What is the fastest way to calculate the average vector from a list of sparse vectors?

Question

What I have:

A Python list of some NumPy sparse csr_matrix matrixes with shape 1x5000. They are actually numerical vectors.

eg:

vec