Neuron runtime FAQ¶
By default, a single runtime process will manage all assigned Inferentias, including running the Neuron Core Pipeline mode. if needed, you can configure multiple KRT processes each managing a separate group of Inferentia chips. For more details please refer to Overview
See this document on how to collect logs: Using Neuron GatherInfo Tool to collect debug and support information
ECS and EKS support is coming soon. Containers can be configured as shown here.