Let's Install Tensorflow 2.4 (with CUDA toolkit) on Ubuntu 20.04 (AGAIN)

I'm back again

in this one we install the latest version of tensorflow (again).
That is version 2.4.1, and we're installing with GPU support for CUDA toolkit 11 and CUDNN 8.

Thanks!

[PlatimaTinkers]

Hey mate cheers for the follow! I know the grind can hurt, but would love to see more content from you if you ever get the time again

Take it easy

[not_a_number]

Your taste on nv/tf related memes is top notch btw

[joshhoneycutt]

Glad to see you're back to making videos.

[gimhara]

Thank you very much. This video was very helpful to me

[amadios]

thank you so much it helped me a lot <3

[karvon.v]

When i run sudo apt install cuda, prompts the following:

The following packages have unmet dependencies:
cuda : Depends: cuda-11-0 (>= 11.0.2) but it is not going to be installed
E: Unable to correct problems, you have held broken packages.

Has anyone else ran into this error?

[RlxRlx]

hi thanks for your video, assuming this video is for ubuntu 20.04 versions. I have been trying to follow this procedure with the new version of ubuntu 21.04! where I successfully install Cuda and Cudnn, upon completion and creating the virtual environment it works! I can see that Cuda is working when importing Tensorflow. My problem is when I reboot, it stops working. could it be the patches? is there a different way than through VIM? I feel that there is the problem

Thank you

[jendafojtik]

Awesome video, you are literally a life-saver only one thing, when I do "sudo cp cuda/include/cudnn.h /usr/local/cuda/include" and the following command, all I get is "cp: cannot create regular file '/usr/local/cuda/include': No such file or directory", any idea why?

[surajojha]

got error: `NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.` could somebody please tell me what i did wrong? i tried disabling secure boot in UEFI mode and then rebooting. i also tried `prime-select nvidia` . But no luck

[Seetosy]

Hi I am trying to install the above but installation went wrong. My OS is Linux-x86_64 and the Nvidia Driver is 495.29.05.
My GPU is Nvidia Grforce RTX 3060 Ti. Can you advise which version of cuda and cudnn should I select.. Thank you.

[achintya-]

Came for just the GitHub sheet
Stayed for the meme

[nilushachamalJ]

Hi! Thanks a lot for your video. I tried to follow your tutorial for my Ubuntu 20.04 OS that I am using within the virtual box. However, upon rebooting, I get the following error: "NVIDIA-SMI has failed because it couldn't communication with NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running". Could you kindly tell me how to resolve this? Thanks

[not_a_number]

Hi Mark, how is this combination working for you? I installed Nvidia driver version 460 cuda 11.2 and cudnn 8.1 on Ubuntu 20.04, but it kept freezing on me randomly, driving me nuts. Did you run into similar situation?

[AbdullahJirjees]

Hi I am getting the error
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

[Seetosy]

Hi I have solved my problem. My nvidia-smi shows that I have cuda-11.4. I used cudnn-11.4 and followed the rest of the instructions. Now it works. Thanks

[not_a_number]

I’m curious why you opted out of anaconda

[shavkat]

I tried running tf with my 960. it had bugs and crashed. I think I follower all the tutorials right, think it was a genuine bug.

[mr.pondering]

NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.
Help plz

[dannyisrael]

Hmm, no one got this?
```
The following packages have unmet dependencies:
nvidia-driver-440 : Depends: nvidia-driver-450 but it is not going to be installed
E: Unable to correct problems, you have held broken packages.
```
Resolved but can’t remember how, installed correct versions of everything as per a table of compatible versions which appeared from a google search.