pipeline.py 7.53 KB
Newer Older
1
2
# -*- coding: utf-8 -*-

3
4
5
6
7
import os
import numpy as np

from mpi4py import MPI

8
9
from keepers import Loggable

10
11
12
13
from imagine.likelihoods import Likelihood
from imagine.magnetic_fields import MagneticFieldFactory
from imagine.observers import Observer
from imagine.priors import Prior
14
from imagine import pymultinest
15
from imagine.sample import Sample
16
17
18
19

comm = MPI.COMM_WORLD
size = comm.size
rank = comm.rank
20

21
22
23
WORK_TAG = 0
DIE_TAG = 1

24
25

class Pipeline(Loggable, object):
26
27
28
29
30
31
32
33
34
35
36
37
    """
    The pipeline
    - posses all the building blocks: magnetic_field, observer,
        likelihood and prior.
    - if multiple log-likelihoods and log-priors are given: sum the result
    - coordinates the repeated observation in order to compute an ensemble
    - controls which parameters of the magnetic field are tested
        (active parameters)


    """
    def __init__(self, magnetic_field_factory, observer, likelihood, prior,
38
                 active_variables=[], ensemble_size=1,
39
                 pymultinest_parameters={}, sample_callback=None):
40
        self.logger.debug("Setting up pipeline.")
41
        self.magnetic_field_factory = magnetic_field_factory
42
43
44
        self.observer = observer
        self.likelihood = likelihood
        self.prior = prior
45
        self.active_variables = active_variables
46
47
        self.ensemble_size = ensemble_size

48
        # setting defaults for pymultinest
49
50
51
52
53
        self.pymultinest_parameters = {'verbose': True,
                                       'n_iter_before_update': 1,
                                       'n_live_points': 100}
        self.pymultinest_parameters.update(pymultinest_parameters)

54
55
        self.sample_callback = sample_callback

56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
    @property
    def observer(self):
        return self._observer

    @observer.setter
    def observer(self, observer):
        if not isinstance(observer, Observer):
            raise TypeError("observer must be an instance of Observer-class.")
        self.logger.debug("Setting observer.")
        self._observer = observer

    @property
    def likelihood(self):
        return self._likelihood

    @likelihood.setter
    def likelihood(self, likelihood):
73
        self.logger.debug("Setting likelihood.")
74
        self._likelihood = ()
75
76
77
        if not (isinstance(likelihood, list) and
                isinstance(likelihood, tuple)):
            likelihood = [likelihood]
78
79
80
81
82
        for l in likelihood:
            if not isinstance(l, Likelihood):
                raise TypeError(
                    "likelihood must be an instance of Likelihood-class.")
            self._likelihood += (l,)
83
84
85
86
87
88
89

    @property
    def prior(self):
        return self._prior

    @prior.setter
    def prior(self, prior):
90
        self.logger.debug("Setting prior.")
91
92
93
94
        if not isinstance(prior, Prior):
            raise TypeError(
                "prior must be an instance of Prior-class.")
        self._prior = prior
95
96

    @property
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
    def magnetic_field_factory(self):
        return self._magnetic_field_factory

    @magnetic_field_factory.setter
    def magnetic_field_factory(self, magnetic_field_factory):
        if not isinstance(magnetic_field_factory, MagneticFieldFactory):
            raise TypeError(
                "magnetic_field_factory must be an instance of the "
                "MagneticFieldFactory-class.")
        self.logger.debug("Setting magnetic_field_factory.")
        self._magnetic_field_factory = magnetic_field_factory

    @property
    def active_variables(self):
        return self._active_variables

    @active_variables.setter
    def active_variables(self, active_variables):
        if not isinstance(active_variables, list):
            raise TypeError(
                    "active_variables must be a list.")
        self.logger.debug("Resetting active_variables to %s" %
                          str(active_variables))
        new_active = []
        for av in active_variables:
            new_active += [str(av)]
        self._active_variables = new_active
124
125
126
127
128
129
130
131
132
133
134
135
136
137

    @property
    def ensemble_size(self):
        return self._ensemble_size

    @ensemble_size.setter
    def ensemble_size(self, ensemble_size):

        ensemble_size = int(ensemble_size)
        if ensemble_size <= 0:
            raise ValueError("ensemble_size must be positive!")
        self.logger.debug("Setting ensemble size to %i." % ensemble_size)
        self._ensemble_size = ensemble_size

138
139
140
141
142
143
144
145
    def _multinest_likelihood(self, cube, ndim, nparams):
        cube_content = np.empty(ndim)
        for i in xrange(ndim):
            cube_content[i] = cube[i]
        if rank != 0:
            raise RuntimeError("_multinest_likelihood must only be called on "
                               "rank==0.")
        for i in xrange(1, size):
146
            comm.send(cube_content, dest=i, tag=WORK_TAG)
Theo Steininger's avatar
Theo Steininger committed
147
        self.logger.debug("Sent multinest-cube to nodes with rank > 0.")
148
149
150
151

        return self._core_likelihood(cube_content)

    def _listen_for_likelihood_calls(self):
152
153
154
155
156
157
158
159
        status = MPI.Status()
        while True:
            cube = comm.recv(source=0, tag=MPI.ANY_TAG, status=status)
            if status == DIE_TAG:
                self.logger.debug("Received DIE_TAG from rank 0.")
                break
            self.logger.debug("Received cube from rank 0.")
            self._core_likelihood(cube)
160
161

    def _core_likelihood(self, cube):
162
163
        self.logger.debug("Beginning Likelihood-calculation for %s." %
                          str(cube))
164
165
166
167
168
169
        # translate cube to variables
        variables = {}
        for i, av in enumerate(self.active_variables):
            variables[av] = cube[i]

        # create magnetic field
170
171
172
173
174
        self.logger.debug("Creating magnetic field.")
        b_field = self.magnetic_field_factory.generate(
                                              variables=variables,
                                              ensemble_size=self.ensemble_size)

175
        # create observables
176
        self.logger.debug("Creating observables.")
177
178
179
        observables = self.observer(b_field)

        # add up individual log-likelihood terms
180
        self.logger.debug("Evaluating likelihood(s).")
181
182
        likelihood = ()
        total_likelihood = 0
183
        for like in self.likelihood:
184
185
186
            current_likelihood = like(observables)
            likelihood += (current_likelihood, )
            total_likelihood += current_likelihood
187

Theo Steininger's avatar
Theo Steininger committed
188
        self.logger.info("Evaluated likelihood: %f for %s" %
189
190
191
192
193
194
195
196
197
198
199
                         (total_likelihood, str(cube)))

        if self.sample_callback is not None:
            self.logger.debug("Creating sample-object.")
            sample = Sample(variables=variables,
                            magnetic_field=b_field,
                            observables=observables,
                            likelihood=likelihood,
                            total_likelihood=total_likelihood)
            self.sample_callback(sample)

200
201
        return likelihood

202
    def __call__(self, variables):
203
204
205

        if rank == 0:
            # kickstart pymultinest
Theo Steininger's avatar
Theo Steininger committed
206
            self.logger.info("Starting pymultinest.")
207
208
209
210
211
            if not os.path.exists("chains"):
                os.mkdir("chains")
            pymultinest.run(self._multinest_likelihood,
                            self.prior,
                            len(self.active_variables),
212
                            **self.pymultinest_parameters)
213
214
215
216
            self.logger.info("pymultinest finished.")
            for i in xrange(1, size):
                self.logger.debug("Sending DIE_TAG to rank %i." % i)
                comm.send(None, dest=i, tag=DIE_TAG)
217
218
219
        else:
            # let all other nodes listen for likelihood evaluations
            self._listen_for_likelihood_calls()