Hi,
I want to install an mlperf test software on our Nvidia Nano development board. Please refer to the following link steps. When creating the object, the attached pictures are generally wrong.
It seems that docker cannot create files, how can I modify it to solve this problem?
github.com
Dockerfile.txt (2.6 KB)
Hi,
The Dockerfile is based on a desktop container which cannot be run on Jetson.
Does that mean I can no longer run docker software on the jetson development board?
Hi,
You can try to use a based container:
Containers for PyTorch, TensorFlow, ETL, AI Training, and Inference. Tuned, tested and optimized by NVIDIA.
For MLPerf, please see the document below for the Jetson environment:
# MLPerf Inference v3.1 NVIDIA-Optimized Inference on Jetson Systems
This is a repository of NVIDIA-optimized implementations for the [MLPerf](https://mlcommons.org/en/) Inference Benchmark.
This README is a quickstart tutorial on how to setup the the Jetson systems as a public / external user.
Please also read README.md for general instructions on how to run the code.
---
NVIDIA Jetson is a platform for AI at the edge. Its high-performance, low-power computing for deep learning and computer vision makes it the ideal platform for compute-intensive projects. The Jetson platform includes a variety of Jetson modules together with NVIDIA JetPack™ SDK.
Each Jetson module is a computing system packaged as a plug-in unit (a System on Module (SOM)). NVIDIA offers a variety of Jetson modules with different capabilities.
JetPack bundles all of the Jetson platform software, starting with NVIDIA Jetson Linux. Jetson Linux provides the Linux kernel, bootloader, NVIDIA drivers, flashing utilities, sample file system, and more for the Jetson platform.
## NVIDIA Submissions
The Jetson AGX Orin / Orin NX submission supports:
- ResNet50 (Offline, Single Stream, and Multistream), at 99% of FP32 accuracy target
- RetinaNet (Offline, Single Stream, and Multistream), at 99% of FP32 accuracy target
- 3D-unet (Offline and Single Stream), at 99% and 99.9% of FP32 accuracy target
- bert (Offline and Single Stream), at 99% of FP32 accuracy target
- rnn-t (Offline and Single Stream), at 99% of FP32 accuracy target
show original
Thanks.
system
December 6, 2023, 5:11am
7
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.
