Skip to content

Commit e0303d8

Browse files
chiang-yuanchiang-yuan
committed
patch mof eos, plot and tex
1 parent 3397e50 commit e0303d8

17 files changed

+1642
-47
lines changed

‎examples/eos_bulk/CHGNet_processed.parquet‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:0416eeed1748994b67e8f6e9768a5f1d2a77c19f9512bc408f9b39ca3c19e3d4
3-
size 358042
2+
oid sha256:bfde7530e6b0d2df5a30e1b7e3ec124fb2a86f6da8e35d2548d37d10a1eff1b1
3+
size 387425

‎examples/eos_bulk/M3GNet_processed.parquet‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:1a34f8148f771f0b751f01ccc9d260fd5ae48b625b979ec2112e008c82c59a08
3-
size 379982
2+
oid sha256:eb43a3c74f3340100b1adb21b3f2d075451e1ffe88ac6d6662741bc4a0576eb8
3+
size 397450

‎examples/eos_bulk/MACE-MP(M)_processed.parquet‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:73e2b8ad6d5e114c1c0fea4697b17810182d0b273185512cb40fa894ea30b4c3
3-
size 371128
2+
oid sha256:7e5507cdc5fe558b5d3fe2ea8f1dd577ac444e82c5347b5fbe738a4f855dffcb
3+
size 397379

‎examples/eos_bulk/MACE-MPA_processed.parquet‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:819bc0c721e99df8dda0a4c6df565deb96736ecc5ceefefe300e5b72b7d6312f
3-
size 365412
2+
oid sha256:0f3032d5a156febdd9580fa3d86cb1a84236374bcac6ccb22d18a948767db502
3+
size 394748

‎examples/eos_bulk/MatterSim_processed.parquet‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:c707ffb285f03a5c7d1486a6998c787088f07a97b206585b17839fff4fab49b4
3-
size 321086
2+
oid sha256:fb1f10a60495f5e88ea8cf737fd7b47d1c471fda422374ee519d14f531c732f8
3+
size 290191

‎examples/eos_bulk/ORBv2_processed.parquet‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:f794da83d1031823577b085c480f7d285520c086bdd0e7e6e7acb7a5a2457329
3-
size 228052
2+
oid sha256:7eb0a3060b8a2d3541b8fb1083176c88aae0a8be0008e84d5770998b01742216
3+
size 402554

‎examples/eos_bulk/SevenNet_processed.parquet‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:a9aab95402aa62169ba6f1e12a7774362b3e5cc027f5c556de734783e6d6f29b
3-
size 364969
2+
oid sha256:9f484928f5086e8d1411a198ac69bfe44313597909b72c8676e9131cce1660f1
3+
size 398295

‎examples/eos_bulk/analyze.py‎

Lines changed: 216 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,216 @@
1+
from pathlib import Path
2+
3+
import numpy as np
4+
import pandas as pd
5+
from ase.db import connect
6+
from scipy import stats
7+
8+
from mlip_arena.models import REGISTRY, MLIPEnum
9+
10+
DATA_DIR = Path(__file__).parent.absolute()
11+
12+
13+
def load_wbm_structures():
14+
"""
15+
Load the WBM structures from a ASE DB file.
16+
"""
17+
with connect(DATA_DIR.parent / "wbm_structures.db") as db:
18+
for row in db.select():
19+
yield row.toatoms(add_additional_information=True)
20+
21+
def gather_results():
22+
for model in MLIPEnum:
23+
if "eos_bulk" not in REGISTRY[model.name].get("gpu-tasks", []):
24+
continue
25+
26+
if (DATA_DIR / f"{model.name}.parquet").exists():
27+
continue
28+
29+
all_data = []
30+
31+
for atoms in load_wbm_structures():
32+
fpath = Path(model.name) / f"{atoms.info['key_value_pairs']['wbm_id']}.pkl"
33+
if not fpath.exists():
34+
continue
35+
36+
all_data.append(pd.read_pickle(fpath))
37+
38+
df = pd.concat(all_data, ignore_index=True)
39+
df.to_parquet(DATA_DIR / f"{model.name}.parquet")
40+
41+
42+
def summarize():
43+
summary_table = pd.DataFrame(
44+
columns=[
45+
"model",
46+
"energy-diff-flip-times",
47+
"tortuosity",
48+
"spearman-compression-energy",
49+
"spearman-compression-derivative",
50+
"spearman-tension-energy",
51+
"missing",
52+
]
53+
)
54+
55+
56+
for model in MLIPEnum:
57+
fpath = DATA_DIR / f"{model.name}.parquet"
58+
if not fpath.exists():
59+
continue
60+
df_raw_results = pd.read_parquet(fpath)
61+
62+
df_analyzed = pd.DataFrame(
63+
columns=[
64+
"model",
65+
"structure",
66+
"formula",
67+
"volume-ratio",
68+
"energy-delta-per-atom",
69+
"energy-diff-flip-times",
70+
"energy-delta-per-volume-b0",
71+
"tortuosity",
72+
"spearman-compression-energy",
73+
"spearman-compression-derivative",
74+
"spearman-tension-energy",
75+
"missing",
76+
]
77+
)
78+
79+
for wbm_struct in load_wbm_structures():
80+
structure_id = wbm_struct.info["key_value_pairs"]["wbm_id"]
81+
82+
try:
83+
results = df_raw_results.loc[df_raw_results["id"] == structure_id]
84+
b0 = results["b0"].values[0]
85+
# vol0 = results["v0"].values[0]
86+
results = results["eos"].values[0]
87+
es = np.array(results["energies"])
88+
vols = np.array(results["volumes"])
89+
90+
indices = np.argsort(vols)
91+
vols = vols[indices]
92+
es = es[indices]
93+
94+
imine = len(es) // 2
95+
# min_center_val = np.min(es[imid - 1 : imid + 2])
96+
# imine = np.where(es == min_center_val)[0][0]
97+
emin = es[imine]
98+
vol0 = vols[imine]
99+
100+
interpolated_volumes = [
101+
(vols[i] + vols[i + 1]) / 2 for i in range(len(vols) - 1)
102+
]
103+
ediff = np.diff(es)
104+
ediff_sign = np.sign(ediff)
105+
mask = ediff_sign != 0
106+
ediff = ediff[mask]
107+
ediff_sign = ediff_sign[mask]
108+
ediff_flip = np.diff(ediff_sign) != 0
109+
110+
etv = np.sum(np.abs(np.diff(es)))
111+
112+
data = {
113+
"model": model.name,
114+
"structure": structure_id,
115+
"formula": wbm_struct.get_chemical_formula(),
116+
"missing": False,
117+
"volume-ratio": vols / vol0,
118+
"energy-delta-per-atom": (es - emin) / len(wbm_struct),
119+
"energy-diff-flip-times": np.sum(ediff_flip).astype(int),
120+
"energy-delta-per-volume-b0": (es - emin) / (b0*vol0),
121+
"tortuosity": etv / (abs(es[0] - emin) + abs(es[-1] - emin)),
122+
"spearman-compression-energy": stats.spearmanr(
123+
vols[:imine], es[:imine]
124+
).statistic,
125+
"spearman-compression-derivative": stats.spearmanr(
126+
interpolated_volumes[:imine], ediff[:imine]
127+
).statistic,
128+
"spearman-tension-energy": stats.spearmanr(
129+
vols[imine:], es[imine:]
130+
).statistic,
131+
}
132+
133+
except Exception as e:
134+
print(e)
135+
data = {
136+
"model": model.name,
137+
"structure": structure_id,
138+
"formula": wbm_struct.get_chemical_formula(),
139+
"missing": True,
140+
"volume-ratio": None,
141+
"energy-delta-per-atom": None,
142+
"energy-delta-per-volume-b0": None,
143+
"energy-diff-flip-times": None,
144+
"tortuosity": None,
145+
"spearman-compression-energy": None,
146+
"spearman-compression-derivative": None,
147+
"spearman-tension-energy": None,
148+
}
149+
150+
df_analyzed = pd.concat([df_analyzed, pd.DataFrame([data])], ignore_index=True)
151+
152+
df_analyzed.to_parquet(DATA_DIR / f"{model.name}_processed.parquet")
153+
# json_fpath = DATA_DIR / f"EV_scan_analyzed_{model.name}.json"
154+
155+
# df_analyzed.to_json(json_fpath, orient="records")
156+
157+
valid_results = df_analyzed[df_analyzed["missing"] == False]
158+
159+
analysis_summary = {
160+
"model": model.name,
161+
"energy-diff-flip-times": valid_results["energy-diff-flip-times"].mean(),
162+
"tortuosity": valid_results["tortuosity"].mean(),
163+
"spearman-compression-energy": valid_results[
164+
"spearman-compression-energy"
165+
].mean(),
166+
"spearman-compression-derivative": valid_results[
167+
"spearman-compression-derivative"
168+
].mean(),
169+
"spearman-tension-energy": valid_results["spearman-tension-energy"].mean(),
170+
"missing": len(df_analyzed[df_analyzed["missing"] == True]),
171+
}
172+
summary_table = pd.concat(
173+
[summary_table, pd.DataFrame([analysis_summary])], ignore_index=True
174+
)
175+
176+
177+
flip_rank = (
178+
(summary_table["energy-diff-flip-times"] - 1)
179+
.abs()
180+
.rank(ascending=True, method="min")
181+
)
182+
tortuosity_rank = summary_table["tortuosity"].rank(ascending=True, method="min")
183+
spearman_compression_energy_rank = summary_table["spearman-compression-energy"].rank(
184+
method="min"
185+
)
186+
spearman_compression_derivative_rank = summary_table[
187+
"spearman-compression-derivative"
188+
].rank(ascending=False, method="min")
189+
spearman_tension_energy_rank = summary_table["spearman-tension-energy"].rank(
190+
ascending=False, method="min"
191+
)
192+
missing_rank = summary_table["missing"].rank(ascending=True, method="min")
193+
194+
rank_aggr = (
195+
flip_rank
196+
+ tortuosity_rank
197+
+ spearman_compression_energy_rank
198+
+ spearman_compression_derivative_rank
199+
+ spearman_tension_energy_rank
200+
+ missing_rank
201+
)
202+
rank = rank_aggr.rank(method="min")
203+
204+
summary_table.insert(1, "rank", rank.astype(int))
205+
summary_table.insert(2, "rank-aggregation", rank_aggr.astype(int))
206+
summary_table = summary_table.sort_values(by="rank", ascending=True)
207+
summary_table = summary_table.reset_index(drop=True)
208+
209+
summary_table.to_csv(DATA_DIR / "summary.csv", index=False)
210+
summary_table.to_latex(DATA_DIR / "summary.tex", index=False)
211+
212+
return summary_table
213+
214+
if __name__ == "__main__":
215+
gather_results()
216+
summarize()

‎examples/eos_bulk/eSEN_processed.parquet‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,3 @@
11
version https://git-lfs.github.com/spec/v1
2-
oid sha256:3d12b36a2bd465e16ada4363e31756d5de5d41dd890d0e88e8ca86b76dd66336
3-
size 313235
2+
oid sha256:d7f754d8e18f645c1608e86286245c11611d5af34f3bd0bbc4a5b63b851a0dee
3+
size 393790

‎examples/eos_bulk/eos-bulk-grid.pdf‎

888 KB
Binary file not shown.

0 commit comments

Comments
 (0)