I am trying to build Tensorflow 2.12 gpu version from source on my personal pcs with Gcc 9.3, CUDA 11.8, cudnn 8.2, ubuntu 22.04.
On my first pc everything went fine. It had 4090 gpu.
On my second pc, which is older and has quadro m4000, it fails with the error
Generating device code failed.
In the TF code it happens in file
kernel_creator.cc, in the
GenerateDeviceCode method. I am wondring if this is coz of my old gpu? If so should the build doc be updated?
Welcome to the TensorFlow Forum!
CUDA works with all Nvidia GPUs from the G8x series onwards, including GeForce, Quadro and the Tesla line. You may need to install the correct Nvidia driver, CUDA and cuDNN libraries to setup GPU with Tensorflow as per this tested build configuration.
Please refer to this TF install official link and follow the step by step instructions mentioned after verifying the Hardware/Software requirements or can refer Build from Source page to install Tensorflow with GPU support in your system.
Let us know if the issue still persists. Thank you.