XLA usage on Xavier in NVIDIA Machine Learning for Jetson/L4T Container

XLA usage on Xavier in NVIDIA Machine Learning for Jetson/L4T Container nvcr.io/nvidia/l4t-ml:r32.4.2-py3.

Is it sufficient to set the flag in the python code running tensorflow?:

config.graph.options.optimizer_options.global_jit_level = tf.OptimizerOptions.ON_1

Is there something else to add to get the full advantage of XLA ?

Hi,

It look like XLA is a software-based optimizer.
So you can use the similar flag as the desktop environment.

Thanks.