fsl.utils.fslsub

This module submits jobs to a computing cluster using FSL’s fsl_sub command line tool. It is assumed that the computing cluster is managed by SGE.

Example usage, building a short pipeline:

from fsl.utils.fslsub import submit, wait

# submits bet to veryshort queue unless <mask_filename> already exists
bet_job = submit('bet <input_filename> -m',
                 queue='veryshort.q',
                 output='<mask_filename>')

# submits another job
other_job = submit('some other pre-processing step', queue='short.q')

# submits cuda job, that should only start after both preparatory jobs are
# finished. This will work if bet_job and other_job are single job-ids
# (i.e., strings) or a sequence of multiple job-ids
cuda_job = submit('expensive job',
                  wait_for=(bet_job, other_job),
                  queue='cuda.q')

# waits for the cuda job to finish
wait(cuda_job)
submit Submits a given command to the cluster
info Gets information on a given job id
output Returns the output of the given job.
wait Wait for one or more jobs to finish
func_to_cmd Defines the command needed to run the function from the command line
fsl.utils.fslsub.submit(*command, minutes=None, queue=None, architecture=None, priority=None, email=None, wait_for=None, job_name=None, ram=None, logdir=None, mail_options=None, output=None, flags=False, multi_threaded=None, verbose=False)[source]

Submits a given command to the cluster

You can pass the command and arguments as a single string, or as a regular or unpacked sequence.

Parameters:
  • command – string or regular/unpacked sequence of strings with the job command
  • minutes – Estimated job length in minutes, used to auto-set queue name
  • queue – Explicitly sets the queue name
  • architecture – e.g., darwin or lx24-amd64
  • priority – Lower priority [0:-1024] default = 0
  • email – Who to email after job completion
  • wait_for – Place a hold on this task until the job-ids in this string or tuple are complete
  • job_name – Specify job name as it will appear on the queue
  • ram – Max total RAM to use for job (integer in MB)
  • logdir – where to output logfiles
  • mail_options – Change the SGE mail options, see qsub for details
  • output – If <output> image or file already exists, do nothing and exit
  • flags – If True, use flags embedded in scripts to set SGE queuing options
  • multi_threaded

    Submit a multi-threaded task - Set to a tuple containing two elements:

    • <pename>: a PE configures for the requested queues
    • <threads>: number of threads to run
  • verbose – If True, use verbose mode
Returns:

string of submitted job id

fsl.utils.fslsub.info(job_id)[source]

Gets information on a given job id

Uses qstat -j <job_id>

Parameters:job_id – string with job id
Returns:dictionary with information on the submitted job (empty if job does not exist)
fsl.utils.fslsub.output(job_id, logdir='.', command=None, name=None)[source]

Returns the output of the given job.

Parameters:
  • job_id – String containing job ID.
  • logdir – Directory containing the log - defaults to the current directory.
  • command – Command that was run. Not currently used.
  • name – Job name if it was specified. Not currently used.
Returns:

A tuple containing the standard output and standard error.

fsl.utils.fslsub.wait(job_ids)[source]

Wait for one or more jobs to finish

Parameters:job_ids – string or tuple of strings with jobs that should finish before continuing
fsl.utils.fslsub._flatten_job_ids(job_ids)[source]

Returns a potentially nested sequence of job ids as a single comma-separated string

Parameters:job_ids – possibly nested sequence of job ids. The job ids themselves should be strings.
Returns:comma-separated string of job ids
fsl.utils.fslsub.func_to_cmd(func, args, kwargs, tmp_dir=None, clean=False)[source]

Defines the command needed to run the function from the command line

WARNING: if submitting a function defined in the __main__ script, the script will be run again to retrieve this function. Make sure there is a “if __name__ == ‘__main__’” guard to prevent the full script from being rerun.

Parameters:
  • func – function to be run
  • args – positional arguments
  • kwargs – keyword arguments
  • tmp_dir – directory where to store the temporary file
  • clean – if True removes the submitted script after running it
Returns:

string which will run the function