This guide provides in-depth documentation of important TensorFlow.js topics. If you're just getting started with TensorFlow.js, you might want to explore the tutorials and then return to this guide to learn more.
TensorFlow.js is an open-source web ML library that can run anywhere JavaScript can. It's based on the original TensorFlow library written in Python and aims to recreate this developer experience and set of APIs for the JavaScript ecosystem.
The topics in this guide will help you understand TensorFlow.js and how TensorFlow APIs work in JavaScript.
Learn about key Tensorflow concepts:
- Tensors and operations – An intro to tensors, data, shapes, and data types: the building blocks of TensorFlow.js.
- Platform and environment – Overview of the different platforms and environments in TensorFlow.js and the tradeoffs between them.
- Custom ops, kernels, and gradients – Outlines the mechanisms for defining custom operations (ops), kernels, and gradients in TensorFlow.js.
Learn about pre-made models:
- Use a pre-made model – Guidance on how to find and select pre-made models for your use case.
Learn more about models and how to use them:
- Models and layers – How to build a model in TensorFlow.js using Layers and the Core API.
- Train models – Introduction to training: models, optimizers, losses, metrics, variables.
- Save and load models – Learn how to save and load TensorFlow.js models.
- Model conversion – See the landscape of model types available in the TensorFlow.js ecosystem and the details behind conversion of models.
- Differences from Python tf.keras – Know the
major differences and capabilities between TensorFlow.js and Python
tf.kerasand the API conventions used in JavaScript.
Learn about TensorFlow.js in Node.js:
- Using TensorFlow.js in Node.js – Understand the tradeoffs between the three available Node.js bindings and the system requirements they have.
- Deploy a TensorFlow.js Node project on cloud – How to
deploy a Node.js process with the
tfjs-nodepackage on cloud platforms.