lab-notes
microfactory-lab-field-notes
Field Notes: Building a Shop-Floor AI on a Small Local Model
I spent ten days building a small local Gemma that learns 3D printing job by
job, and these are the notes from doing it. The one moment worth watching: it
reads today's room, pulls up the closest prior jobs, and either applies what they
taught — "humidity is higher than the job where this overhang sagged, so I'm
raising retraction and adding support" — or says, plainly, "no close
precedent." Two named agents keep it… See the full description on the dataset page: https://huggingface.co/datasets/build-small-hackathon/microfactory-lab-field-notes.atomic-lab-notes
atomic-lab-notes
Synthetic lab-notebook entries split into sections.
Fields
id (int)
section (string) — one of: intro, methods, results, discussion
text (string) — the section text
Split: train.
EMPIRICAL subset
The EMPIRICAL subset is defined as rows whose section is methods or results.
Tokenizer
Token counts for this dataset should use the gpt2 tokenizer.