Running CellProfiler on computing clusters. This is aimed towards the University of Edinburgh's Eddie3 cluster with the staging workflow.
python setup.py install --user
For more details see the wiki.
cptools2 can use a configuration file to list the jobs parameters.
For example if we have a a file named awesome_experiment-1.yml containing:
experiment : path/to/imageXpress/experiment
chunk : 46
pipeline : /path/to/cellprofiler/pipeline.cppipe
location : /path/to/scratch/space
commands location : /home/userWe could run this as cptools2 awesome_experiment-1.yml
This produces a directory containing a loaddata file for each task, and three text files containing staging commands, cellprofiler commands, and de-staging commands that can be run as three concurrent array jobs.
A script named YYYY-MM-DD-h:m:s_SUBMIT_JOBS.sh is created which will submit
the three jobs in the correct order. This can be run with
./YYYY-MM-DD-h:m:s_SUBMIT_JOBS.sh or bash YYYY-MM-DD-h:m:s_SUBMIT_JOBS.sh
on a login node (the ones that can submit jobs to the queue).
There are configuration details you can add to a job:
experiment: path to an ImageXpress experimentchunknumber of imagesets per jobpipeline: path to a cellprofiler pipelinelocation: path to where to store the loaddata modules, staged data and outputcommands location: path to where to store the qsub array commandsremove plate: plate names inexperimentto be removedadd plate:experiment: path to another ImageXpress experimentplates: plate name(s) in the above experiment
new_ix: if true/yes then will treat filepaths as from the new ImageXpress
i.e we could remove some plates from an experiment, and also include some plates from a different experiment
experiment: /path/to/imageXpress/experiment
# remove two plates from current experiment
remove plate:
- plate_1
- plate_2
# add plates from a separate experiment
add plate:
- experiment : /path/to/different/experiment
- plates:
- plate_exp2_1
- plate_exp2_2
chunk: 96
pipeline: /pipelines/pipeline_1.cppipe
location: /path/to/scratch/space
commands location: /path/to/scratch spaceWe can also use add plate without a top-level experiment tag to add a few plates
from a large experiment.
add plate:
- experiment: /path/to/large/experiment
- plates:
- plate_1
- plate_2
- plate_3
chunk : 46
pipeline : /path/to/cellprofiler/pipeline.cppipe
location : /path/to/scratch/space
commands location : /home/userNOTE: The default is expecting old ImageXpress filepaths (for backwarsd
compatibility), in order to correctly parse metadata from the new IX paths you
can specify in the config file with new_ix.
experiment : path/to/imageXpress/experiment
chunk : 46
pipeline : /path/to/cellprofiler/pipeline.cppipe
location : /path/to/scratch/space
commands location : /home/user
new_ix: truePrevious version for the AFM filesystem is available here.