Hi,
I try running onnx model using NPU in IMX8MPPLUS-BB. Download the image "Real-time_Edge_v2.5_IMX8MP-LPDDR4-EVK.zip"from office website. But I find this image didn't include onnxtime or tensorflow in /usr/bin. Could you tell me how to build an image with Machine Learning and How to build a toolchain with onnxtime and tensorlfow-lite lib.
I also what to know how to run onnx model using NPU in IMX8MPPLUS-BB.
Hi @dongyang12!
Thank you for contacting NXP Support!
You can refer to our i.MX Machine Learning User's Guide, but You should use the full image to get access to machine learning packages.
https://www.nxp.com/docs/en/user-guide/IMX-MACHINE-LEARNING-UG.pdf
Best Regards.
Chavira
Hi, Chavira
Could you tell me how to get the full image or how to build the full image?
Hi @dongyang12!
You can download the software for iMX8MP by following the link:
https://www.nxp.com/design/design-center/software/embedded-software/i-mx-software/embedded-linux-for...
Best Regards!
Chavira
@Chavira , Thanks your information.
After study file "IMX-MACHINE-LEARNING-UG.pdf", I have some question.
1, From below picture, It seems ONNX only support running in CPU on NXP? Does ONNX support running in NPU? Which EPs are supported?
2, Build file C_Api_Sample.cpp, I need cross-compiling tool, How to generate this toolchain?
Hi @dongyang12!
1 >> Yes, Onxx only runs on the CPU.
2 >> You have to follow the next steps to generate the toolchain:
Install the repo utility:
$: mkdir ~/bin
$: curl http://commondatastorage.googleapis.com/git-repo-downloads/repo > ~/bin/repo
$: chmod a+x ~/bin/repo
$: PATH=${PATH}:~/bin
Download the Yocto Project BSP
$: mkdir <release>
$: cd <release>
$: repo init -u https://github.com/nxp-imx/imx-manifest -b <branch name> [ -m <release manifest>]
$: repo sync
Setup the build folder for a BSP release:
$: [MACHINE=<machine>] [DISTRO=fsl-imx-<backend>] source ./imx-setup-release.sh -b bld-<backend>
<machine> defaults to `imx6qsabresd`
<backend> Graphics backend type xwayland
Wayland with X11 support - default distro
wayland Wayland
fb Framebuffer (not supported for mx8)
Build SDK:
bitbake -c populate_sdk imx-image-full
Best Regards!
Chavira
@Chavira Thanks, I can generate the toolchain.
And I have a question, I get the information from the relase note in file "IMX-MACHINE-LEARNING-UG.pdf". Onnx can run VSI-NPU Execution provider and NNAPI Execution Provider in the previous version. But in the latest version, VSI-NPU and NNAPI been remove. Could you tell me why they been removed. Is it because of poor performance?
@Chavira Could you answer this question?
And I have a question, I get the information from the relase note in file "IMX-MACHINE-LEARNING-UG.pdf". Onnx can run VSI-NPU Execution provider and NNAPI Execution Provider in the previous version. But in the latest version, VSI-NPU and NNAPI been remove. Could you tell me why they been removed. Is it because of poor performance?