[开源库]TensorFlow backend for TensorFlow.js via Node.js

时间:2021-1-8 作者:admin

This repo is under active development and is not production-ready. We will be actively developing this in open source. Stay tuned for an official release.

Currently, we only support developing TensorFlow.js code directly inside of this repository, however we plan on publishing on NPM soon.

Trying it out

# Download and install JS depencies, including libtensorflow 1.8.
yarn

# Publish the NPM locally for usage with other packages.
yarn publish-local

See the demo directory that trains MNIST using TensorFlow.js with the TensorFlow C backend.

cd demo/
yarn

# Link the tfjs-node NPM we published above.
yarn link-local tfjs-node

# Run the training script. See demo/package.json for this script.
yarn mnist

The important line to note is at the top of mnist.ts, which sets the backend to TensorFlow.

bindTensorFlowBackend();

Optional: Build libtensorflow From TensorFlow source

This requires installing bazel first.

bazel build //tensorflow/tools/lib_package:libtensorflow

Supported Platforms

  • Mac OS
  • Linux
  • Windows coming soon
声明:本文内容由互联网用户自发贡献自行上传,本网站不拥有所有权,未作人工编辑处理,也不承担相关法律责任。如果您发现有涉嫌版权的内容,欢迎进行举报,并提供相关证据,工作人员会在5个工作日内联系你,一经查实,本站将立刻删除涉嫌侵权内容。