Different results between Tensorflow model and Numpy model

Why do we have different results although we have used the same weights in both models


There might be a problem with the implementation of the Numpy version of the model, it seems to only predict 0.

But it passed all tests!