Slurm Configuration Basics

Tensorflow with GPU support configuration

Spydur Interative session procedure

 

Spydur Tensorflow Configuration

After creating a Conda environment with the suitable version of Python (see  https://www.tensorflow.org/install/source?hl=pt-br#gpu):

a) Install Nvidia cuda tools: conda install nvidia/label/cuda-12.2.0::cuda-tools (see compatible version with Nvidia Drivers – check with nvidia-smi or nvcc -V. See if CUDA is on path with whereis cuda).

b) Install tensorflow using pip: pip intsall tensorflow==version (see compatible version with nvidia)