Check Nvidia GPU (Name / Memory / Version)

nvidia-smi --query-gpu=name,memory.total,driver_version --format=csv,noheader
# Example Output: NVIDIA GeForce RTX 4050 Laptop GPU, 6141 MiB, 580.65.06

Watch nvidia gpu status

watch nvidia-smi
# or
watch -n 2 'nvidia-smi'

Install CUDA on WSL2

sudo apt-key del 7fa2af80
curl -L https://gist.githubusercontent.com/moracabanas/63070d9012ad4b95e86a4004caac5a9d/raw/install_cuda12.2_wsl2.sh | bash

Cuda Toolkit 12.3.2

https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=WSL-Ubuntu&target_version=2.0&target_type=runfile_local

Check if CUDA Available in Python

Install tensorflow with: uv add tensorflow[and-cuda]

Install torch with:

import os
import sys

# Silence TensorFlow logging
os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"

import torch
torch.cuda.init()

print(f'Using python in {sys.executable}\n')
print(f'CHECKING TORCH CUDA:')
print(torch.cuda.get_device_name(0))
print(f'Cuda Available Check: {torch.cuda.is_available()}')
print(f'Torch cuda version: {torch.version.cuda}\n')

import tensorflow as tf
print(f'CHECKING TENSORFLOW CUDA:')
print(tf.config.list_physical_devices('GPU'))

CUDNN

https://developer.nvidia.com/rdp/cudnn-archive

https://docs.nvidia.com/deeplearning/cudnn/installation/linux.html

https://portal.databasemart.com/kb/a2128/6-steps-to-install-gpu-accelerated-tensorflow.aspx

https://developer.nvidia.com/cudnn-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=22.04&target_type=deb_local

https://developer.nvidia.com/blog/updating-the-cuda-linux-gpg-repository-key/

https://medium.com/@gokul.a.krishnan/how-to-install-cuda-cudnn-and-tensorflow-on-ubuntu-22-04-2023-20fdfdb96907