Torch.cuda.is_available() gives False

Your locally installed CUDA toolkit won’t be used unless you build PyTorch from source or a custom CUDA extension. You would need to properly install a current NVIDIA driver and PyTorch via the provided install instructions from here.