Unexpected behavior for float64 with InferenceSession sklearn

[maulberto3]

The following code won't work, onnx will complain that is expecting float32 when everything was float64: data, model and convert [ONNXRuntimeError] : 1 : FAIL : Load model from rf_iris.onnx failed:Type Error: Type (tensor(double)) of output arg (variable) of node (TreeEnsembleRegressor) does not match expected type (tensor(float)). .

Is this ok? Or maybe I am missing something?

from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor

from skl2onnx import convert_sklearn, to_onnx
from skl2onnx.common.data_types import DoubleTensorType

import onnxruntime as rt


# Train
X, y = make_regression(100, 10, bias=1, noise=1, random_state=16)  # float64
X.dtype, y.dtype
X_train, X_test, y_train, y_test = train_test_split(X, y)
rf = RandomForestRegressor(n_jobs=-1, random_state=16)
_ = rf.fit(X_train, y_train)
rf.predict(X_test).dtype  # float64

# Convert
init_types = [('double_input', DoubleTensorType([None, 10]))]
onx = convert_sklearn(rf, initial_types=init_types)
with open("rf_iris.onnx", "wb") as f:
    _ = f.write(onx.SerializeToString())

# Call
sess = rt.InferenceSession("rf_iris.onnx", )
input_name = sess.get_inputs()[0].name
label_name = sess.get_outputs()[0].name
sess.run(None, {input_name: X_test})[0]
sess.run([label_name], {input_name: X_test})[0]