RAPIDS Notebooks

Intro

These notebooks provide examples of how to use RAPIDS libraries, like cuML and cuGraph. These notebooks are designed to be self contained with the RAPIDS Docker Container and RAPIDS Nightly Docker Containers and can run on airgapped systems. You can quickly get this continaer using the install guide from the RAPIDS.ai Getting Started page

More notebooks in Notebooks Contrib Repo

For additional, community driven notebooks, which will include our blogs, tutorials, workflows, and more intricate examples, please see the Notebooks Contrib Repo. These notebooks use generally use real world data sets.

If you want to include these notebooks with your RAPIDS Docker pull's Jupyter Lab, please follow these instructions.

If you're using Conda, you can just git clone https://github.com/rapidsai/notebooks-contrib.git

RAPIDS Notebooks and Utilities

XGBoost Notebook

Folder	Notebook Title	Description
XGBoost	XGBoost Demo	This notebook shows the acceleration one can gain by using GPUs with XGBoost in RAPIDS.

CuML Notebooks

The cuML notebooks showcase how to use the machine learning algorithms implemented in cuML along with the advantages of using cuML over scikit-learn. These notebooks compare the time required and the performance of the algorithms. Below are a list of such algorithms:

Folder	Notebook Title	Description
cuML	Coordinate Descent	This notebook includes code examples of lasso and elastic net models. These models are placed together so a comparison between the two can also be made in addition to their sklearn equivalent.
cuML	DBSCAN Demo	This notebook showcases density-based spatial clustering of applications with noise (dbscan) algorithm using the fit and predict functions
cuML	Forest Inference	This notebook shows how to use the forest inference library to load saved models and perform prediction using them. In addition, it also shows how to perform training and prediction using xgboost and lightgbm models.
cuML	HoltWinters Demo	This notebook includes code example for the holt-winters algorithm and it showcases the fit and forecast functions.
cuML	K-Means Demo	This notebook includes code example for the k-means algorithm and it showcases the fit and predict functions.
cuML	K-Means MNMG Demo	This notebook includes code example for the k-means multi-node multi-GPU algorithm and it showcases the fit and predict functions.
cuML	Linear Regression Demo	This notebook includes code example for linear regression algorithm and it showcases the fit and predict functions.
cuML	Metrics Demo	This notebook includes code examples showcasing the different metrics provided in cuML. The results are compared with their scikit learn counterparts.
cuML	Mini Batch SGD Demo	This notebook includes code example for mbsgd classifier and regressor algorithms and it showcases their fit and predict functions.
cuML	Nearest Neighbors_demo	This notebook showcases k-nearest neighbors (knn) algorithm using the fit and kneighbors functions
cuML	PCA Demo	This notebook showcases principal component analysis (PCA) algorithm where the model can be used for prediction (using fit_transform) as well as converting the transformed data into the original dataset (using inverse_transform).
cuML	Random Forest Classification and Pickling	Demonstrates how to fit cuML and scikit-learn Random Forest Classification models. Then we save the cuML model for future use with Python's pickling mechanism and demonstrate how to re-load it for prediction.
cuML	Random Forest Multi-node / Multi-GPU	Demonstrates how to fit Random Forest models using multiple GPUs via Dask.
cuML	Ridge Regression Demo	This notebook includes code examples of ridge regression and it showcases the fit and predict functions.
cuML	SGD_Demo	The stochastic gradient descent algorithm is demonstrated in the notebook using fit and predict functions
cuML	SVM_Demo	Binary Support Vector Machine classification is demonstrated in this notebook using fit and predict functions.
cuML	TSNE_Demo	In this notebook, T-Distributed Stochastic Neighborhood Embedding is demonstrated applying the Barnes Hut method on the Fashion MNIST dataset using our fit_transform function
cuML	TSVD_Demo	This notebook showcases truncated singular value decomposition (tsvd) algorithm which like PCA performs both prediction and transformation of the converted dataset into the original data using fit_transform and inverse_transform functions respectively
cuML	UMAP_Demo	The uniform manifold approximation & projection algorithm is compared with the original author's equivalent non-GPU Python implementation using fit and transform functions
cuML	UMAP_Demo_Graphed	Demonstration of cuML uniform manifold approximation & projection algorithm's supervised approach against mortgage dataset and comparison of results against the original author's equivalent non-GPU \Python implementation.
cuML	UMAP_Demo_Supervised	Demonstration of UMAP supervised training. Uses a set of labels to perform supervised dimensionality reduction. UMAP can also be trained on datasets with incomplete labels, by using a label of "-1" for unlabeled samples.

CuDF Notebooks

Folder	Notebook Title	Description
cuDF	notebooks_Apply_Operations_in_cuDF	This notebook showcases two special methods where cuDF goes beyond the Pandas library: apply_rows and apply_chunk functions. They utilized the Numba library to accelerate the data transformation via GPU in parallel.
cuDF	notebooks_numba_cuDF_integration	This notebook showcases how to use Numba CUDA to accelerate cuDF data transformation and how to step by step accelerate it using CUDA programming tricks

CuGraph Notebooks

Folder	Notebook Title	Description
cuGraph -> centrality	Katz	Compute the Katz centrality for every vertex
cuGraph -> community	Louvain	Demonstration of using cuGraph to identify clusters in a test graph using the Louvain algorithm
cuGraph -> community	Spectral-Clustering	Demonstration of using cuGraph to identify clusters in a test graph using Spectral Clustering using both the (A) Balance Cut and (B) the Modularity Maximization quality metrics
cuGraph -> community	Subgraph Extraction	Compute a subgraph of the existing graph including only the specified vertices
cuGraph -> community	Triangle Counting	Demonstration of using both NetworkX and cuGraph to compute the the number of Triangles in our test dataset
cuGraph -> components	Connected Components	Find weakly and strongly connected components in a graph
cuGraph -> cores	K-Core	Extracts the K-core cluster
cuGraph -> cores	Core Number	Computer the Core number for each vertex in a graph
cuGraph -> link_analysis	Pagerank	Demonstration of using both NetworkX and cuGraph to compute the PageRank of each vertex in our test dataset
cuGraph -> link_prediction	Jacard Similarity	Compute vertex similarity score using both: - Jaccard Similarity - Weighted Jaccard
cuGraph -> link_prediction	Overlap Similarity	Compute vertex similarity score using the Overlap Coefficient
cuGraph -> traversal	BFS	Demonstration of using cuGraph to computer the Breadth First Search space from a given vertex to all other in our training graph
cuGraph -> traversal	SSSP	Demonstration of using cuGraph to computer the The Shortest Path from a given vertex to all other in our training graph
cuGraph -> structure	Renumber	Demonstrate of using the renumbering features to assigned new vertex IDs to the test graph. This is useful for when the data sets is non-contiguous or not integer values
cuGraph -> structure	Symmetrize	Symmetrize the edges in a graph

Tutorial with an End to End workflow

Folder	Notebook Title	Description
Tutorials	DBSCAN_demo_full	Demonstration of how to use DBSCAN - a popular clustering algorithm - and how to use the GPU accelerated implementation of this algorithm in RAPIDS.
Tutorials	HoltWinters_demo_full	Demonstration of how to use Holt-Winters, a time-series forecasting algorithm, on a dataset to make GPU accelerated out-of-sample predictions.

Utils Scripts

Folder	Script Title	Description
utils	dask-cluster.py	launches a configured Dask cluster (a set of nodes) for use within a notebook
utils	dask-setup.sh	a low-level script for constructing a set of Dask workers on a single node
utils	dask.conf	a Dask user configuration and settings file for the RAPIDS docker container
utils	split-data-mortgage.sh	splits mortgage data files into smaller parts, and saves them for use with the mortgage notebook
utils	start-jupyter.sh	starts a JupyterLab environment for interacting with, and running, notebooks
utils	stop-jupyter.sh	identifies all process IDs associated with Jupyter and kills them

Documentation

Folder	Document Title	Description
docs	ngc-readme
docs	dockerhub-readme

Additional Information

• The cuml folder also includes a small subset of the Mortgage Dataset used in the notebooks and the full image set from the Fashion MNIST dataset.

• utils: contains a set of useful scripts for interacting with RAPIDS