GPU not Detected

I am having a problem that my GPU is not detected. The following is my GPU information:

  • NVIDIA RTX A1000 Laptop GPU
  • Driver Version: 535.171.04
  • CUDA Version: 12.2

When I try to compile a model into HEF file, GPU is not detected with this error message:
“Reducing optimization level to 0 because there’s no available GPU”

But then, after I downgrade CUDA to 11.8, GPU is detected. But now I faced this error:
“Could not load library libcublasLt.so.12. Error: libcublasLt.so.12: cannot open shared object file: No such file or direcctory”
Seems like it required packages from CUDA 12.

So is CUDA 12 supported by Hailo? Or what is wrong here?

Hi ShinHin, currently the supported CUDA version is 11.8. We strongly suggest to use the Dockerized environment, all the packages would be already aligned :slight_smile:

I further installed CuDnn 8.9 and the error goes away, GPU successfully detected. But thanks for your advice.

Hi,
I have a similar problem. I’m having problems with these warning messages:

[warning] Reducing optimization level to 0 (the accuracy won’t be optimized and compression won’t be used) because there’s no available GPU
[warning] Running model optimization with zero level of optimization is not recommended for production use and might lead to suboptimal accuracy results

I have installed nvidia drivers and cuda 11.8, although when I run nvidia-smi, it says it couldn’t detect any cuda, but it is installed. There is a problem with installing older versions of cuda and it always defaults to cuda 12.2, which is not supported. Here is a thread: discussion.

I installed the latest Nvidia drivers and cuda 11.8 with their runfile and only cuda selected link to cuda 11.8.

BTW, I’m running the code in a docker container created with the image in the hailo_model_zoo github repo.

Do you have any ideas on how to fix it?

I’m running the Hailo SW suit as docker.

I’m facing this warning.

image

GPU is Nvidia GTX Geforce 1650

@ShingHin.Hung how you installed cuda and cudnn by using conda in virtual env or globaly

@vinaygouda.ttssl I installed CUDA and CuDnn globally. I simply followed the official installation guide from NVIDIA.