-
Notifications
You must be signed in to change notification settings - Fork 0
/
LE_ps_old.py
170 lines (138 loc) · 6.01 KB
/
LE_ps_old.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
import string
from typing import Counter
import numpy as np
import torch
from equistore import TensorMap, Labels, TensorBlock
from rascaline import SphericalExpansion
def cut_to_LE(map: TensorMap, E_nl, E_max) -> TensorMap:
LE_blocks = []
for idx, block in map:
l = idx[0]
counter = 0
for n in block.properties["n"]:
if E_nl[n, l] <= E_max: counter += 1
LE_values = torch.zeros((block.values.shape[0], block.values.shape[1], counter))
counter_LE = 0
counter_total = 0
labels_LE = []
for n in block.properties["n"]:
if E_nl[n, l] <= E_max:
LE_values[:, :, counter_LE] = torch.tensor(block.values[:, :, counter_total])
labels_LE.append([block.properties["species_neighbor"][counter_total], n, l])
counter_LE += 1
counter_total += 1
LE_block = TensorBlock(
values=LE_values,
samples=block.samples,
components=block.components,
properties=Labels(
names = ("a1", "n1", "l1"),
values = np.array(labels_LE),
),
)
LE_blocks.append(LE_block)
return TensorMap(
keys = Labels(
names = ("lam", "a_i"),
values = map.keys.asarray(),
),
blocks = LE_blocks
)
def get_LE_expansion(structures, spline_file: string, E_nl, E_max, rcut) -> TensorMap:
n_max = np.where(E_nl[:, 0] <= E_max)[0][-1] + 1
l_max = np.where(E_nl[0, :] <= E_max)[0][-1]
hypers_spherical_expansion = {
"cutoff": rcut,
"max_radial": int(n_max),
"max_angular": int(l_max),
"center_atom_weight": 0.0,
"radial_basis": {"Tabulated": {"file": spline_file}},
"atomic_gaussian_width": 100.0,
"cutoff_function": {"Step": {}},
}
calculator = SphericalExpansion(**hypers_spherical_expansion)
spherical_expansion_coefficients = calculator.compute(structures)
all_species = np.unique(spherical_expansion_coefficients.keys["species_center"])
all_neighbor_species = Labels(
names=["species_neighbor"],
values=np.array(all_species, dtype=np.int32).reshape(-1, 1),
)
spherical_expansion_coefficients.keys_to_properties(all_neighbor_species)
n_max_l = []
for l in range(l_max+1):
n_max_l.append(np.where(E_nl[:, l] <= E_max)[0][-1]+1)
LE_spherical = cut_to_LE(spherical_expansion_coefficients, E_nl, E_max)
return LE_spherical
def get_LE_ps(structures, spline_file: string, E_nl, E_max_2, rcut) -> TensorMap:
E_max_1 = E_max_2 - E_nl[0, 0]
spherical_expansion = get_LE_expansion(structures, spline_file, E_nl, E_max_1, rcut)
all_species = np.unique(np.concatenate([spherical_expansion.keys["a_i"], spherical_expansion.keys["a_i"]])) # This may actually need to come from outside
l_max = 0
for idx, block in spherical_expansion:
l_max = max(l_max, idx[0])
n_max_l = []
a_max = 0
a_i = all_species[0]
for l in range(l_max+1):
old_block = spherical_expansion.block(lam=l, a_i=a_i)
a = old_block.properties["a1"]
n = old_block.properties["n1"]
a_max = np.max(a) + 1
n_max_l.append(np.max(n)+1)
combined_anl = {}
anl_counter = 0
for a in range(a_max):
for l in range(l_max+1):
for n in range(n_max_l[l]):
combined_anl[(a, n, l,)] = anl_counter
anl_counter += 1
blocks = []
for a_i in all_species:
soap_count = 0
for l in range(l_max+1):
old_block = spherical_expansion.block(lam=l, a_i=a_i)
a = old_block.properties["a1"]
n = old_block.properties["n1"]
for i in range(old_block.values.shape[-1]):
for j in range(old_block.values.shape[-1]):
if combined_anl[(a[i], n[i], l)] > combined_anl[(a[j], n[j], l)]: continue # Lexicographic
if E_nl[n[i], l] + E_nl[n[j], l] > E_max_2: continue # LE eigenvalue
soap_count += 1
data = torch.empty((len(old_block.samples), soap_count), device=old_block.values.device)
soap_count = 0 # reset counter
properties_names = (
[f"{name[:-1]}1" for name in old_block.properties.names]
+ [f"{name[:-1]}2" for name in old_block.properties.names]
)
properties_values = []
for l in range(l_max + 1): # loops over l to ensure consistent order, independent on key storage
old_block = spherical_expansion.block(lam=l, a_i=a_i)
a = old_block.properties["a1"]
n = old_block.properties["n1"]
soap_prefactor = 1.0 / np.sqrt(2 * l + 1)
for i in range(old_block.values.shape[-1]):
for j in range(old_block.values.shape[-1]):
if combined_anl[(a[i], n[i], l)] > combined_anl[(a[j], n[j], l)]: continue # Lexicographic
if E_nl[n[i], l] + E_nl[n[j], l] > E_max_2: continue # LE eigenvalue
multiplicity_factor = np.sqrt(2.0)
if combined_anl[(a[i], n[i], l)] == combined_anl[(a[j], n[j], l)]: multiplicity_factor = 1.0
properties_values.append([a[i], n[i], l, a[j], n[j], l])
data[:, soap_count] = multiplicity_factor*soap_prefactor*torch.sum(old_block.values[:, :, i]*old_block.values[:, :, j], dim = 1, keepdim = False)
soap_count += 1
block = TensorBlock(
values=data,
samples=old_block.samples,
components=[],
properties=Labels(
names=properties_names,
values=np.asarray(np.vstack(properties_values), dtype=np.int32),
),
)
blocks.append(block)
LE_ps = TensorMap(
keys = Labels(
names = ("a_i",),
values = np.array(all_species).reshape((-1, 1)),
),
blocks = blocks)
return LE_ps