[ARM] aot_arm_compiler.py does not quantize Conv2d for cortex-m55+int8 target

[jonasdaugalas]

🐛 Describe the bug

Model:

import torch

ModelUnderTest = torch.nn.Conv2d(in_channels=2, out_channels=4, kernel_size=3)
ModelInputs = (torch.randn(1, 2, 8, 8),)

Log:

$ /home/wsluser/proj/nn-deploy-kit/.venv/bin/python -m examples.arm.aot_arm_compiler \
--model_name /home/wsluser/proj/nn-deploy-kit/repro/aot_compiler_cortex_m_conv/conv2d.py \
--target cortex-m55+int8 \
--output /home/wsluser/proj/nn-deploy-kit/repro/aot_compiler_cortex_m_conv/out_conv2d__cortex-m__Delegate=N__Quantize=Y/model.pte \
--intermediates /home/wsluser/proj/nn-deploy-kit/repro/aot_compiler_cortex_m_conv/out_conv2d__cortex-m__Delegate=N__Quantize=Y/intermediates \
--quantize
[WARNING 2026-03-05 15:40:35,398 aot_arm_compiler.py:812] Converting input tensor with shape [1, 2, 8, 8] to channels_last
/usr/lib/python3.12/copyreg.py:99: FutureWarning: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead.
  return cls.__new__(cls, *args)
/usr/lib/python3.12/copyreg.py:99: FutureWarning: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead.
  return cls.__new__(cls, *args)
/usr/lib/python3.12/copyreg.py:99: FutureWarning: `isinstance(treespec, LeafSpec)` is deprecated, use `isinstance(treespec, TreeSpec) and treespec.is_leaf()` instead.
  return cls.__new__(cls, *args)
/home/wsluser/proj/nn-deploy-kit/.venv/lib/python3.12/site-packages/executorch/exir/tensor.py:83: FutureWarning: guard_size_oblivious will be removed. Consider using explicit unbacked handling     potentially utilizing guard_or_false, guard_or_true, or statically_known_true
  return guard_size_oblivious(self.stride < other.stride)
PTE file saved as /home/wsluser/proj/nn-deploy-kit/repro/aot_compiler_cortex_m_conv/out_conv2d__cortex-m__Delegate=N__Quantize=Y/model.pte

Visualization model-explorer --extensions=pte_adapter_model_explorer:

Observed behavior: the convolution op is floating-point.
Expected behavior: the convolution to be int8x8 quantized.

Versions

Collecting environment information...
PyTorch version: 2.11.0.dev20260215+cpu
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A

OS: Ubuntu 24.04.3 LTS (x86_64)
GCC version: (Ubuntu 13.3.0-6ubuntu2~24.04.1) 13.3.0
Clang version: Could not collect
CMake version: version 3.31.10
Libc version: glibc-2.39

Python version: 3.12.3 (main, Jan 22 2026, 20:57:42) [GCC 13.3.0] (64-bit runtime)
Python platform: Linux-6.6.87.2-microsoft-standard-WSL2-x86_64-with-glibc2.39
Is CUDA available: False
CUDA runtime version: No CUDA
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
Is XPU available: False
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True
Caching allocator config: N/A

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 39 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 16
On-line CPU(s) list: 0-15
Vendor ID: GenuineIntel
Model name: Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz
CPU family: 6
Model: 165
Thread(s) per core: 2
Core(s) per socket: 8
Socket(s): 1
Stepping: 5
BogoMIPS: 5808.01
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon rep_good nopl xtopology cpuid tsc_known_freq pni pclmulqdq vmx ssse3 fma cx16 pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow ept vpid ept_ad fsgsbase bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt xsaveopt xsavec xgetbv1 xsaves vnmi md_clear flush_l1d arch_capabilities
Virtualization: VT-x
Hypervisor vendor: Microsoft
Virtualization type: full
L1d cache: 256 KiB (8 instances)
L1i cache: 256 KiB (8 instances)
L2 cache: 2 MiB (8 instances)
L3 cache: 16 MiB (1 instance)
NUMA node(s): 1
NUMA node0 CPU(s): 0-15
Vulnerability Gather data sampling: Unknown: Dependent on hypervisor status
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT Host state unknown
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Mitigation; Enhanced IBRS
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; Enhanced / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI SW loop, KVM SW loop
Vulnerability Srbds: Unknown: Dependent on hypervisor status
Vulnerability Tsx async abort: Not affected

Versions of relevant libraries:
[pip3] executorch==1.2.0a0+1619308
[pip3] numpy==2.4.2
[pip3] pytorch_tokenizers==1.1.0
[pip3] torch==2.11.0.dev20260215+cpu
[pip3] torchao==0.16.0+git026b76d1
[pip3] torchaudio==2.11.0.dev20260215+cpu
[pip3] torchdata==0.11.0
[pip3] torchsr==1.0.4
[pip3] torchtune==0.0.0
[pip3] torchvision==0.26.0.dev20260215+cpu
[conda] Could not collect

cc @digantdesai @SS-JIA @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell @psiddh @AdrianLundell

[AdrianLundell]

Hi, many of the cortex-m kernels (i.e convolution) require channels-last input, and the quantizer is designed to not quantize any non-supported kernels, hence the floating point operator. Make sure to export the graph with channels-last input using ModelInputs = (torch.randn(1, 2, 8, 8).to(memory_format=torch.channels_last,). If you run with DEBUG log level you should see a full quantization report where this is logged, have you seen this?

[jonasdaugalas]

Thanks @AdrianLundell , the memory format change does the job 👍