Draft: [New] first conv with cmsis_nn working

Context

In the context of the integration of cmsis_nn for aidge. This merge request is a first version focusing on the ConvScaling metoperator.

Modified files

• operator.py : Now using aidge_quantization.Quantizer instead of aidge_core.scaling. Only including "arm_nn_functions.h" and copying kernels in export.py. Using my own function to compute fixed_point shift and multiplier;
• export.py, now copying folder of cmsis_nn and two options between "aidge_arm" or "cmsis_nn";
• arm_cmsis_nn_conv2d.jinja, both for configuration and forward call with debug options

TODO

• Dataformat different for CMSIS_NN (HWC), need to change ImplSpec and apply recipes
• Only one _mem_offset.jinja, _meminfo.jinja and _def_io.jinja
• Other Kernels (Max/AvgPooling, Relu, FC, PaddedConv, DW) that are available in CMSIS_NN library
• Good method to choose the dtype

assigned to @wboussella

requested review from @cmoineau

marked this merge request as draft

mentioned in merge request !5 (closed)

added 1 commit

• cff2ffbe - export_registry.py : adding an init for ExportLib

Compare with previous version

added 22 commits

• cff2ffbe...b920a41f - 18 commits from branch eclipse/aidge:dev
• 467eb8ea - Merge branch 'dev' into 'main'
• 5a253419 - Merge branch 'dev' into 'main'
• 52fdeccc - UPD: version 0.1.0 -> 0.1.1
• 2ff196bd - Merge

Compare with previous version

currently adapt to backend is not working which block the goals of this MR. I suggest to do it without adapt_to_backend by transposing parameters while exporting like it's done in export_cpp, and by adding a transpose node for the first input. It's allow me to confirm my work on templates, and in a future version it will be easy to use it with adapt to backend. @cmoineau @olivierbichler

What is the non working behavior of adaptToBackend?

[ERROR] - Assertion failed: mImpl != nullptr in /local2/is148265/wb274724/STM32_dev/dev/superpoint/aidge_may/aidge/aidge/aidge_core/src/operator/Operator.cpp:64
[FATAL] - resetConsummerProducer(): an implementation is required for ReLU!
Traceback (most recent call last):
  File "/local2/is148265/wb274724/STM32_dev/dev/superpoint/aidge_may/aidge/aidge/aidge_export_arm_cortexm/ba.py", line 201, in <module>
    scheduler = aidge_core.SequentialScheduler(model)
RuntimeError: resetConsummerProducer(): an implementation is required for ReLU!

The problem is that i connot generate scheduling with a model after an adapt to backend containing metaop in this case i haven't a ReLU in my model only a metaop named PaddedConvScalingRelu which is defined in operator.py from aidge_export_arm_cortexm

Seems like a minor bug, reserConsummerProducer must have a bad behavior if MetaOp has an implementation, it shouldn't call resetConsummerProducer on the sub graph.

added StatusDelayed label

added StatusWork in Progress label and removed StatusDelayed label

added StatusDelayed label and removed StatusWork in Progress label