The World’s Fastest Deep Learning Framework
neon is an open source Python-based language and set of libraries for developing deep learning models.
neon is super fast, super powerful, and super easy to use.


latest release: neon v1.8.0

Highlights from this release include:

• Skip Thought Vectors example
• Dilated convolution support
• Nesterov Accelerated Gradient option to SGD optimizer
• MultiMetric class to allow wrapping Metric classes
• Support for serializing and deserializing encoder-decoder models
• Allow specifying the number of time steps to evaluate during beam search
• A new community-contributed Docker image
• Improved error messages when a tensor is created with an invalid shape or reshaped to an incompatible size
• Fix bugs in MultiCost support
• Documentation fixes [#331]

Skip Thought Vector Model


neon features

Committed to best performance on all hardware

neon’s performance advantage is the result of assembler-level optimization, multi-GPU support, optimized data-loading, and use of the Winograd algorithm for computing convolutions.

Enlightening visualizations

neon’s nviz utility generates histograms and other visualizations to help you track progress and better understand what’s going on in the deep learning process.

Powerful simplicity

neon’s Python-like syntax includes object-oriented implementations of all the deep learning components, including layers, learning rules, activations, optimizers, initializers, and costs functions. And out-of-the-box examples cover all the standard deep learning use cases, including image recognition, speech, video, and natural language processing. So even those new to deep learning can easily implement all the common deep learning models in neon, including convnets, MLPs, RNNs, LSTMs and autoencoders. Meanwhile, more experienced deep learning data scientists can create their own novel algorithms using linear algebra, auto-differentiation, and other advanced capabilities with a numpy-like syntax.

neon is free to use

neon is available to download for free under an open source Apache 2.0 License. You can easily install neon on your own machine since it runs on CPU, GPU, and TX1. Got questions? neon is supported by an active user group

Model Zoo

neon comes equipped with Model Zoo containing pre-trained weights and example scripts for start-of-the-art models including: VGGReinforcement learningDeep Residual NetworksImage Captioning, and Sentiment analysis.


neon demos

Check out neon in action below and more demos here.

Video Recognition

Automatic Speech Recognition

Sentiment Analysis

Recent posts featuring neon

Load More Posts