[Bug]: ALIKED on GPU

[mattiasmar]

OpenVINO Version

2023.3

Operating System

Other (Please specify in description)

Device used for inference

GPU

Framework

PyTorch

Model used

ALIKED

Issue description

Can't compile for GPU

Step-by-step reproduction

core = ov.Core()
device = 'GPU'
config = {"PERFORMANCE_HINT": "LATENCY",
            "PERFORMANCE_HINT_NUM_REQUESTS": "1",
            "INFERENCE_PRECISION_HINT":"f16"}
from nets.aliked import ALIKED    # Inference Model from: https://github.com/Shiaoming/ALIKED
m = ALIKED(model_name=model_name, scores_th = 0.1, top_k=1000, device="cpu")
compiled_model = core.compile_model(openvino_model_path, device, config)

Relevant log output

[Step 1/11] Parsing and validating input arguments
[ INFO ] Parsing input parameters
[Step 2/11] Loading OpenVINO Runtime
[ INFO ] OpenVINO:
[ INFO ] Build ................................. 2024.0.0-000--
[ INFO ] 
[ INFO ] Device info:
[ INFO ] GPU
[ INFO ] Build ................................. 2024.0.0-000--
[ INFO ] 
[ INFO ] 
[Step 3/11] Setting device configuration
[ WARNING ] Performance hint was not explicitly specified in command line. Device(GPU) performance hint will be set to THROUGHPUT.
[Step 4/11] Reading model files
[ INFO ] Loading model files
[ INFO ] Read model took 7.55 ms
[ INFO ] Original model I/O parameters:
[ INFO ] Network inputs:
[ INFO ]     image (node: image) : f32 / [...] / [1,3,240,320]
[ INFO ] Network outputs:
[ INFO ]     keypoints , 149 (node: aten::stack_3) : f32 / [...] / [1,1000,2]
[ INFO ]     descriptors , 152 (node: aten::stack_2) : f32 / [...] / [1,1000,64]
[ INFO ]     scores , 155 (node: aten::stack_1) : f32 / [...] / [1,1000]
[ INFO ]     score_dispersity , 158 (node: aten::stack) : f32 / [...] / [1,1000]
[ INFO ]     scores_map , score_map , 141 (node: aten::slice/Slice_1) : f32 / [...] / [1,1,240,320]
[Step 5/11] Resizing model to match image sizes and given batch
[ WARNING ] image: layout is not set explicitly, so it is defaulted to NCHW. It is STRONGLY recommended to set layout manually to avoid further issues.
[Step 6/11] Configuring input of the model
[ INFO ] Model batch size: 1
[ INFO ] Network inputs:
[ INFO ]     image (node: image) : u8 / [N,C,H,W] / [1,3,240,320]
[ INFO ] Network outputs:
[ INFO ]     keypoints , 149 (node: aten::stack_3) : f32 / [...] / [1,1000,2]
[ INFO ]     descriptors , 152 (node: aten::stack_2) : f32 / [...] / [1,1000,64]
[ INFO ]     scores , 155 (node: aten::stack_1) : f32 / [...] / [1,1000]
[ INFO ]     score_dispersity , 158 (node: aten::stack) : f32 / [...] / [1,1000]
[ INFO ]     scores_map , score_map , 141 (node: aten::slice/Slice_1) : f32 / [...] / [1,1,240,320]
[Step 7/11] Loading the model to the device
[ ERROR ] Exception from src/inference/src/core.cpp:99:                                            16:21 27-May-24
[ GENERAL_ERROR ] Check 'false' failed at src/plugins/intel_gpu/src/plugin/program_builder.cpp:179:
[GPU] ProgramBuilder build failed!
Exception from src/plugins/intel_gpu/src/graph/include/primitive_type_base.h:58:
[GPU] Can't choose implementation for convolution:__module.score_head.4/aten::_convolution/Convolution node (type=convolution)
[GPU] Original name: __module.score_head.4/aten::_convolution/Convolution
[GPU] Original type: Convolution
[GPU] Reason: Unsupported onednn dnnl::memory::desc find_format. ndims: 4, inner_nblks: 2, inner_blks: (blk 4, idx 0) (blk 2, idx 1) , strides_order : 0 1 2 3 , stride_value : 144 72 24 8 

Issue submission checklist

• I'm reporting an issue. It's not a question.
• I checked the problem with the documentation, FAQ, open issues, Stack Overflow, etc., and have not found a solution.
• There is reproducer code and related data files such as images, videos, models, etc.

[ilya-lavrenov]

What is the error message ? Exit code?

[mattiasmar]

I missed that part in the log above. Updated the original message now.
The output above is when calling:
/root/openvino_cpp_samples_build/intel64/Release/benchmark_app -m aliked-t16/serialized.xml -d GPU

[mattiasmar]

serialized.zip

    core = ov.Core()
    device = 'GPU'
    device = device if 'GPU' in core.available_devices else 'CPU'
    config = {"PERFORMANCE_HINT": "LATENCY",
                "PERFORMANCE_HINT_NUM_REQUESTS": "1",
                "INFERENCE_PRECISION_HINT":"f16"}
 compiled_model = core.compile_model(openvino_model_path, device, config)

This gives:

 Traceback (most recent call last):
  File "python/ALIKED/compile_openvino_model.py", line 195, inion node (type= <module>
    keypoints_Ex0_Model0, keypoints_Ex1_Model0, descriptors_Ex0_Model0, _ = test(openvino_model_pat               convolution)h[h0][top_k], example_inputs[h0]['img1'], example_inputs[h0]['img2'])
  File "python/ALIKED/compile_openvino_model.py", line 131, inks: (blk 4, idx test                                                                                                              0) (blk 2, i
    compiled_model = core.compile_model(openvino_model_path, device,config)
  File "/opt/intel/openvino_2024/python/openvino/runtime/ie_api.py", line 492, in compile_model
    super().compile_model(model, device_name, {} if config is None else config),
RuntimeError: Exception from src/inference/src/core.cpp:116:
[ GENERAL_ERROR ] Check 'false' failed at src/plugins/intel_gpu/src/plugin/program_builder.cpp:179:16:21 27-May-24
[GPU] ProgramBuilder build failed!
Exception from src/plugins/intel_gpu/src/graph/include/primitive_type_base.h:58:
[GPU] Can't choose implementation for convolution:__module.score_head.4/aten::_convolution/Convolution node (type=convolution)
[GPU] Original name: __module.score_head.4/aten::_convolution/Convolution
[GPU] Original type: Convolution
[GPU] Reason: Unsupported onednn dnnl::memory::desc find_format. ndims: 4, inner_nblks: 2, inner_blks: (blk 4, idx 0) (blk 2, idx 1) , strides_order : 0 1 2 3 , stride_value : 144 72 24 8 

by setting INFERENCE_PRECISION_HINT to f32 avoids this error.

[andrei-kochin]

@mattiasmar thank you for reaching the OpenVINO!

Could you please try the newer version of the OpenVINO?
I haven't found any issues with the latest openvino-nightly package:

benchmark_app -m .\AlikedGPU\serialized.xml -d GPU
[ INFO ] OpenVINO:
[ INFO ] Build ................................. 2024.3.0-15533-31fccc801fc
...
[Step 11/11] Dumping statistics report
[ INFO ] Execution Devices:['GPU.0']
[ INFO ] Count:            184 iterations

[avitial]

Closing this as it can't be reproduced, I hope previous responses were sufficient to help you proceed or resolve the issue. Feel free to reopen and ask additional questions related to this topic if issue persists.