Cuda 10.1 issues

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,
I am facing some problems with CUDA Toolkit 10.1. I downloaded the latest version of my GPU’s(GeForce GT 740M) driver 425.31, then Visual Studio 2019 and at last CUDA Toolkit 10.1 Update 2. I managed to run a CUDA 10.1 template (kernel ) in Visual Studio 2019 and i took a screenshot of the results. The main problem that keeps being unsolved is when i try to run a script in VS Code. I created a virtual environment with Python 3.8, downloaded pytorch with relevance to cuda 10.1 according to pytorch documentation and although cuda.is_available() is True when i pass a random tensor to cuda ( torch.tensor([1.0,2.0,3.0

Screenshot (12)
Screenshot (12)1920×1080 169 KB

Screenshot (11)
Screenshot (11)1920×1080 175 KB

Screenshot (10)
Screenshot (10)931×985 34.8 KB

Screenshot (9)
Screenshot (9)931×1037 35.3 KB

]).cuda() ), there is an error “no kernel image is available for execution on the device”. Can you guess what is the problem? Thank you in advance.

2

Your pytorch version is not compiled with support for your GPU. This is covered in many forum posts. GT 740M has a compute capability of 3.x and Pytorch hasn’t supported that compute capabiliity in a long time.

I won’t be able to help further here, this isn’t a pytorch forum. But there are probably hundreds or thousands of questions like this on various forums such as stackoverflow and discuss.pytorch.org