GPT-J-6B – A 6 billion parameter, autoregressive text generation model

Re:GPT-J-6B – A 6 billion parameter, autoregressive text gen... by gantagonist on Sunday July 04, 2021 @06:04AM (#27728513)

This is fun! It took some prompt crafting to get actual "ASCII art" out of the model (with settings suggested elsewhere in these comments):

My name is GPT-J-6B and between the six backticks below you can see my face in 400 characters of ASCII art:

``` ┌─────┐ │←┤ ││(◣◡)│ │└─◡◢┘ │─┆┌─◢┌─┈┐ ││││(|)││ │││(O)││ │└─└─┘─┘──┐ │└─└─┘─└─┘ │└─└─┘─└─┘ │└─└─┘─└─┘ └─└─┘─└─┘ ```

[ Parent | on HN ]

Re:GPT-J-6B – A 6 billion parameter, autoregressive text gen... by mdaniel on Sunday July 04, 2021 @02:10PM (#27731521)

HN markup(sic) is just beyond dumb (and your edit window has likely expired) but in the future you'll want to indent preformatted things by 4 spaces because the backticks are unconditionally ignored

Ironically, view-source on your comment still had the correct line layout, although it seems to have eaten the intra-element spacing, so:

       ┌─────┐
       │←┤
       ││(◣◡)│
       │└─◡◢┘
       │─┆┌─◢┌─┈┐
       ││││(|)││
       │││(O)││
       │└─└─┘─┘──┐
       │└─└─┘─└─┘
       │└─└─┘─└─┘
       │└─└─┘─└─┘
       └─└─┘─└─┘

[ Parent | on HN ]