All Tools
C
Dev ToolsFreeOpen Source
CUML
GPU-accelerated ML algorithms with scikit-learn-compatible APIs
Apache-2.0
ABOUT
Traditional ML libraries like scikit-learn are CPU-bound, limiting training and inference speed on large datasets even with multi-core processors. cuML eliminates this bottleneck by porting the most common ML algorithms — random forests, gradient boosting, PCA, UMAP, KNN, and linear models — to run on NVIDIA GPUs with a familiar, scikit-learn-compatible API so data scientists get GPU-speed training and inference without learning new frameworks or rewriting their pipelines.
INSTALL
pip install cuml-cu12INTEGRATION GUIDE
1. Train random forest and gradient boosting models on large datasets at GPU speed instead of CPU-bound scikit-learn
2. Accelerate dimensionality reduction (PCA, UMAP, t-SNE) and clustering (DBSCAN, K-Means) on GPU hardware
3. Run GPU-powered nearest neighbor search and similarity matching in real-time ML inference pipelines
4. Replace CPU-based feature engineering with GPU-parallel transformations for faster experimentation
5. Build end-to-end GPU-accelerated ML training pipelines that keep data on-GPU across preprocessing and modeling
TAGS
gpumachine-learningscikit-learnrapidstraininginferencenvidiaparallel