README.md

---
base_model: Qwen/Qwen2.5-Coder-7B-Instruct
tags:
- text-to-sql
- lora
- qlora
- spider
- peft
---

# Qwen2.5-Coder-7B — Text-to-SQL (Spider)

LoRA adapter fine-tuned on the [Spider benchmark](https://yale-lily.github.io/spider).

- **77.7% execution accuracy** on Spider dev set (official Yale-LILY evaluation)
- **+24.5% over zero-shot Llama 3.3-70B** baseline
- Training: 4-bit QLoRA, r=16, 2 epochs, Kaggle T4 GPU

**GitHub:** [text-to-sql-spider](https://github.com/Gansaw98/text-to-sql-spider) | **Demo:** [Gradio Space](https://huggingface.co/spaces/Gansaw98/text-to-sql-demo)

## Usage
```python
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-Coder-7B-Instruct")
base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-Coder-7B-Instruct", torch_dtype=torch.float16)
model = PeftModel.from_pretrained(base, "Gansaw98/qwen2.5-coder-7b-text2sql-spider")