Very simple torch.tensor().to("cuda") gives CUDA error: device-side assert triggered

Question

All I do is,

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
torch.tensor([123,123]).to(device)

And I get:

RuntimeError: CUDA error: device-side assert triggered

I really cannot see why.

Edit: Weirdly enough, I see that this happens only I after I run some code before hand. Restarting the kernel solves it. But why, some code can cause a problem like this? I cannot share the code because it's not mine but I still welcome any guesses. Thanks!

Does this answer your question? CUDA runtime error (59) : device-side assert triggered — Harshit Kumar
– Harshit Kumar, Commented Dec 3, 2020 at 4:55
perhaps it can be of help to someone: stackoverflow.com/questions/78037880 — Dan M
– Dan M, Commented Feb 27, 2024 at 17:32

Vigna Tej · Accepted Answer · 2024-01-08 06:14:12Z

-1

device is a string, not a tensor. change your code to

device = "cuda" if torch.cuda.is_available() else "cpu"

now your code will work fine.

answered Jan 8, 2024 at 6:14

Vigna Tej

92 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Ivan Over a year ago

torch.device doesn't return a tensor so the code in the question is valid.

Collectives™ on Stack Overflow

Very simple torch.tensor().to("cuda") gives CUDA error: device-side assert triggered

1 Answer 1

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related