Some readers are likely to be more familiar with the concept of machine learning than others. In short, it's a way for an A.I. to be constantly improved that works by allowing the system to look over data it's compiled in the past and by figuring out what's related and appropriate, learn how to handle situations involving new data. Machine learning is used extensively in things like A.I. voice assistants, security suites and search engines. Alphabet, in fact, makes extensive use of machine learning. So extensive, in fact, that they developed their own system for it, called TensorFlow. The deployment module that allows TensorFlow systems to receive new data, be implemented in existing systems and be improved upon while still working is a vital part of TensorFlow. It is this part that Google announced on February 17 has now gone open-source.
Google previously made the main part of TensorFlow open-source, but now that Serving has joined the main piece of the puzzle on GitHub, those who want to implement machine learning but couldn't make their own system for whatever reason, will be able to use Google's in-house system for their own operations. TensorFlow serving is also compatible with home brewed and commercial machine learning solutions, allowing for fairly simple in-place deployment of just about any type of machine learning system within a service framework, as well as iteration on the solution's core and feeding it new data.
TensorFlow Serving, even with tweaks to make the open-source version compatible with systems outside of Google's stable, is extremely efficient; On a top of the line machine and pushing the main TensorFlow program, it managed to output up to 100,000 operations per core per second in testing. TensorFlow serving is written in C++, making it extremely versatile and easy to port to just about any architecture or machine. The blog post from Google also touts Linux compatibility. TensorFlow Serving is up on GitHub right now under the Apache 2.0 open source license. If you're curious about the new system, about machine learning in general, or you know exactly what you're doing and you'd like to see how TensorFlow serving can improve your machine learning system, hit up the source link to read the detailed blog post from Google in full and complete with GitHub link.