CReferenceEvaluator

Class CReferenceEvaluator extends onnx.reference.ReferenceEvaluator with custom operators implemented in C++ in order to speed up the evaluation of this python runtime. This class works as a replacement of ReferenceEvaluator. It rewrites the following kernels:

<<<

import pprint
from onnx_extended.reference import CReferenceEvaluator

pprint.pprint(CReferenceEvaluator.default_ops())

>>>

    [<class 'onnx_extended.reference.c_ops.c_op_conv.Conv'>,
     <class 'onnx_extended.reference.c_ops.c_op_tree_ensemble_classifier.TreeEnsembleClassifier_1'>,
     <class 'onnx_extended.reference.c_ops.c_op_tree_ensemble_classifier.TreeEnsembleClassifier_3'>,
     <class 'onnx_extended.reference.c_ops.c_op_tree_ensemble_regressor.TreeEnsembleRegressor_1'>,
     <class 'onnx_extended.reference.c_ops.c_op_tree_ensemble_regressor.TreeEnsembleRegressor_3'>,
     <class 'onnx_extended.reference.c_custom_ops.custom_op_tree_ensemble_regressor.TreeEnsembleRegressor_1'>,
     <class 'onnx_extended.reference.c_custom_ops.custom_op_tree_ensemble_regressor.TreeEnsembleRegressor_3'>]

Full example:

<<<

import numpy as np
from onnx import TensorProto
from onnx.helper import (
    make_graph,
    make_model,
    make_node,
    make_opsetid,
    make_tensor_value_info,
)
from onnx.reference import ReferenceEvaluator
from onnxruntime import InferenceSession
from onnx_extended.ext_test_case import measure_time
from onnx_extended.reference import CReferenceEvaluator

X = make_tensor_value_info("X", TensorProto.FLOAT, [None, None, None, None])
Y = make_tensor_value_info("Y", TensorProto.FLOAT, [None, None, None, None])
B = make_tensor_value_info("B", TensorProto.FLOAT, [None, None, None, None])
W = make_tensor_value_info("W", TensorProto.FLOAT, [None, None, None, None])
node = make_node(
    "Conv",
    ["X", "W", "B"],
    ["Y"],
    pads=[1, 1, 1, 1],
    dilations=[1, 1],
    strides=[2, 2],
)
graph = make_graph([node], "g", [X, W, B], [Y])
onnx_model = make_model(graph, opset_imports=[make_opsetid("", 16)])

sH, sW = 64, 64
X = np.arange(sW * sH).reshape((1, 1, sH, sW)).astype(np.float32)
W = np.ones((1, 1, 3, 3), dtype=np.float32)
B = np.array([[[[0]]]], dtype=np.float32)

sess1 = ReferenceEvaluator(onnx_model)
sess2 = CReferenceEvaluator(onnx_model)  # 100 times faster

expected = sess1.run(None, {"X": X, "W": W, "B": B})[0]
got = sess2.run(None, {"X": X, "W": W, "B": B})[0]
diff = np.abs(expected - got).max()
print(f"difference: {diff}")

>>>

    difference: 0.0