RuntimeError: cublas runtime error : resource allocation failed at /opt/conda/conda-bld/pytorch_1524584710464/work/aten/src/THC/THCGeneral.cpp:411

[Xinzhu-Fang]

Some people get the error on the second GPU here: Problem of running on the second GPU on Ubuntu
I am using a single tesla k-80. I get this error when doing a linear operation on a 4 by 2048 matrix

Why I am getting this error?

[Xinzhu-Fang]

I already screwed up the environment here that this error “RuntimeError: cuda runtime error (77) : an illegal memory access was encountered” was returned on everything. I screwed it up by not following the guideline here:

Converting numpy array to tensor on GPU

import torch from skimage import io img = io.imread('input.png') device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") print(device) img =torch.tensor(img, device=device).float() print(img.device) Output cuda:0 cpu I am not able to convert a numpy array into a torch tensor on GPU. What is the right way to do this?