Anybody who thinks running Windows is easier hasn’t tried to get Tensorflow working on Windows with GPU support.
In theory, it could run on a straight Windows build of Python, but nobody seems to have given that serious consideration. It must go through WSL, but that means passing through the GPU to WSL. When you Google how to do it, you’ll find three different approaches that have been taken over the years, only one of which is valid on modern setups. If you take one of the old approaches, you will likely twist your system in knots and need a complete reinstall to fix.
On Linux, you install the GPU drivers, compile Tensorflow with the GPU flags, and you’re done.
Anybody who thinks running Windows is easier hasn’t tried to get Tensorflow working on Windows with GPU support.
In theory, it could run on a straight Windows build of Python, but nobody seems to have given that serious consideration. It must go through WSL, but that means passing through the GPU to WSL. When you Google how to do it, you’ll find three different approaches that have been taken over the years, only one of which is valid on modern setups. If you take one of the old approaches, you will likely twist your system in knots and need a complete reinstall to fix.
On Linux, you install the GPU drivers, compile Tensorflow with the GPU flags, and you’re done.