Output of Round is confusing

[tungld]

Ask a Question

Question

In the definition of Round, it said it finds the nearest integer for each value, but according to the example

round([0.9]) = [1.0]
round([2.5]) = [2.0]
round([2.3]) = [2.0]
round([1.5]) = [2.0]
round([-4.5]) = [-4.0]

I am confused whether 2.0 is the nearest integer of 2.5 or 1.5. How was the nearest integer computed? Thanks!

Further information

Similar thing happened with the reference output of Round also: https://github.com/onnx/onnx/blob/master/docs/Operators.md#examples-110

• Is this issue related to a specific model?
  Model name (e.g. mnist):
  Model opset (e.g. 7):

Notes

Any additional information, code snippets.

[jcwchen]

Hi @tungld,
Actually the doc mentions this kind of scenario: In case of halfs, the rule is to round them to the nearest "even" integer. The output tensor has the same shape and type as the input. And this behavior is similar to numpy.around: https://numpy.org/doc/stable/reference/generated/numpy.around.html

[tungld]

Oh, I see, I was not aware of even. It would be helpful if we mention numpy.around in the doc. Thanks @jcwchen!