Skip to content

how to distribute stitching over slurm

You can launch the stitching on a computer having access to the slurm sbatch command then you can distribution stitching over slurm.

The stitching configuration file contains a dedicated slurm section which allow you to define the distribution to be done.

It contains the following keys:

  • partition: name of the partition to use for launching the jobs
  • mem: memory to allocate for each job.
  • n_jobs: define on how many jobs the stitching should be splitted.
  • python_venv: python virtual environment to source before applying stitching. It must contain nabu.

  • modules: modules to be loaded before applying stitching. One must contain nabu. If provided then python_venv should be empty. Examples:

    modules = "tomotools"

    modules = ("tomotools/dev", "second_module")

Distribution will be done as follow (presudo-code):

compute_shifts [locally or remotly] (option locally will enforce to do this calculation locally)
for each part of the stitching:
    apply_stitching_on_this_part()
concatenate_the_different_part  [locally or remotly] (option locally will enforce to do this calculation locally)