Smart-City-Sample icon indicating copy to clipboard operation
Smart-City-Sample copied to clipboard
verified publisher icon OpenVisualCloud

"GPU" name is not registered in the InferenceEngine

Open Gsarg18 opened this issue 4 years ago • 7 comments

I am not able to run inference model of smart city on the GPU device

Gsarg18 avatar Feb 05 '21 10:02 Gsarg18

More details please. What GPU device do you use? Currently the sample only supported the VCAC-A accelerator. To support other GPU devices, you have to extend the sample to use different base analytics images, for example, replace xeon-ubuntu1804-analytics-gst with xeone3-ubuntu1804-analytics-gst under analytics/common. There are other changes to the pipeline configuration to utilize VAAPI for decoding.

xwu2git avatar Feb 06 '21 17:02 xwu2git

Currently I am using Intel® UHD Graphics 630 (CFL GT2) graphics card

Gsarg18 avatar Feb 07 '21 06:02 Gsarg18

Is xeone3-ubuntu1804-analytics-gst image support Intel® UHD Graphics 630 (CFL GT2) graphics card? I am using Intel® Core™ i7-8700 CPU @ 3.20GHz × 12 processor.

Gsarg18 avatar Feb 09 '21 04:02 Gsarg18

You have to try it. It should work.

xwu2git avatar Feb 09 '21 23:02 xwu2git

I have a branch here. Not yet successful enabling inference. Something is missing.

xwu2git avatar Feb 10 '21 06:02 xwu2git

@xwu2git Do you have a system that reproduces the issue? We recently ran into an issue with Comet Late GPUs and the XeonE3 container. I will send the details to you offline.

nnshah1 avatar Feb 12 '21 05:02 nnshah1

See issue #662 I posted on OVC Dockerhub repo, but it is specific to Comet Lake. Issue shows how to check for OpenCL device.

whbruce avatar Feb 12 '21 05:02 whbruce