Introduction to TinyML on Arm using PyTorch and ExecuTorch: Failed to generate .pte file

[avptin]

Hi!

I am following https://learn.arm.com/learning-paths/embedded-and-microcontrollers/introduction-to-tinyml-on-arm/ on Ubuntu 24.04 WSL and getting the following error on the .pte file generation stage:

python -m examples.arm.aot_arm_compiler --model_name=examples/arm/simple_nn.py \
--delegate --quantize --target=ethos-u85-256 \
--so_library=cmake-out-aot-lib/kernels/quantized/libquantized_ops_aot_lib.so \
--system_config=Ethos_U85_SYS_DRAM_Mid --memory_mode=Sram_Only

Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/home/peutin/executorch/examples/arm/aot_arm_compiler.py", line 819, in <module>
    args = get_args()
           ^^^^^^^^^^
  File "/home/user/executorch/examples/arm/aot_arm_compiler.py", line 624, in get_args
    torch.ops.load_library(args.so_library)
  File "/home/user/executorch-venv/lib/python3.12/site-packages/torch/_ops.py", line 1490, in load_library
    raise OSError(f"Could not load this library: {path}") from e
OSError: Could not load this library: /home/user/executorch/cmake-out-aot-lib/kernels/quantized/libquantized_ops_aot_lib.so

The entire "cmake-out-aot-lib" directory is missing.

The libquantized_ops_aot_lib.so library is being generated to a different location, but the operations fails too:

python -m examples.arm.aot_arm_compiler --model_name=examples/arm/simple_nn.py --delegate --quantize --target=ethos-u85-256 --so_library=pip-out/lib.linux-x86_64-cpython-312/executorch/kernels/quantized/libquantized_ops_aot_lib.so --system_config=Ethos_U85_SYS_DRAM_Mid --memory_mode=Sram_Only
[operator_registry.cpp:89] Re-registering quantized_decomposed::add.out, from NOT_SUPPORTED
[operator_registry.cpp:90] key: (null), is_fallback: true
[operator_registry.cpp:111] In function register_kernels(), assert failed (false): Kernel registration failed with error 22, see error log for details.
Aborted (core dumped)

executorch_full_log.txt

[perheld]

Hi! Annoying that the example doesnt work. Seems like this has not been updated since the library started to be loaded by default when importing the quantizer.

So, what has happened is that you dont need to specify the so library manually anymore.

So the line:
python -m examples.arm.aot_arm_compiler --model_name=examples/arm/simple_nn.py --delegate --quantize --target=ethos-u85-256 --system_config=Ethos_U85_SYS_DRAM_Mid --memory_mode=Sram_Only

Should be enough, I tested it.

[pareenaverma]

Thank you @avptin for reporting the issue and @perheld for addressing the issue. It hadn't been updating since the change was made where the library is loaded by default now. Updated the instructions and will be published on the site on Friday.

[avptin]

Hi @pareenaverma! The https://learn.arm.com/learning-paths/embedded-and-microcontrollers/introduction-to-tinyml-on-arm/4-build-model/ has been updated - closing the issue now. Thank you!