ARM-software / armnn

Arm NN ML Software. The code here is a read-only mirror of https://review.mlplatform.org/admin/repos/ml/armnn
https://developer.arm.com/products/processors/machine-learning/arm-nn
MIT License
1.15k stars 307 forks source link

delegate test failed #589

Closed steven9046 closed 2 years ago

steven9046 commented 2 years ago

I followed BuildGuideNative.md in delegate/ and successfully build armnn and delegate. When I come to this:

To ensure that the build was successful you can run the unit tests for the delegate that can be found in the build directory for the delegate. Doctest was used to create those tests. Using test filters you can filter out tests that your build is not configured for. In this case, because Arm NN was only built for Cpu acceleration (CpuAcc), we filter for all test suites that have CpuAcc in their name.

cd $BASEDIR/armnn/delegate/build
./DelegateUnitTests --test-suite=*CpuAcc* 

If you have built for Gpu acceleration as well you might want to change your test-suite filter:

./DelegateUnitTests --test-suite=*CpuAcc*,*GpuAcc*

I build this project on an aarch64 machine. The cpu test is ok but GpuAcc did't work.

2102125884

When I test with tflite another error came up.

temi@temi:~/ArmNNDelegate/benchmarking$ LD_LIBRARY_PATH=../armnn/build ./benchmark_model --graph=mobilenet_v2_1.0_224_quantized_1_default_1.tflite --external_delegate_path="../armnn/build/delegate/libarmnnDelegate.so" --external_delegate_options="backends:CpuAcc;logging-severity:info" STARTING! Log parameter values verbosely: [0] Graph: [mobilenet_v2_1.0_224_quantized_1_default_1.tflite] External delegate path: [../armnn/build/delegate/libarmnnDelegate.so] External delegate options: [backends:CpuAcc;logging-severity:info] Loaded model mobilenet_v2_1.0_224_quantized_1_default_1.tflite EXTERNAL delegate created

When I type clinfo it prints"Number of platforms 0 " Is this means my driver for Mali GPU is missing?

How to solve this? Any suggestions? Thanks!

steven9046 commented 2 years ago

I followed BuildGuideNative.md in delegate/ and successfully build armnn and delegate. When I come to this:

To ensure that the build was successful you can run the unit tests for the delegate that can be found in the build directory for the delegate. Doctest was used to create those tests. Using test filters you can filter out tests that your build is not configured for. In this case, because Arm NN was only built for Cpu acceleration (CpuAcc), we filter for all test suites that have CpuAcc in their name.

cd $BASEDIR/armnn/delegate/build
./DelegateUnitTests --test-suite=*CpuAcc* 

If you have built for Gpu acceleration as well you might want to change your test-suite filter:

./DelegateUnitTests --test-suite=*CpuAcc*,*GpuAcc*

I build this project on an aarch64 machine. The cpu test is ok but GpuAcc did't work.

2102125884

When I test with tflite another error came up.

temi@temi:~/ArmNNDelegate/benchmarking$ LD_LIBRARY_PATH=../armnn/build ./benchmark_model --graph=mobilenet_v2_1.0_224_quantized_1_default_1.tflite --external_delegate_path="../armnn/build/delegate/libarmnnDelegate.so" --external_delegate_options="backends:CpuAcc;logging-severity:info" STARTING! Log parameter values verbosely: [0] Graph: [mobilenet_v2_1.0_224_quantized_1_default_1.tflite] External delegate path: [../armnn/build/delegate/libarmnnDelegate.so] External delegate options: [backends:CpuAcc;logging-severity:info] Loaded model mobilenet_v2_1.0_224_quantized_1_default_1.tflite EXTERNAL delegate created

When I type clinfo it prints"Number of platforms 0 " Is this means my driver for Mali GPU is missing?

How to solve this? Any suggestions? Thanks!

I have fix the mali driver problem. I donwload the mali.so from here and now when I run

sudo clinfo

It prints

arm_release_ver of this libmali is 'r18p0-01rel0', rk_so_ver is '4'.Number of platforms                               1                                                                                                                                                                                                        
  Platform Name                                   ARM Platform                                                                                                                                                                                                                                                                 
  Platform Vendor                                 ARM                                                                                                                                                                                                                                                                          
  Platform Version                                OpenCL 1.2 v1.r18p0-01rel0.5cb5681058e8e076ff89747c20c32578                                                                                                                                                                                                                  
  Platform Profile                                FULL_PROFILE                                                                                                                                                                                                                                                                 
  Platform Extensions                             cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_exte                                                                                                                                                nded_atomics cl_khr_byte_addressable_store cl_khr_3d_image_writes cl_khr_fp64 cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_fp16 cl_khr_icd cl_khr_egl_image c                                                                                                                                                l_khr_image2d_from_buffer cl_arm_core_id cl_arm_printf cl_arm_thread_limit_hint cl_arm_non_uniform_work_group_size cl_arm_import_memory                                                                                                                                                                                        
  Platform Extensions function suffix             ARM                                                                                                                                                                                                                                                                          

  Platform Name                                   ARM Platform                                                                                                                                                                                                                                                                 
Number of devices                                 1                                                                                                                                                                                                                                                                            
  Device Name                                     Mali-T860                                                                                                                                                                                                                                                                    
  Device Vendor                                   ARM                                                                                                                                                                                                                                                                          
  Device Vendor ID                                0x8602000                                                                                                                                                                                                                                                                    
  Device Version                                  OpenCL 1.2 v1.r18p0-01rel0.5cb5681058e8e076ff89747c20c32578                                                                                                                                                                                                                  
  Driver Version                                  1.2                                                                                                                                                                                                                                                                          
  Device OpenCL C Version                         OpenCL C 1.2 v1.r18p0-01rel0.5cb5681058e8e076ff89747c20c32578                                                                                                                                                                                                                
  Device Type                                     GPU                                                                                                                                                                                                                                                                          
  Device Profile                                  FULL_PROFILE                                                                                                                                                                                                                                                                 
  Device Available                                Yes                                                                                                                                                                                                                                                                          
  Compiler Available                              Yes                                                                                                                                                                                                                                                                          
  Linker Available                                Yes                                                                                                                                                                                                                                                                          
  Max compute units                               4                                                                                                                                                                                                                                                                            
  Max clock frequency                             5MHz                                                                                                                                                                                                                                                                         
  Device Partition                                (core)                                                                                                                                                                                                                                                                       
    Max number of sub-devices                     0                                                                                                                                                                                                                                                                            
    Supported partition types                     None                                                                                                                                                                                                                                                                         
    Supported affinity domains                    (n/a)                                                                                                                                                                                                                                                                        
  Max work item dimensions                        3                                                                                                                                                                                                                                                                            
  Max work item sizes                             256x256x256                                                                                                                                                                                                                                                                  
  Max work group size                             256                                                                                                                                                                                                                                                                          
  Preferred work group size multiple              4                                                                                                                                                                                                                                                                            
  Preferred / native vector sizes                                                                                                                                                                                                                                                                                              
    char                                                16 / 16                                                                                                                                                                                                                                                                
    short                                                8 / 8                                                                                                                                                                                                                                                                 
    int                                                  4 / 4                                                                                                                                                                                                                                                                 
    long                                                 2 / 2                                                                                                                                                                                                                                                                 
    half                                                 8 / 8        (cl_khr_fp16)                                                                                                                                                                                                                                            
    float                                                4 / 4                                                                                                                                                                                                                                                                 
    double                                               2 / 2        (cl_khr_fp64)                                                                                                                                                                                                                                            
  Half-precision Floating-point support           (cl_khr_fp16)                                                                                                                                                                                                                                                                
    Denormals                                     Yes                                                                                                                                                                                                                                                                          
    Infinity and NANs                             Yes                                                                                                                                                                                                                                                                          
    Round to nearest                              Yes                                                                                                                                                                                                                                                                          
    Round to zero                                 Yes                                                                                                                                                                                                                                                                          
    Round to infinity                             Yes                                                                                                                                                                                                                                                                          
    IEEE754-2008 fused multiply-add               Yes                                                                                                                                                                                                                                                                          
    Support is emulated in software               No                                                                                                                                                                                                                                                                           
  Single-precision Floating-point support         (core)                                                                                                                                                                                                                                                                       
    Denormals                                     Yes                                                                                                                                                                                                                                                                          
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
    Correctly-rounded divide and sqrt operations  No
  Double-precision Floating-point support         (cl_khr_fp64)
    Denormals                                     Yes
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
  Address bits                                    64, Little-Endian
  Global memory size                              4030976000 (3.754GiB)
  Error Correction support                        No
  Max memory allocation                           1007744000 (961.1MiB)
  Unified memory for Host and Device              Yes
  Minimum alignment for any data type             128 bytes
  Alignment of base address                       1024 bits (128 bytes)
  Global Memory cache type                        Read/Write
  Global Memory cache size                        262144 (256KiB)
  Global Memory cache line size                   64 bytes
  Image support                                   Yes
    Max number of samplers per kernel             16
    Max size for 1D images from buffer            65536 pixels
    Max 1D or 2D image array size                 2048 images
    Base address alignment for 2D image buffers   32 bytes
    Pitch alignment for 2D image buffers          16 pixels
    Max 2D image size                             65536x65536 pixels
    Max 3D image size                             65536x65536x65536 pixels
    Max number of read image args                 128
    Max number of write image args                8
  Local memory type                               Global
  Local memory size                               32768 (32KiB)
  Max number of constant args                     8
  Max constant buffer size                        65536 (64KiB)
  Max size of kernel argument                     1024
  Queue properties                                
    Out-of-order execution                        Yes
    Profiling                                     Yes
  Prefer user sync for interop                    No
  Profiling timer resolution                      1000ns
  Execution capabilities                          
    Run OpenCL kernels                            Yes
    Run native kernels                            No
  printf() buffer size                            1048576 (1024KiB)
  Built-in kernels                                (n/a)
  Device Extensions                               cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_byte_addressable_store cl_khr_3d_image_writes cl_khr_fp64 cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_fp16 cl_khr_icd cl_khr_egl_image cl_khr_image2d_from_buffer cl_arm_core_id cl_arm_printf cl_arm_thread_limit_hint cl_arm_non_uniform_work_group_size cl_arm_import_memory

NULL platform behavior
  clGetPlatformInfo(NULL, CL_PLATFORM_NAME, ...)  ARM Platform
  clGetDeviceIDs(NULL, CL_DEVICE_TYPE_ALL, ...)   Success [ARM]
  clCreateContext(NULL, ...) [default]            Success [ARM]
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_DEFAULT)  Success (1)
    Platform Name                                 ARM Platform
    Device Name                                   Mali-T860
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_CPU)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_GPU)  Success (1)
    Platform Name                                 ARM Platform
    Device Name                                   Mali-T860
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_ACCELERATOR)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_CUSTOM)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_ALL)  Success (1)
    Platform Name                                 ARM Platform
    Device Name                                   Mali-T860

ICD loader properties
  ICD loader Name                                 OpenCL ICD Loader
  ICD loader Vendor                               OCL Icd free software
  ICD loader Version                              2.2.11

But when I run the test, it still failed.

/home/temi/ArmNNDelegate/armnn/delegate/src/test/UnpackTest.cpp:163:
TEST SUITE: Unpack_GpuAccTests
TEST CASE:  Unpack_Uint8_Axis0_Num4_GpuAcc_Test

/home/temi/ArmNNDelegate/armnn/delegate/src/test/UnpackTest.cpp:163: ERROR: test case THREW exception: TfLiteArmnnDelegate: No known backend specified.

Failed creating base context during opening of kernel driver.
Kernel module may not have been loaded
INFO: TfLiteArmnnDelegate: Requested unknown backend GpuAcc
arm_release_ver of this libmali is 'r18p0-01rel0', rk_so_ver is '4'.arm_release_ver of this libmali is 'r18p0-01rel0', rk_so_ver is '4'.===============================================================================
/home/temi/ArmNNDelegate/armnn/delegate/src/test/UnpackTest.cpp:169:
TEST SUITE: Unpack_GpuAccTests
TEST CASE:  Unpack_Uint8_Axis2_Num6_GpuAcc_Test

/home/temi/ArmNNDelegate/armnn/delegate/src/test/UnpackTest.cpp:169: ERROR: test case THREW exception: TfLiteArmnnDelegate: No known backend specified.

===============================================================================
[doctest] test cases:    185 |      3 passed |    182 failed |    396 skipped
[doctest] assertions:   1240 |   1240 passed |      0 failed |
[doctest] Status: FAILURE!

If I run the test with sudo:

temi@temi:~/ArmNNDelegate/armnn/delegate/build$ sudo ./DelegateUnitTests --test-suite=*GpuAcc*
[doctest] doctest version is "2.4.0"
[doctest] run with "--help" for options
INFO: TfLiteArmnnDelegate: Created TfLite ArmNN delegate.
arm_release_ver of this libmali is 'r18p0-01rel0', rk_so_ver is '4'.===============================================================================
/home/temi/ArmNNDelegate/armnn/delegate/src/test/ActivationTest.cpp:261:
TEST SUITE: Activation_GpuAccTests
TEST CASE:  Activation_ReLu_GpuAcc_Test

/home/temi/ArmNNDelegate/armnn/delegate/src/test/ActivationTest.cpp:261: FATAL ERROR: test case CRASHED: SIGABRT - Abort (abnormal termination) signal

===============================================================================
[doctest] test cases:      1 |      0 passed |      1 failed |    580 skipped
[doctest] assertions:      9 |      9 passed |      0 failed |
[doctest] Status: FAILURE!
Aborted

Does this mean that I need to fix the linux kernel? I'm not very familiar with these linux things. Any suggestion?Thanks a lot!

janeil01 commented 2 years ago

Hey @sunforever1990

it looks to me like you haven't built Acl and ArmNN with Gpu support. Have you added the Gpu flags as mentioned in this comment in the guide?

The machine used for this guide only has a Neon CPU which is why I only have "neon=1" but if your machine has an arm Gpu you can enable that by adding opencl=1 embed_kernels=1 to the command below: scons arch=arm64-v8a neon=1 extra_cxx_flags="-fPIC" benchmark_tests=0 validation_tests=0

and the same when building ArmNN: if you've got an arm Gpu add -DARMCOMPUTECL=1 to the command below cmake .. -DARMCOMPUTE_ROOT=$BASEDIR/ComputeLibrary -DARMCOMPUTENEON=1 -DBUILD_UNIT_TESTS=0

Let me know how you are getting on! Jan

steven9046 commented 2 years ago

Hey @sunforever1990

it looks to me like you haven't built Acl and ArmNN with Gpu support. Have you added the Gpu flags as mentioned in this comment in the guide?

The machine used for this guide only has a Neon CPU which is why I only have "neon=1" but if your machine has an arm Gpu you can enable that by adding opencl=1 embed_kernels=1 to the command below: scons arch=arm64-v8a neon=1 extra_cxx_flags="-fPIC" benchmark_tests=0 validation_tests=0

and the same when building ArmNN: if you've got an arm Gpu add -DARMCOMPUTECL=1 to the command below cmake .. -DARMCOMPUTE_ROOT=$BASEDIR/ComputeLibrary -DARMCOMPUTENEON=1 -DBUILD_UNIT_TESTS=0

Let me know how you are getting on! Jan

Sorry, did't repay at weekends. I'm sure I have add these flags

if you've got an arm Gpu add -DARMCOMPUTECL=1 to the command below
cmake .. -DARMCOMPUTE_ROOT=$BASEDIR/ComputeLibrary -DARMCOMPUTENEON=1 -DBUILD_UNIT_TESTS=0
The machine used for this guide only has a Neon CPU which is why I only have "neon=1" but if your machine has an arm Gpu you can enable that by adding opencl=1 embed_kernels=1 to the command below:
scons arch=arm64-v8a neon=1 extra_cxx_flags="-fPIC" benchmark_tests=0 validation_tests=0

I follwed BuildGuideNative.md step by step.

janeil01 commented 2 years ago

These are exactly the kind of errors that you see when you haven't built armnn for the backend that you try to run tests on. Have you tried only running CpuAcc tests? ./DelegateUnitTests --test-suite=*CpuAcc*

If that works I suggest to rebuild armnn and ensure that the correct flags are set.

steven9046 commented 2 years ago

These are exactly the kind of errors that you see when you haven't built armnn for the backend that you try to run tests on. Have you tried only running CpuAcc tests? ./DelegateUnitTests --test-suite=*CpuAcc*

If that works I suggest to rebuild armnn and ensure that the correct flags are set.

CpuAcc is OK. I rebuilt ACL and Armnn. This error it still there. I'm sure the flags have been set correctlly this time. I'm afraid the problem may due to my hardware? I will test it on another hardware later. Thanks very much!

janeil01 commented 2 years ago

Can I ask what hardware you are running on?

steven9046 commented 2 years ago

Can I ask what hardware you are running on?

RK3399 with ubuntu20.04. The linux image is made by another guy. I'm not sure whether the image is OK or not...

janeil01 commented 2 years ago

I've been getting some feedback from the team. It's very likely that your image is causing the issue. I suggest you try an official image for the board if you can.

There are other issues in our github repo that also mention this board. Maybe one of them says something that can help you out.

gfa99 commented 2 years ago

@sunforever1990: you can try like follow:

  1. place the libmails.so into /usr/lib

    git clone -b 29mirror https://github.com/rockchip-linux/libmali
    sudo cp -R libmail/include/CL/ /usr/include/
    sudo cp libmail/lib/aarch64-linux-gnu/libmali-midgard-t86x-r14p0.so /usr/lib/
    sudo ln -s /usr/lib/libmali-midgard-t86x-r14p0.so /usr/lib/libOpenCL.so
  2. config udev rules:

    sudo cat > /etc/udev/rules.d/50-mail400.rules << EOF
    KERNEL=="mali", MODE="0666", GROUP="video"
    EOF
    sudo cat > /etc/udev/rules.d/50-mail.rules << EOF
    KERNEL=="mali0", MODE="0666", GROUP="video"
    EOF
  3. reboot your system

  4. you could do sample test with opencv: opencv-demo.tar.gz refer to https://www.cnblogs.com/releases/p/10308741.html

steven9046 commented 2 years ago

KERNEL=="mali", MODE="0666", GROUP="video" EOF sudo cat > /etc/udev/rules.d/50-mail.rules << EOF KERNEL=="mali0", MODE="0666", GROUP="video" EOF

This actually solved my problem. There was no rules file under /etc/udev/rules.d/ in my system. Thank you a lot!