Running linux container from windows host with GPU support

[miccio-dk]

I would like to run a Linux-based container image (currently tried openvino_cgvh_data_dev_2021.4.2) from a Windows host and would like to run inference on GPU.

I tried passing --device /dev/dri which resulted in:

docker: Error response from daemon: error gathering device information while adding custom device "/dev/dri": no such file or directory.

I also tried passing --gpus=all which allowed me to run the container, but as soon as I tried one of the demos with -d GPU I got the following:

###################################################
Run Inference Engine classification sample
Run ./classification_sample_async -d GPU -i /opt/intel/openvino/deployment_tools/demo/car.png -m /root/openvino_models/ir/public/squeezenet1.1/FP16/squeezenet1.1.xml
...
[ INFO ] Device info:
[ ERROR ] Failed to create plugin /opt/intel/openvino/deployment_tools/inference_engine/lib/intel64/libclDNNPlugin.so for device GPU
Please, check your environment
[CLDNN ERROR]. clGetPlatformIDs error -1001
Error on or near line 217; exiting with status 1

Is it possible to map the GPU from windows (host) to linux (container)?
Thanks in advance.

[ryanloney]

@miccio-dk Please see the instructions here about running iGPU using Windows Subsystem for Linux 2 (WSL2): https://github.com/openvinotoolkit/docker_ci/blob/master/get-started.md#run-a-container

[miccio-dk]

Thanks for the prompt answer, unfortunately adding the arguments mentioned above doesn't seem to help.
Here are my steps:

# start a new container (ubuntu20_dev:2021.4.2 is my image based on the openvino_cgvh_data_dev_2021.4.2 dockerfile)
docker run -itu root:root --rm --device /dev/dxg --volume /usr/lib/wsl:/usr/lib/wsl ubuntu20_dev:2021.4.2
# once inside the container, install sudo
apt-get update && apt-get install -y sudo
# navigate inside demo directory
cd /opt/intel/openvino/deployment_tools/demo
# run demo
./demo_security_barrier_camera.sh -d GPU

This is the output message

###################################################

Run Inference Engine security_barrier_camera demo

Run ./security_barrier_camera_demo -d GPU -d_va GPU -d_lpr GPU -i /opt/intel/openvino_2021.4.752/deployment_tools/demo/car_1.bmp -m /root/openvino_models/ir/intel/vehicle-license-plate-detection-barrier-0106/FP16/vehicle-license-plate-detection-barrier-0106.xml -m_lpr /root/openvino_models/ir/intel/license-plate-recognition-barrier-0001/FP16/license-plate-recognition-barrier-0001.xml -m_va /root/openvino_models/ir/intel/vehicle-attributes-recognition-barrier-0039/FP16/vehicle-attributes-recognition-barrier-0039.xml

[ INFO ] InferenceEngine:       IE version ......... 2021.4
        Build ........... 2
[ INFO ] Files were added: 1
[ INFO ]     /opt/intel/openvino_2021.4.752/deployment_tools/demo/car_1.bmp
[ INFO ] Loading device GPU
[ ERROR ] Failed to create plugin /opt/intel/openvino_2021.4.752/deployment_tools/inference_engine/lib/intel64/libclDNNPlugin.so for device GPU
Please, check your environment
[CLDNN ERROR]. clGetPlatformIDs error -1001

Error on or near line 198; exiting with status 1

This time, when trying to run nvidia-smi inside the container, i get a "command not found error". If I run /usr/lib/wsl/lib/nvidia-smi or if I add the directory to $PATH, i get:

NVIDIA-SMI couldn't find libnvidia-ml.so library in your system. Please make sure that the NVIDIA Display Driver is properly installed and present in your system.
Please also try adding directory that contains libnvidia-ml.so to your system PATH.

Note that /usr/lib/wsl/lib/ contains libnvidia-ml.so.1.

[ryanloney]

@miccio-dk the GPU support is for Intel GPUs, not Nvidia. OpenVINO support for Nvidia GPU is limited. An open source module is available here: https://github.com/openvinotoolkit/openvino_contrib/tree/releases/2021/4/modules/cuda_plugin