Source code for smt_optim.utils.inject_doe
import numpy as np
from smt_optim.core import Sample
from smt_optim.utils.constraints import compute_rscv
[docs]
def inject_data(state, xt, yt, ct=None):
"""
Injects matrix-form data (X, Y, and optionally Constraints) into the State.
"""
# 1. Ensure inputs are 2D arrays
xt = np.atleast_2d(xt)
yt = np.atleast_2d(yt)
# Handle constraints
num_samples = xt.shape[0]
if ct is None:
ct = np.empty((num_samples, 0))
else:
ct = np.atleast_2d(ct)
# 2. Add to dataset
for i in range(num_samples):
# We compute RSCV because export_as_dict will crash without it
rscv_val = compute_rscv(ct[i : i + 1, :], state.problem.cstr_configs).item()
sample = Sample(
x=xt[i],
fidelity=0,
obj=yt[i],
cstr=ct[i] if ct.size > 0 else np.array([]),
eval_time=np.zeros(yt.shape[1] + ct.shape[1]),
metadata={"rscv": rscv_val, "iter": 0, "budget": 0.0},
)
state.dataset.add(sample)
# Update state metadata if necessary
state.iter = 0
# Backward compatibility alias
InjectData = inject_data