def retrieveLikelihoods(gene_rhos, back_rhos, off_rhos, off_rate, like_fun,
nums, gene_span, I_step):
# Checks if single processor
if nums == 1:
single = True
else:
single = False
# Calculates gaussian kernel estimations of the background distributions
back_dist = st.gaussian_kde(back_rhos)
off_dist = st.gaussian_kde(off_rhos)
# Checks and creates parallel environment
try:
if not single:
import pp
except ImportError:
print(
'Parallel Python package pp not found. Defaulting to single core')
single = True
if not single:
# Initiates parallel server
job_server = pp.Server(nums)
fn = pp.Template(job_server, trial, (
like_fun,
findInterval,
), ("numpy", "math", "scipy.misc"))
# Single core version
if single:
geneI, geneL, geneInterval, geneDist = trial(gene_rhos, back_dist,
off_dist, off_rate,
like_fun, gene_span,
I_step)
# Parallel version
if not single:
geneI = {}
geneL = {}
geneInterval = {}
geneDist = {}
GeneRhosSplit = []
keys = gene_rhos.keys()
# Determines the number of genes in each task
n = int(len(keys) / nums) + 1
keysSplit = [keys[i:i + n] for i in range(0, len(keys), n)]
# Splits dictionary
for keyList in keysSplit:
dictChunk = {}
for key in keyList:
dictChunk[key] = gene_rhos[key]
GeneRhosSplit.append(dictChunk)
# Submits individual jobs for each gene list
jobs = []
for splits in GeneRhosSplit:
jobs.append(
fn.submit(splits, back_dist, off_dist, off_rate, like_fun,
gene_span, I_step))
# Retrieves results from individual jobs
for job in jobs:
val = job()
try:
geneIi, geneLi, geneIntervali, geneDisti = val
# Catches subprocess errors
except TypeError:
sys.exit('Subprocess failed')
# Saves results
geneI.update(geneIi)
geneL.update(geneLi)
geneInterval.update(geneIntervali)
geneDist.update(geneDisti)
return geneI, geneL, geneInterval, geneDist
def retrievePvals(gene_rhos, back_rhos, nums):
# Checks if parallel process requested
if nums == 1:
single = True
else:
single = False
# Checks and creates parallel environment
try:
if not single:
import pp
except ImportError:
print(
'Parallel Python package pp not found. Defaulting to single core')
single = True
# Initiates parallel server
if not single:
job_server = pp.Server(nums)
fn = pp.Template(job_server, calculatePvals, (),
("numpy", "scipy.stats.mstats"))
# Single core version
if single:
rhoMW, rhoKS = calculatePvals(gene_rhos, back_rhos)
# Parallel version
if not single:
rhoMW = {}
rhoKS = {}
GeneRhosSplit = []
keys = gene_rhos.keys()
# Determines the number of genes in each task
n = int(len(keys) / nums) + 1
keysSplit = [keys[i:i + n] for i in range(0, len(keys), n)]
# Splits dictionary
for keyList in keysSplit:
dictChunk = {}
for key in keyList:
dictChunk[key] = gene_rhos[key]
GeneRhosSplit.append(dictChunk)
# Submits individual jobs for each gene list
jobs = []
for splits in GeneRhosSplit:
jobs.append(fn.submit(splits, back_rhos))
# Retrieves results from individual jobs
for job in jobs:
val = job()
try:
rhoMWi, rhoKSi = val
# Catches subprocess errors
except TypeError:
sys.exit('Subprocess failed')
# Saves results
rhoMW.update(rhoMWi)
rhoKS.update(rhoKSi)
return rhoMW, rhoKS
def parallel_evaluation_pp(candidates, args):
"""Evaluate the candidates in parallel using Parallel Python.
This function allows parallel evaluation of candidate solutions.
It uses the `Parallel Python <http://www.parallelpython.com>`_ (pp)
library to accomplish the parallelization. This library must already
be installed in order to use this function. The function assigns the
evaluation of each candidate to its own job, all of which are then
distributed to the available processing units.
.. note::
All arguments to the evaluation function must be pickleable.
Those that are not will not be sent through the ``args`` variable
and will be unavailable to your function.
.. Arguments:
candidates -- the candidate solutions
args -- a dictionary of keyword arguments
Required keyword arguments in args:
- *pp_evaluator* -- actual evaluation function to be used (This function
should have the same signature as any other inspyred evaluation function.)
Optional keyword arguments in args:
- *pp_dependencies* -- tuple of functional dependencies of the serial
evaluator (default ())
- *pp_modules* -- tuple of modules that must be imported for the
functional dependencies (default ())
- *pp_servers* -- tuple of servers (on a cluster) that will be used
for parallel processing (default ("*",))
- *pp_secret* -- string representing the secret key needed to authenticate
on a worker node (default "inspyred")
- *pp_nprocs* -- integer representing the number of worker processes to
start on the local machine (default "autodetect", which sets it to the
number of processors in the system)
For more information about these arguments, please consult the
documentation for `Parallel Python <http://www.parallelpython.com>`_.
"""
import pp
logger = args['_ec'].logger
try:
evaluator = args['pp_evaluator']
except KeyError:
logger.error(
'parallel_evaluation_pp requires \'pp_evaluator\' be defined in the keyword arguments list'
)
raise
secret_key = args.setdefault('pp_secret', 'inspyred')
try:
job_server = args['_pp_job_server']
except KeyError:
pp_servers = args.get('pp_servers', ("*", ))
pp_nprocs = args.get('pp_nprocs', 'autodetect')
job_server = pp.Server(ncpus=pp_nprocs,
ppservers=pp_servers,
secret=secret_key)
args['_pp_job_server'] = job_server
pp_depends = args.setdefault('pp_dependencies', ())
pp_modules = args.setdefault('pp_modules', ())
pickled_args = {}
for key in args:
try:
pickle.dumps(args[key])
pickled_args[key] = args[key]
except (TypeError, pickle.PickleError, pickle.PicklingError):
logger.debug(
'unable to pickle args parameter {0} in parallel_evaluation_pp'
.format(key))
pass
func_template = pp.Template(job_server, evaluator, pp_depends, pp_modules)
jobs = [func_template.submit([c], pickled_args) for c in candidates]
fitness = []
for i, job in enumerate(jobs):
r = job()
try:
fitness.append(r[0])
except TypeError:
logger.warning(
'parallel_evaluation_pp generated an invalid fitness for candidate {0}'
.format(candidates[i]))
fitness.append(None)
return fitness
def sum_primes(n):
# Calculates sum of all primes until given integer n
return sum([x for x in xrange(2,n+1) if isprime(x)])
if len(sys.argv) > 1:
ncpus = int(sys.argv[1])
# Creates jobserver with ncpus workers
job_server = pp.Server(ncpus)
else:
# Creates jobserver with automatically detected number of workers
job_server = pp.Server()
print "Starting pp with", job_server.get_ncpus(), "workers"
# The following submits 8 jobs and then retrieves the results
inputs = (1000, 1001, 1002, 1003, 1004, 1005, 1006, 100700)
start_time = time.time()
jobn = pp.Template(job_server, sum_primes, (isprime,), ("math",))
jobs = [(input, jobn.submit(input)) for input in inputs]
for input, job in jobs:
print "Sum of primes until", input, "is", job()
print "Time elapsed: ", time.time() - start_time, "s"
job_server.print_stats()
mut = 10**-4 # Forward Mutation Rate
bmut = 10**-5 # Backward Mutation Rate
wfit, hfit, dfit = 1, 0.95, 0.95 # Fitness for wildtype, heterozygous, and dominant organisms
output = 'output.csv' # Name of output file
nums = 32 # Number of times to repeat simulation
# To run in command line: move to containing directory, enter command
# python balFinPop.py
# Outputs as comma-seperated values (csv), which can be opened in excel.
################################################################
# Code initiates simulation
job_server = pp.Server()
fn = pp.Template(job_server, trial, (), (), None, (), 'default', globals())
for pop in [100, 500, 1000, 3000, 10000]:
for ctr in [0, 1]:
for en in [0, 5, 25, 50, 75]:
if ctr and en == 0:
continue
jobs = []
for t in range(nums):
jobs.append(
fn.submit(pop, wfit, hfit, dfit, en, ctr, mut, bmut))
qss = []
avgs = []
for job in jobs:
import pp
import fortran
from numpy import array, zeros
jobserver = pp.Server(3)
def _normcall2(*args):
return fortran._ffnet.normcall2(*args)
tnormcall2 = pp.Template(jobserver, _normcall2, modules=('fortran', ))
def _func(*args):
return fortran._ffnet.func(*args)
tfunc = pp.Template(jobserver, _func, modules=('fortran', ))
def _grad(*args):
return fortran._ffnet.grad(*args)
tgrad = pp.Template(jobserver, _grad, modules=('fortran', ))
#######################################3
def _divide(inp):
n = jobserver.get_ncpus()
# sum_primes - the function
# (100,) - tuple with arguments for sum_primes
# (isprime,) - tuple with functions on which function sum_primes depends
# ("math",) - tuple with module names which must be imported
# before sum_primes execution
# Execution starts as soon as one of the workers will become available
#job1 = job_server.submit(sum_primes, (10,), (isprime,), ("math",), callback=callback)
# Retrieves the result calculated by job1
# The value of job1() is the same as sum_primes(100)
# If the job has not been finished yet,
# execution will wait here until result is available
#result = job1()
#print "Sum of primes below 10 is", result
inputs = range(10000000, 10100000)
ajo = pp.Template(job_server, isprime, (), ("math", ))
jobs = [(input, ajo.submit(input)) for input in inputs]
for input, job in jobs:
if (job() == True):
print "Number %i is a prime number" % input
print "Run done with ", job_server.get_ncpus(), "workers"
job_server.print_stats()
# Parallel Python Software: http://www.parallelpython.com
