Large sparse matrix, svd with sparks, python

Question

I want to analyze spark data. I need an svd matrix to get a recommendation algorithm using python or scala if python is not working. But the data is large and rare.

there are two columns in the data. one username, the other is the name of the element. If the user and the item are on the same line, this means that the user likes the item. There are 700 thousand elements and 20 thousand users. So this is a matter of memory.

How to deal with it