Why do I get RuntimeError: CUDA error: invalid argument in pytorch?

Question

Recently I've frequently been getting RuntimeError: CUDA error: invalid argument when calling functions like torch.cholesky e.g.:

import torch
a = torch.randn(3, 3, device="cuda:0")
a = torch.mm(a, a.t()) # make symmetric positive-definite
torch.cholesky(a)

This works fine if I use device="cpu" instead. This error isn't very descriptive, so I'm not sure what's wrong here.

Nathan · Accepted Answer · 2019-05-15 19:00:31Z

4

I discovered that this error was because the machine I'm running things on has CUDA 10 installed now, but I just installed pytorch as pip install torch. From their website, the proper way to install with pip and CUDA 10 is pip install https://download.pytorch.org/whl/cu100/torch-1.1.0-cp37-cp37m-linux_x86_64.whl.

answered May 15, 2019 at 19:00

Nathan

10.5k4 gold badges51 silver badges72 bronze badges

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

Collectives™ on Stack Overflow

Why do I get RuntimeError: CUDA error: invalid argument in pytorch?

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related