Explore GPU-accelerated data science with RAPIDS in this conference talk from ODSC East 2019. Learn how RAPIDS leverages GPUs to accelerate Python-based data science workflows, including libraries like cuDF, cuML, and cuGraph. Discover the platform's integration with Dask and Numba for multi-GPU scaling and JIT compilation of User Defined Functions. Gain insights into RAPIDS' installation methods, interoperability with other libraries and deep learning frameworks, and its impact on large-scale data science problems. Understand the evolution of RAPIDS libraries, upcoming features, and long-term direction for GPU-powered data science.
RAPIDS - The Platform Inside and Outside - Joshua Patterson - ODSC East 2019
Overview
Syllabus
Intro
Apache Arrow
Traditional Data Science
Best Data Scientist
Best API
Python ecosystem
How people spend time
COOTF
open ucx
string
CSV reader
JSON lines
Show of Hands
DLPack
WildAsk
Arrays
GPUs
Scaling
Machine Learning
Building the library
GPU Size
Umap
What we started
What we are today
What we dont have
Data sets
Python vs GraphX
Scale
MultiGPU
One Data
Features
User experience
Integration
Documentation
Getting Started Guides
How to Get Started
Active Users
GPU Support
Taught by
Open Data Science
