How to install TensorFlow with GPU support in 2024? | ONLY method that works | Techy Man's Solutions

[@TechyMansSolutions]

If you have any problem do comment it and I will solve it. Also links may not be clickable so you can google search these with version numbers and you will find the page. Thank You

[@kaiohenrique3122]

Just want to say thank you. I'm on the final steps of my reserach for my final work of University. Your video is the only one that worked for me, so thanks a lot!

[@TechyMansSolutions]

@@kaiohenrique3122 thanks for watching 😊

[@linkduda]

Hello tahnks for the video, but I got the same error of other versions: A module that was compiled using NumPy 1.x cannot be run in NumPy 2.0.2 as it may crash. To support both 1.x and 2.x versions of NumPy, modules must be compiled with NumPy 2.0. Some module may need to rebuild instead e.g. with 'pybind11>=2.12'. 😅

[@TechyMansSolutions]

@@linkduda you should then try to use WSL or create a different conda environment for it

[@Latsyuha]

I can't get TensorRT to work with FaceFusion. This is their instruction for installing the accelerator: conda install cudatoolkit=11.8 cudnn=8.9.2.26 conda-forge::gputil=1.4.0 python=3.10 If I understand you correctly, this combination cannot work on Windows?

[@TechyMansSolutions]

Did you install the whl file in tensorrt zip?

[@Latsyuha]

@@TechyMansSolutions I followed these instructions kzbin.info/www/bejne/gaCmnqZsq7J7bKM Maybe I did something wrong.

[@TechyMansSolutions]

@@Latsyuha You can message me on reddit - Certain-Phrase-4721 we will clear it all there

[@gerardoasaellopezalfaro8639]

I man, do you have some solution using google colab or workstation with ubuntu?

[@TechyMansSolutions]

@@gerardoasaellopezalfaro8639 in Google Colab GPU support is already configured. For Ubuntu, Open build from source guide where in the end there will be compatible version mentioned for linux. Right now the latest version is 2.17 which requires cudnn 8.9 and cuda 12.3 . Install these in a conda environment (it's better) by using: conda install nvidia/label/cuda-12.3.2::cuda-toolkit conda install -c conda-forge cudnn=8.9 Then you can go ahead and install version 2.17 by: pip install tensorflow[and-cuda]==2.17 This will install TensorFlow with GPU. TensorRT is usually installed but if you are new, don't mess with it as it takes a lot of trial and error to install the right whl for it.

[@JOGUANDREWS]

Thanks im gay (happy)

[@_ProdiptoChakraborty_IT_A]

Hey man do you have a discord link where i can show my problem it will be really helpful i got a rtx 2050 gpu and tensorflow is not able to recognise it am stuck in it for like 2 days now

[@TechyMansSolutions]

@@_ProdiptoChakraborty_IT_A discord.com/invite/JTDst4aJ

[@Zico-y9e]

I have Nvidia T1200 it is work for me??

[@TechyMansSolutions]

yes ofcourse

[@sarthakjain1021]

Will it work for window 11 ?

[@TechyMansSolutions]

@@sarthakjain1021 yes, I also did on windows 11

[@sarthakjain1021]

i got the solution thanks dude i can now finally run tensorflow on gpu its the only video that worked for me

[@TechyMansSolutions]

@@sarthakjain1021 I am glad I could help you 😊

[@farazfitness]

so i have driver version 555.99 and i downloaded cuda 11.2 and cnn 9.2 and python 3.10 tensor flow 2.10 and nvidia smi shows cuda 12.5 too confused and straight up in a mess

[@TechyMansSolutions]

That nvidia-smi version doesn't matter, we need to have 11.2 CUDA toolkit and we have it. If you have errors then write here

[@farazfitness]

@@TechyMansSolutions just want to ask the tensor rt shows winows 10 but i guess you are using windows 11 right?? hopefully that doesnt make a difference

[@TechyMansSolutions]

​​@@farazfitness yes, cause the cuda 11.2 only supports till windows 10 so it shows windows 10 Windows 11 will cause no problem You can ignore that

[@farazfitness]

@@TechyMansSolutions thanks for the help Will keep you posted

[@TechyMansSolutions]

@@farazfitness welcome