在 Jetson AGX Orin 上用 cuDLA 运行原生 DLA 推理

边缘 AI 应用往往需要在有限的算力与功耗预算内运行多个神经网络。NVIDIA Deep Learning Accelerator (DLA) 平台为深度学习推理提供专用硬件。在 NVIDIA Jetson AGX Orin 上,DLA 执行受支持的神经网络运算,使应用能够将推理工作从 GPU 卸载。这可以提高能效,并让……
译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(Run Native DLA Inference in Python on NVIDIA Jetson AGX Orin Using cuDLA)
AI applications at the edge often need to run multiple neural networks within limited compute and power budgets. NVIDIA Deep Learning Accelerator (DLA)...
AI applications at the edge often need to run multiple neural networks within limited compute and power budgets. NVIDIA Deep Learning Accelerator (DLA) platforms provide dedicated hardware for deep learning inference. On NVIDIA Jetson AGX Orin, DLA executes supported neural network operations, allowing applications to offload inference work from the GPU. This can improve power efficiency and leave…
正文保留英文原文,标题与摘要由机器翻译。
本文由采集器自动同步
阅读原文 ↗
原文发布于 NVIDIA Developer(Run Native DLA Inference in Python on NVIDIA Jetson AGX Orin Using cuDLA),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。