Fmllm: 4mb training data, 100mb model, Fibonacci embeddings, near-coherent. WTF?

(github.com)

37 points | by gabriel666smith 3 days ago ago

27 comments