To use this model, craft your prompt to start with your database schema in the form of CREATE TABLE, followed by your natural language query preceded by --. Make sure your prompt ends with SELECT in order for the model to finish the query for you. There is no need to set other parameters like temperature or max token limit.

python

from vllm import LLM, SamplingParams

llm = LLM(model="onekq-ai/OneSQL-v0.1-Qwen-32B-AWQ")
sampling_params = SamplingParams(temperature=0.7, max_tokens=200)

prompt="CREATE TABLE students (
    id INTEGER PRIMARY KEY,
    name TEXT,
    age INTEGER,
    grade TEXT
);

-- Find the three youngest students
SELECT "

outputs = llm.generate(f"<|im_start|>system\nYou are a SQL expert. Return code only.<|im_end|>\n<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant\n", sampling_params)
print(outputs[0].outputs[0].text.strip())

The model response is the finished SQL query without SELECT

* FROM students ORDER BY age ASC LIMIT 3

Caveats

The performance drop from the original model is due to quantization itself, and the lack of beam search support in the vLLM framework. Use at your own discretion.