yinliang420/LDH-OER: v1.0.0 - Initial Release: Models, Training Pipelines, and Datasets

🚀 Overview We are pleased to announce the initial release of the LDH-OER repository. This release provides the complete codebase implementation for our research on high-entropy catalysts.

This package includes the Attention-Enhanced models, training pipelines for overpotential and doping prediction, and scripts for inference ("yuce").

✨ Key Features 🧠 Model & Training Core Architecture: model.py contains the implementation of the deep learning model.

Task-Specific Training: Specialized training scripts for different targets:

overpotential.py: For training the overpotential prediction model.

doping.py: For training the doping prediction model.

Fine-tuning: Scripts for layer-specific fine-tuning (finallayer_finetuning.py, firstlayer_fintuning.py).

🔮 Inference & Analysis Prediction Scripts: Ready-to-use scripts for making predictions:

yuce_overpotential.py: Predict overpotential for new structures.

yuce_doping.py: Predict doping effects.

Interpretability: attention_map.py to generate attention maps for analyzing active sites.

📊 Datasets The datasets and model checkpoints used in this study are hosted on Hugging Face.

Hugging Face Hub: https://huggingface.co/datasets/yinliang22/oer_dataset

Content: Processed input data, trained model checkpoints, and prediction datasets.

📂 File Structure Highlights The repository follows a flat structure for ease of use:

model.py: Main model definition.

overpotential.py / doping.py: Main training scripts.

yuce_overpotential.py / yuce_doping.py: Inference (prediction) scripts.

dataprocess.ipynb: Jupyter notebook for data preprocessing and exploration.

requirements.txt: Python dependencies.

🔧 Getting Started

1. Installation Clone the repository and install the required dependencies:

Bash

git clone https://github.com/yinliang420/LDH-OER.git

cd LDH-OER

pip install -r requirements.txt

      2. Download Data Download the necessary datasets and checkpoints from Hugging Face:

Python

#You can use the huggingface_hub library

from huggingface_hub import snapshot_download

snapshot_download( repo_id="yinliang22/oer_dataset", local_dir="./oer_dataset", repo_type="dataset" )

      3. Usage Examples Training a model:

To train the model for overpotential prediction:

python overpotential.py

To train the model for doping prediction:

python doping.py

Making Predictions (Inference): 

To predict overpotential using a trained model:

python yuce_overpotential.py

📝 Citation If you use this code or data in your research, please cite our paper:

Decoding Active Sites in High-Entropy Catalysts via Attention-Enhanced Model