Commit Graph

13 Commits

Author SHA1 Message Date
Javier Viguera 70ec1a8630 eiq-examples: update for NXP's lf-6.6.52-2.2.2 release
Removed not needed patch and refreshed the rest.

https://onedigi.atlassian.net/browse/DEL-9905

Signed-off-by: Javier Viguera <javier.viguera@digi.com>
2026-01-21 12:19:25 +01:00
Isaac Hermida 36ba80b3ea imx-machine-learning: use local ssd_mobilenet_v1 model
after commit 8ad093c37b we are detecting
if a model conversion fails.
Currently the mode ssd_mobilenet_v1 fails and cannot be converted by
vela.
This is reported to NXP and will be fixed in next releases.
Meanwhile, for the current version, use a local copy.

Signed-off-by: Isaac Hermida <isaac.hermida@digi.com>
2025-02-25 16:34:14 +01:00
Gabriel Valcazar 007c84d684 imx-machine-learning: onnxruntime: fix compilation error
This mirrors the change done for the STM version of the recipe in meta-digi
commit b7c15294a2, re-using the same patch that was used to fix the
compilation error introduced by poky commit 19fce77a5f13.

Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
2025-01-08 15:32:20 +01:00
Isaac Hermida 8ad093c37b imx-machine-learning: check vela return code
Return an error in case any of the models could not be converted.

https://onedigi.atlassian.net/browse/DEL-9424

Signed-off-by: Isaac Hermida <isaac.hermida@digi.com>
2024-12-20 11:54:27 +01:00
Javier Viguera 16c281ba74 ccimx93: update/fix ethos-u-firmware for NXP release 6.6.36-2.1.0
This imports ethos-u firmware from NXP's MCUXpresso SDK (version
SDK_2_16_000_MIMX9352xxxxM) + some patches provided by NXP to fix the
communication with the TFlite delegate (support case 00663302).

Repo: emp/ethos_u_firmware.git
Revision: 2a98a3899f56830c45fadc708229824e9fee3a01

https://onedigi.atlassian.net/browse/DEL-9355

Signed-off-by: Javier Viguera <javier.viguera@digi.com>
2024-12-17 17:42:36 +01:00
Gabriel Valcazar c7b2c5e94c Generalize WAYLAND_DISPLAY variable at a platform level and set a default value
The value of this variable is always the same within a platform, so it doesn't
make sense to have different variables for it in each recipe that requires it.
While at it, it seems like all currently supported platforms in DEY 5.0 use the
same value of "wayland-1", presumably because they all use similar versions of
wayland/weston. Set this as the default value for all platforms for now. As we
support more platforms, we can adjust this value if needed, but if it ends up
being the same for all platforms, we can just hardcode it.

https://onedigi.atlassian.net/browse/DEL-9404

Co-authored-by: Isaac Hermida <isaac.hermida@digi.com>
Signed-off-by: Isaac Hermida <isaac.hermida@digi.com>
Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
2024-12-11 13:18:03 +01:00
Javier Viguera d83cb9b914 ccimx93: update ML package for NXP release 6.6.36-2.1.0
https://onedigi.atlassian.net/browse/DEL-9257

Signed-off-by: Javier Viguera <javier.viguera@digi.com>
2024-11-05 18:17:02 +01:00
Gabriel Valcazar 26e7365005 imx-machine-learning: ethos-u-firmware: sync with meta-imx
Sync with the scarthgap-6.6.23-2.0.0 branch of meta-imx by transforming the bb
file into a bbappend that only contains our customizations

https://onedigi.atlassian.net/browse/DEL-9011
https://onedigi.atlassian.net/browse/DEL-9081

Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
2024-10-17 13:03:37 +02:00
Gabriel Valcazar f812cbb9b8 imx-machine-learning: tensorflow-lite-ethosu-delegate: remove backported patch
The patch is now included in the latest available version, 2.15.

https://onedigi.atlassian.net/browse/DEL-9011
https://onedigi.atlassian.net/browse/DEL-9081

Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
2024-10-17 13:03:37 +02:00
Gabriel Valcazar 8a5b1594b0 imx-machine-learning: remove recipe for cmake-native 3.25.2
Use the version currently available in poky, which is 3.28.3

https://onedigi.atlassian.net/browse/DEL-9011
https://onedigi.atlassian.net/browse/DEL-9081

Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
2024-10-17 13:03:37 +02:00
Gabriel Valcazar 3316853054 imx-machine-learning: remove code related to deepview-rt
Starting in rel_imx_6.1.36_2.1.0, this functionality is no longer provided by
meta-imx-ml

https://onedigi.atlassian.net/browse/DEL-9011
https://onedigi.atlassian.net/browse/DEL-9081

Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
2024-10-17 13:03:37 +02:00
Gabriel Valcazar 2c14fd36a5 imx-machine-learning: remove backported meta-imx-ml recipes
Back in Yocto 4.0, we backported some ml recipes exclusively for the ccimx93
from meta-imx's rel_imx_6.1.55_2.2.0 tag, but now we simply use what's provided
in the scarthgap-6.6.23-2.0.0 branch of meta-imx. Reflect this in
imx-digi-base.inc.

Remove the recipes for:
    * arm-compute-library
    * eiq-examples
    * ethos-u-driver-stack
    * ethos-u-vela
    * flatbuffers
    * onnxruntime
    * tensorflow-lite

https://onedigi.atlassian.net/browse/DEL-9011
https://onedigi.atlassian.net/browse/DEL-9081

Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
2024-10-17 13:03:37 +02:00
Gabriel Valcazar bd7e84d8b3 global: reflect changes in community layer names after Yocto 5.0 migration
* meta-ml was renamed to meta-imx-ml, and its BBFILE_COLLECTIONS was renamed to
  imx-machine-learning
* meta-st-stm32mpu-ai was renamed to meta-st-x-linux-ai

https://onedigi.atlassian.net/browse/DEL-9011

Signed-off-by: Gabriel Valcazar <gabriel.valcazar@digi.com>
2024-10-17 11:28:26 +02:00