You are not logged in.
System specs:
~ nvidia-smi
Tue Oct 20 00:04:24 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.28 Driver Version: 455.28 CUDA Version: 11.1 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 GeForce RTX 2080 Off | 00000000:01:00.0 On | N/A |
| 0% 43C P8 19W / 245W | 375MiB / 7979MiB | 3% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
Ive installed the following.
`https://www.archlinux.org/packages/community/x86_64/python-tensorflow-cuda/` which (i assume) is the supported cuda GPU package.
if I `pip freeze` python seems to show the correct package.
tensorflow-gpu @ file:///build/tensorflow/src/tmpcuda/tensorflow_gpu-2.3.1-cp38-cp38-linux_x86_64.whl
However when I run the following script I only see my CPU
from tensorflow.python.client import device_lib
def get_devices():
return [x.name for x in device_lib.list_local_devices()]
print (get_devices())
['/device:CPU:0', '/device:XLA_CPU:0']
Are there any suggestions for how to solve this issue?
Thank you in advance.
Last edited by AmazingMarks (2020-10-20 17:55:47)
Why doesn't this site support markdown?
Offline
CUDA is currently broken with the 5.9 kernel, due to a change in the kernel that needs an adjustment on nvidia's side that's slated to be released in november.
Either use the LTS kernel or downgrade to a 5.8 variant (... or patch the kernel to remove the code that prevents the relevant nvidia module from loading), see https://bbs.archlinux.org/viewtopic.php?id=259936 and https://bugs.archlinux.org/task/68312 for more information
Last edited by V1del (2020-10-20 08:05:10)
Offline
Hi, I am having the same problem.
Offline