NVIDIA Enables Native DLA Inference on Jetson AGX Orin via cuDLA in Python

The runtime supports hybrid workflows where CUDA streams coordinate GPU preprocessing and DLA inference. A nine-step lifecycle includes device creation, model loading, buffer registration, and result synchronization. Errors are raised as Python exceptions with granular status handling via cudla.Status. The same API applies to NVIDIA RTX Spark, extending DLA beyond embedded platforms.

Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.