def test_pdf_primitive_vs_supercell(self):
test_doc, success = res2dict(REAL_PATH + "data/KP_primitive.res",
db=False)
test_doc["text_id"] = ["primitive", "cell"]
test_doc["lattice_cart"] = abc2cart(test_doc["lattice_abc"])
test_doc["cell_volume"] = cart2volume(test_doc["lattice_cart"])
supercell_doc, success = res2dict(REAL_PATH + "data/KP_supercell.res",
db=False)
supercell_doc["text_id"] = ["supercell", "cell"]
supercell_doc["lattice_cart"] = abc2cart(supercell_doc["lattice_abc"])
supercell_doc["cell_volume"] = cart2volume(
supercell_doc["lattice_cart"])
test_doc["pdf"] = PDF(test_doc,
dr=0.01,
low_mem=True,
rmax=10,
num_images="auto",
debug=DEBUG)
supercell_doc["pdf"] = PDF(
supercell_doc,
dr=0.01,
low_mem=True,
rmax=10,
num_images="auto",
debug=DEBUG,
)
overlap = PDFOverlap(test_doc["pdf"], supercell_doc["pdf"])
self.assertLessEqual(overlap.similarity_distance, 1e-3)
self.assertGreaterEqual(overlap.similarity_distance, 0.0)
def pdf_sim_dist(doc_test, doc_supercell):
doc_test["text_id"] = ["test", "cell"]
doc_supercell["text_id"] = ["super", "cell"]
pdf_test = PDF(doc_test, low_mem=True)
pdf_supercell = PDF(doc_supercell, low_mem=True)
overlap = PDFOverlap(pdf_test, pdf_supercell)
return overlap.similarity_distance
def test_pdf_from_projected(self):
doc, success = res2dict(REAL_PATH + "data/LiPZn-r57des.res")
doc["lattice_cart"] = abc2cart(doc["lattice_abc"])
doc["text_id"] = ["unprojected", "test"]
doc["pdf_unprojected"] = PDF(doc, dr=0.01, **{"debug": False})
doc["text_id"] = ["projected", "test"]
doc["pdf_projected"] = PDF(doc, dr=0.01, **{"debug": False})
np.testing.assert_array_almost_equal(doc["pdf_unprojected"].gr,
doc["pdf_projected"].gr)
def test_single_atom_pdf(self):
from math import ceil
doc = dict()
box_size = 20
rmax = 41
dr = 0.1
num_images = "auto"
doc["positions_frac"] = [[0.5, 0.5, 0.5]]
doc["atom_types"] = ["C"]
doc["lattice_cart"] = np.asarray([[box_size, 0, 0], [0, box_size, 0],
[0, 0, box_size]])
doc["cell_volume"] = box_size**3
doc["text_id"] = ["hist", "ogram"]
doc["pdf"] = PDF(
doc,
num_images=num_images,
dr=dr,
rmax=rmax,
lazy=True,
style="histogram",
debug=True,
)
doc["pdf"].calc_pdf()
doc["text_id"] = ["smear"]
doc["pdf_smear"] = PDF(
doc,
num_images=num_images,
gaussian_width=0.01,
dr=0.1,
rmax=rmax,
lazy=True,
style="smear",
debug=True,
)
doc["pdf_smear"].calc_pdf()
doc["text_id"] = ["low"]
doc["pdf_low"] = PDF(
doc,
low_mem=True,
num_images=num_images,
gaussian_width=0.01,
dr=0.1,
rmax=rmax,
lazy=True,
style="smear",
debug=True,
)
doc["pdf_low"].calc_pdf()
peaks = [20, np.sqrt(2) * 20, np.sqrt(3) * 20, 40]
indices = [ceil(peak / dr) for peak in peaks]
self.assertListEqual(
np.where(doc["pdf_low"].gr > 1e-8)[0].tolist(), indices)
self.assertListEqual(
np.where(doc["pdf"].gr > 1e-8)[0].tolist(), indices)
self.assertListEqual(
np.where(doc["pdf_smear"].gr > 1e-8)[0].tolist(), indices)
def is_dupe(self, doc, sim_tol=5e-2, extra_pdfs=None):
""" Compare doc with all other structures at same stoichiometry via PDF overlap.
Parameters:
doc (dict): structure to compare.
Keyword Arguments:
sim_tol (float): similarity tolerance to compare to
extra_pdfs (list(dict)): list of structures with extra pdfs
to compare against
"""
new_pdf = PDF(doc, projected=True)
for ind, pdf in enumerate(self.pdfs):
if sorted(doc["stoichiometry"]) == self._stoichs[ind]:
dist = new_pdf.get_sim_distance(pdf, projected=True)
if dist < sim_tol:
return True
if extra_pdfs is not None:
for ind, _doc in enumerate(extra_pdfs):
pdf = _doc["pdf"]
if sorted(doc["stoichiometry"]) == sorted(
_doc["stoichiometry"]):
dist = new_pdf.get_sim_distance(pdf,
projected=pdf.projected)
if dist < sim_tol:
return True
return False
def calc_pdfs(self):
""" Compute PDFs for each structure in the generation. """
self._pdfs = []
self._stoichs = []
for structure in self.populace:
self._pdfs.append(PDF(structure, projected=True))
self._stoichs.append(sorted(structure["stoichiometry"]))
def test_concurrent_pdf(self):
import glob
import numpy as np
import time
from copy import deepcopy
files = glob.glob(REAL_PATH + "data/hull-KPSn-KP/*.res")[0:24]
cursor = [res2dict(file, db=False)[0] for file in files]
serial_cursor = deepcopy(cursor)
pdf_args = {
"dr": 0.1,
"num_images": "auto",
"gaussian_width": 0.1,
"lazy": False,
"projected": False,
}
start = time.time()
pdf_factory = PDFFactory(cursor, **pdf_args)
factory_elapsed = time.time() - start
start = time.time()
for doc in serial_cursor:
doc["pdf"] = PDF(doc, **pdf_args, timing=False)
serial_elapsed = time.time() - start
print("{:.2f} s over {} processes vs {:.2f} s in serial".format(
factory_elapsed, pdf_factory.nprocs, serial_elapsed))
print("Corresponding to a speedup of {:.1f} vs ideal {:.1f}".format(
serial_elapsed / factory_elapsed, pdf_factory.nprocs))
for ind, doc in enumerate(serial_cursor):
np.testing.assert_array_almost_equal(doc["pdf"].gr,
cursor[ind]["pdf"].gr,
decimal=6)
def calculate_pdf(self, **kwargs):
""" Calculate and set the PDF with the passed parameters. """
from matador.fingerprints.pdf import PDF
if 'pdf' not in self._data:
self._data['pdf'] = PDF(self._data,
label=self.formula_tex,
**kwargs)
return self._data['pdf']
def pdf(self):
""" Returns a PDF object (pair distribution function) for the
structure, calculated with default PDF settings.
"""
from matador.fingerprints.pdf import PDF
if 'pdf' not in self._data:
self._data['pdf'] = PDF(self._data, label=self.formula_tex)
return self._data['pdf']
def test_overlap_smear_vs_hist(self):
doc, success = res2dict(REAL_PATH + "data/LiPZn-r57des.res")
doc["lattice_cart"] = abc2cart(doc["lattice_abc"])
doc["text_id"] = ["smear", "test"]
doc["pdf_smear"] = PDF(
doc,
num_images=3,
dr=0.01,
gaussian_width=0.1,
projected=False,
style="smear",
low_mem=True,
)
doc["text_id"] = ["hist", "test"]
doc["pdf_hist"] = PDF(doc,
num_images=3,
dr=0.1,
projected=False,
style="histogram")
overlap = PDFOverlap(doc["pdf_smear"], doc["pdf_hist"])
self.assertLessEqual(overlap.similarity_distance, 0.02)
self.assertGreater(overlap.similarity_distance, 0.0)
def test_identity_overlap(self):
doc, success = res2dict(REAL_PATH + "data/LiPZn-r57des.res")
doc["lattice_cart"] = abc2cart(doc["lattice_abc"])
doc["text_id"] = ["pdf", "test"]
doc["pdf_smear"] = PDF(
doc,
num_images=3,
dr=0.001,
gaussian_width=0.1,
style="smear",
debug=False,
low_mem=True,
)
overlap = PDFOverlap(doc["pdf_smear"], doc["pdf_smear"])
self.assertEqual(overlap.similarity_distance, 0.0)
def test_auto_images_vs_large(self):
doc, success = res2dict(REAL_PATH + "data/LiPZn-r57des.res")
doc["lattice_cart"] = abc2cart(doc["lattice_abc"])
doc["text_id"] = ["pdf", "test"]
doc["pdf_num_images"] = PDF(doc,
low_mem=True,
num_images=5,
rmax=15,
dr=0.1,
**{
"debug": True,
"projected": False
})
doc["pdf_auto_images"] = PDF(doc,
low_mem=True,
num_images="auto",
rmax=15,
dr=0.1,
**{
"debug": True,
"projected": False
})
np.testing.assert_array_almost_equal(doc["pdf_num_images"].gr,
doc["pdf_auto_images"].gr)
def test_concurrent_pdf_stoichs(self):
import glob
import numpy as np
import time
from copy import deepcopy
from matador.hull import QueryConvexHull
files = glob.glob(REAL_PATH + "data/hull-KPSn-KP/*.res")
cursor = [res2dict(file, db=True)[0] for file in files]
hull = QueryConvexHull(
cursor=cursor,
no_plot=True,
hull_cutoff=0.5,
summary=True,
elements=["K", "Sn", "P"],
quiet=True,
)
serial_cursor = deepcopy(hull.cursor)
pdf_args = {
"dr": 0.1,
"num_images": "auto",
"gaussian_width": 0.1,
"lazy": False,
"projected": False,
}
start = time.time()
pdf_factory = PDFFactory(hull.cursor, **pdf_args)
factory_elapsed = time.time() - start
start = time.time()
for doc in serial_cursor:
doc["pdf"] = PDF(doc, **pdf_args, timing=False)
serial_elapsed = time.time() - start
print("{:.2f} s over {} processes vs {:.2f} s in serial".format(
factory_elapsed, pdf_factory.nprocs, serial_elapsed))
print("Corresponding to a speedup of {:.1f} vs ideal {:.1f}".format(
serial_elapsed / factory_elapsed, pdf_factory.nprocs))
for ind, doc in enumerate(serial_cursor):
np.testing.assert_array_almost_equal(doc["pdf"].gr,
hull.cursor[ind]["pdf"].gr,
decimal=6)
def test_ideal_gas_pdf(self, retry=0):
""" DEPRECATED.
Slow, and not very useful.
"""
# create fake matador doc
doc = dict()
max_retries = 1
self.assertLess(
retry,
max_retries,
msg="After {} attempts, PDF still failed.".format(retry))
num_atoms = 100
box_size = 10
num_samples = 1
rmax = 15
dr = 0.01
num_images = "auto"
i = 0
doc["atom_types"] = num_atoms * ["C"]
doc["lattice_cart"] = np.asarray([[box_size, 0, 0], [0, box_size, 0],
[0, 0, box_size]])
doc["cell_volume"] = box_size**3
doc["text_id"] = ["ideal", "gas"]
while i < num_samples:
doc["positions_frac"] = np.random.rand(num_atoms, 3)
doc["text_id"] = "hist"
doc["pdf"] = PDF(
doc,
num_images=num_images,
dr=dr,
rmax=rmax,
lazy=True,
style="histogram",
debug=True,
)
doc["pdf"].calc_pdf()
self.assertAlmostEqual(np.mean(doc["pdf"].gr[50:]), 1.0, places=1)
