Help with cuDNN installation: GPU With Tensorflow on Jupyter Notebook

TensorFlow General Discussion
1

Hi guys, i’m trying to use my GPU(Nvidia GeForce 540M) with Tensorflow.
I’ve installed CUDA 6.5 but it seems not be supported by tensorflow.
I’ve seen that is mandatory to have cuDNN
Can anyone help me?

1 Like
2

Hello @Luca_Zanetti You may need to create an NVIDIA Developer Account, if you don’t have one already, before downloading cuDNN. (Requirements: 使用 pip 安裝 TensorFlow) This tutorial showing how to install CUDA, cuDNN, etc by a TensorFlow user seems to be getting some good feedback, you may find this useful: https://youtu.be/hHWkvEcDBO0?t=144 Keep us posted and good luck!

1 Like
3

I wrote a blog few months ago. It may help you. Here it is.

1 Like
4

One more thing @Luca_Zanetti - for TensorFlow 2.5 with GPU support, you’ll need:

(The docs are being updated.)

Thanks for the writing a blog post about this, @ashutosh1919. Can you check the link you shared?

1 Like
5

https://tensorflow-prod.ospodiscourse.com/t/tensorflow-2-5-with-gpu-device-python-3-9-cuda-11-2-2-cudnn-8-1-1-conda-environment-windows-10/1385?u=8bitmp3

(TF 2.5 with conda on Windows 10 - by @Ricardo)

2 Likes
6

It seems the problem is with the GPU CUDA compute capability. My GPU have a compute capability of 2.1 but probably tensorflow need at least 3.0
can anyone confirm this?

1 Like
7

Hi @Luca_Zanetti - here are the requirements: Install TensorFlow with pip

Is your GPU listed here? CUDA GPUs - Compute Capability | NVIDIA Developer

1 Like
8

Yes, GeForce GT 540M

1 Like
9

Based on this information, it looks like it’s not supported @Luca_Zanetti

Have you considered Google Colab (free) or Colab Paid Services Pricing (more details: Google Colab)?

Great graphics card :smiley: 540m on Alienware-M11x | Dell :+1:

2 Likes
10

Yes, i’m using Google Colab now.
I was trying to set up also my pc with GPU in order to not depend only on network connection and run locally sometimes.
Thank you

1 Like