Tensorflow can't find available GPUs (win11)

Question

I have a pretty common issue: tensorflow framework can't recognize available GPUs on my laptop. When i try to list available GPUs in VS Code using Python (v3.11.0):

import tensorflow as tf
print("Num GPUs Available: ", len(tf.config.list_physical_devices('GPU')))
print("Built with CUDA:", tf.test.is_built_with_cuda())
print("Built with GPU device support:", tf.test.is_built_with_gpu_support())

output: 
Num GPUs Available: 0
Built with CUDA: False
Built with GPU device support: False

Laptop specs:

• System: Windows 11 Pro (23H2)
• GPU: GeForce GTX 1660Ti 6Gb
• CPU: Intel Core i7 9750h

What i've done before:

• Download Tensorflow framework (2.16.1) via pip

pip install tensorflow[and-cuda]

• Updated NVIDIA driver (v560.70)
• Downloaded NVIDIA GPU Computing Toolkit (CUDA v12.3)
• Downloaded cuDNN driver (v8.9) and paste files from bin, lib, include to appropriate folders in CUDA root path
• Manually added environment variables in Windows system properties (Path, CUDA_PATH, CUDA_HOME) Env variables config
• Added custom graphic settings to VS Code IDE Settings -> System -> Display -> Graphics
• Checked: does the system detect the video card using cmd: nvidia-smi terminal output
• Checked drivers compatibility (i guess everything ok)
• Restart system

Hope we'll find the solution :), if you have specific questions, please ask

Answer 1

TensorFlow 2.10 was the last TensorFlow release that supported GPUs on native-Windows. If you need to run on a Windows system with a newer version of TensorFlow you'll need to set up WSL2. Here's the user guide from Nvidia on setting up WSL2 for Cuda: https://docs.nvidia.com/cuda/wsl-user-guide/index.html