Google has updated its TensorFlow open source machine learning code to enable it to be deployed on cloud platforms and across hundreds of distributed machines.
TensorFlow 0.8 now gives developers the option to build machine learning models that can run operations on distributed clusters of computers.
Machine learning software gets smarter the more data it ingests and analyses, so deploying TensorFlow across operations running on multiple machines gives the framework far greater scope to learn at scale.
This ability to scale across distributed machines allows TensorFlow models to be trained and deployed on the Google Cloud Platform, and complements Google Cloud Machine Learning, a new smart service offered to companies and developers looking to build Google cloud-based machine learning software and apps.
Equally, having access to the combined power of even a small cluster of computers, rather than relying on one machine, means that the overall data throughput of machine learning models and the speed at which they deliver accurate results can be accelerated.
Google has also released new libraries of machine learning model architecture, which it uses to train its Inception neural network and give developers a way to define their own TensorFlow models for distributed machine learning.
“This architecture makes it easier to scale up a single-process job to use a cluster, and to experiment with novel architectures for distributed training,” said Google’s research team.
The update should be a boon for developers looking to scale out the use of TensorFlow, which is currently the most popular machine learning framework on GitHub.
Google already uses TensorFlow at the machine learning heart of some of its popular services, such as Translate and Photos.
The company made the code available to open source communities in 2015, allowing developers to build new code, apps and services around TensorFlow. Google takes some of the innovations and developments and integrates them with its own machine learning services, thereby creating an ecosystem that serves the firm and the open source world.
Machine learning can be a fairly complex aspect of computing, so tapping into the combined knowledge of the open source community is a way to take some of the development strain off a single organisation.
However, it is worth keeping in mind that the nature of open source means that major bugs and code flaws can creep in unnoticed.
And the widely adopted nature of such code and frameworks means that an unconsidered approach to adopting open source could expose companies to security vulnerabilities without even realising it.
14nm Cavium ThunderX2 CPUs deployed in HPE Apollo 70 supercomputer for US National Nuclear Security Administration
MWR's Countercept platform and phishd technologies key to F-Secure acquisition
Brexit labour shortages will lead to higher adoption of robotics
Newbies will be thrown in with the big boys on Sanhok as Kar98 fodder