rsspolymlp 0.4.6__tar.gz → 0.4.8__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {rsspolymlp-0.4.6/src/rsspolymlp.egg-info → rsspolymlp-0.4.8}/PKG-INFO +1 -1
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/pyproject.toml +1 -1
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/rss_summarize.py +23 -22
- rsspolymlp-0.4.8/src/rsspolymlp/analysis/struct_matcher/assignment_algo.py +187 -0
- rsspolymlp-0.4.8/src/rsspolymlp/analysis/struct_matcher/dataclass.py +24 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/struct_matcher/gen_struct_rep.py +72 -129
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/struct_matcher/identify_uniqstr.py +22 -22
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/struct_matcher/niggli_reduce.py +28 -23
- rsspolymlp-0.4.6/src/rsspolymlp/analysis/struct_matcher/redrep_position.py → rsspolymlp-0.4.8/src/rsspolymlp/analysis/struct_matcher/position_redrep.py +148 -152
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/struct_matcher/primitive_transform.py +87 -85
- rsspolymlp-0.4.8/src/rsspolymlp/analysis/struct_matcher/struct_match.py +379 -0
- rsspolymlp-0.4.6/src/rsspolymlp/analysis/struct_matcher/test/gen_lattice.py → rsspolymlp-0.4.8/src/rsspolymlp/analysis/struct_matcher/test/gen_rand_lattice.py +24 -1
- rsspolymlp-0.4.8/src/rsspolymlp/analysis/struct_matcher/test/gen_rand_position.py +65 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/cli_rsspolymlp.py +4 -5
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/cli_rsspolymlp_utils.py +2 -3
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/rsspolymlp.py +3 -2
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/rsspolymlp_utils.py +20 -14
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/common/interface_spglib.py +10 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/rss/eliminate_duplicates.py +1 -1
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/rss/random_struct.py +1 -1
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8/src/rsspolymlp.egg-info}/PKG-INFO +1 -1
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp.egg-info/SOURCES.txt +4 -2
- rsspolymlp-0.4.6/src/rsspolymlp/analysis/struct_matcher/assignment_algo.py +0 -245
- rsspolymlp-0.4.6/src/rsspolymlp/analysis/struct_matcher/struct_match.py +0 -293
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/LICENSE +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/README.md +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/setup.cfg +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/eos.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/ghost_minima.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/phase_analysis.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/struct_matcher/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/struct_matcher/chiral_spg.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/analysis/utils/convert_data_format.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/cli_rsspolymlp_devkit.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/cli_rsspolymlp_plot.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/rsspolymlp_devkit.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/api/rsspolymlp_plot.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/common/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/common/atomic_energy.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/common/composition.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/common/interface_pypolymlp.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/common/interface_vasp.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/common/property.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/dataset/compress_vasprun.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/dataset/gen_mlp_dataset.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/dataset/split_dataset.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/estimate_cost.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/model_selection/binary/gen_single_model.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/model_selection/elemental/gen_hybrid_model.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/model_selection/elemental/gen_single_model.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/model_selection/n_feature.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/model_selection/pypolymlp_gridsearch.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/model_selection/reduce_grid.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/pareto_opt_mlp.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/mlp_dev/polymlp_dev.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/rss/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/rss/load_logfile.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/rss/optimization_mlp.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/lammps_utils.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/matplot_util/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/matplot_util/custom_plt.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/matplot_util/examples/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/matplot_util/examples/example.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/matplot_util/examples/template.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/matplot_util/make_plot.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/opt_sscha/opt_geometry_sscha.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/opt_sscha/opt_position_sscha.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/opt_sscha/run_opt_sscha.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/opt_sscha/sscha_property.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/pymatgen_utils.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/vasp_util/__init__.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/vasp_util/api.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/vasp_util/gen_incar.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp/utils/vasp_util/gen_script.py +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp.egg-info/dependency_links.txt +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp.egg-info/entry_points.txt +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp.egg-info/requires.txt +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/src/rsspolymlp.egg-info/top_level.txt +0 -0
- {rsspolymlp-0.4.6 → rsspolymlp-0.4.8}/tests/test_cli_help.py +0 -0
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: rsspolymlp
|
|
3
|
-
Version: 0.4.
|
|
3
|
+
Version: 0.4.8
|
|
4
4
|
Summary: A framework for random structure search using polynomial MLPs
|
|
5
5
|
Author-email: Hayato Wakai <wakai@cms.mtl.kyoto-u.ac.jp>
|
|
6
6
|
Maintainer-email: Hayato Wakai <wakai@cms.mtl.kyoto-u.ac.jp>
|
|
@@ -14,7 +14,7 @@ from pypolymlp.utils.vasp_utils import write_poscar_file
|
|
|
14
14
|
from rsspolymlp.analysis.ghost_minima import detect_ghost_minima
|
|
15
15
|
from rsspolymlp.analysis.struct_matcher.gen_struct_rep import (
|
|
16
16
|
StructRepGenerator,
|
|
17
|
-
|
|
17
|
+
StructRep,
|
|
18
18
|
)
|
|
19
19
|
from rsspolymlp.analysis.struct_matcher.identify_uniqstr import (
|
|
20
20
|
UniqueStructIdentifier,
|
|
@@ -38,7 +38,7 @@ class RSSResultSummarizer:
|
|
|
38
38
|
num_process: int = -1,
|
|
39
39
|
backend: str = "loky",
|
|
40
40
|
assign_method: Literal["redrep", "greedy", "HK", "pymatgen"] = "redrep",
|
|
41
|
-
tol_sm:
|
|
41
|
+
tol_sm: float = 1e-2,
|
|
42
42
|
output_poscar: bool = False,
|
|
43
43
|
thresholds: list[float] = None,
|
|
44
44
|
parse_vasp: bool = False,
|
|
@@ -91,12 +91,13 @@ class RSSResultSummarizer:
|
|
|
91
91
|
axis_tol = 0.1
|
|
92
92
|
pos_tol = 0.1
|
|
93
93
|
|
|
94
|
-
self.analyzer = UniqueStructIdentifier(
|
|
95
|
-
axis_tol=axis_tol,
|
|
96
|
-
pos_tol=pos_tol,
|
|
97
|
-
assign_method=self.assign_method,
|
|
98
|
-
)
|
|
99
94
|
for comp_tag, res_paths in json_paths_c.items():
|
|
95
|
+
self.analyzer = UniqueStructIdentifier(
|
|
96
|
+
axis_tol=axis_tol,
|
|
97
|
+
pos_tol=pos_tol,
|
|
98
|
+
assign_method=self.assign_method,
|
|
99
|
+
)
|
|
100
|
+
|
|
100
101
|
print(f"Composition {comp_tag}: summarizing...")
|
|
101
102
|
time_start = time()
|
|
102
103
|
|
|
@@ -283,22 +284,22 @@ class RSSResultSummarizer:
|
|
|
283
284
|
generator = StructRepGenerator(
|
|
284
285
|
tol_sm=self.tol_sm,
|
|
285
286
|
)
|
|
286
|
-
|
|
287
|
+
struct_rep_list = generator.generate_struct_rep_parallel(
|
|
287
288
|
rss_results,
|
|
288
289
|
num_process=self.num_process,
|
|
289
290
|
backend=self.backend,
|
|
290
291
|
)
|
|
291
292
|
if self.update_parent:
|
|
292
293
|
print(" - Eliminating duplicate structures...")
|
|
293
|
-
for idx,
|
|
294
|
+
for idx, struct_rep in enumerate(struct_rep_list):
|
|
294
295
|
self.analyzer.identify_duplicate_struct(
|
|
295
|
-
|
|
296
|
+
struct_rep=struct_rep,
|
|
296
297
|
other_properties=rss_results[idx],
|
|
297
298
|
axis_tol=0.03,
|
|
298
299
|
pos_tol=0.03,
|
|
299
300
|
)
|
|
300
301
|
else:
|
|
301
|
-
self.analyzer._initialize_unique_structs(
|
|
302
|
+
self.analyzer._initialize_unique_structs(struct_rep_list, rss_results)
|
|
302
303
|
|
|
303
304
|
return set(processed_paths)
|
|
304
305
|
|
|
@@ -441,7 +442,7 @@ class RSSResultSummarizer:
|
|
|
441
442
|
continue
|
|
442
443
|
|
|
443
444
|
energy_dft /= len(polymlp_st.elements)
|
|
444
|
-
if energy_dft < -
|
|
445
|
+
if energy_dft < -20:
|
|
445
446
|
print(path_name, "exhibits an unphysically low energy. Skipping.")
|
|
446
447
|
continue
|
|
447
448
|
|
|
@@ -489,16 +490,16 @@ class RSSResultSummarizer:
|
|
|
489
490
|
|
|
490
491
|
def dump_json_rss_results(
|
|
491
492
|
file_name: str,
|
|
492
|
-
|
|
493
|
+
struct_rep_list: list[StructRep],
|
|
493
494
|
enthalpy_lists: list[float],
|
|
494
495
|
is_ghost_minima=None,
|
|
495
496
|
pressure=None,
|
|
496
497
|
):
|
|
497
498
|
if is_ghost_minima is None:
|
|
498
|
-
is_ghost_minima = np.full_like(
|
|
499
|
+
is_ghost_minima = np.full_like(struct_rep_list, False, dtype=bool)
|
|
499
500
|
|
|
500
501
|
rss_results = []
|
|
501
|
-
for idx, _str in enumerate(
|
|
502
|
+
for idx, _str in enumerate(struct_rep_list):
|
|
502
503
|
axis = _str.original_axis
|
|
503
504
|
positions = _str.original_positions
|
|
504
505
|
elements = _str.original_elements
|
|
@@ -520,7 +521,7 @@ def dump_json_rss_results(
|
|
|
520
521
|
"is_ghost_minima": bool(is_ghost_minima[idx]),
|
|
521
522
|
}
|
|
522
523
|
)
|
|
523
|
-
if len(
|
|
524
|
+
if len(struct_rep_list) > 0:
|
|
524
525
|
comp_res = compute_composition(elements)
|
|
525
526
|
rss_result_all = {
|
|
526
527
|
"elements": comp_res.unique_elements.tolist(),
|
|
@@ -540,14 +541,14 @@ def dump_json_rss_results(
|
|
|
540
541
|
|
|
541
542
|
def log_unique_structures(
|
|
542
543
|
file_name: str,
|
|
543
|
-
|
|
544
|
+
struct_rep_list: list[StructRep],
|
|
544
545
|
enthalpy_lists: Optional[list[float]] = None,
|
|
545
546
|
is_ghost_minima=None,
|
|
546
547
|
unique_struct_iters=None,
|
|
547
548
|
print_duplicates: bool = False,
|
|
548
549
|
):
|
|
549
550
|
if is_ghost_minima is None:
|
|
550
|
-
is_ghost_minima = np.full_like(
|
|
551
|
+
is_ghost_minima = np.full_like(struct_rep_list, False, dtype=bool)
|
|
551
552
|
energy_min = None
|
|
552
553
|
if enthalpy_lists is not None:
|
|
553
554
|
for i in range(len(enthalpy_lists)):
|
|
@@ -556,10 +557,10 @@ def log_unique_structures(
|
|
|
556
557
|
break
|
|
557
558
|
|
|
558
559
|
struct_num_max = max(
|
|
559
|
-
(_s.struct_tag for _s in
|
|
560
|
+
(_s.struct_tag for _s in struct_rep_list if _s.struct_tag is not None),
|
|
560
561
|
default=0,
|
|
561
562
|
)
|
|
562
|
-
for _s in
|
|
563
|
+
for _s in struct_rep_list:
|
|
563
564
|
if _s.struct_tag is None:
|
|
564
565
|
struct_num_max += 1
|
|
565
566
|
_s.struct_tag = struct_num_max
|
|
@@ -567,7 +568,7 @@ def log_unique_structures(
|
|
|
567
568
|
with open(file_name, "a") as f:
|
|
568
569
|
print("unique_structures:", file=f)
|
|
569
570
|
for is_ghost in [False, True]:
|
|
570
|
-
for idx, _str in enumerate(
|
|
571
|
+
for idx, _str in enumerate(struct_rep_list):
|
|
571
572
|
if energy_min is not None:
|
|
572
573
|
e_diff = round((enthalpy_lists[idx] - energy_min) * 1000, 2)
|
|
573
574
|
if (not is_ghost and e_diff < -300) or (
|
|
@@ -617,7 +618,7 @@ def log_unique_structures(
|
|
|
617
618
|
|
|
618
619
|
def log_all_unique_structures(
|
|
619
620
|
file_name,
|
|
620
|
-
unique_structs: list[list[
|
|
621
|
+
unique_structs: list[list[StructRep]],
|
|
621
622
|
unique_str_prop: Optional[list[list[dict]]] = None,
|
|
622
623
|
unique_other_prop=None,
|
|
623
624
|
):
|
|
@@ -0,0 +1,187 @@
|
|
|
1
|
+
import time
|
|
2
|
+
from typing import Literal
|
|
3
|
+
|
|
4
|
+
import numpy as np
|
|
5
|
+
from scipy.optimize import minimize
|
|
6
|
+
from scipy.sparse import csr_array
|
|
7
|
+
from scipy.sparse.csgraph import maximum_bipartite_matching
|
|
8
|
+
|
|
9
|
+
|
|
10
|
+
def sort_match_screen(
|
|
11
|
+
pos_reps1: np.ndarray,
|
|
12
|
+
pos_reps2: np.ndarray,
|
|
13
|
+
elements1: np.ndarray,
|
|
14
|
+
elements2: np.ndarray,
|
|
15
|
+
axis_abc: np.ndarray,
|
|
16
|
+
pos_tol: float,
|
|
17
|
+
verbose: bool = False,
|
|
18
|
+
):
|
|
19
|
+
# pos1: (N1, n_atoms, 3)
|
|
20
|
+
# pos2: (N2, n_atoms, 3)
|
|
21
|
+
if len(elements1) != len(elements2):
|
|
22
|
+
return np.zeros((len(pos_reps1), len(pos_reps2)), dtype=bool)
|
|
23
|
+
unique_elements = np.unique(elements1)
|
|
24
|
+
|
|
25
|
+
def sort_positions_axis(pos_reps, elements, axis):
|
|
26
|
+
pos_reps = np.asarray(pos_reps)
|
|
27
|
+
reps_diff = np.empty_like(pos_reps[:, :, axis])
|
|
28
|
+
start = 0
|
|
29
|
+
for el in unique_elements:
|
|
30
|
+
idx = np.flatnonzero(elements == el)
|
|
31
|
+
sorted_reps = np.sort(pos_reps[:, idx, axis], axis=1)
|
|
32
|
+
diff = np.roll(sorted_reps, -1, axis=1) - sorted_reps
|
|
33
|
+
diff[:, -1] += axis_abc[axis]
|
|
34
|
+
reps_diff[:, start : start + len(idx)] = np.sort(diff, axis=1)
|
|
35
|
+
start += len(idx)
|
|
36
|
+
return reps_diff
|
|
37
|
+
|
|
38
|
+
matched = np.ones(
|
|
39
|
+
(len(pos_reps1), len(pos_reps2)),
|
|
40
|
+
dtype=bool,
|
|
41
|
+
)
|
|
42
|
+
if verbose:
|
|
43
|
+
print("code: sort_match_screen (all candidates)")
|
|
44
|
+
print(f" Candidate matches: {np.count_nonzero(matched)}")
|
|
45
|
+
for axis in range(3):
|
|
46
|
+
rows, cols = np.where(matched)
|
|
47
|
+
if len(rows) == 0:
|
|
48
|
+
break
|
|
49
|
+
|
|
50
|
+
t_start_s = time.time()
|
|
51
|
+
pos1_axis = sort_positions_axis(pos_reps1, elements1, axis)
|
|
52
|
+
pos2_axis = sort_positions_axis(pos_reps2, elements2, axis)
|
|
53
|
+
if verbose:
|
|
54
|
+
print(f"code: sort_match_screen (axis {axis})")
|
|
55
|
+
print(f" Sorting: {round((time.time() - t_start_s) * 1000, 2)} ms")
|
|
56
|
+
|
|
57
|
+
t_start_m = time.time()
|
|
58
|
+
diff = pos1_axis[rows] - pos2_axis[cols]
|
|
59
|
+
diff -= axis_abc[axis] * np.round(diff / axis_abc[axis])
|
|
60
|
+
matched[rows, cols] = np.all(np.abs(diff) <= pos_tol, axis=1)
|
|
61
|
+
if verbose:
|
|
62
|
+
print(f" Matching: {round((time.time() - t_start_m) * 1000, 2)} ms")
|
|
63
|
+
print(f" Candidate matches: {np.count_nonzero(matched)}")
|
|
64
|
+
return matched
|
|
65
|
+
|
|
66
|
+
|
|
67
|
+
def position_matching(
|
|
68
|
+
pos1: np.ndarray,
|
|
69
|
+
pos2: np.ndarray,
|
|
70
|
+
elements1: np.ndarray,
|
|
71
|
+
elements2: np.ndarray,
|
|
72
|
+
axis_abc: np.ndarray,
|
|
73
|
+
pos_tol: float,
|
|
74
|
+
assign_method: Literal["greedy", "HK"] = "greedy",
|
|
75
|
+
verbose: bool = False,
|
|
76
|
+
):
|
|
77
|
+
max_diff = 0
|
|
78
|
+
t_start_g = time.time()
|
|
79
|
+
for el in np.unique(elements1):
|
|
80
|
+
targets1 = np.where(elements1 == el)[0]
|
|
81
|
+
targets2 = np.where(elements2 == el)[0]
|
|
82
|
+
pos1_sub = pos1[targets1]
|
|
83
|
+
pos2_sub = pos2[targets2]
|
|
84
|
+
|
|
85
|
+
col_ind = []
|
|
86
|
+
inv_axis_abc = 1 / axis_abc
|
|
87
|
+
pos_d = pos1_sub[:, None, :] - pos2_sub[None, :, :]
|
|
88
|
+
pos_d -= axis_abc * np.round(pos_d * inv_axis_abc)
|
|
89
|
+
for _ in range(10):
|
|
90
|
+
if assign_method == "greedy":
|
|
91
|
+
pos_d_res, matching = greedy_matching(pos_d, pos_tol=pos_tol * 2)
|
|
92
|
+
elif assign_method == "HK":
|
|
93
|
+
pos_d_res, matching = HK_matching(pos_d, pos_tol=pos_tol * 2)
|
|
94
|
+
else:
|
|
95
|
+
raise ValueError(f"Unknown assignment method: {assign_method}")
|
|
96
|
+
|
|
97
|
+
if matching is None:
|
|
98
|
+
return False, None
|
|
99
|
+
|
|
100
|
+
if np.max(np.abs(pos_d_res)) < pos_tol:
|
|
101
|
+
if max_diff < np.max(np.abs(pos_d_res)):
|
|
102
|
+
max_diff = np.max(np.abs(pos_d_res))
|
|
103
|
+
break
|
|
104
|
+
if np.array_equal(matching, col_ind):
|
|
105
|
+
return False, None
|
|
106
|
+
|
|
107
|
+
col_ind = matching
|
|
108
|
+
translation_correction, max_displacement = optimize_translation(pos_d_res)
|
|
109
|
+
pos_d = pos_d - translation_correction
|
|
110
|
+
|
|
111
|
+
if verbose:
|
|
112
|
+
print("code: position_matching")
|
|
113
|
+
print(f" Matching: {round((time.time() - t_start_g) * 1000, 2)} ms")
|
|
114
|
+
|
|
115
|
+
return True, max_diff
|
|
116
|
+
|
|
117
|
+
|
|
118
|
+
def greedy_matching(pos_d, pos_tol):
|
|
119
|
+
pos_d_res = pos_d.copy()
|
|
120
|
+
pos_d_max = np.max(np.abs(pos_d), axis=2)
|
|
121
|
+
|
|
122
|
+
sorted_pos_d = np.sort(pos_d, axis=0)
|
|
123
|
+
pos_d_minmax = np.max(sorted_pos_d[0])
|
|
124
|
+
if pos_d_minmax > pos_tol:
|
|
125
|
+
return None, None
|
|
126
|
+
|
|
127
|
+
used1 = np.zeros(pos_d_max.shape[0], dtype=bool)
|
|
128
|
+
matching = np.full(len(pos_d_max), -1, dtype=int)
|
|
129
|
+
while not np.all(used1):
|
|
130
|
+
i, j = np.unravel_index(
|
|
131
|
+
np.argmin(pos_d_max),
|
|
132
|
+
pos_d_max.shape,
|
|
133
|
+
)
|
|
134
|
+
if pos_d_max[i, j] >= pos_tol:
|
|
135
|
+
return None, None
|
|
136
|
+
matching[i] = j
|
|
137
|
+
pos_d_max[i, :] = np.inf
|
|
138
|
+
pos_d_max[:, j] = np.inf
|
|
139
|
+
used1[i] = True
|
|
140
|
+
pos_d_res = pos_d_res[np.arange(len(matching)), matching]
|
|
141
|
+
return pos_d_res, matching
|
|
142
|
+
|
|
143
|
+
|
|
144
|
+
def HK_matching(pos_d, pos_tol):
|
|
145
|
+
pos_d_res = pos_d.copy()
|
|
146
|
+
pos_d_max = np.max(np.abs(pos_d), axis=2)
|
|
147
|
+
|
|
148
|
+
sorted_pos_d = np.sort(pos_d, axis=0)
|
|
149
|
+
pos_d_minmax = np.max(sorted_pos_d[0])
|
|
150
|
+
if pos_d_minmax > pos_tol:
|
|
151
|
+
return None, None
|
|
152
|
+
|
|
153
|
+
rows, cols = np.where(pos_d_max <= pos_tol)
|
|
154
|
+
graph = csr_array(
|
|
155
|
+
(
|
|
156
|
+
np.ones(len(rows), dtype=np.int8),
|
|
157
|
+
(rows, cols),
|
|
158
|
+
),
|
|
159
|
+
shape=pos_d_max.shape,
|
|
160
|
+
)
|
|
161
|
+
matching = maximum_bipartite_matching(graph, perm_type="column")
|
|
162
|
+
if not np.all(matching >= 0):
|
|
163
|
+
return None, None
|
|
164
|
+
|
|
165
|
+
pos_d_res = pos_d_res[np.arange(len(matching)), matching]
|
|
166
|
+
return pos_d_res, matching
|
|
167
|
+
|
|
168
|
+
|
|
169
|
+
def optimize_translation(diffs):
|
|
170
|
+
"""Find a translation minimizing the maximum displacement."""
|
|
171
|
+
|
|
172
|
+
def objective(translation):
|
|
173
|
+
corrected = diffs - translation
|
|
174
|
+
return np.max(np.linalg.norm(corrected, axis=1))
|
|
175
|
+
|
|
176
|
+
diffs = np.asarray(diffs, dtype=float)
|
|
177
|
+
initial_translation = np.mean(diffs, axis=0)
|
|
178
|
+
result = minimize(
|
|
179
|
+
objective,
|
|
180
|
+
initial_translation,
|
|
181
|
+
method="Nelder-Mead",
|
|
182
|
+
options={"xatol": 1e-10, "maxiter": 10000},
|
|
183
|
+
)
|
|
184
|
+
translation = result.x
|
|
185
|
+
max_displacement = objective(translation)
|
|
186
|
+
|
|
187
|
+
return translation, max_displacement
|
|
@@ -0,0 +1,24 @@
|
|
|
1
|
+
from dataclasses import dataclass
|
|
2
|
+
from typing import Optional
|
|
3
|
+
|
|
4
|
+
import numpy as np
|
|
5
|
+
|
|
6
|
+
|
|
7
|
+
@dataclass
|
|
8
|
+
class StructRep:
|
|
9
|
+
axis: np.ndarray
|
|
10
|
+
metric_tensor: np.ndarray
|
|
11
|
+
reduced_posreps: dict
|
|
12
|
+
all_posreps: dict
|
|
13
|
+
elements: np.ndarray
|
|
14
|
+
spg_number: int
|
|
15
|
+
spg_list: list[str]
|
|
16
|
+
original_axis: np.ndarray
|
|
17
|
+
original_positions: np.ndarray
|
|
18
|
+
original_elements: np.ndarray
|
|
19
|
+
tol_sm: float
|
|
20
|
+
tol_redpos: list[float]
|
|
21
|
+
normalize_axis: bool
|
|
22
|
+
struct_path: Optional[str]
|
|
23
|
+
struct_tag: Optional[int]
|
|
24
|
+
dupstr_paths: set[str]
|