Despite being primarily trained on text, these models achieved remarkable compression rates on image and audio data.
  • akrot@lemmy.worldEnglish
    31·
    1 year ago

    I wonder how consistent is the decompression and how much information is lost in the process.