Closed chuangchuang-wang closed 1 year ago
Hi there, the Spider dataset uses SQLite while our model is trained to generate PostgreSQL. Additionally, our model is optimized for enterprise use-cases when columns descriptions are well defined. You can read more about the differences in the Spider dataset and our SQL-Eval here: https://defog.ai/blog/open-sourcing-sqleval/
Has anyone tested the effect on the spider dataset? I have tested the result and the accuracy rate is only 30%, is it normal