On February 15th, 2017 in Mountain View, California Google conducted their inaugural TensorFlow Dev Summit.During the event, they officially announced Brand New Version of TensorFlow 1.0 which was live streamed around the world.TensorFlow is an Open source framework which makes use of artificial intelligence for deep learning.TensorFlow has open-source repositories around 6000 plus within a year of its launch.The earlier version made a huge impact in almost everything from early detection of skin cancer to preventing blindness in diabetics and also Language translation.
The major three highlights of TensorFlow 1.0 is that Firstly it’s incredibly fast, making performance improvement in future and it includes tips & tricks for tuning on achieve maximum speed for respective models.
Secondly, it’s highly Flexible.The new TensorFlow 1.0 has high-level API.It includes the new tf.keras module which has full compatibility with a popular neural network named Keras and with that it includes tf.losses modules,tf.metrics, and tf.layers.
The third Feature is that it makes an assurance that Python API stability will be made easier without any breakage in your existing code.There are more addition to the features of TensorFlow 1.0 like it has API for Java and Go, there is camera based image styling, For object detection and localization new Android Demos is available, In the installation part docker images of Python 3 has been added and we can install TensorFlow by simply typing pip install tensorflow, Introduction to TensorFlow Debugger and for TensorFlow Graphs releasing a domain-specific compiler’s experimental edition XLA.
Through Cloud Machine learning Google makes TensorFlow run through Cloud infrastructure. The director Megan Kacholia said, by march end, Google will release new benchmarks performance models on TensorFlow compared to different deep learning Framework. Amy McDonald Sandjideh, Technical Program Manager of TensorFlow posted the details of this release on the blog post.