def test_parallel():
x = arange(10.)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
parallel, p_func, n_jobs = parallel_func(sqrt, n_jobs=-1, verbose=0)
y = parallel(p_func(i**2) for i in range(10))
testing.assert_equal(x, y)
def test_parallel():
x = arange(10.)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
parallel, p_func, n_jobs = parallel_func(sqrt, n_jobs=-1, verbose=0)
y = parallel(p_func(i**2) for i in range(10))
testing.assert_equal(x,y)
def fit_joblib(self,
data_generator,
fit_kwds,
parallel_backend,
init_kwds_generator=None):
"""Performs the distributed estimation in parallel using joblib
Parameters
----------
data_generator : generator
A generator that produces a sequence of tuples where the first
element in the tuple corresponds to an endog array and the
element corresponds to an exog array.
fit_kwds : dict-like
Keywords needed for the model fitting.
parallel_backend : None or joblib parallel_backend object
used to allow support for more complicated backends,
ex: dask.distributed
init_kwds_generator : generator or None
Additional keyword generator that produces model init_kwds
that may vary based on data partition. The current usecase
is for WLS and GLS
Returns
-------
join_method result. For the default, _join_debiased, it returns a
p length array.
"""
from statsmodels.tools.parallel import parallel_func
par, f, n_jobs = parallel_func(_helper_fit_partition, self.partitions)
if parallel_backend is None and init_kwds_generator is None:
results_l = par(
f(self, pnum, endog, exog, fit_kwds)
for pnum, (endog, exog) in enumerate(data_generator))
elif parallel_backend is not None and init_kwds_generator is None:
with parallel_backend:
results_l = par(
f(self, pnum, endog, exog, fit_kwds)
for pnum, (endog, exog) in enumerate(data_generator))
elif parallel_backend is None and init_kwds_generator is not None:
tup_gen = enumerate(zip(data_generator, init_kwds_generator))
results_l = par(
f(self, pnum, endog, exog, fit_kwds, init_kwds)
for pnum, ((endog, exog), init_kwds) in tup_gen)
elif parallel_backend is not None and init_kwds_generator is not None:
tup_gen = enumerate(zip(data_generator, init_kwds_generator))
with parallel_backend:
results_l = par(
f(self, pnum, endog, exog, fit_kwds, init_kwds)
for pnum, ((endog, exog), init_kwds) in tup_gen)
return results_l
def _run_batch(
self, input_data: t.Union["ext.NpNDArray", "ext.PdDataFrame"]
) -> t.Any: # type: ignore[override] # noqa
# TODO: type hint return type.
parallel, p_func, _ = parallel_func( # type: ignore[arg-type]
self._predict_fn,
n_jobs=self._num_threads,
verbose=0,
)
return parallel(p_func(i) for i in input_data)[0] # type: ignore
def fit_joblib(self, data_generator, fit_kwds, parallel_backend,
init_kwds_generator=None):
"""Performs the distributed estimation in parallel using joblib
Parameters
----------
data_generator : generator
A generator that produces a sequence of tuples where the first
element in the tuple corresponds to an endog array and the
element corresponds to an exog array.
fit_kwds : dict-like
Keywords needed for the model fitting.
parallel_backend : None or joblib parallel_backend object
used to allow support for more complicated backends,
ex: dask.distributed
init_kwds_generator : generator or None
Additional keyword generator that produces model init_kwds
that may vary based on data partition. The current usecase
is for WLS and GLS
Returns
-------
join_method result. For the default, _join_debiased, it returns a
p length array.
"""
from statsmodels.tools.parallel import parallel_func
par, f, n_jobs = parallel_func(_helper_fit_partition, self.partitions)
if parallel_backend is None and init_kwds_generator is None:
results_l = par(f(self, pnum, endog, exog, fit_kwds)
for pnum, (endog, exog)
in enumerate(data_generator))
elif parallel_backend is not None and init_kwds_generator is None:
with parallel_backend:
results_l = par(f(self, pnum, endog, exog, fit_kwds)
for pnum, (endog, exog)
in enumerate(data_generator))
elif parallel_backend is None and init_kwds_generator is not None:
tup_gen = enumerate(zip(data_generator, init_kwds_generator))
results_l = par(f(self, pnum, endog, exog, fit_kwds, init_kwds)
for pnum, ((endog, exog), init_kwds)
in tup_gen)
elif parallel_backend is not None and init_kwds_generator is not None:
tup_gen = enumerate(zip(data_generator, init_kwds_generator))
with parallel_backend:
results_l = par(f(self, pnum, endog, exog, fit_kwds, init_kwds)
for pnum, ((endog, exog), init_kwds)
in tup_gen)
return results_l
trends = ('c', 'ct')
T = np.array((20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100, 120, 140,
160, 180, 200, 250, 300, 350, 400, 450, 500, 600, 700, 800,
900, 1000, 1200, 1400, 2000))
T = T[::-1]
percentiles = list(np.arange(0.5, 100.0, 0.5))
seeds = np.arange(0, 2**32, step=2**23)
for tr in trends:
results = np.zeros((len(percentiles), len(T), EX_NUM))
for i in range(EX_NUM):
print("Experiment Number {0} of {1} (trend {2})".format(
i + 1, EX_NUM, tr))
now = datetime.datetime.now()
parallel, p_func, n_jobs = parallel_func(wrapper,
n_jobs=NUM_JOBS,
verbose=2)
out = parallel(p_func(t, tr, EX_SIZE, seed=seeds[i]) for t in T)
q = lambda x: np.percentile(x, percentiles)
quantiles = map(q, out)
results[:, :, i] = np.array(quantiles).T
print('Elapsed time {0} seconds'.format(datetime.datetime.now() -
now))
if i % 50 == 0:
np.savez('dfgls_' + tr + '.npz',
trend=tr,
results=results,
percentiles=percentiles,
T=T)
if __name__ == '__main__':
trends = ('nc', 'c', 'ct', 'ctt')
T = array(
(20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100, 120, 140, 160,
180, 200, 250, 300, 350, 400, 450, 500, 600, 700, 800, 900,
1000, 1200, 1400, 2000))
T = T[::-1]
m = T.shape[0]
percentiles = list(arange(0.5, 100.0, 0.5))
seed(0)
seeds = random_integers(0, 2 ** 31 - 2, size=EX_NUM)
parallel, p_func, n_jobs = parallel_func(wrapper,
n_jobs=NUM_JOBS,
verbose=2)
parallel.pre_dispatch = NUM_JOBS
for tr in trends:
results = zeros((len(percentiles), len(T), EX_NUM)) * nan
filename = 'adf_z_' + tr + '.npz'
for i in range(EX_NUM):
print("Experiment Number {0} for Trend {1}".format(i + 1, tr))
# Non parallel version
# out = lmap(wrapper, T, [tr] * m, [EX_SIZE] * m, [seeds[i]] * m))
now = datetime.datetime.now()
out = parallel(p_func(t, tr, EX_SIZE, seed=seeds[i]) for t in T)
quantiles = map(lambda x: percentile(x, percentiles), out)
results[:, :, i] = array(quantiles).T
elapsed = datetime.datetime.now() - now
def test_parallel():
x = arange(10.)
parallel, p_func, n_jobs = parallel_func(sqrt, n_jobs=-1, verbose=0)
y = parallel(p_func(i**2) for i in range(10))
testing.assert_equal(x,y)
