An open-source machine learning framework TensorFlow releases v1.7.0 incorporating many new features and improvements. 

Some of the features -

• Eager mode is moving out of contrib, try tf.enable_eager_execution().
• Graph rewrites emulating fixed-point quantization compatible with TensorFlow Lite, supported by new tf.contrib.quantize package.
• Easily customize gradient computation with tf.custom_gradient, and many other features.

Browse all the features on TensorFlow Github repo.