From badd446e949022a9b6d1e21d9b90f62165849805 Mon Sep 17 00:00:00 2001
From: Michael Osthege <m.osthege@fz-juelich.de>
Date: Sun, 21 Jul 2019 21:26:10 +0200
Subject: [PATCH 1/4] test with and without parallelization across cores
demonstrates issue #3555
---
pymc3/tests/test_step.py | 15 ++++++++++++++-
1 file changed, 14 insertions(+), 1 deletion(-)
diff --git a/pymc3/tests/test_step.py b/pymc3/tests/test_step.py
index 32547b1d88..ed9ed0e99e 100644
--- a/pymc3/tests/test_step.py
+++ b/pymc3/tests/test_step.py
@@ -915,7 +915,7 @@ def test_checks_population_size(self):
trace = sample(draws=100, chains=4, step=step)
pass
- def test_parallelized_chains_are_random(self):
+ def test_nonparallelized_chains_are_random(self):
with Model() as model:
x = Normal("x", 0, 1)
for stepper in TestPopulationSamplers.steppers:
@@ -928,6 +928,19 @@ def test_parallelized_chains_are_random(self):
)
pass
+ def test_parallelized_chains_are_random(self):
+ with Model() as model:
+ x = Normal("x", 0, 1)
+ for stepper in TestPopulationSamplers.steppers:
+ step = stepper()
+ trace = sample(chains=4, draws=20, tune=0, step=DEMetropolis(), parallelize=True)
+ samples = np.array(trace.get_values("x", combine=False))[:, 5]
+
+ assert len(set(samples)) == 4, "Parallelized {} " "chains are identical.".format(
+ stepper
+ )
+ pass
+
@pytest.mark.xfail(condition=(theano.config.floatX == "float32"), reason="Fails on float32")
class TestNutsCheckTrace:
From 6324408b48fdf7f65336505d690ea93031403417 Mon Sep 17 00:00:00 2001
From: Michael Osthege <m.osthege@fz-juelich.de>
Date: Sun, 21 Jul 2019 22:05:16 +0200
Subject: [PATCH 2/4] replace parallelize kwarg by reliance on cores setting
closes #3555
---
pymc3/sampling.py | 6 +++---
pymc3/tests/test_step.py | 4 ++--
2 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/pymc3/sampling.py b/pymc3/sampling.py
index 8fb17df4f5..b3628d857f 100644
--- a/pymc3/sampling.py
+++ b/pymc3/sampling.py
@@ -452,7 +452,7 @@ def sample(draws=500, step=None, init='auto', n_init=200000, start=None, trace=N
if has_population_samplers:
_log.info('Population sampling ({} chains)'.format(chains))
_print_step_hierarchy(step)
- trace = _sample_population(**sample_args)
+ trace = _sample_population(**sample_args, parallelize=cores > 1)
else:
_log.info('Sequential sampling ({} chains in 1 job)'.format(chains))
_print_step_hierarchy(step)
@@ -689,7 +689,7 @@ def __init__(self, steppers, parallelize):
if parallelize:
try:
# configure a child process for each stepper
- _log.info('Attempting to parallelize chains.')
+ _log.info('Attempting to parallelize chains to all cores. You can turn this off with `pm.sample(cores=1)`.')
import multiprocessing
for c, stepper in enumerate(tqdm(steppers)):
slave_end, master_end = multiprocessing.Pipe()
@@ -714,7 +714,7 @@ def __init__(self, steppers, parallelize):
_log.debug('Error was: ', exec_info=True)
else:
_log.info('Chains are not parallelized. You can enable this by passing '
- 'pm.sample(parallelize=True).')
+ 'pm.sample(cores=2).')
return super().__init__()
def __enter__(self):
diff --git a/pymc3/tests/test_step.py b/pymc3/tests/test_step.py
index ed9ed0e99e..d67c0ef1d2 100644
--- a/pymc3/tests/test_step.py
+++ b/pymc3/tests/test_step.py
@@ -920,7 +920,7 @@ def test_nonparallelized_chains_are_random(self):
x = Normal("x", 0, 1)
for stepper in TestPopulationSamplers.steppers:
step = stepper()
- trace = sample(chains=4, draws=20, tune=0, step=DEMetropolis())
+ trace = sample(chains=4, cores=1, draws=20, tune=0, step=DEMetropolis())
samples = np.array(trace.get_values("x", combine=False))[:, 5]
assert len(set(samples)) == 4, "Parallelized {} " "chains are identical.".format(
@@ -933,7 +933,7 @@ def test_parallelized_chains_are_random(self):
x = Normal("x", 0, 1)
for stepper in TestPopulationSamplers.steppers:
step = stepper()
- trace = sample(chains=4, draws=20, tune=0, step=DEMetropolis(), parallelize=True)
+ trace = sample(chains=4, cores=4, draws=20, tune=0, step=DEMetropolis())
samples = np.array(trace.get_values("x", combine=False))[:, 5]
assert len(set(samples)) == 4, "Parallelized {} " "chains are identical.".format(
From 7ac416a1ee12caf41a50e958293e270eeb0dd57e Mon Sep 17 00:00:00 2001
From: Michael Osthege <m.osthege@fz-juelich.de>
Date: Sun, 21 Jul 2019 22:08:07 +0200
Subject: [PATCH 3/4] add the changes from pull 3559
---
RELEASE-NOTES.md | 1 +
1 file changed, 1 insertion(+)
diff --git a/RELEASE-NOTES.md b/RELEASE-NOTES.md
index 37c8366838..dfc7eae382 100644
--- a/RELEASE-NOTES.md
+++ b/RELEASE-NOTES.md
@@ -11,6 +11,7 @@
### Maintenance
- Moved math operations out of `Rice`, `TruncatedNormal`, `Triangular` and `ZeroInflatedNegativeBinomial` `random` methods. Math operations on values returned by `draw_values` might not broadcast well, and all the `size` aware broadcasting is left to `generate_samples`. Fixes [#3481](https://github.com/pymc-devs/pymc3/issues/3481) and [#3508](https://github.com/pymc-devs/pymc3/issues/3508)
+- Parallelization of population steppers (`DEMetropolis`) is now set via the `cores` argument. ([#3559](https://github.com/pymc-devs/pymc3/pull/3559))
## PyMC3 3.7 (May 29 2019)
From 971c8e0008fc87631b4dd0687c10a0618fd9d53a Mon Sep 17 00:00:00 2001
From: michaelosthege <thecakedev@hotmail.com>
Date: Sun, 21 Jul 2019 23:01:13 +0200
Subject: [PATCH 4/4] use more general suggestion in the log message
Co-Authored-By: Colin <ColCarroll@users.noreply.github.com>
---
pymc3/sampling.py | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/pymc3/sampling.py b/pymc3/sampling.py
index b3628d857f..1c9b515b13 100644
--- a/pymc3/sampling.py
+++ b/pymc3/sampling.py
@@ -714,7 +714,7 @@ def __init__(self, steppers, parallelize):
_log.debug('Error was: ', exec_info=True)
else:
_log.info('Chains are not parallelized. You can enable this by passing '
- 'pm.sample(cores=2).')
+ '`pm.sample(cores=n)`, where n > 1.')
return super().__init__()
def __enter__(self):