simrun¶

The kinds of the run layer as task types of simrun.

simrun runs the tasks of a workflow on a cluster; this module registers the thirteen kinds of python -m sbmlsim.run (sbmlsim.run) as its task types sbmlsim.<kind>, entry points of the group simrun.task_types in the pyproject.toml of sbmlsim, see the plugins of simrun. The settings of a task are the spec of its kind, and plan is the plan of the kind, which simrun's client calls before it submits a task.

A splittable kind (fit, identifiability, scan, sensitivity, uncertainty) is split: simrun's workers compute its units with sbmlsim.run.compute_unit and its merge merges them with sbmlsim.run.merge. Every other kind (simulate, fisher, report, petab, adopt, factory, credibility, release) runs as the command (simrun-exec runs it, {settings}, {out} and {input} filled by simrun)

python -m sbmlsim.run <kind> {settings} --out {out}/out
    --summary {out}/summary.json --resume --input={input}

in which --resume lets a retry after an eviction continue the run it finds and starts a run in an empty output.

A task writes the output of its run into its output port out, the directory out/out of the task (OUTPUT), which a task downstream reads as the input <task>.out, e.g. the fit of an identifiability analysis. The run, the command of an unsplit task or the merge of a split one, also writes its summary to out/summary.json, the summary of the task which simrun compare reads.

The resource hints are the maximum resident memory of a run on one worker (/usr/bin/time -v, python 3.14 on linux, the HCTZ project examples.hctz_fitting.project, --workers 1) times 1.5, rounded up to 0.5 GiB: the fit of the subset PK with 4 repeats 0.36 GiB, the scan dose 0.37 GiB, the sensitivity sobol 0.38 GiB, the uncertainty cl 0.38 GiB, the simulation of the three experiments 0.38 GiB and the report of the fit with its figures 0.48 GiB, so 1 GiB per worker for every kind. The build of a release (release) runs the analyses of the release in its own processes and is hinted alike.

The kind verify is no task type: it compares the dumps of two environments, a task type comes when a workflow needs it. The base memory is the process of simrun's worker besides the processes of its units, 0.10 GiB measured with this module imported, 0.5 GiB.

This module imports simrun; import sbmlsim and sbmlsim.run do not import it, only simrun's loading of the entry points and the tests do.

simulate_plan(settings)

fit_plan(settings)

fit_count_units(settings)

fit_compute_unit(settings, index, paths)

fit_merge(settings, paths)

identifiability_plan(settings)

identifiability_count_units(settings)

identifiability_compute_unit(settings, index, paths)

identifiability_merge(settings, paths)

fisher_plan(settings)

report_plan(settings)

adopt_plan(settings)

factory_plan(settings)

credibility_plan(settings)

release_plan(settings)

petab_plan(settings)

scan_plan(settings)

scan_count_units(settings)

scan_compute_unit(settings, index, paths)

scan_merge(settings, paths)

sensitivity_plan(settings)

sensitivity_count_units(settings)

sensitivity_compute_unit(settings, index, paths)

sensitivity_merge(settings, paths)

uncertainty_plan(settings)

uncertainty_count_units(settings)

uncertainty_compute_unit(settings, index, paths)

uncertainty_merge(settings, paths)