def testAnyAll(self):
self.assertTrue(mpi.mpi_all(True))
self.assertFalse(mpi.mpi_all(False))
self.assertEqual(mpi.mpi_all(mpi.RANK == 0), mpi.SIZE == 1)
self.assertTrue(mpi.mpi_any(True))
self.assertFalse(mpi.mpi_any(False))
self.assertTrue(mpi.mpi_any(mpi.RANK == 0))
def testAnyAll(self):
self.assertTrue(mpi.mpi_all(True))
self.assertFalse(mpi.mpi_all(False))
self.assertEqual(mpi.mpi_all(mpi.RANK == 0), mpi.SIZE == 1)
self.assertTrue(mpi.mpi_any(True))
self.assertFalse(mpi.mpi_any(False))
self.assertTrue(mpi.mpi_any(mpi.RANK == 0))
def __init__(self, **kwargs):
"""Initializes the Puff sampling layer.
kwargs:
minibatch: the minibatch size.
puff: a list of puff files to be read. These files should have the
same number of data points, and the sampler will return data
points from different points with the same index.
use_mpi: if set True, when the code is run with mpirun, each mpi
node will only deal with the part of file that has index range
[N * RANK / SIZE, N * (RANK+1) / SIZE). Note that in this
case, one need to make sure that the minibatch size is smaller
than the number of data points in the local range on every mpi
node. Default True.
"""
base.DataLayer.__init__(self, **kwargs)
self._filenames = self.spec['puff']
self._use_mpi = self.spec.get('use_mpi', True)
self._minibatch = self.spec['minibatch']
self._puffs = [base.Puff(filename) for filename in self._filenames]
num_data = [puff.num_data() for puff in self._puffs]
if len(set(num_data)) == 1:
raise ValueError('The puff files have different number of data.')
if self._use_mpi:
local_start = int(num_data[0] * mpi.RANK / mpi.SIZE)
local_end = int(num_data[0] * (mpi.RANK + 1) / mpi.SIZE)
if mpi.mpi_any(local_end - local_start < self._minibatch):
raise ValueError('Local range smaller than minibatch.')
for puff in self._puffs:
puff.set_range(local_start, local_end)
return
