I’ve been using TVM to deploy pre-trained models to Bare Metal embedded systems (specifically, ARM 32 bits).
I am wondering if there is any way to cross-compile a model (e.g. .*.tflite, *.onnx) or even an operator to an ARM 32 bits binary with AoT executor. I do not need the flash it to any board – Currently, I just want to check the generated binary.
Thank you in advance and any suggestion is very appreciated!
please take a look at tests/micro/zephyr/test_zephyr_aot.py. we don’t have this properly exposed yet in e.g. tvmc or other more user-facing scripts such as tutorials just yet. apologies for that–development is still ongoing here particularly to the firmware-facing APIs.