neon™

Jun 28, 2017   |   Jayaram Bobba

neon™ 2.0: Optimized for Intel® Architectures

neon™ is a deep learning framework created by Nervana Systems with industry leading performance on GPUs thanks to its custom assembly kernels and optimized algorithms. After Nervana joined Intel, we have been working together to bring superior performance to CPU platforms as well. Today, after the result of a great collaboration between the teams, we…

Read more

#neon

Intel®'s reference deep learning framework committed to best performance on all hardware

Get started

Powerful Simplicity

The simplicity of neon’s Python-like syntax makes it an extraordinarily powerful, out-of-the-box tool for anyone from the novice, to the experienced scientist.

Best Performance

Assembler-level optimization, multi-GPU support, optimized data-loading, and the Winograd algorithm all add-up to stellar performance.

Enlightening Visuals

neon’s nviz utility helps you track progress and better understand the deep learning process with inspired visuals and histograms.

Model Zoo

neon™ comes with its own Model Zoo featuring pre-trained weights and example scripts for state-of-the-art models—the ones you see below are just a few.

Get started with Model Zoo

Image Classification

ImageNet
- Googlenet
- VGG
- Alexnet

CIFAR-10
- Deep Residual Network
- All CNN

Learn more

Object localization

Fast-RCNN

Learn more

Scene Classification

Deep Residual Network

Learn more

Image Captioning

LSTM

Learn more

Video Games

Deep Q-network

Learn more

NLP

bAbI (Question Answering)
-GRU/LSTM model
Sentiment classification
-LSTM (IMDB dataset)

Learn more

Video

C3D (UCF101 dataset)

Learn more

Speech

Deep Speech 2 (Librispeech dataset)

Learn more

neon Demos

Video Recognition

A fast 3-D convolution network (C3D) based model for activity recognition using neon on the UCF-101 dataset.

Automatic Speech Recognition

Neon’s implementation of Baidu’s "Deep Speech 2" model for speech recognition trained on audio-books from the Librispeech corpus.
Spectrogram (top left), raw audio (top right), and FFT spectrum (bottom).

Sentiment Analysis

A LSTM-based model constructed in neon to perform sentiment analysis on live Twitter stream.

Stay Connected

Keep tabs on all the latest news with our monthly newsletter.