VoErik/cypher-gemma

Model Card

This model is a fine-tuned version of google/gemma-3-270m-it. Its purpose is turning natural language queries into CypherQueryLanguage.

It has been trained using TRL.

Quick start

from transformers import pipeline
from schemas import MOVIE_SCHEMA # you need to define this yourself!

query = "Which actors played a role in the movie Titanic?"

pipe = pipeline("text-generation", model="VoErik/cypher-gemma", device="cuda")
output = pipe([{"role": "user", "content": f"Question: {question} \n Schema: {MOVIE_SCHEMA}"}], max_new_tokens=256, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with SFT on the text2cypher-2025v1 dataset from Neo4j. It was trained for roughly 3500 steps.

Framework versions

• TRL: 0.23.1
• Transformers: 4.57.0
• Pytorch: 2.8.0
• Datasets: 4.2.0
• Tokenizers: 0.22.1

Citations

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}