Well, dont know why TensorFlow 2.10+ stopped supporting GPUs on Windows… maybe, time to move to PyTorch…What a pity ?
Hopefully the Tensorflow team will soon reverse this decision.
Don’t you just have to install WSL2?
Say one uses Tensorflow at his/her workplace. For a set of different reasons (admin, technical, security, …), having a new (call it) software installed on your machine at the organization you work for and having it working properly can turn out to be challenging.
And generally speaking, I’m not in the camp of adding extra layers (no ML puns intended!) to my setup.
WSL2…then install whatever linux distribution… [wsl2 install -d…] … maybe, make it the default…
but you won’'t get far beyond TF v2.10…cause TF stopped supporting GPU under windows beyond that version…sad…but a reality…
so i went ahead and went with pytorch(for my current project) and even made sure it uses my shared memory(under windows 10+) for the GPU model/data…
NOTE: 50% of your RAM is reserved as shared memory VRAM under windows…mostly for the onboard Video/GPU like intel. which doesnt have their own memory.