---
title: AIDD
canonical_url: "https://www.modelscope.cn/models/BAAI/AIDD"
md_url: "https://www.modelscope.cn/models/BAAI/AIDD.md"
repository: BAAI/AIDD
last_updated: 2026-09-10
license: cc-by-4.0
library_name:
  - pytorch
frameworks:
  - pytorch
downloads: 1
stars: 0
---

# AIDD

> AIDD - BAAI 在 ModelScope 开源的模型。AIDD Open-Source Resources

BAAI/AIDD 是 ModelScope 魔搭社区上的机器学习模型，采用 cc-by-4.0 许可。

- **Repository**: BAAI/AIDD
- **License**: cc-by-4.0
- **Downloads**: 1
- **Stars**: 0
- **Last updated**: 2026-09-10

Source: https://www.modelscope.cn/models/BAAI/AIDD

---

# AIDD Open-Source Resources

This repository brings together open-source resources for three projects: **EPT**, **UniPath**, and **MiSI**. The EPT and UniPath source trees are mirrored here from their official GitHub repositories, including their original README documentation. The MiSI benchmark is released separately on Hugging Face.

## Contents

1. [EPT](#ept)
2. [UniPath](#unipath)
3. [MiSI](#misi)

## EPT

### An Equivariant Pretrained Transformer for Unified 3D Molecular Representation Learning

EPT provides an equivariant pretrained transformer for unified 3D molecular representation learning. The release includes environment setup, preprocessing pipelines for molecular and protein data, multi-domain pretraining, and downstream fine-tuning/evaluation on LBA, MSP, and MPP tasks.

- [Source code and full README](./EPT/README.md)
- [Original GitHub repository](https://github.com/jiaor17/EPT)
- [Preprocessed datasets](https://doi.org/10.6084/m9.figshare.c.8197622)

The complete upstream README is preserved at [`EPT/README.md`](./EPT/README.md), including setup, data preparation, training, evaluation, citation, and contact information.

## UniPath

### One Path to Model Them All: Learnable-Time Flow Matching for Crystal Structure and Energy Prediction

UniPath presents learnable-time flow matching for crystal structure and energy prediction. The release includes setup instructions, checkpoint and dataset download guidance, sampling and evaluation commands, multi-stage training pipelines, and reproducibility notes.

<div align="center">
  <a href="https://huggingface.co/BAAI/AIDD/resolve/main/UniPath/assets/UniPath_Poster.png">
    <img src="https://huggingface.co/BAAI/AIDD/resolve/main/UniPath/assets/UniPath_Poster.png" alt="UniPath KDD 2026 poster" width="900">
  </a>
</div>

- [Source code and full README](./UniPath/README.md)
- [Original GitHub repository](https://github.com/GLAD-RUC/UniPath)
- [Paper](https://dl.acm.org/doi/pdf/10.1145/3770855.3818902)
- [Checkpoints](https://huggingface.co/Wthinker/UniPath)
- [Datasets](https://huggingface.co/datasets/Wthinker/UniPath)

The complete upstream README is preserved at [`UniPath/README.md`](./UniPath/README.md). Its poster link was rewritten to an absolute Hugging Face `resolve/main` URL so that it renders correctly from the repository landing page. The upstream README referenced a poster PDF that was not present in the source repository, so the migrated link opens the available full-resolution PNG instead.

## MiSI

The **MiSI benchmark** has been open-sourced on Hugging Face:

- [MiSI-bench](https://huggingface.co/datasets/zongzhao/MiSI-bench)

## Provenance

| Project | Upstream source | Mirrored revision |
|---|---|---|
| EPT | [jiaor17/EPT](https://github.com/jiaor17/EPT) | `main` (`481a55016c777d19ae06587813f0569b360abe0b`) |
| UniPath | [GLAD-RUC/UniPath](https://github.com/GLAD-RUC/UniPath) | `main` (`4ce59f81cc06c75b597f62074f59d736b6d997aa`) |

Please consult each project directory for its documentation and any project-specific licensing information. The `cc-by-4.0` metadata above applies to this Hugging Face repository's Model Card metadata and does not replace upstream notices.