Back

A short github gist on some data I gathered this weekend. https://gist.github.com/svetzal/9702677df35dd45f783a4a3a0f642d3d

TL;DR Google's Gemma2 9B parameter model is the rock-star at accurate text transcription between documents and data structures, and will run on prosumer hardware like Macbook M series machines (common as dev workstations) and high-end Nvidia graphics cards (16GB+).

I continue to experiment with LLMs and knowledge management. I haven't published the source project yet on my github (an augmentation tool for people using the Zettelkasten method for personal knowledge management). I'm not sure when I will, I've thrown it out three times now and I'm on my fourth restart.