README.md

metadata

title: DataInventor
emoji: 🧪
colorFrom: indigo
colorTo: pink
sdk: static
pinned: false
license: mit
short_description: Invent training rows, or curate your pretraining data

DataInventor

Open-source training data creation and curation. Describe the data you want, pick a format, domains and languages, preview a blueprint, and download instruction or preference rows as CSV or JSONL. Or curate the pretraining data you already have: clean, score, rephrase and mix it on a sample here, then run the same recipe on all of your data on your own machines. It works with any OpenAI-compatible model, and the default is a free model on Hugging Face Inference Providers.

This page embeds the app, which runs at codelion/data-inventor. Not affiliated with Adaption Labs.