Open Issues: ML Layer Template + Simulation Configuration

[ReinhardKeil]

https://open-cmsis-pack.github.io/cmsis-toolbox/Experimental-Features/#mlops-management defines how project parameters are passed to an MLOps system. These are example files

However there are open ToDo's

I would address now these:

• Need templates for clayer's, basically how is this ai_layer structured for ExecuTorch (or LiteRT).
• The pseudocode does not specify expected file names; the files should have also a namespace and must ensure that no symbol duplicates are exposed; there should be also a C++ version of the pseudocode
• vela memory configuration may have implications on linker script and section names

---

Another topic is the configuration for simulation. Is it possible to use superset configurations for validation (i.e. ethos-u85-2048 to validate also for ethos-u85-256). Are system and/or memory configurations relevant for simulation.

---

How to upgrade the Alif DFP to the latest version of the Ethos driver?

---

Potentially useful:

• https://docs.pytorch.org/executorch/1.0/backends-arm-ethos-u.html - Arm® Ethos™-U NPU Backend — ExecuTorch 1.0 documentation