onnx--onnx/onnx/reference

ReferenceEvaluator

This is a first attempt to provide an implementation for all operators defined by onnx. This is a pure python implementation. Mismatches may remain between the official specification and the implementation here. In the case of such a mismatch, the official spec overrides this implementation. The class can use any implementation available in folder ops. It covers most of the tests defined in onnx/backend/test/case and reported on ONNX Backend Scoreboard.

The class ReferenceEvaluator is used as follows:

import numpy as np
from onnx.reference import ReferenceEvaluator

X = np.array(...)
sess = ReferenceEvaluator("model.onnx")
results = sess.run(None, {"X": X})
print(results[0])

In addition to the implementation of every operator, it can be used to display intermediate results and help debugging a model.

import numpy as np
from onnx.reference import ReferenceEvaluator

X = np.array(...)
sess = ReferenceEvaluator("model.onnx", verbose=1)
results = sess.run(None, {"X": X})
print(results[0])