GaussianProcessClassifier fails during ONNX runtime inference with "com.microsoft:Solve(-1) is not a registered function/op

[vassilismin]

Describe the issue

When converting a scikit-learn GaussianProcessClassifier to ONNX format and attempting to run inference with ONNX Runtime, the following error occurs:

Any help or feedback with the following issue will be much appreciated!

onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : Fatal error: com.microsoft:Solve(-1) is not a registered function/op

Environment

• Python version: 3.10
• scikit-learn version: 1.5.0
• skl2onnx version: 1.17.0
• onnxruntime version: 1.19.2
• OS: macOS 15.3.2

To reproduce

Steps to reproduce

import numpy as np
from sklearn.gaussian_process import GaussianProcessClassifier
from sklearn.gaussian_process.kernels import RBF
from skl2onnx import convert_sklearn
from skl2onnx.common.data_types import FloatTensorType
import onnxruntime as rt

# Generate a simple binary classification dataset with 20 points
np.random.seed(42)
X = np.random.randn(20, 2)  # 20 points with 2 features
y = (X[:, 0] + X[:, 1] > 0).astype(int)  # Simple decision boundary: x + y > 0

# Train a Gaussian Process Classifier
kernel = 1.0 * RBF(length_scale=1.0)
gpc = GaussianProcessClassifier(kernel=kernel, random_state=42)
gpc.fit(X, y)

# Convert the model to ONNX
initial_type = [("float_input", FloatTensorType([None, 2]))]
onx = convert_sklearn(gpc, initial_types=initial_type)

# Create a single data point for prediction
test_point = np.array([[0.5, 0.5]], dtype=np.float32)

# Print the sklearn prediction for comparison
sklearn_pred = gpc.predict(test_point)
sklearn_proba = gpc.predict_proba(test_point)
print(f"Sklearn prediction: {sklearn_pred}")
print(f"Sklearn probabilities: {sklearn_proba}")

# Make prediction with ONNX runtime
# Create an ONNX runtime session
sess = rt.InferenceSession(onx.SerializeToString())
input_name = sess.get_inputs()[0].name
pred_onx = sess.run(None, {input_name: test_point})
print(f"ONNX prediction: {pred_onx[0]}")
print(f"ONNX probabilities: {pred_onx[1]}")

Urgency

No response

Platform

Mac

OS Version

15.3.2

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.19.2

ONNX Runtime API

Python

Architecture

X64

Execution Provider

Default CPU

Execution Provider Library Version

No response