So I had like 200 old shopping lists in a notes app from the past 4 years. Anyway I was bored at 3am and threw them into a tiny llm I was messing with. Idk why I do this.
It... learned something. Not useful something. But something.
First outputs were normal. Then it started getting weird. "milk (apologetic)" showed up. Then "aggressive potatoes". Then just "regret (medium)" with no category.
I'll paste some generations if anyone wants. It's mostly nonsense but sometimes it nails my actual shopping patterns in a way that feels like being judged by a very small brain.
Python
# 3am bad idea
Import torch
From grocery_gpt import GroceryGPT # don't ask
Model = GroceryGPT.from_pretrained("notes_export.txt")
# yes I know this is wrong
builds at 3AM, sleeps at noon