1 comments

  • ricardobeat 18 hours ago

    Love this kind of experiment. Would the model perform better with word tokens?