Hacker News new | past | comments | ask | show | jobs | submit login

In case you might be trying this with Ollama (given the model:tag syntax) – I haven't added this to the readme yet (seeing some issues with the prompt template) but check out llama2-uncensored:

  ollama run llama2-uncensored



  % ollama run llama2-uncensored
  >>> Generate a json object describing the most stunning part of the movie: "The Matrix"
  {   
    "title": "The Matrix",
    "director": "The Wachowskis",
    "release_year": 1999,
    "description": "Set in a dystopian future, The Matrix follows the story of Neo (Keanu Reeves), a computer hacker who is drawn into an underground war against machines that have enslaved humanity. Along with his allies Morpheus (Laurence Fishburne) and Trinity (Carrie-Anne Moss), Neo learns to harness the power of the Matrix, a virtual reality system designed to keep humans obedient and ignorant. With the help of a rogue program named Agent Smith (Hugo Weaving), Neo discovers the true nature of the Matrix and sets out on a journey to free humankind from their digital prison.",
    "stunning scenes": [
      "bulletproof kung-fu fighting",
      "bullet time",
      "dream-like imagery",
      "special effects",
    ]
  }
edit: formatting


Are you going to add a "stop" argument to the API?


Looking into this! Since quite a few of the recommended prompts don't always work https://github.com/jmorganca/ollama/issues/217


This argument is vital even when using openAI’s api. The LLMs want to continue writing… even when they shouldn’t.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: