An open-source benchmark infrastructure providing curated datasets, baseline models, standardized preprocessing pipelines, and interactive evaluation tools for Chemical Engineering ML research.
Machine Learning research in Chemical Engineering suffers from:
- Scattered datasets
- Repetitive preprocessing
- No standardized benchmarks
ChemBench provides a unified framework for:
- Dataset loading
- Data preprocessing
- Standardized train/test splits
- Baseline benchmarking
- Evaluation metrics
- Interactive visualization
- 📦 Curated benchmark datasets
- 🤖 Traditional ML + Deep Learning baselines
- 🧹 Standardized preprocessing pipelines
- 📊 Streamlit dashboard
- ⚡ Unified API for experiments
git clone https://github.com/IAyaanHere/ChemBench.git
cd ChemBench
pip install -e .from chembench.data.loader import ChemBenchDataLoader
from chembench.models import RandomForestModel
loader = ChemBenchDataLoader("esol")
X_train, X_test, y_train, y_test = loader.get_splits()
model = RandomForestModel()
model.fit(X_train, y_train)
metrics = model.evaluate(X_test, y_test)
print(metrics)streamlit run app.py| Dataset | Domain | Task |
|---|---|---|
| Tennessee Eastman Process | Process Control | Classification |
| QM9 | Quantum Chemistry | Regression |
| ESOL | Molecular Solubility | Regression |
| MoleculeNet | Molecular ML | Multi-task |
| CheMixHub | Mixture Properties | Regression |
- Python
- PyTorch
- Scikit-Learn
- Pandas
- RDKit
- Streamlit
Contributions are welcome.
You can contribute by:
- Adding datasets
- Implementing models
- Improving pipelines
- Fixing bugs
- Improving documentation
MIT License