Open utterances-bot opened 7 months ago
This is amazing, thanks !
great job!
Great overview of the different concepts, discovered many! thanks @osanseviero
great job! marking this post
Good stuff. Would be nice to have a dive into Embeddings and tooling around it.
Good post! One comment is that Flash Attention is not an approximation of attention but it is exact, meaning it computes the exact attention calculation. It achieves the speedup through optimized memory access and parallel processing techniques.
This is an incredibly useful article. Thank you @osanseviero for maintaining this.
hackerllama - The Llama Hitchiking Guide to Local LLMs
https://osanseviero.github.io/hackerllama/blog/posts/hitchhiker_guide/