---
license: apache-2.0
tags:
- deep-learning
- chemistry
- computational-chemistry
- vibrational-spectroscopy
- molecular-structure
---
[![](https://img.shields.io/badge/arXiv-2503.07014-c72c2c.svg)](https://arxiv.org/abs/2503.07014)
[![](https://img.shields.io/badge/huggingface-ckpt--of--vib2mol-dd9029)](https://huggingface.co/xinyulu/vib2mol)
[![](https://img.shields.io/badge/huggingface-vibench-dd9029)](https://huggingface.co/datasets/xinyulu/vibench)
[![](https://img.shields.io/badge/figshare-10.6084/m9.figshare.28579832-2243da)](https://doi.org/10.6084/m9.figshare.28579832)


# Model Overview
Vib2Mol is a versatile deep learning model designed for two key tasks in computational chemistry: **spectrum-structure retrieval** and ***de novo* spectrum-to-structure generation**. It leverages a novel architecture to build a robust relationship between molecular structures and their corresponding vibrational spectra (specifically, Infrared and Raman).

This model is a core component of the research presented in our paper, ***Vib2Mol: from vibrational spectra to molecular structures—a versatile deep learning model***. More details can be found in our [arXiv paper](https://arxiv.org/abs/2503.07014) and [GitHub repository](https://github.com/X1nyuLu/vib2mol).

# How to Use
To use this model for inference, you can load the model checkpoint and use the `infer_retrieval.sh` or `infer_generation.sh` scripts from our GitHub repository.

# Acknowledgements
This work was supported by the National Natural Science Foundation (Grant No: 22227802, 22021001, 22474117 and 22272139) of China and the Fundamental Research Funds for the Central Universities (20720220009 and 20720250005) and Shanghai Innovation Institute.

# Contact
Welcome to contact us or raise issues if you have any questions. 
Email: xinyulu@stu.xmu.edu.cn