gridjug package¶
Submodules¶
gridjug.grid_jug module¶
-
gridjug.grid_jug.
grid_jug
(jugfile, jugdir=None, jug_args=None, jug_nworkers=4, name='gridjug', keep_going=False, verbose=False, capture_jug_stdout=False, **kwargs)[source]¶ A light-weight wrapper to run Jug with GridMap on a Grid Engine cluster
From their own description, GridMap is a package that allows to easily create jobs on a Grid Engine powered cluster directly from Python. This wrapper lets GridMap simply spawn several jug-execute workers on a Grid Engine cluster. Thus we have the benefit of programmatic (reproducible) execution of Jug processes. Furthermore, GridMap adds a convenient monitoring and reporting layer. Under the hood, of course, Jug keeps doing the actual work.
Parameters: - jugfile (path) – Path to the jugfile
- jugdir (path) – Where to save intermediate results
- jug_args (list) –
Other jug command-line arguments. Note that
'execute'
is already included. The command line is roughly equivalent to:‘jug execute {jugfile} ‘ + ‘ ‘.join(jug_args) - jug_nworkers (int, optional) – number of Grid Engine tasks to start (i.e. number of times ‘jug execute’ is run)
- name (str, optional) – base name of the Grid Engine task
- keep_going (bool, optional) – Strongly recommended! Defaults to
False
: if a single Jug task fails, GridMap will cancel all jobs! IfTrue
, Jug does not raise an exception but keeps retrying the task. - verbose (bool, optional) – If
True
, Jug logsINFO
events - capture_jug_stdout (bool, optional) – Defaults to
False
. IfTrue
, captures Jug’s task summary printed to stdout. - **kwargs –
additional options passed through to
gridmap.grid_map
Module contents¶
Notes
For some reason, pickle does not work on some cluster environments if the
submodule has the same name as the package.
Hence, we call the submodule grid_jug
instead of gridjug
.