Via conda
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
|
# Check the version of the CUDA:
$ conda list cudatookit
# Name Version Build Channel
cudatoolkit 10.1.243 h8cb64d8_10 conda-forge
$ conda list cudnn:
# Name Version Build Channel
cudnn 7.6.5.32 hc0a50b0_1 conda-forge
# install/update CUDA and CUDNN through conda:
$ conda install -c anaconda cudatoolkit
$ conda install -c anaconda cudnn
# show the highest supported CUDA version?
$ nvidia-smi
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.47.03 Driver Version: 510.47.03 CUDA Version: 11.6 |
#
$ nvcc --version
|
Refer to 1; 3
Another way to check the version of cuDNN 4:
1
|
cat /usr/include/cudnn.h | grep CUDNN_MAJOR -A 2
|
Via PyTorch
1
2
3
|
import torch
print(torch.version.cuda)
>>> 10.2
|
Via TensorFlow
1
2
3
4
5
|
import tensorflow as tf
sys_details = tf.sysconfig.get_build_info()
cuda_version = sys_details["cuda_version"]
print(cuda_version)
|
Not quiet right.
Refer to How to check CUDA version in TensorFlow - gcptutorials
Compatible combinations for TF
Refer to Docs; 2;
Compatible CUDA for cards
Determine the driver version first and then determine the cuda version. Genearlly, cuda is backward compatible with the driver. Nvidia-page
Ref
-
searched by DDG: "What is the version of CUDA and cuDNN"
-
-
-