kernel 'executorch_prim::et_view.default' not found.

[dinusha94]

🐛 Describe the bug

I am trying to use mobilenet v2 pte with the executorch module extension. but getting the following error. I have followed the setting up tutorial and set up the executorch environment.

I am trying to incorporate the executorch into my simple application using the following CMakeList.txt

cmake_minimum_required(VERSION 3.19)
project(SimpleProject)

# Specify the C++ standard
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
set(CMAKE_CXX_EXTENSIONS OFF)

# Add the main executable
add_executable(SimpleApp src/main.cpp)

# Link against the ExecuTorch library and additional dependencies
target_link_libraries(SimpleApp PRIVATE
    /usr/local/lib/libexecutorch.a
    /usr/local/lib/libextension_module_static.a
    /usr/local/lib/libexecutorch_no_prim_ops.a
    /usr/local/lib/libextension_data_loader.a
    /usr/local/lib/libextension_tensor.a
    -Wl,--whole-archive /usr/local/lib/libportable_ops_lib.a -Wl,--no-whole-archive
    /usr/local/lib/libportable_kernels.a
)

# Include directories for ExecuTorch
target_include_directories(SimpleApp PRIVATE
    /usr/local/include/executorch
)

I exported the mobilent v2 into a pte file using the following python script

import torch
import torchvision
import executorch.exir as exir

from torch._export import capture_pre_autograd_graph
from torch.export import export, ExportedProgram

m = torchvision.models.MobileNetV2()
m.eval()

example_args = (torch.randn(1, 3, 224, 224),)
pre_autograd_aten_dialect = capture_pre_autograd_graph(m, example_args)

aten_dialect: ExportedProgram = export(pre_autograd_aten_dialect, example_args)
edge_program: exir.EdgeProgramManager = exir.to_edge(aten_dialect)

executorch_program: exir.ExecutorchProgramManager = edge_program.to_executorch(
    exir.ExecutorchBackendConfig(
        passes=[],  # User-defined passes
    )
)

with open("./models/mobilenet_v2.pte", "wb") as file:
    file.write(executorch_program.buffer)

Following in my main.cpp script where I use the executorch module extension to run an inference.

#include <iostream>
#include <executorch/extension/module/module.h>
#include <executorch/extension/tensor/tensor.h>
#include <executorch/runtime/core/evalue.h>

using namespace ::executorch::extension;
using namespace ::executorch::runtime;

using executorch::runtime::Result;


int main() {

    // Create a Module.
    Module module("../models/mobilenet_v2.pte");

    // Wrap the input data with a Tensor.
    const int size = 1 * 3 * 224 * 224; // Total size of the array
    float input[size]; // Declare the array

    // Populate the array with a constant value of 2
    for (int i = 0; i < size; ++i) {
        input[i] = 2.0f;
    }

    auto tensor = from_blob(input, {1, 3, 224, 224});

    // Perform an inference.
    const auto result = module.forward(tensor);

    // Check for success or failure.
    if (result.ok()) {

        const auto& vec = *result;
        size_t length = vec.size();
        std::cout << "Vector length: " << length << std::endl;

        const auto num_dim = result->at(0).toTensor().dim();
        std::cout << "num dim : " << num_dim << std::endl; 

        const auto ze_dim_ = result->at(0).toTensor().size(1);
        std::cout << "size dim : " << ze_dim_ << std::endl;   
        
        const auto num_elements = result->at(0).toTensor().numel();
        std::cout << "num elements in bits : " << num_elements << std::endl; 

        // Retrieve the output data.
        const auto output = result->at(0).toTensor().const_data_ptr<float>();

        std::cout << "Tensor data: ";
        for (int i = 0; i < num_elements; ++i) {
            std::cout << output[i] << " ";
        }
        std::cout << std::endl;
    }

    return 0;
}

after make and run ./SimpleApp I get the following error. when I try the add.pte and the SimpleConv model shown in the exporting tutorial I get no such issues and works fine.

when I inspect the libexcutorch.a with nm tool I see the following et_view definition, not the one shown in the error

nm -C libexecutorch.a | grep et_view

0000000000000000 T torch::executor::function::et_view(executorch::runtime::KernelRuntimeContext&, executorch::runtime::EValue**)
                 U torch::executor::function::et_view(executorch::runtime::KernelRuntimeContext&, executorch::runtime::EValue**)

E 00:00:00.031327 executorch:operator_registry.cpp:185] kernel 'executorch_prim::et_view.default' not found.
E 00:00:00.031451 executorch:operator_registry.cpp:186] dtype: 6 | dim order: [
E 00:00:00.031467 executorch:operator_registry.cpp:186] 0,
E 00:00:00.031476 executorch:operator_registry.cpp:186] 1,
E 00:00:00.031483 executorch:operator_registry.cpp:186] 2,
E 00:00:00.031491 executorch:operator_registry.cpp:186] 3,
E 00:00:00.031498 executorch:operator_registry.cpp:186] ]
E 00:00:00.031506 executorch:operator_registry.cpp:186] dtype: 6 | dim order: [
E 00:00:00.031513 executorch:operator_registry.cpp:186] 0,
E 00:00:00.031521 executorch:operator_registry.cpp:186] 1,
E 00:00:00.031528 executorch:operator_registry.cpp:186] ]
E 00:00:00.031536 executorch:method.cpp:542] Missing operator: [5] executorch_prim::et_view.default
E 00:00:00.031557 executorch:method.cpp:736] There are 1 instructions don't have corresponding operator registered. See logs for details

Any help on this would be highly appreciated
Thank you

Versions

Collecting environment information...
PyTorch version: 2.5.0+cpu
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A

OS: Ubuntu 20.04.6 LTS (x86_64)
GCC version: (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0
Clang version: Could not collect
CMake version: version 3.31.2
Libc version: glibc-2.31

Python version: 3.10.0 (default, Mar  3 2022, 09:58:08) [GCC 7.5.0] (64-bit runtime)
Python platform: Linux-5.15.0-127-generic-x86_64-with-glibc2.31
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
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture:                         x86_64
CPU op-mode(s):                       32-bit, 64-bit
Byte Order:                           Little Endian
Address sizes:                        45 bits physical, 48 bits virtual
CPU(s):                               4
On-line CPU(s) list:                  0-3
Thread(s) per core:                   1
Core(s) per socket:                   1
Socket(s):                            4
NUMA node(s):                         1
Vendor ID:                            GenuineIntel
CPU family:                           6
Model:                                154
Model name:                           12th Gen Intel(R) Core(TM) i7-1255U
Stepping:                             4
CPU MHz:                              2611.209
BogoMIPS:                             5222.41
Hypervisor vendor:                    VMware
Virtualization type:                  full
L1d cache:                            192 KiB
L1i cache:                            128 KiB
L2 cache:                             5 MiB
L3 cache:                             48 MiB
NUMA node0 CPU(s):                    0-3
Vulnerability Gather data sampling:   Not affected
Vulnerability Itlb multihit:          Not affected
Vulnerability L1tf:                   Mitigation; PTE Inversion
Vulnerability Mds:                    Mitigation; Clear CPU buffers; SMT Host state unknown
Vulnerability Meltdown:               Mitigation; PTI
Vulnerability Mmio stale data:        Not affected
Vulnerability Reg file data sampling: Vulnerable: No microcode
Vulnerability Retbleed:               Mitigation; IBRS
Vulnerability Spec rstack overflow:   Not affected
Vulnerability Spec store bypass:      Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:             Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:             Mitigation; IBRS; IBPB conditional; STIBP disabled; RSB filling; PBRSB-eIBRS Not affected; BHI SW loop, KVM SW loop
Vulnerability Srbds:                  Not affected
Vulnerability Tsx async abort:        Not affected
Flags:                                fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon rep_good nopl xtopology tsc_reliable nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch pti ssbd ibrs ibpb stibp fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves avx_vnni arat umip gfni vaes vpclmulqdq rdpid movdiri movdir64b fsrm md_clear serialize flush_l1d arch_capabilities

Versions of relevant libraries:
[pip3] executorch==0.4.0a0+6a085ff
[pip3] numpy==1.21.3
[pip3] torch==2.5.0+cpu
[pip3] torchaudio==2.5.0+cpu
[pip3] torchsr==1.0.4
[pip3] torchvision==0.20.0+cpu
[conda] executorch                0.4.0a0+6a085ff          pypi_0    pypi
[conda] numpy                     1.21.3                   pypi_0    pypi
[conda] torch                     2.5.0+cpu                pypi_0    pypi
[conda] torchaudio                2.5.0+cpu                pypi_0    pypi
[conda] torchsr                   1.0.4                    pypi_0    pypi
[conda] torchvision               0.20.0+cpu               pypi_0    pypi

[dinusha94]

After I use another implementation of mobilenetv2 it worked, don't know why

[kimishpatel]

So if you add /usr/local/lib/libexecutorch.a to be linked whole as well, similar to /usr/local/lib/libportable_ops_lib.a, it might work. Can you try?

[kimishpatel]

@dbort to update our docs to showcase how to use executorch libs into your custom cpp apps

[jpeng2012]

@dinusha94
I am not able to find
/usr/local/lib/libextension_module_static.a
/usr/local/lib/libextension_data_loader.a
/usr/local/lib/libextension_tensor.a
following the setup build method from the tutorial.

[SS-JIA]

@jpeng2012 in order for those libraries to be built, you need to enable them in your cmake command. For example:

(rm -rf cmake-out && \
  cmake . \
  -DCMAKE_INSTALL_PREFIX=cmake-out \
  -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
  -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
  -DEXECUTORCH_BUILD_EXTENSION_RUNNER_UTIL=ON \
  -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
  -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
  -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \
  -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
  -DEXECUTORCH_BUILD_TESTS=ON \
  -Bcmake-out && \
  cmake --build cmake-out -j64 --target install)

[SS-JIA]

After I use another implementation of mobilenetv2 it worked, don't know why

@dinusha94 just curious, what other implementation of mobilenetv2 did you try that resulted in it working?