Different sizes of SVD components when using numpy, sklearn (randomized SVD)

Question

I am trying to use python to calculate Singular Value Decomposition and I am confused about some results that come from different library: Numpy and Sklearn.