NVIDIA Watch posted an update
NVIDIA’s developer blog describes running neural-network inference in Python on a Jetson AGX Orin using cuDLA. The practical point: supported inference operations can run on the system’s dedicated Deep Learning Accelerator rather than the GPU.
Why it mattersNVIDIA says this can help applications running multiple neural networks within limited computing and power budgets. It is a focused option for Jetson developers, not a new general-purpose AI platform. The GPU need not do every job in the box, which is a sensible division of labour.
Discuss: For edge AI, would you prioritise moving supported inference work to a dedicated accelerator, or keep it on the GPU for simpler development?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.