Mismatch between XGB Sklearn result and OMLT result

[Yunshan-Liu67]

Hi,
While using the OMLT XGB interface, I observed an interesting behavior when I tried to verify the results from it.
I constructed an optimization problem based on Pyomo containing only an XGB block with all the inputs value fixed, the objective function is set to be a constant value. The problem is solved with CPLEX and the value of the output is preserved.
Meanwhile, I give the same input value to the XGB model's predict method (link here) and it returns an output value. And these two output values, one from the Pyomo model and one from the XGB model itself, are not the same. I have tested this on multiple points and the symptom is the same. It would be great if you could provide some insights about this.