MSc thesis · Univ. of Bucharest
Text2EF
Llama 3 8BFine-tuningText-to-SQLEF
- Role
- MSc thesis · Univ. of Bucharest
- Year
- 2024
- Stack
- Llama 3 8B · PyTorch · Hugging Face
- Visit
- GitHub ↗
Text-to-SQL and text-to-Entity Framework. Fine-tuned Llama 3 8B, raising accuracy from 31% to 51%.
- Built the dataset by translating 87% of Spider’s SQL into Entity Framework.
- Then fine-tuned Llama 3 8B on it.