CUDA initialize error

1

When I try to use GPU, I got the error as follow:

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-6-52891be9f0e3> in <module>
----> 1 torch.ones((3,4)).to("cuda")

e:\miniconda3\lib\site-packages\torch\cuda\__init__.py in _lazy_init()
    161             "Cannot re-initialize CUDA in forked subprocess. " + msg)
    162     _check_driver()
--> 163     torch._C._cuda_init()
    164     _cudart = _load_cudart()
    165     _cudart.cudaGetErrorName.restype = ctypes.c_char_p

RuntimeError: CUDA error: unknown error

This is my environment info.

PyTorch version: 1.1.0
Is debug build: No
CUDA used to build PyTorch: 10.0

OS: Microsoft Windows 10 Pro
GCC version: Could not collect
CMake version: Could not collect

Python version: 3.7
Is CUDA available: Yes
CUDA runtime version: 10.0.130
GPU models and configuration: GPU 0: GeForce GTX 960M
Nvidia driver version: 430.39
cuDNN version: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin\cudnn64_7.dll

Versions of relevant libraries:
[pip3] numpy==1.16.3
[conda] blas                      1.0                         mkl
[conda] mkl                       2019.3                      203
[conda] mkl_fft                   1.0.12           py37h14836fe_0
[conda] mkl_random                1.0.2            py37h343c172_0
[conda] pytorch                   1.1.0           py3.7_cuda100_cudnn7_1    pytorch
[conda] torchvision               0.2.2                      py_3    pytorch

How to fix this error

2 Likes
2

Did you find any solution to this?
I was able to use the GPU with pytorch a couple of times before this error appeared magically one day.

Edit: I was able to solve the problem using the discussion here https://github.com/pytorch/pytorch/issues/17108

What worked for me was calling torch.cuda.current_device() before any cuda calls. For eg.

import torch
torch.cuda.current_device()
torch.cuda.is_available()

works but the opposite order of calls will not. Hope it helps!

3

it’s a version match problem i meet it too. now solved as follows
driver: NVIDIA-Linux-x86_64-440.118.02.run
cuda: cuda_10.2.89_440.33.01_linux.run
cuda patch: cuda_10.2.2_linux.run

libcudnn:
libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb
libcudnn7-dev_7.6.5.32-1+cuda10.2_amd64.deb
libcudnn7-doc_7.6.5.32-1+cuda10.2_amd64.deb

pytorch:
pip3 install torch==1.7.1 torchvision==0.8.2 torchaudio==0.7.2

#do not install torch 1.8.0