llmsql 0.1.3

A Text2SQL benchmark for evaluation of Large Language Models

LLMSQL

Patched and improved version of the original large crowd-sourced dataset for developing natural language interfaces for relational databases, WikiSQL.

Our datasets are available for different scenarios on our HuggingFace page.

Overview

Install

pip3 install llmsql

This repository provides the LLMSQL Benchmark — a modernized, cleaned, and extended version of WikiSQL, designed for evaluating and fine-tuning large language models (LLMs) on Text-to-SQL tasks.

Note

The package doesn't have the dataset, it is stored on our HuggingFace page.

This package contains

• Support for modern LLMs.
• Tools for evaluation, inference, and finetuning.
• Support for Hugging Face models out-of-the-box.
• Structured for reproducibility and benchmarking.

---

Usage Recommendations

Modern LLMs are already strong at producing SQL queries without finetuning.
We therefore recommend that most users:

1. Run inference directly on the full benchmark:

   • Use llmsql.LLMSQLVLLMInference (the main inference class) for generation of SQL predictions with your LLM from HF.
   • Evaluate results against the benchmark with the llmsql.LLMSQLEvaluator evaluator class.

2. Optional finetuning:

   • For research or domain adaptation, we provide finetuning script for HF models. Use llmsql finetune --help or read Finetune Readme to find more about finetuning.

[!Tip] You can find additional manuals in the README files of each folder(Inferece Readme, Evaluation Readme, Finetune Readme)

---

Repository Structure

llmsql/
├── evaluation/          # Scripts for downloading DB + evaluating predictions
├── inference/           # Generate SQL queries with your LLM
└── finetune/            # Fine-tuning with TRL's SFTTrainer

Quickstart

Install

Make sure you have the package installed (we used python3.11):

pip3 install llmsql

1. Run Inference

from llmsql import LLMSQLVLLMInference

# Initialize inference engine
inference = LLMSQLVLLMInference(
    model_name="Qwen/Qwen2.5-1.5B-Instruct",  # or any Hugging Face causal LM
    tensor_parallel_size=1,
)

# Run generation
results = inference.generate(
    output_file="path_to_your_outputs.jsonl",
    questions_path="data/questions.jsonl",
    tables_path="data/tables.jsonl",
    shots=5,
    batch_size=8,
    max_new_tokens=256,
    temperature=0.7,
)

2. Evaluate Results

from llmsql import LLMSQLEvaluator

evaluator = LLMSQLEvaluator(workdir_path="llmsql_workdir")
report = evaluator.evaluate(outputs_path="path_to_your_outputs.jsonl")
print(report)

Finetuning (Optional)

If you want to adapt a base model on LLMSQL:

llmsql finetune --config_file examples/example_finetune_args.yaml

This will train a model on the train/val splits with the parameters provided in the config file. You can find example config file here.

Suggested Workflow

• Primary: Run inference on dataset/questions.jsonl → Evaluate with evaluation/.
• Secondary (optional): Fine-tune on train/val → Test on test_questions.jsonl.

License & Citation

Please cite LLMSQL if you use it in your work:

@inproceedings{llmsql_bench,
  title={LLMSQL: Upgrading WikiSQL for the LLM Era of Text-to-SQL},
  author={Pihulski, Dzmitry and  Charchut, Karol and Novogrodskaia, Viktoria and Koco{'n}, Jan},
  booktitle={2025 IEEE International Conference on Data Mining Workshops (ICDMW)},
  year={2025},
  organization={IEEE}
}