HinglishDB-Qwen: QLoRA Fine-tuned on HinglishDB

Qwen-2.5-Coder-3B fine-tuned on HinglishDB using QLoRA for Hinglish text-to-SQL generation.

Performance (Hinglish test set, 381 pairs)

Metric	Score
BLEU	20.39
SQAM	0.319
TSED	0.308

How to use

from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from peft import PeftModel
import torch

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True, bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.float16
)
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-Coder-3B")
base_model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-Coder-3B",
    quantization_config=bnb_config, device_map="auto"
)
model = PeftModel.from_pretrained(base_model,
    "Alex-24816/HinglishDB")
model.eval()

Dataset

huggingface.co/datasets/Alex-24816/HinglishDB

Paper

Singh, V., Bhagat, A. (2026). Generalizing Data Retrieval through a Hinglish-to-SQL Framework. Language Resources and Evaluation, Springer Nature.