OpenVino classification_sample stuck

[Lin__Aaron]

I'm using OpenVino 2019 R1.1 on CentOS 7.4.

I followed this guide https://docs.openvinotoolkit.org/latest/_docs_install_guides_PAC_Configure_2019R1.html 

When I try to run classification_sample, I get the following, I tried the HETERO plugin with just CPU and it works fine, but when I add FPGA, it just seems to stop running when inferencing starts (no error messages were recived).

./classification_sample -m /home/merg/1/squeezenet1.1.xml -i /opt/intel/openvino_2019.1.133/deployment_tools/demo/car.png -d HETERO:FPGA,CPU
[ INFO ] InferenceEngine:
        API version ............ 1.6
        Build .................. custom_releases/2019/R1_ebce9728578ef3131f2f282b3fbc3232109c598e
[ INFO ] Parsing input parameters
[ INFO ] Files were added: 1
[ INFO ]     /opt/intel/openvino_2019.1.133/deployment_tools/demo/car.png
[ INFO ] Loading plugin

        API version ............ 1.6
        Build .................. heteroPlugin
        Description ....... heteroPlugin
[ INFO ] Loading network files:
        /home/merg/1/squeezenet1.1.xml
        /home/merg/1/squeezenet1.1.bin
[ INFO ] Preparing input blobs
[ WARNING ] Image is resized from (787, 259) to (227, 227)
[ INFO ] Batch size is 1
[ INFO ] Preparing output blobs
[ INFO ] Loading model to the plugin
[ INFO ] Starting inference (1 iterations)

Here's the fme info of my FPGA:

//****** FME ******//
Object Id                     : 0xF500000
PCIe s:b:d:f                  : 0000:01:00:0
Device Id                     : 0x09C4
Socket Id                     : 0x00
Ports Num                     : 01
Bitstream Id                  : 0x123000200000185
Bitstream Version         : 0x556900030201
Pr Interface Id               : 69528db6-eb31-577a-8c36-68f9faa081f6
 

[Shubha_R_Intel]

Dear Lin, Aaron,

I assume you're following this documentation ?

In the aforementioned documentation this command-line example is given :

./classification_sample_async -i <path_to_image>/cat.bmp -m <path_to_model>/alexnet_fp32.xml -nt 5 -d HETERO:FPGA,CPU -nireq 2 -ni 200

Note that it is using classification_sample_async instead of just classification_sample. Can you kindly try the async version and let me know if it works ?

Thanks kindly,

Shubha

[Lin__Aaron]

Hi, Shubha, Thanks for the reply,

I tried both async and sync samples and they both hang at the same place, I look into the source code of these examples and found that the problem was with this line infer_request.infer() in the do inference section (this is the classification_sample, and its the same for the async version).

 for (size_t iter = 0; iter < FLAGS_ni; ++iter) {
            auto t0 = Time::now();
            infer_request.Infer();
            slog::info << "finished infer" << slog::endl;
            auto t1 = Time::now();
            fsec fs = t1 - t0;
            ms d = std::chrono::duration_cast<ms>(fs);
            total += d.count();
        }

And when I tried the Hello World application in the guide mentioned before, the program also hangs. 

./bin/host
Querying platform for info:
==========================
CL_PLATFORM_NAME                         = Intel(R) FPGA SDK for OpenCL(TM)
CL_PLATFORM_VENDOR                       = Intel(R) Corporation
CL_PLATFORM_VERSION                      = OpenCL 1.0 Intel(R) FPGA SDK for Open                                                                      CL(TM), Version 18.1

Querying device for info:
========================
CL_DEVICE_NAME                           = pac_a10 : PAC Arria 10 Platform (pac_                                                                      a10_f500000)
CL_DEVICE_VENDOR                         = Intel Corp
CL_DEVICE_VENDOR_ID                      = 4466
CL_DEVICE_VERSION                        = OpenCL 1.0 Intel(R) FPGA SDK for Open                                                                      CL(TM), Version 18.1
CL_DRIVER_VERSION                        = 18.1
CL_DEVICE_ADDRESS_BITS                   = 64
CL_DEVICE_AVAILABLE                      = true
CL_DEVICE_ENDIAN_LITTLE                  = true
CL_DEVICE_GLOBAL_MEM_CACHE_SIZE          = 32768
CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE      = 0
CL_DEVICE_GLOBAL_MEM_SIZE                = 8589934592
CL_DEVICE_IMAGE_SUPPORT                  = false
CL_DEVICE_LOCAL_MEM_SIZE                 = 16384
CL_DEVICE_MAX_CLOCK_FREQUENCY            = 1000
CL_DEVICE_MAX_COMPUTE_UNITS              = 1
CL_DEVICE_MAX_CONSTANT_ARGS              = 8
CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE       = 2147483648
CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS       = 3
CL_DEVICE_MEM_BASE_ADDR_ALIGN            = 8192
CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE       = 1024
CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR    = 4
CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT   = 2
CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT     = 1
CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG    = 1
CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT   = 1
CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE  = 0
Command queue out of order?              = false
Command queue profiling enabled?         = true
Using AOCX: hello_world.aocx

Kernel initialization is complete.
Launching the kernel... (It just stops here)
 

Did I miss any steps?

Thanks for your help!

[VDemc2]

I think we have the same problem as you. But for us it hangs randomly https://software.intel.com/en-us/forums/computer-vision/topic/814697