Despite being primarily trained on text, these models achieved remarkable compression rates on image and audio data.
  • Hexagon
    5·
    2 years ago

    … and this is why I should actually read the articles before commenting lol