add 100BT trained models. remove 64x expansion models for now

add top-128-k models for 32x and 64x expansion

add some links to the readme

Update README.md

initial trained models

initial commit