1 comments

  • ricardobeat 6 hours ago ago

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