Spaces:
Running
Running
File size: 7,405 Bytes
79edee4 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 |
from pathlib import Path
import numpy as np
import pandas as pd
from ase.db import connect
from scipy import stats
from mlip_arena.models import REGISTRY, MLIPEnum
DATA_DIR = Path(__file__).parent.absolute()
def load_wbm_structures():
"""
Load the WBM structures from a ASE DB file.
"""
with connect(DATA_DIR.parent / "wbm_structures.db") as db:
for row in db.select():
yield row.toatoms(add_additional_information=True)
def gather_results():
for model in MLIPEnum:
if "wbm_ev" not in REGISTRY[model.name].get("gpu-tasks", []):
continue
if (DATA_DIR / f"{model.name}.parquet").exists():
continue
all_data = []
for atoms in load_wbm_structures():
fpath = Path(model.name) / f"{atoms.info['key_value_pairs']['wbm_id']}.json"
if not fpath.exists():
continue
all_data.append(pd.read_json(fpath))
df = pd.concat(all_data, ignore_index=True)
df.to_parquet(DATA_DIR / f"{model.name}.parquet")
def summarize():
summary_table = pd.DataFrame(
columns=[
"model",
"energy-diff-flip-times",
"tortuosity",
"spearman-compression-energy",
"spearman-compression-derivative",
"spearman-tension-energy",
"missing",
]
)
for model in MLIPEnum:
fpath = DATA_DIR / f"{model.name}.parquet"
if not fpath.exists():
continue
df_raw_results = pd.read_parquet(fpath)
df_analyzed = pd.DataFrame(
columns=[
"model",
"structure",
"formula",
"volume-ratio",
"energy-delta-per-atom",
"energy-diff-flip-times",
"tortuosity",
"spearman-compression-energy",
"spearman-compression-derivative",
"spearman-tension-energy",
"missing",
]
)
for wbm_struct in load_wbm_structures():
structure_id = wbm_struct.info["key_value_pairs"]["wbm_id"]
try:
results = df_raw_results.loc[df_raw_results["id"] == structure_id]
results = results["eos"].values[0]
es = np.array(results["energies"])
vols = np.array(results["volumes"])
vol0 = wbm_struct.get_volume()
indices = np.argsort(vols)
vols = vols[indices]
es = es[indices]
imine = len(es) // 2
# min_center_val = np.min(es[imid - 1 : imid + 2])
# imine = np.where(es == min_center_val)[0][0]
emin = es[imine]
interpolated_volumes = [
(vols[i] + vols[i + 1]) / 2 for i in range(len(vols) - 1)
]
ediff = np.diff(es)
ediff_sign = np.sign(ediff)
mask = ediff_sign != 0
ediff = ediff[mask]
ediff_sign = ediff_sign[mask]
ediff_flip = np.diff(ediff_sign) != 0
etv = np.sum(np.abs(np.diff(es)))
data = {
"model": model.name,
"structure": structure_id,
"formula": wbm_struct.get_chemical_formula(),
"missing": False,
"volume-ratio": vols / vol0,
"energy-delta-per-atom": (es - emin) / len(wbm_struct),
"energy-diff-flip-times": np.sum(ediff_flip).astype(int),
"tortuosity": etv / (abs(es[0] - emin) + abs(es[-1] - emin)),
"spearman-compression-energy": stats.spearmanr(
vols[:imine], es[:imine]
).statistic,
"spearman-compression-derivative": stats.spearmanr(
interpolated_volumes[:imine], ediff[:imine]
).statistic,
"spearman-tension-energy": stats.spearmanr(
vols[imine:], es[imine:]
).statistic,
}
except Exception:
data = {
"model": model.name,
"structure": structure_id,
"formula": wbm_struct.get_chemical_formula(),
"missing": True,
"volume-ratio": None,
"energy-delta-per-atom": None,
"energy-diff-flip-times": None,
"tortuosity": None,
"spearman-compression-energy": None,
"spearman-compression-derivative": None,
"spearman-tension-energy": None,
}
df_analyzed = pd.concat([df_analyzed, pd.DataFrame([data])], ignore_index=True)
df_analyzed.to_parquet(DATA_DIR / f"{model.name}_processed.parquet")
# json_fpath = DATA_DIR / f"EV_scan_analyzed_{model.name}.json"
# df_analyzed.to_json(json_fpath, orient="records")
valid_results = df_analyzed[df_analyzed["missing"] == False]
analysis_summary = {
"model": model.name,
"energy-diff-flip-times": valid_results["energy-diff-flip-times"].mean(),
"tortuosity": valid_results["tortuosity"].mean(),
"spearman-compression-energy": valid_results[
"spearman-compression-energy"
].mean(),
"spearman-compression-derivative": valid_results[
"spearman-compression-derivative"
].mean(),
"spearman-tension-energy": valid_results["spearman-tension-energy"].mean(),
"missing": len(df_analyzed[df_analyzed["missing"] == True]),
}
summary_table = pd.concat(
[summary_table, pd.DataFrame([analysis_summary])], ignore_index=True
)
flip_rank = (
(summary_table["energy-diff-flip-times"] - 1)
.abs()
.rank(ascending=True, method="min")
)
tortuosity_rank = summary_table["tortuosity"].rank(ascending=True, method="min")
spearman_compression_energy_rank = summary_table["spearman-compression-energy"].rank(
method="min"
)
spearman_compression_derivative_rank = summary_table[
"spearman-compression-derivative"
].rank(ascending=False, method="min")
spearman_tension_energy_rank = summary_table["spearman-tension-energy"].rank(
ascending=False, method="min"
)
missing_rank = summary_table["missing"].rank(ascending=True, method="min")
rank_aggr = (
flip_rank
+ tortuosity_rank
+ spearman_compression_energy_rank
+ spearman_compression_derivative_rank
+ spearman_tension_energy_rank
+ missing_rank
)
rank = rank_aggr.rank(method="min")
summary_table.insert(1, "rank", rank.astype(int))
summary_table.insert(2, "rank-aggregation", rank_aggr.astype(int))
summary_table = summary_table.sort_values(by="rank", ascending=True)
summary_table = summary_table.reset_index(drop=True)
summary_table.to_csv(DATA_DIR / "summary.csv", index=False)
summary_table.to_latex(DATA_DIR / "summary.tex", index=False)
return summary_table
if __name__ == "__main__":
gather_results()
summarize()
|