Fine tuning an LLM on a 2M word diary (jl33-ai.github.io)

(and accidentally curing my memory hoarding ocd)

Interesting. why didn’t you train on gpt2-do or a more recent model out of interest?