Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

cudaErrorSystemNotReady: system not yet initialized

See original GitHub issue

output of python -c 'import cupy; cupy.show_config()':

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/berceanu/miniconda3/envs/fbpic/lib/python3.7/site-packages/cupy/__init__.py", line 812, in show_config
    sys.stdout.write(str(_cupyx.get_runtime_info()))
  File "/home/berceanu/miniconda3/envs/fbpic/lib/python3.7/site-packages/cupyx/runtime.py", line 174, in get_runtime_info
    return _RuntimeInfo()
  File "/home/berceanu/miniconda3/envs/fbpic/lib/python3.7/site-packages/cupyx/runtime.py", line 107, in __init__
    cupy.cuda.cublas.CUBLASError)
  File "/home/berceanu/miniconda3/envs/fbpic/lib/python3.7/site-packages/cupyx/runtime.py", line 23, in _eval_or_error
    return func()
  File "/home/berceanu/miniconda3/envs/fbpic/lib/python3.7/site-packages/cupyx/runtime.py", line 106, in <lambda>
    cupy.cuda.device.get_cublas_handle()),
  File "cupy/cuda/device.pyx", line 44, in cupy.cuda.device.get_cublas_handle
  File "cupy/cuda/device.pyx", line 45, in cupy.cuda.device.get_cublas_handle
  File "cupy/cuda/device.pyx", line 27, in cupy.cuda.device._get_device
  File "cupy/cuda/runtime.pyx", line 229, in cupy.cuda.runtime.getDevice
  File "cupy/cuda/runtime.pyx", line 201, in cupy.cuda.runtime.check_status
cupy.cuda.runtime.CUDARuntimeError: cudaErrorSystemNotReady: system not yet initialized

Conditions
- OS/Platform: Ubuntu 16.04.6 LTS
- CUDA version: 10.0, installed via conda install cudatoolkit=10.0
- CuPy version: 7.0.0, installed via pip install cupy-cuda100 in my conda env

nvidia-smi works properly, and shows NVIDIA-SMI 430.64 Driver Version: 430.64 CUDA Version: 10.1

Issue Analytics

State:
Created 4 years ago
Comments:13 (5 by maintainers)

Top GitHub Comments

2reactions

berceanucommented, Apr 1, 2020

Did you try https://www.supermicro.com/support/faqs/faq.cfm?faq=31029

0reactions

JordanCheneycommented, Apr 1, 2020

@berceanu I did, that changed the return code from an 802 to an 83 on my machine. 83 is not documented in the CUDA API so I’m not sure what that means. I opened an issue on the developer forum that details everything I’ve tried.

Top Results From Across the Web

Error 802: system not yet initialized CUDA 11.3

Error 802 generally indicates your system requires the NVLink/NVSwitch fabric manager, and you have not installed it.

NVIDIA GPU driver fails to initialize - IBM

Solution. This issue can occur due to your GPU driver library not being successfully installed when you first created your GPU device plug-in....

Bug#1016595: cuda: CUDA Error : system not yet initialized

CUDA Error : system not yet initialized - Unable to enum CUDA GPUs: unknown error - No OpenCL platforms found - No avaiable...

cuda runtime error (802) : system not yet initialized ...

cudaGetDeviceProperties returned 802 -> system not yet initialized CUDA error at bandwidthTest.cu:256 code=802(cudaErrorSystemNotReady) ...

PyTorch-v1.7.1+cu110 - CUDA initialization error

Did you run some cuda functions before calling NumCudaDevices() that might have already set an error? Error 802: system not yet initialized (Triggered ......