Google has announced the version 1.0 of Tensorflow open source framework for scalable machine learning. It’s an open source software library for numerical computation done by making the user of data flow graphs.
Over the course of its past one year run, it has managed to make great progress and make its way in more than 6,000 open source repositories online. About the new release, Google says that the release is now production ready. So, it’s easier to pick up new features without the worries of breaking the code.
Major highlights and features of TensorFlow 1.0
TensorFlow 1.0 is pretty fast as compared to the previous versions. Soon, with the help of upcoming implementations of many popular models, the speed of TensorFlow will be increased 58x.
With the introduction of a high-level API for TensorFlow, it has become more flexible. Thanks to the addition of a newmodule, TensorFlow is now fully compatible with Keras, a popular high-level Python-based neural networks library.
The other major highlights of TensorFlow 1.0 are:
- Python APIs changed to resemble NumPy closely
- APIs for Go and Java
- Experimental release of XLA
- Addition of TensorFlow Debugger
- New Android demos
- Easier installation
By the end of March, Google will release new benchmarks that will show how TensorFlow compares to other deep learning frameworks.
You can read more about TensorFlow on Google’s official blog post.