Parallel class function calls using python joblib

Question

It is possible to make multiple calls to a function in python using joblib.

from joblib import Parallel, delayed 

def normal(x):
    print "Normal", x
    return x**2

if  __name__ == '__main__':

    results = Parallel(n_jobs=2)(delayed(normal)(x) for x in range(20))
    print results

Gives: [0, 1, 4, 9, 16, 25, 36, 49, 64, 81, 100, 121, 144, 169, 196, 225, 256, 289, 324, 361]

However, what I really want is to call a class function on a list of class instances in parallel. The function simply stores a class variable. Then later I will access this variable.

from joblib import Parallel, delayed 

class A(object):
    def __init__(self, x):
        self.x = x
    def p(self):
        self.y = self.x**2

if  __name__ == '__main__':

    runs = [A(x) for x in range(20)]
    Parallel(n_jobs=4)(delayed(run.p() for run in runs))
    for run in runs:
        print run.y

This gives an error:

Traceback (most recent call last):

File "", line 1, in runfile('G:/My Drive/CODE/stackoverflow/parallel_classfunc/parallel_classfunc.py', wdir='G:/My Drive/CODE/stackoverflow/parallel_classfunc')

File "C:\ProgramData\Anaconda2\lib\site-packages\spyder\utils\site\sitecustomize.py", line 710, in runfile execfile(filename, namespace)

File "C:\ProgramData\Anaconda2\lib\site-packages\spyder\utils\site\sitecustomize.py", line 86, in execfile exec(compile(scripttext, filename, 'exec'), glob, loc)

File "G:/My Drive/CODE/stackoverflow/parallel_classfunc/parallel_classfunc.py", line 12, in Parallel(n_jobs=4)(delayed(run.p() for run in runs))

File "C:\ProgramData\Anaconda2\lib\site-packages\joblib\parallel.py", line 183, in delayed pickle.dumps(function)

File "C:\ProgramData\Anaconda2\lib\copy_reg.py", line 70, in _reduce_ex raise TypeError, "can't pickle %s objects" % base.name

TypeError: can't pickle generator objects

How is it possible to use joblib with classes like this? Or is there a better approach?

user3666197 user3666197 · Accepted Answer · 2018-06-05T15:57:19

How is it possible to use joblib with classes like this ?

Let's propose some code polishing first :

Not all things will fit the joblib.Parallel()( delayed() ) call-signature capabilities to swallow:

# >>> type( runs )                        <type 'list'>
# >>> type( runs[0] )                     <class '__main__.A'>
# >>> type( run.p() for run in runs )     <type 'generator'>

so, let's make the DEMO-objects to pass "through" aContainerFUN():

StackOverflow_DEMO_joblib.Parallel.py :

from sklearn.externals.joblib import Parallel, delayed
import time

class A( object ):

    def __init__( self, x ):
        self.x = x
        self.y = "Defined on .__init__()"

    def p(        self ):
        self.y = self.x**2

def aNormalFUN( aValueOfX ):
    time.sleep( float( aValueOfX ) / 10. )
    print ": aNormalFUN() has got aValueOfX == {0:} to process.".format( aValueOfX )
    return aValueOfX * aValueOfX

def aContainerFUN( aPayloadOBJECT ):
    time.sleep( float( aPayloadOBJECT.x ) / 10. )
    # try: except: finally:
    pass;  aPayloadOBJECT.p()
    print  "| aContainerFUN: has got aPayloadOBJECT.id({0:}) to process. [ Has made .y == {1:}, given .x == {2: } ]".format( id( aPayloadOBJECT ), aPayloadOBJECT.y, aPayloadOBJECT.x )
    time.sleep( 1 )

if __name__ == '__main__':
     # ------------------------------------------------------------------
     results = Parallel( n_jobs = 2
                         )(       delayed( aNormalFUN )( aParameterX )
                         for                             aParameterX in range( 11, 21 )
                         )
     print results
     print '.'
     # ------------------------------------------------------------------
     pass;       runs = [ A( x ) for x in range( 11, 21 ) ]
     # >>> type( runs )                        <type 'list'>
     # >>> type( runs[0] )                     <class '__main__.A'>
     # >>> type( run.p() for run in runs )     <type 'generator'>

     Parallel( verbose = 10,
               n_jobs  = 2
               )(        delayed( aContainerFUN )( run )
               for                                 run in runs
               )

Results ? Works as charm !

C:\Python27.anaconda> python StackOverflow_DEMO_joblib.Parallel.py

: aNormalFUN() has got aValueOfX == 11 to process.
: aNormalFUN() has got aValueOfX == 12 to process.
: aNormalFUN() has got aValueOfX == 13 to process.
: aNormalFUN() has got aValueOfX == 14 to process.
: aNormalFUN() has got aValueOfX == 15 to process.
: aNormalFUN() has got aValueOfX == 16 to process.
: aNormalFUN() has got aValueOfX == 17 to process.
: aNormalFUN() has got aValueOfX == 18 to process.
: aNormalFUN() has got aValueOfX == 19 to process.
: aNormalFUN() has got aValueOfX == 20 to process.
[121, 144, 169, 196, 225, 256, 289, 324, 361, 400]
.
| aContainerFUN: has got aPayloadOBJECT.id(50369168) to process. [ Has made .y == 121, given .x ==  11 ]
| aContainerFUN: has got aPayloadOBJECT.id(50369168) to process. [ Has made .y == 144, given .x ==  12 ]
[Parallel(n_jobs=2)]: Done   1 tasks      | elapsed:    2.4s
| aContainerFUN: has got aPayloadOBJECT.id(12896752) to process. [ Has made .y == 169, given .x ==  13 ]
| aContainerFUN: has got aPayloadOBJECT.id(12896752) to process. [ Has made .y == 196, given .x ==  14 ]
[Parallel(n_jobs=2)]: Done   4 tasks      | elapsed:    4.9s
| aContainerFUN: has got aPayloadOBJECT.id(12856464) to process. [ Has made .y == 225, given .x ==  15 ]
| aContainerFUN: has got aPayloadOBJECT.id(12856464) to process. [ Has made .y == 256, given .x ==  16 ]
| aContainerFUN: has got aPayloadOBJECT.id(50368592) to process. [ Has made .y == 289, given .x ==  17 ]
| aContainerFUN: has got aPayloadOBJECT.id(50368592) to process. [ Has made .y == 324, given .x ==  18 ]
| aContainerFUN: has got aPayloadOBJECT.id(12856528) to process. [ Has made .y == 361, given .x ==  19 ]
| aContainerFUN: has got aPayloadOBJECT.id(12856528) to process. [ Has made .y == 400, given .x ==  20 ]
[Parallel(n_jobs=2)]: Done  10 out of  10 | elapsed:   13.3s finished

Parallel class function calls using python joblib

1 Answers

Results ? Works as charm !