Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Jianjun_G_Intel
Employee
116 Views

OpenVINO 2019R1: super_resolution_demo failed with GPU

the sample runs ok with CPU and MYRIAD, but failed with GPU,  

With NCS2 : OK

    root@drd-HP-ENVY-x360-m6-Convertible:/opt/intel/workshop/AI# ./super_resolution_demo -i ball.jpg -m single-image-super-resolution-1033-fp16.xml -d MYRIAD
    [ INFO ] InferenceEngine: 
        API version ............ 1.6
        Build .................. custom_releases/2019/R1_c9b66a26e4d65bb986bb740e73f58c6e9e84c7c2
    [ INFO ] Parsing input parameters
    [ INFO ] Files were added: 1
    [ INFO ]     ball.jpg
    [ INFO ] Loading plugin

        API version ............ 1.6
        Build .................. 22443
        Description ....... myriadPlugin
    [ INFO ] Loading network files
    [ INFO ] Preparing input blobs
    [ INFO ] Batch size is 1
    [ INFO ] Preparing output blobs
    [ INFO ] Loading model to the plugin
    [ INFO ] Create infer request
    [ INFO ] Start inference (1 iterations)

    Average running time of one iteration: 2100.95 ms
    [ INFO ] Output size [N,C,H,W]: 1, 3, 1080, 1920
    [ INFO ] Execution successful

 

With GPU: failed in FP32/FP16
    root@drd-HP-ENVY-x360-m6-Convertible:/opt/intel/workshop/AI# ./super_resolution_demo -i ball.jpg -m single-image-super-resolution-1033-fp16.xml -d GPU   
    [ INFO ] InferenceEngine: 
        API version ............ 1.6
        Build .................. custom_releases/2019/R1_c9b66a26e4d65bb986bb740e73f58c6e9e84c7c2
    [ INFO ] Parsing input parameters
    [ INFO ] Files were added: 1
    [ INFO ]     ball.jpg
    [ INFO ] Loading plugin

        API version ............ 1.6
        Build .................. 22443
        Description ....... clDNNPlugin
    [ INFO ] Loading network files
    [ INFO ] Preparing input blobs
    [ INFO ] Batch size is 1
    [ INFO ] Preparing output blobs
    [ INFO ] Loading model to the plugin
    [ ERROR ] network allocation failed: /teamcity/work/scoring_engine_build/releases_2019_R1/thirdparty/clDNN/src/reshape.cpp at line: 87
    Error has occured for: Reshape:86
    Output layout count(=16588800) is not equal to: input layout count(=72)
    Output layout of reshape

 

Another SR model with GPU: Failed in FP32/FP16
    root@drd-HP-ENVY-x360-m6-Convertible:/opt/intel/workshop/AI# ./super_resolution_demo -i tree1.jpg -m single-image-super-resolution-1032-fp16.xml -d GPU   
    [ INFO ] InferenceEngine: 
        API version ............ 1.6
        Build .................. custom_releases/2019/R1_c9b66a26e4d65bb986bb740e73f58c6e9e84c7c2
    [ INFO ] Parsing input parameters
    [ INFO ] Files were added: 1
    [ INFO ]     tree1.jpg
    [ INFO ] Loading plugin

        API version ............ 1.6
        Build .................. 22443
        Description ....... clDNNPlugin
    [ INFO ] Loading network files
    [ INFO ] Preparing input blobs
    [ INFO ] Batch size is 1
    [ INFO ] Preparing output blobs
    [ INFO ] Loading model to the plugin
    [ ERROR ] network allocation failed: /teamcity/work/scoring_engine_build/releases_2019_R1/thirdparty/clDNN/src/reshape.cpp at line: 87
    Error has occured for: Reshape:125
    Output layout count(=16588800) is not equal to: input layout count(=32)
    Output layout of reshap

 

I tried both FP32 and FP16 model and with Windows release and Linux releasee, and I also tried the OpenVINO 2018R5 version, all failed with no luck .

 

Any solution ?

0 Kudos
3 Replies
Shubha_R_Intel
Employee
116 Views

Dear Jianjun Gu,

Since you are an intel employee I have sent you an email directly.

Thanks,

Shubha

116 Views

Hi Shubha,

I also have the same issue .

I have tried using openvino 2019 R1.1 . on ubuntu 18.04.

can you please help me

Thanks

Aamna.

Shubha_R_Intel
Employee
116 Views

Dear Jianjun Gu and Syed, Aamna Iqra,

Sorry for the trouble you are having. I can assure you that it works fine on OpenVino R2.

Please see the following where I downloaded the model (FP16) from here: https://download.01.org/opencv/2019/open_model_zoo/R1/2019/single-image-super-resolution-1032/FP16/

If you are executing the demo similarly to me and if you are using an FP16 model on the GPU, then it must be a bug on 2019R1 which has been fixed in R2.

Unfortunately R2 is not officially released yet.


C:\Users\sdramani\Documents\Intel\OpenVINO\inference_engine_samples_build\intel64\Release>super_resolution_demo.exe -m c:\users\sdramani\Downloads\single-image-super-resolution-1032.xml -i c:\users\sdramani\Downloads\Jack-olantern-mushroom-480x270.jpg -d GPU
[ INFO ] InferenceEngine:
        API version ............ 1.6
        Build .................. 25506
        Description ....... API
[ INFO ] Parsing input parameters
[ INFO ] Files were added: 1
[ INFO ]     c:\users\sdramani\Downloads\Jack-olantern-mushroom-480x270.jpg
[ INFO ] Loading Inference Engine
[ INFO ] Device info:
        GPU
        clDNNPlugin version ......... 1.6
        Build ........... 25506

[ INFO ] Loading network files
[ INFO ] Preparing input blobs
[ INFO ] Batch size is 1
[ INFO ] Preparing output blobs
[ INFO ] Loading model to the device
[ INFO ] Create infer request
[ INFO ] Start inference (1 iterations)
To close the application, press 'CTRL+C' here

Average running time of one iteration: 233.742 ms
[ INFO ] Output size [N,C,H,W]: 1, 3, 1080, 1920
[ INFO ] Execution successful
 

Sorry for the inconvenience,

Shubha

Reply