ChemRANG 🧪

ChemRANG (Chemical Reaction-Role Annotation and Negative-Sample Generation) is a two-stage data augmentation framework to building AI-Ready datasets.

📂 Project Structure

ChemRANG/
├── core/
│   ├── downstream_task/
│   ├── utils/
│   ├── inefficient_agents.py
│   ├── settings.py
│   └── *_prompt.md
├── modules/
│   └── data_manger.py
├── Suzuki_RANG/
│   └── suzuki_samples_*.jsonl
├── raw_data_filter_single_file.py
├── run_chemical_model_train_and_test.py
├── start_filter.sh
└── README.md

🚀 Getting Started

1. Prerequisites

Ensure your environment has Python 3.8+ installed. You also need standard data science requirements and RDKit for chemoinformatics functionalities.

# Example conda environment creation
conda create -n chemrang python=3.9
conda activate chemrang
conda install -c conda-forge rdkit

2. Data Filtering and Preprocessing

To clean and filter your raw chemical dataset with the framework, you can use the elegant batch shell script:

bash start_filter.sh

3. Model Training & Testing

Once data has been filtered and processed into JSONL formats (such as those in Suzuki_RANG), you can trigger the downstream training and evaluation processes:

python run_chemical_model_train_and_test.py

📄 License

This project is dual-licensed:

• Codebase: All source code (.py, .sh, etc.) is distributed under the MIT License.
• Datasets: All data files and datasets (such as the JSONL files in Suzuki_RANG/) are distributed under the Creative Commons Attribution 4.0 International License (CC BY 4.0).

See the respective LICENSE and DATA_LICENSE files for more details.