RuntimeError: CUDA error: out of memoryy

1

Hey.

I started working on a new workplace virtual system and did all the necessary installations. But currently, I am unable to create a simple cuda variable. I don’t know what requirement is I am lacking.

I am using Windows Server 2016 and just installed latest official pytorch

Attached is the screenshot and any guidance is highly appreciated.

image
image849×254 9.14 KB

image
image879×407 11.7 KB

thanks alot.

2

Could you check the memory usage on the server via nvidia-smi and make sure you have enough device memory left?
If you are using multiple GPUs, you can specify a specific via
x = torch.randn(1, device='cuda:id'), where you would have to replace id with the device ID.