Available on Hugging Face and released under the Apache 2.0 license, Gretel’s dataset consists of over 100,000 high-quality synthetic Text-to-SQL samples with SQL metadata and spans 100 verticals.

I posted about this a while ago but I can't find the thread because search is borked. Anyway, the basic idea is that this program imports a fixed width text file which has ~9 pieces of information ...

In tests, all baselines achieved no more than 20% accuracy, while table-augmented generation did far better with 40% or better accuracy.

Gretel releases world's largest open source Text-to-SQL dataset, empowering enterprises to harness AI's potential with high-quality synthetic data while balancing privacy and accessibility.

SAN FRANCISCO, April 5, 2024 — Gretel, a leader in synthetic data, released the world’s largest open source Text-to-SQL dataset to unlock new possibilities for AI in the enterprise. Available on ...