Let every developer use machine learning technology

Issued by ：TensorFlow Team Laurence Moroney and Josh Gordon

Recently carried out Stack Overflow The developer survey has now revealed the results , We are glad to see ,TensorFlow It has grown into the most widely used Machine Learning Tools , There are 300 Million software developers use it to use machine learning technology to enhance their products and solutions . More importantly , This survey shows that ,TensorFlow It is also the preferred framework for developers , Estimated to have 400 Wan developers hope to adopt it in the near future .

• Stack Overflow Developer research
  

  https://survey.stackoverflow.co/2022/#most-loved-dreaded-and-wanted-misc-tech-want

• The most widely used Machine Learning Tools
  

  https://survey.stackoverflow.co/2022/#most-popular-technologies-misc-tech

TensorFlow Currently, the number of downloads per month exceeds 1,800 Ten thousand times , stay GitHub Upper cumulative standard star 16.6 Ten thousand times , Far more than anything else Machine Learning frame . stay Google Inside , Almost all AI The workflow of production system relies on this framework , Include Google Search for 、Google Ads、YouTube、Gmail、Google Map 、Google Play、Google Album, etc . Many production systems of world-class giants also adopt this framework , Such as apple 、 Netflix 、Stripe、 tencent 、 Uber 、 Roche 、 Led the 、 twitter 、 Baidu 、 Orange、 Mo Yue · Hennessy - Louis · Vuitton (LVMH) etc. , beyond count . Besides , More than every month 3,000 Species involved TensorFlow or Keras Of the new scientific publications are Google Academic search indexed , Including those dedicated to exploring anti-cancer codes CANDLE Research and other major Applied Sciences .

• CANDLE
  

  https://candle.cels.anl.gov/

constitute Google Machine Learning Ecosystem products and open source services are constantly expanding their lineup . In recent years , We realize that there is no common framework for all scenarios —— In particular, there are often conflicts between the needs of production systems and the needs of cutting-edge research . In view of this , We developed JAX, This is a minimalist for distributed numerical computation API, It aims to add wings to the next wave of scientific computing research .JAX It is very suitable for opening up new frontiers ： Using it, the scale of parallel computing can reach a new level 、 Promote new algorithms and architectures , And developing new compilers and systems . The researchers JAX The widespread adoption of is exciting , and AlphaFold and Imagen The progress made by such projects highlights its value .

• AlphaFold
  

  https://www.deepmind.com/research/highlighted-research/alphafold

• Imagen
  

  https://imagen.research.google/

In this new era of various frameworks , We go through TensorFlow To meet the application Machine Learning The needs of developers —— Where it is necessary to build and deploy reliably 、 Stable 、 High performance Machine Learning System Engineer , Whatever Machine Learning How about the scale of the system 、 Which platform is suitable , All belong to this category . Our vision is to create an ecosystem that enables components to collaborate better , Ensure that researchers and engineers can make full use of these components , No matter what framework it comes from . In the realization of JAX And TensorFlow Interoperability aspects of , We have made great progress , especially jax2tf " . Development JAX The researchers of the model will be able to pass TensorFlow Various tools of the platform , Introduce this model into the production environment .

• jax2tf
  

  https://github.com/google/jax/tree/main/jax/experimental/jax2tf

In the future , We intend to do both , take TensorFlow As an excellent application Machine Learning Continue to develop the platform , At the same time, continuous improvement JAX To break through Machine Learning The boundaries of research . We will continue to Machine Learning Input on the framework , Help millions of users promote research and Application .

We still have many wonderful products under development and can't wait to share with you , Please follow our platform for more details ！