Convert a pipeline with a XGBoost model

sklearn-onnx only converts scikit-learn models into ONNX but many libraries implement scikit-learn API so that their models can be included in a scikit-learn pipeline. This example considers a pipeline including a XGBoost model. sklearn-onnx can convert the whole pipeline as long as it knows the converter associated to a XGBClassifier. Let’s see how to do it.

Train a XGBoost classifier

from pyquickhelper.helpgen.graphviz_helper import plot_graphviz
from mlprodict.onnxrt import OnnxInference
import numpy
import onnxruntime as rt
from sklearn.datasets import load_iris, load_diabetes, make_classification
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from xgboost import XGBClassifier, XGBRegressor, DMatrix, train as train_xgb
from skl2onnx.common.data_types import FloatTensorType
from skl2onnx import convert_sklearn, to_onnx, update_registered_converter
from skl2onnx.common.shape_calculator import (
    calculate_linear_classifier_output_shapes,
    calculate_linear_regressor_output_shapes)
from onnxmltools.convert.xgboost.operator_converters.XGBoost import (
    convert_xgboost)
from onnxmltools.convert import convert_xgboost as convert_xgboost_booster


data = load_iris()
X = data.data[:, :2]
y = data.target

ind = numpy.arange(X.shape[0])
numpy.random.shuffle(ind)
X = X[ind, :].copy()
y = y[ind].copy()

pipe = Pipeline([('scaler', StandardScaler()),
                 ('xgb', XGBClassifier(n_estimators=3))])
pipe.fit(X, y)

# The conversion fails but it is expected.

try:
    convert_sklearn(pipe, 'pipeline_xgboost',
                    [('input', FloatTensorType([None, 2]))],
                    target_opset={'': 12, 'ai.onnx.ml': 2})
except Exception as e:
    print(e)

# The error message tells no converter was found
# for :epkg:`XGBoost` models. By default, :epkg:`sklearn-onnx`
# only handles models from :epkg:`scikit-learn` but it can
# be extended to every model following :epkg:`scikit-learn`
# API as long as the module knows there exists a converter
# for every model used in a pipeline. That's why
# we need to register a converter.

Register the converter for XGBClassifier

The converter is implemented in onnxmltools: onnxmltools…XGBoost.py. and the shape calculator: onnxmltools…Classifier.py.

update_registered_converter(
    XGBClassifier, 'XGBoostXGBClassifier',
    calculate_linear_classifier_output_shapes, convert_xgboost,
    options={'nocl': [True, False], 'zipmap': [True, False, 'columns']})

Convert again

model_onnx = convert_sklearn(
    pipe, 'pipeline_xgboost',
    [('input', FloatTensorType([None, 2]))],
    target_opset={'': 12, 'ai.onnx.ml': 2})

# And save.
with open("pipeline_xgboost.onnx", "wb") as f:
    f.write(model_onnx.SerializeToString())

Compare the predictions

Predictions with XGBoost.

print("predict", pipe.predict(X[:5]))
print("predict_proba", pipe.predict_proba(X[:1]))

predict [1 1 1 2 1]
predict_proba [[0.1758379  0.43438542 0.3897767 ]]

Predictions with onnxruntime.

sess = rt.InferenceSession("pipeline_xgboost.onnx")
pred_onx = sess.run(None, {"input": X[:5].astype(numpy.float32)})
print("predict", pred_onx[0])
print("predict_proba", pred_onx[1][:1])

predict [1 1 1 2 1]
predict_proba [{0: 0.175837904214859, 1: 0.43438541889190674, 2: 0.38977670669555664}]

Final graph

oinf = OnnxInference(model_onnx)
ax = plot_graphviz(oinf.to_dot())
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)

Same example with XGBRegressor

update_registered_converter(
    XGBRegressor, 'XGBoostXGBRegressor',
    calculate_linear_regressor_output_shapes, convert_xgboost)


data = load_diabetes()
x = data.data
y = data.target
X_train, X_test, y_train, _ = train_test_split(x, y, test_size=0.5)

pipe = Pipeline([('scaler', StandardScaler()),
                 ('xgb', XGBRegressor(n_estimators=3))])
pipe.fit(X_train, y_train)

print("predict", pipe.predict(X_test[:5]))

predict [73.99487 44.04601 77.94355 75.82603 44.04601]

ONNX

onx = to_onnx(pipe, X_train.astype(numpy.float32),
              target_opset={'': 12, 'ai.onnx.ml': 2})

sess = rt.InferenceSession(onx.SerializeToString())
pred_onx = sess.run(None, {"X": X_test[:5].astype(numpy.float32)})
print("predict", pred_onx[0].ravel())

predict [73.99487 44.04601 77.94355 75.82603 44.04601]

Some discrepencies may appear. In that case, you should read Issues when switching to float.

Same with a Booster

A booster cannot be inserted in a pipeline. It requires a different conversion function because it does not follow scikit-learn API.

x, y = make_classification(n_classes=2, n_features=5,
                           n_samples=100,
                           random_state=42, n_informative=3)
X_train, X_test, y_train, _ = train_test_split(x, y, test_size=0.5,
                                               random_state=42)

dtrain = DMatrix(X_train, label=y_train)

param = {'objective': 'multi:softmax', 'num_class': 3}
bst = train_xgb(param, dtrain, 10)

initial_type = [('float_input', FloatTensorType([None, X_train.shape[1]]))]

try:
    onx = convert_xgboost_booster(bst, "name", initial_types=initial_type)
    cont = True
except AssertionError as e:
    print("XGBoost is too recent or onnxmltools too old.", e)
    cont = False

if cont:
    sess = rt.InferenceSession(onx.SerializeToString())
    input_name = sess.get_inputs()[0].name
    label_name = sess.get_outputs()[0].name
    pred_onx = sess.run(
        [label_name], {input_name: X_test.astype(numpy.float32)})[0]
    print(pred_onx)

[0 0 1 1 0 1 0 1 0 1 0 0 1 1 1 0 0 1 1 1 1 0 0 1 0 0 0 1 1 1 0 1 1 0 1 1 1
 0 1 1 1 0 0 1 1 0 0 0 1 0]