Uni-Mol Tools Independent Release: Build a Few-Shot, No-Threshold, Fast-Deployment Molecular Modeling Toolkit!

Posted on 2025-05-29 In Uni-Mol Word count in article: 953 Reading time ≈ 3 mins.

We are pleased to announce to the DeepModeling community that Uni-Mol Tools [1] is officially and independently released! As an important part of the Uni-Mol ecosystem, Uni-Mol Tools provides more flexible and efficient tool support for molecular AI research and applications with its characteristics of lightweight, out-of-the-box, and scenario-oriented.

What is Uni-Mol Tools?

Uni-Mol Tools is a toolset built based on the Uni-Mol ecosystem, focusing on common needs such as "molecular property prediction", "molecular representation processing", and "downstream task debugging". It is designed for actual scientific research and engineering implementation scenarios, allowing for one-click fine-tuning and modeling.

PyPI Installation: pip install unimol-tools
Documentation Address: https://unimol-tools.readthedocs.io/en/latest/
GitHub Address: https://github.com/deepmodeling/unimol_tools

Uni-Mol vs Uni-Mol Tools: Why the Split?

As the Uni-Mol project has gradually matured, the code volume and functional complexity of the main repository have continued to increase. To improve module reusability, reduce user entry barriers, and enhance community collaboration efficiency, we decided to extract some tool modules with "strong versatility", "model independence", and "standalone operation" from the main repository and maintain them separately as the unimol-tools package.

This split brings the following benefits:

Dimension	Uni-Mol Main Repository	Uni-Mol Tools
Functional Positioning	Model Training & Evaluation	Tools & Downstream Tasks
Dependency Volume	Heavy, Requires GPU Environment	Lightweight, Can Run on CPU
User Threshold	Researchers, Developers	Beginners, Teaching, Application Scenarios
Usage Method	Training + Inference	Scenario Fine-Tuning, Feature Extraction

Community Feedback: The Toolkit Has Served Thousands of Users

We are glad to see that during the early silent testing phase, unimol-tools has achieved stable download volumes on PyPI, demonstrating good user stickiness and practicality. We also thank the large number of users who provided feedback on suggestions and bug reports during this period.

Target Users and Tool Positioning

Who is suitable for using Uni-Mol Tools?

We have developed Uni-Mol Tools into a general-purpose toolkit covering scientific research, teaching, and engineering implementation, suitable for the following user groups:

Academic Researchers: In fields such as chemistry, biology, drug design, and materials science, researchers can use Uni-Mol Tools for tasks such as molecular property prediction, PK curve fitting, and atomic system structure modeling to quickly complete the closed-loop process from data preprocessing to result analysis.
Students and Educators: During teaching, this tool can be used to demonstrate core concepts such as molecular modeling, machine learning regression, and multi-task learning, helping students gain hands-on experience in a real research environment.
Industrial R&D Personnel: For enterprises' internal specific molecular modeling scenarios, this tool can be applied to extract molecular representations, build model pipelines, achieve fast and efficient application deployment, and enable efficient iteration of the wet-dry closed loop.

Our Positioning: More than just a toolkit

Universal Molecular Property Prediction Platform: Uni-Mol Tools provides rich support for molecular scenario tasks, with high scalability and reusability, adapting to various application scenarios.
Easy-to-Use Scientific Research Tool: Emphasizing a "zero-threshold" user experience, it is equipped with documentation, examples, and preset configurations to help users focus on scientific research itself.
Bridge for Interdisciplinary Collaboration: The design concept focuses on serving interdisciplinary fields such as chemistry, materials, pharmacy, and computational science, providing a high-efficiency collaborative basic platform for interdisciplinary research.
Open Source-Driven Community Innovation Tool: We uphold the spirit of open source, continue to iterate, and welcome community contributions, committed to building an open molecular AI ecosystem.

Quick Installation

The toolkit has been released on PyPI and supports one-click installation via pip:

1	pip install unimol-tools

After installation, you can import the core modules and apply unimol-tools to your projects!

Whether you are a novice or an experienced modeler, Uni-Mol Tools strives to be "plug-and-play". Now, you can quickly get started through the Bohrium online Notebook and try running your own molecular data through the model!

Online Documentation & Examples

We have built a complete documentation website for this project, including:

Installation Guide
Usage Tutorials
API Reference
Example Projects (Continuously Expanding)

Access Address: https://unimol-tools.readthedocs.io/en/latest/

Future Plans

UniMol Tools will be maintained for the long term and is committed to becoming:

A low-threshold entry toolkit for molecular AI
An easily extensible molecular task component library
A community-built open-source ecosystem project

Subsequent plans include:

Support for MCP service protocols
Improvement of DDP distributed training/inference processes
Multi-platform compatibility and performance optimization
HuggingFace deployment support
...

We also welcome you to propose functional suggestions or directly submit PRs!

How to Contribute

Uni-Mol Tools is one of the open-source projects of the DeepModeling[2] community, and developers interested in participating are always welcome:

GitHub Repo: https://github.com/deepmodeling/unimol_tools
The Issue section welcomes feedback on problems, suggestions, and feature requests.
New users can refer to the Readme and documentation to get started quickly.

If you encounter any problems during use, welcome to submit an Issue on GitHub or contact us via email.

About Uni-Mol

Uni-Mol[3] is a widely concerned molecular pre-training model in recent years, dedicated to building a universal 3D molecular modeling framework. As its derivative toolkit, Uni-Mol Tools aims to make model applications have lower thresholds and higher development efficiency.

We believe that the open-source sharing of tools is an important cornerstone for promoting the development of scientific research.

Welcome to forward and share with interested friends to jointly accelerate the implementation of AI for Molecule!

Relevant Links

[1]https://github.com/deepmodeling/unimol_tools

[2]https://github.com/deepmodeling

[3]https://github.com/dptech-corp/UniMol