Live data from Hacker News

Viewing profile — johhns4

johhns4

HN member
Joined
Fri, Sep 15, 2023, 12:01 PM UTC
HN karma
87
Public activity
45 items

About johhns4

No profile information was provided.

Recent public activity

  1. story
  2. story
  3. story
    Ask HN: Options to Deploy LLMs?

    To deploy fine-tuned LLMs there are these options yes? 1. HuggingFace 2. Runpod 3. Bedrock 4. FireworksAI 5. TogetherAI 6. Replicate 7. Modal Any other that I'm missing? What do yo…

  4. comment
    Comment #40773744

    Do please say when this will work as this could make my workflows a lot easier to visualize and work with.

  5. comment
    Comment #40726819

    Wow amazing work! How does the inputs work, are they created for you or does it support custom as well?

  6. comment
    Comment #40631358

    Funny, probably have no communication between these departments either, the people hiring them think they are excellent and then the ones firing them don't. But the problem never g…

  7. comment
    Comment #40583042

    And me for SQL. I'm really dumb at SQL.

  8. comment
    Comment #40561646

    It's not a lot of money they get, it's probably just to survive. Without kids and a career women have a much better life. Women with kids really have to work hard, women knows this…

  9. comment
    Comment #40409724

    What happens when renters go and shame landlords for getting kicked out for not paying rent and such? Most people leave reviews when they are angry, not when they are satisfied. La…

  10. story
    Have you successfully fine-tuned an LLM for anything useful?

    I'm interested to see if people are fine-tuning the larger LLMs just for kicks or if there are situation where it makes sense (not including high-stakes decision making like medica…

  11. comment
    Comment #39976943

    There is a difference between having all the facts and intelligence. Most LLMs will repeat what others have already said in a different way.

  12. comment
    Comment #39824842

    It's hardly that black and white. If people are burnt out there is a reason for it, blaming them won't make the problem go away. I've seen this in a few where there is certainly an…

  13. comment
    Comment #39818967

    If it is a constant issue where dedicated employees feel burnt out, then of course it is his business. I don't think the solution is to find less engaged people to work. It's not j…

  14. comment
    Comment #39676964

    Thanks! The Gemini one doesn't work though. Claude refused my request. Will have to try Mistral as well.

  15. comment
    Comment #39520767

    However, you are forgetting one thing most people who are productive may not be loud enough to show that they are productive. It's not until they leave that you understand the work…

  16. comment
    Comment #39417074

    You can do more with less. The more you introduce the more complicated it will be. Every person will introduce a new dynamic. Focus on creating small, strong and resistant teams in…

  17. comment
    Comment #39417063

    How are you going to identify the ones that are productive though through a layoff? If you just do a general cut across the board based on 'opinions' how do you know that the ones …

  18. story
    Hugging Face Inference Endpoints – 500 errors – anyone else?

    Been deploying a few models to a Nvidia Tesla T4 · 1x GPU · 16 GB via Hugging Face but keep getting back 500 errors every 2 minutes. Hasn't happened before today, just wondering if…

  19. comment
    Comment #39025812

    I think you made the point for him, "the entire game falls apart if you let [AI] play by the same rules as a human."

  20. comment
    Comment #39025459

    If anyone is experiencing the same

  21. story
  22. comment
    Comment #38810812

    Here it's good to point out that most of what is published is published within a platform that owns that content. Say you post a blog post on medium, or you write something on a su…

  23. comment
    Comment #38748031

    Oh god, when I build a dataset I manually go a lot of it to make sure it is in good shape. I will go through thousands of rows of data otherwise I don't get the result I need. Thin…

  24. story
  25. comment
    Comment #38672743

    What exactly did they do? Just transform data or generate completely new data for training? I've seen plenty of people that have used GPT-4 to help transform data, i.e. using GPT-4…