[resolved] Cuda Runtime Error(30)

21

Win10, NVIDIA GTX 1050 Ti, Cuda toolkit 9.2, Python 3.6, Torch 1.0.1
RuntimeError (cuda runtime error (30) ) for the Cuda driver version 10.1 (419.67). I reinstalled an old NVIDIA driver 397.93 (i.e. Cuda driver verison 9.2). OK!. Tried twice.

22

Same problem here with Win10, NVIDIA GTX 1050 Ti, Cuda 10.0, Python 3.6, Torch 1.1.0

~\Anaconda3\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

Can someone help me fix this. I’m stuck with this issue for a while.

23

Are you using multiprocessing and try to initialize the CUDA context multiple times?
Could you post a small code snippet to reproduce this issue?

24

Hi @ptrblck
Thanks for the response.
I’m trying to follow along this notebook from Udacity’s github repository and I’m getting the error on the code cell 19.
https://github.com/udacity/deep-learning-v2-pytorch/blob/master/sentiment-rnn/Sentiment_RNN_Solution.ipynb

25

Ubuntu 16.04.6
Cuda 10.1
Driver Version: 418.87.00
Pytorch 1.1.0
Python 3.6

Updated driver to listed version and got following:

RuntimeError: cuda runtime error (30) : unknown error at /pytorch/aten/src/THC/THCGeneral.cpp:51

Resolved after reboot.

26

Yeah, same thing was happened to me. So i think after reboot your PC the problem will solved.

27

The same here. The error and the location of error occur in the file are all the same.