Text2SQL using Hugging Face Dataset Viewer API and Motherduck DuckDB-NSQL-7B
Hugging Face 2 years ago 38
MotherDuck and Numbers Station released DuckDB-NSQL-7B, a 7-billion-parameter language model fine-tuned to convert natural language questions into DuckDB SQL queries. The model was built on Meta's Llama-2-7b and further trained on DuckDB text-to-SQL pairs, enabling it to generate not just SELECT statements but any valid DuckDB SQL command. Users can now query datasets from Hugging Face's 120,000+ open datasets by describing their information needs in plain language, with the model translating requests into executable SQL without requiring SQL expertise.