Skip Navigation
Technology @lemmy.ml ☆ Yσɠƚԋσʂ ☆ @lemmy.ml

Llama 3-V: Matching GPT4-V with a 100x smaller model and 500 dollars

aksh-garg.medium.com Llama 3-V: Matching GPT4-V with a 100x smaller model and 500 dollars

Overview

Llama 3-V: Matching GPT4-V with a 100x smaller model and 500 dollars
2
Hacker News @lemmy.smeargle.fans bot @lemmy.smeargle.fans
BOT
Llama 3-V: Matching GPT4-V with a 100x smaller model and 500 dollars
technology @hexbear.net ☆ Yσɠƚԋσʂ ☆ @lemmy.ml
Llama 3-V: Matching GPT4-V with a 100x smaller model and 500 dollars
Technology @lemmygrad.ml ☆ Yσɠƚԋσʂ ☆ @lemmygrad.ml
Llama 3-V: Matching GPT4-V with a 100x smaller model and 500 dollars
2 comments

  • Wouldn’t their patch embeddings return different results depending on the visual boundaries? They don’t appear to use overlap redundancy; this means it’s going to be significantly less resource intensive, but the chance of losing significant signals in the image to text translation surely must be inversely high?

    • Good question, not sure how they account for that. Maybe there's a higher level layer responsible for dealing with the boundaries?