tianyouli / inference-library-for-javascript

Unified JavaScript API for scoring via various DL framework

Inference Library for Javascript

An unified JavaScript API for scoring via various DL framework

This project is a JS framework that provide the handy API for client side scoring/reference, built on top of several existing DL/ML framework. Inference Library for Javascript is designed to lower the bar for programmers to adopt the hottest AI technology to their applications, reduce the cross-platform deployment headache, specifically optimized for Intel hardware platform and minimize the migration cost from one DL/ML framework to another.

Installation

# Preparation

npm -g install node-gyp

git clone https://github.com/TianyouLi/Inference-Library-for-JavaScript.git

Follow the instuctions under bagua-* folder and build the libraries.

Demos are provided under demo folder.

Currently Supported Backends

1. bagua-caffe
2. bagua-tensorflow
3. bagua-mxnet
4. bagua-movidius

inference-library-for-javascript's Issues

tensorflow: detecting CPU instruction set for sse4/fma/avx/avx2 in runtime

make deb distribution for Linux

statically linking caffe to reduce library dependencies