FS#65961 - [tensorflow-opt-cuda] This version requires cuda-10.1, but the latest version is cuda-10.2
Attached to Project:
Community Packages
Opened by Maksym Ganenko (brut) - Tuesday, 24 March 2020, 14:11 GMT
Last edited by Konstantin Gizdov (kgizdov) - Tuesday, 24 March 2020, 18:53 GMT
Opened by Maksym Ganenko (brut) - Tuesday, 24 March 2020, 14:11 GMT
Last edited by Konstantin Gizdov (kgizdov) - Tuesday, 24 March 2020, 18:53 GMT
|
Details
This package is expected to provide GPU for Tensorflow
library, but it doesn't provide GPU because it requires
cuda-10.1.
The current version of cuda is 10.2 and so the package doesn't work without downgrading cuda to 10.1: `sudo pacman -U /var/cache/pacman/pkg/cuda-10.1.168-4-x86_64.pkg.tar.xz` |
This task depends upon
Closed by Konstantin Gizdov (kgizdov)
Tuesday, 24 March 2020, 18:53 GMT
Reason for closing: Not a bug
Tuesday, 24 March 2020, 18:53 GMT
Reason for closing: Not a bug
```
$ python
Python 3.8.2 (default, Feb 26 2020, 22:21:03)
[GCC 9.2.1 20200130] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow
2020-03-24 17:48:40.734076: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.2
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.compat.v2.summary API due to missing TensorBoard installation.
WARNING:root:Limited tf.summary API due to missing TensorBoard installation.
>>>
```
which is as expected
I uninstalled [tensorflow-gpu] from pip and updated all related packages: `sudo pacman -S cuda cudnn tensorflow-opt-cuda python-tensorflow-opt-cuda`
Currently I observe another error, but obviously it's not related to invalid version of cuda.
UPDATE:
When I removed `sudo rm -rf /usr/lib/python3.8/site-packages/tensorflow*` and reinstalled all related packages `sudo pacman -S cuda cudnn tensorflow-opt-cuda python-tensorflow-opt-cuda` the problem had been fixed.
How can I close the task? "Request closure"?