Fix weights by putting the right value in `lm_head.weight` (#2)

Adding generation config file(s)

Initial commit

initial commit