xla stability and performance

#XLA (Accelerated Linear Algebra) is a domain-specific linear algebra compiler that optimizes TensorFlow calculations. Improve memory usage and portability. The XLA framework is experimental and still under active development.                                                                                       (Recommended learning: phpstorm)

TensorFlow 1.12.0-rc2 has been released.

TensorFlow is Google’s second-generation machine learning system and the most popular machine learning framework today.

The main updates are as follows:

Improve XLA stability and performance.

Fix single replica TensorBoard summary statistics in Cloud ML Engine.

Main features and improvements:

Keras models can now be exported directly to SavedModel format (tf.contrib.saved_model.save_keras_model()) and used with Tensorflow serving.

Keras models now support evaluation using tf.data.Dataset.

TensorFlow binaries are built with XLA support linked by default.

Ignite Dataset added to contrib/ignite, allowing use of Apache Ignite.

The above is the detailed content of xla stability and performance. For more information, please follow other related articles on the PHP Chinese website!