Fine Tune + Embeddings

  • Resolved insiderperks

    (@insiderperks)


    2 years, 1 month ago

    I am struggling with blank answers. I have used this fine-tune model before and I know it works. I attempted to take those Q&A’s that I generated and use them as embeddings instead, which seemed to work at first, then started generating blank responses for data I know is there.

    Why would it ever generate a blank answer? Even if it can’t find the data in embeddings shouldn’t it say I don’t know?

    I’ve tried this with the already fine-tuned model from davinci plus the embeddings, using excerpts only, using davinci plus the embeddings, nothing works and it returns blank answers for tons of things.

    I upgraded to a paid Pinecone account, everything I can think of.

    The page I need help with: [log in to see the link]

Viewing 3 replies - 1 through 3 (of 3 total)
  • Plugin Author senols

    (@senols)

    Hello @insiderperks

    Can you send me an email here please: [email protected]

    I would like to check and troubleshoot. Please get in touch.

    Plugin Author senols

    (@senols)

    Also had a quick check on your website and noticed this error message:

    “This model’s maximum context length is 4097 tokens, however you requested 5707 tokens (4407 in your prompt; 1300 for the completion). Please reduce your prompt; or completion length.”

    Hit me an email and lets discuss: [email protected]

    Plugin Author senols

    (@senols)

    For some reasons if you can not email me, here is my suggestion;

    based on the error message I am assuming that you have embedded huge content at once. Pls try to split content and embed it separately to see how it goes.

Viewing 3 replies - 1 through 3 (of 3 total)
  • The topic ‘Fine Tune + Embeddings’ is closed to new replies.