AnsweredAssumed Answered

eIQ Tensorflow on i.MX8M Mini yocto

Question asked by David Vescovi on Jun 3, 2019
Latest reply on Jun 4, 2019 by Bio_TICFSL

I was following the new eIQ directions https://www.nxp.com/docs/en/nxp/user-guides/UM11226.pdf 

My target is iMX8Mmini so I slightly changed the yocto build for the mini target.

Everything builds except when it come to including tensorflow in the image.

 

when I have local.conf line:

IMAGE_INSTALL_append = " arm-compute-library tensorflow-lite armnn onnxruntime"

 

everything builds fine but when I add the "tesorflow" component like the app note describes:

IMAGE_INSTALL_append = " arm-compute-library tensorflow tensorflow-lite armnn onnxruntime"

 

I get this error:

david@ubuntu18:/datadrive/imx_yocto-bspml/bld8mm_xwayland$ bitbake fsl-image-qt5
WARNING: Host distribution "ubuntu-18.04" has not been validated with this version of the build system; you may possibly experience unexpected failures. It is recommended that you use a tested distribution.
Loading cache: 100% |####################################################################################| Time: 0:00:00
Loaded 3499 entries from dependency cache.
NOTE: Resolving any missing task queue dependencies
ERROR: Nothing RPROVIDES 'tensorflow' (but /datadrive/imx_yocto-bspml/sources/meta-fsl-bsp-release/imx/meta-sdk/dynamic-layers/qt5-layer/recipes-fsl/images/fsl-image-qt5.bb RDEPENDS on or otherwise requires it)
tensorflow was skipped: incompatible with machine imx8mmevk (not in COMPATIBLE_MACHINE)
NOTE: Runtime target 'tensorflow' is unbuildable, removing...
Missing or unbuildable dependency chain was: ['tensorflow']
ERROR: Required build target 'fsl-image-qt5' has no buildable providers.
Missing or unbuildable dependency chain was: ['fsl-image-qt5', 'tensorflow']
Summary: There was 1 WARNING message shown.
Summary: There were 2 ERROR messages shown, returning a non-zero exit code.
david@ubuntu18:/datadrive/imx_yocto-bspml/bld8mm_xwayland$

 

It looks like it has an issue with the COMPATIBLE_MACHINE. Is this enumerated somewhere in the recipes somewhere?

 

Looks like it does not complain when using the iMX8QM as the target.

Any advice? 

Outcomes