search

tensorflow / tflite-micro

Infrastructure to enable deployment of ML models to low-power resource-constrained embedded targets (including microcontrollers and digital signal processors).
Apache License 2.0
1.93k stars 826 forks source link

Update Ethos-U driver. #2626

Closed iabdalkader closed 3 months ago

iabdalkader commented 4 months ago

Note I tested with TARGET=cortex_m_corstone_300 test_network_tester_test and on real hardware as well built with:

make -j12 -f tensorflow/lite/micro/tools/make/Makefile \
TARGET=cortex_m_generic TARGET_ARCH=cortex-m55 \
CO_PROCESSOR=ethos_u ETHOSU_ARCH=u55 OPTIMIZED_KERNEL_DIR=ethos_u  \
CORE_OPTIMIZATION_LEVEL=-O2 KERNEL_OPTIMIZATION_LEVEL=-O2 \
THIRD_PARTY_KERNEL_OPTIMIZATION_LEVEL=-O2 \
TARGET_TOOLCHAIN_ROOT=/opt/arm-none-eabi/bin/ \ 
TARGET_TOOLCHAIN_PREFIX=arm-none-eabi- BUILD_TYPE=release microlite

BUG=#2619

google-cla[bot] commented 4 months ago

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

iabdalkader commented 4 months ago

Rebased.. unsure how to fix that CI entry test error.

rascani commented 4 months ago

Rebased.. unsure how to fix that CI entry test error.

Don't worry about this, it's just an issue I haven't had time to fix in the CI system. Basically, it is erroring because it doesn't have the ready to merge label applied, but that shouldn't be needed until after the PR is approved.

Thank you for the PR! It is greatly appreciated.

iabdalkader commented 3 months ago

@rascani Seems my patch might have broken some test here: https://github.com/tensorflow/tflite-micro/actions/runs/10087978372/job/27892985767

I'll try to fix it.

rascani commented 3 months ago

Thanks @iabdalkader

mansnils commented 3 months ago

Thanks for this @iabdalkader

iabdalkader commented 3 months ago

You're welcome. Note that there were some issue that I've missed because the workflow runs on schedule. The workflow for Cortex-M doesn't take much time, just a little over 10 minutes, so perhaps it's best to change it to trigger on PRs. This could save a lot of time in the future.

mansnils commented 3 months ago

Yes agreed. We are looking into that.