# NVIDIA cuML Documentation

NVIDIA cuML is a suite of fast, GPU-accelerated machine learning algorithms
designed for data science and analytical tasks. Our API mirrors scikit-learn,
providing practitioners with the familiar fit-predict-transform paradigm
without requiring GPU programming expertise. With [`cuml.accel`](api/cuml.accel/index.html.md#module-cuml.accel), cuML can also
automatically accelerate existing code with zero code changes.

cuML delivers on average **10-50x faster performance** than CPU-based
alternatives for realistic workloads and supports **50+ algorithms** across all
major machine learning categories, including clustering, regression,
classification, and dimensionality reduction. With comprehensive **multi-GPU
and multi-node support** via Dask, cuML scales from single workstations to
large clusters.

Especially if your scikit-learn, umap-learn, or hdbscan workflows take many
minutes to complete, you will likely benefit from using cuML. The equivalent
cuML estimators often run in seconds.

# Quick Start

```python
from cuml.datasets import make_blobs
from cuml.cluster import DBSCAN

# Create sample data
X, y = make_blobs(n_samples=100, centers=3, n_features=2, random_state=42)

# Fit clustering model
dbscan = DBSCAN(eps=1.0, min_samples=5)
dbscan.fit(X)
print(dbscan.labels_)
```

# Key Features

* **GPU Acceleration**: 10-50x faster than CPU-based alternatives
* **Scikit-learn Compatible**: Drop-in replacement for most sklearn algorithms
* **Multi-GPU Support**: Scale across multiple GPUs and nodes with Dask
* **Comprehensive Coverage**: 50+ algorithms across all major ML categories
* **Flexible Input**: Works with NumPy, cuDF, cuPy, and PyTorch tensors
* **Production Ready**: Battle-tested in enterprise environments

# Installation

cuML is available through conda and pip. For detailed installation instructions,
visit the [install guide](https://docs.nvidia.com/datascience/install#selector).

#### NOTE
cuML is only supported on Linux operating systems and WSL 2. See
[the install page](https://docs.nvidia.com/datascience/install/#system-req)
for details on system and hardware requirements.

# CUDA-X Data Science

NVIDIA cuML is an open-source library for GPU-accelerated machine learning. It
integrates with libraries in the broader CUDA-X Data Science ecosystem,
including cuDF for data manipulation and cuGraph for graph analytics.

# Community & Support

* [User Guide](user_guide/index.html.md) - Comprehensive usage documentation
* [Python API Reference](api/index.html.md) - Supported user-facing API documentation
* [Developer Guide](developer_guide/index.html.md) - Contributor guidance and internal C++ reference
* [GitHub Issues](https://github.com/NVIDIA/cuml/issues) - Report bugs and request features
* [CUDA-X Data Science Community](https://developer.nvidia.com/topics/ai/data-science/cuda-x-for-data-science#join-the-community) - Join our community