Nvidia Rapids

Nvidia Rapids is the platform for acceleration using GPU, the analysis of Big Data and machine learning.

2018

On October 10, 2018 the NVIDIA company submitted the platform of GPU acceleration for processing of data bulks and machine learning which got broad support at leaders of the industry. According to the company, the platform allows even to the large companies to analyze huge data arrays and to do forecasts for business.

The open source software of RAPIDS provides to analysts large increase of performance in business challenges of high complexity, such as prediction of fraud in transactions with credit cards, the forecast of a stock of goods in a warehouse, forecasting of a consumer consumer behavior.

Data analysis and machine learning are largest segments of the market of high-performance computing which till October, 2018 did not receive acceleration. The largest world companies start the algorithms created using machine learning on numerous servers to reveal difficult patterns in segments where they work, and to do the fast and exact forecasts rendering a direct effect on results of their activity. Having taken as a basis CUDA with its global ecosystem, we created the platform of GPU acceleration RAPIDS in close cooperation with developers of the open source software. It is easily integrated into the most widespread libraries of data processing and the existing processes for acceleration of machine learning. Jensen Huang, founder and CEO of NVIDIA

RAPIDS includes a set of open libraries for the analysis, machine learning and, very soon, data visualization with GPU acceleration.

Specialists for the first time receive necessary tools entirely to start a processing pipeline of data on GPU. For October, 2018 tests of RAPIDS with an algorithm of machine learning XGBoost for training at the NVIDIA DGX-2 system showed a 50-fold gain of performance in comparison with systems based on CPU.

The RAPIDS platform is based on open projects, including Apache Arrow, pandas and scikit-learn, allocating with GPU acceleration tools for data processing on Python. To add in RAPIDS of library and a feature for machine learning, NVIDIA cooperates with such players of the market of the open source software as Anaconda, BlazingDB, Databricks, Quansight and scikit-learn. To accelerate distribution of the platform, NVIDIA integrates RAPIDS into Apache Spark.