Why models need to get fully uploaded in memory? (Noob question)

Hi, I am new to the field of deep learning and I got basically overwhelmed by the amount of memory those models require.

However, having a CS background, I remembered about an OS course where they talked about Pagination and (for those who don't know) it enables programs that are larger than available RAM to be executed with some tables that reference memory address on the disk, so those can be loaded into memory only when the program needs them …

course deep learning deeplearning however memory question