.sof: parameters file
• -list identifiers: input file list
• -machine_list machine_list.sof: list of machine to distribute the jobs
• -accumulator_list accum_list.sof: list of accumulator files (note: there should be a unique accumulator file for each machine in the machine list)
• -log_list log_list.sof: list of log files corresponding to each machine
• -identifier_list: list of identifier files (note: there should be a unique identifier file for each machine in the machine list)
• -database_file database.sof: database that specifies the machine attributes
  
  For Cluster mode we can use the following options:
  
  
• -cpus: this specifies the number of cpus required
• -command_to_execute: "some command"
• -list: A list file which has to be split and submitted to the cluster with the appropriate command.
• -walltime: specifies the maximum time we expect a single job to run

• file1.raw file2.sof *.* ... : speech files or any other file format which will work with the command given by the user.

• This example shows how to use isip_run in the server mode. Server mode is used only for parallel training on servers:

  isip_run -parameter_file params.sof -list identifiers.sof -identifier_list identifier_list.sof -machine_list machine_list.sof -accumulator_list accumulator_list.sof [-database_file machine_database.sof] [-verbose brief]

• This example shows how to use the cluster mode. Cluster mode is more flexible as we can do training, decoding, feature extraction and run shell commands in parallel:

  isip_run -mode cluster -cpus 63 -command_to_execute "isip_transform -param param.sof -verbosity all" -list list.sof -walltime 09:00:00"
  

None.

This utility is written entirely in C++.

None.

Command line options always override parameter file values.