A GPU-native, all-Python platform for tree-based machine learning.

Note: OpenBoost is in active development. APIs may change between releases. Use at your own risk.

For standard GBDT, use XGBoost/LightGBM — they're highly optimized C++.

For GBDT variants (probabilistic predictions, interpretable GAMs, custom algorithms), OpenBoost brings GPU acceleration to methods that were previously CPU-only and slow:

• NaturalBoost: 1.3-2x faster than NGBoost
• OpenBoostGAM: 10-40x faster than InterpretML EBM

Plus: ~20K lines of readable Python. Modify, extend, and build on — no C++ required.

OpenBoost provides primitives (histograms, binning, tree fitting) that you combine into algorithms:

• Standard GBDT — drop-in gradient boosting with multiple growth strategies, early stopping, and callbacks
• Distributional GBDT — predict full probability distributions with NGBoost-style natural gradient boosting
• Interpretable GAMs — explainable feature effects inspired by EBM
• DART — dropout regularization for reduced overfitting
• Linear-leaf models — linear models in tree leaves for better extrapolation
• Your own algorithms — custom losses, distributions, or entirely new methods

All run on GPU with the same Python code. All models support save()/load() persistence, and most support callbacks and early stopping.

High-level API:

import openboost as ob

model = ob.GradientBoosting(n_trees=100, max_depth=6, random_state=42)
model.fit(X_train, y_train,
          eval_set=[(X_val, y_val)],
          callbacks=[ob.EarlyStopping(patience=10)])
predictions = model.predict(X_test)

sklearn-compatible:

from openboost import OpenBoostRegressor
from sklearn.model_selection import GridSearchCV

# Works with GridSearchCV, Pipeline, cross_val_score, etc.
model = OpenBoostRegressor(n_estimators=100, random_state=42)
search = GridSearchCV(model, {"max_depth": [4, 6, 8]}, cv=5)
search.fit(X_train, y_train)

# Also available: OpenBoostClassifier, OpenBoostDARTRegressor,
# OpenBoostGAMRegressor, OpenBoostDistributionalRegressor

Hyperparameter suggestions:

# Auto-suggest params based on dataset characteristics
params = ob.suggest_params(X_train, y_train, task='regression', style='core')
model = ob.GradientBoosting(**params)

Low-level API (full control over the training loop):

import openboost as ob

X_binned = ob.array(X_train)
pred = np.zeros(len(y_train), dtype=np.float32)

for round in range(100):
    grad = 2 * (pred - y_train)  # your gradients
    hess = np.ones_like(grad) * 2
    tree = ob.fit_tree(X_binned, grad, hess, max_depth=6)
    pred += 0.1 * tree(X_binned)

pip install openboost

# With GPU support
pip install openboost[cuda]

# With sklearn integration
pip install openboost[sklearn]

Full docs, tutorials, and API reference: jxucoder.github.io/openboost

• Getting Started
• User Guide
• API Reference
• Examples

On standard GBDT, OpenBoost's GPU-native tree builder is 3-4x faster than XGBoost's GPU histogram method on an A100, with comparable accuracy:

# Local (requires CUDA GPU)
uv run python benchmarks/bench_gpu.py --task all --scale medium

# On Modal A100
uv run modal run benchmarks/bench_gpu.py --task all --scale medium

Where OpenBoost really shines is on GBDT variants that don't exist in XGBoost/LightGBM:

Note: Benchmarks reflect the current state of development and may change as both OpenBoost and comparison libraries evolve.

Train-many optimization: Industry workloads often train many models (hyperparameter tuning, CV, per-segment models). XGBoost optimizes for one model fast. OpenBoost plans to enable native optimization for training many models efficiently.

OpenBoost implements and builds on ideas from these papers:

• Gradient Boosting: Friedman, J. H. (2001). Greedy Function Approximation: A Gradient Boosting Machine. Annals of Statistics.
• XGBoost: Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. KDD.
• LightGBM: Ke, G., et al. (2017). LightGBM: A Highly Efficient Gradient Boosting Decision Tree. NeurIPS.
• CatBoost: Prokhorenkova, L., et al. (2018). CatBoost: Unbiased Boosting with Categorical Features. NeurIPS.
• NGBoost: Duan, T., et al. (2020). NGBoost: Natural Gradient Boosting for Probabilistic Prediction. ICML.
• EBM: Nori, H., et al. (2019). InterpretML: A Unified Framework for Machine Learning Interpretability.
• DART: Rashmi, K. V., & Gilad-Bachrach, R. (2015). DART: Dropouts meet Multiple Additive Regression Trees. AISTATS.

Apache 2.0