IMP logo
IMP Reference Guide  develop.cb469e0f11,2024/05/28
The Integrative Modeling Platform
IMP.parallel.Manager Class Reference

Manages workers and contexts. More...

Inherits object.

Detailed Description

Manages workers and contexts.

This class is only available in Python.

Definition at line 370 of file parallel/

Public Member Functions

def __init__
 Constructor. More...
def add_worker
 Add a Worker object. More...
def get_context
 Create and return a new Context in which tasks can be run. More...

Constructor & Destructor Documentation

def IMP.parallel.Manager.__init__ (   self,
  python = None,
  host = None,
  output = 'worker%d.output' 


pythonIf not None, the command to run to start a Python interpreter that can import the IMP module. Otherwise, the same interpreter that the manager is currently using is used. This is passed to the shell, so a full command line (including multiple words separated by spaces) can be used if necessary.
hostThe hostname that workers use to connect back to the manager. If not specified, the manager machine's primary IP address is used. On multi-homed machines, such as compute cluster headnodes, this may need to be changed to allow all workers to reach the manager (typically the name of the machine's internal network address is needed). If only running local workers, 'localhost' can be used to prohibit connections across the network.
outputA format string used to name worker output files. It is given the numeric worker id, so for example the default value 'worker%d.output' will yield output files called worker0.output, worker1.output, etc.

Definition at line 391 of file parallel/

Member Function Documentation

def IMP.parallel.Manager.add_worker (   self,

Add a Worker object.

Definition at line 417 of file parallel/

def IMP.parallel.Manager.get_context (   self,
  startup = None 

Create and return a new Context in which tasks can be run.

startupIf not None, a callable (Python function or class that implements the __call__ method) that sets up the worker to run tasks. This method is only called once per worker. The return values from this method are passed to the task object when it runs on the worker.
A new Context object.

Definition at line 427 of file parallel/

The documentation for this class was generated from the following file: