Home > Storage > PowerScale (Isilon) > Industry Solutions and Verticals > Analytics > Deep Learning with Dell EMC Isilon > Machine Learning with RAPIDS
The RAPIDS suite of software libraries can accelerate some machine learning algorithms and analytics pipelines. Like TensorFlow, it uses NVIDIA GPUs for acceleration. RAPIDS requires CUDA 9.2 or 10.0 so you must ensure that the NVIDIA driver on the host supports this.
The following steps show how to run a simple RAPIDS benchmark using the mortgage demo.
mkdir -p /mnt/isilon/data/mortgage
cd /mnt/isilon/data/mortgage
# Use below for a quick test with 1 year of data.
wget http://rapidsai-data.s3-website.us-east-2.amazonaws.com/notebook-mortgage-data/mortgage_2000.tgz
tar -xzvf mortgage_2000.tgz
# Use below for the full benchmark with 17 years of data.
wget http://rapidsai-data.s3-website.us-east-2.amazonaws.com/notebook-mortgage-data/mortgage_2000-2016.tgz
tar -xzvf mortgage_2000-2016.tgz
docker run --runtime=nvidia \
-it \
-p 8888:8888 \
-p 8787:8787 \
-p 8786:8786 \
-v /mnt:/mnt \
--name rapidsai \
nvcr.io/nvidia/rapidsai/rapidsai:0.5-cuda10.0-runtime-ubuntu18.04-gcc7-py3.7
jupyter@12b8e8800eef:/rapids/notebooks$
source activate rapids
(rapids) jupyter@12b8e8800eef:/rapids/notebooks$
bash utils/start-jupyter.sh