Description
AutoGraph is one of the most exciting new features of Tensorflow 2.0: it allows transforming a subset of Python syntax into its portable, high- performance and language agnostic graph representation bridging the gap between Tensorflow 1.x and the 2.0 release based on eager execution.
Using AutoGraph with the @tf.fuction decorator seems easy, but in practice, writing efficient and correctly graph-convertible code requires to know in detail how AutoGraph and tf.function work.
In particular, knowing how:
- A graph is created and when it is re-used;
- To deal with functions that create a state;
- To correctly use the Tensorflow tf.Tensor object instead of
using the Python native types to speed-up the computation;
defines the minimum skill-set required to write correct graph-accelerable code.
The talk will guide you trough AutoGraph and tf.function highlighting all the peculiarities that are worth knowing to build the right skill-set.