TensorFlow, developed by Google Brain team, is an open source software library for a building machine learning models for range of tasks in data science. It is written in Python, C++, CUDA and is mainly used for machine learning applications such as neural networks.
The open source software library has had many releases, each one with more improvements and fixes. Google recently announced the release of TensorFlow 1.5.0 in January, which included Eager Execution of TensorFlow and developer preview of TensorFlow Lite as major additions to the library. Google’s contribution to the world of Machine Learning and Data Science did not end with that. A new version, TensorFlow 1.6.0 has been released, with significant changes over the previously existing versions.
TensorFlow 1.6.0 has introduced a second version of Getting started for Machine Learning newcomers. It provides a document explaining Machine Learning fundamentals along with TensorFlow programming. Here is the document on Getting Started for ML Beginners.
Furthermore, TensorFlow 1.6.0 has a number of Bug fixes, few of which are:
- Added a clarification document
resize_images.align_cornersparameter and an additional document on TPUs.
- A client side throttle is added on Google Cloud Storage.
FlushCaches()method to the FileSystem interface is added with an implementation for GcsFileSystem.
Apart from that, we have some improvements made in TensorFlow 1.6.0:
- Prebuilt binaries are now built against CUDA 9.0 and cuDNN 7.
- Prebuilt binaries will use AVX instructions. But be wary that this may break TensorFlow on older CPUs.
Additionally, there are some technical advancements in API:
- Introducing prepare_variance boolean with default setting to False for backward compatibility.
For more information, visit TensorFlow 1.6.0
Our take on this:
The newer version of TensorFlow has made an effort to emerge as a beginner friendly software library. It has introduced the Getting started version for machine learning beginners which provides a detailed explanation of the machine learning fundamentals. Also, the updated version has support for CUDA 9.0 and cuDNN 7, bringing the glimpse of state-of-the-art and cutting edge technologies both at the same time.