Surrogate

Core model class — train, predict, export.

from prandtl import Surrogate

surrogate = Surrogate(
    params=["alpha", "mach"],
    outputs=["CL", "CD"],
    method="gp"
)

Parameter	Type	Default	Description
params	list[str]	—	Input parameter names
outputs	list[str]	—	Output variable names
method	str	"gp"	"gp" or "mlp"

Methods

fit(X, Y, **kwargs)

Train the surrogate.

surrogate.fit(X, Y)
surrogate.fit(X, Y, n_iter=500, lr=0.01, physics=constraints)  # MLP opts

predict(X)

Predict on new inputs.

Y_pred = surrogate.predict(X_test)
# → {"CL": array([...]), "CD": array([...])}

validate(X, Y)

Quick validation on a test set.

report = surrogate.validate(X_test, Y_test)
# → {"CL": {"r2": 0.999, "rmse": 0.001, ...}, ...}

export(path)

Export to ONNX (MLP only).

surrogate.export("model")
# → model__CL.onnx, model__CD.onnx