extract LorA from pervious Checkpoint and based on the merged model uploaded,

try complex images