When I install the tensorflow, they told me to inspect the Nvidia drive and So I did.
I install the Nvidia's drive on Nividia wibset. And when I input 'nvidia-smi' in CMD, it says " CUDA Version：10.1 " So I thought the cuda was installed.but after I install tensorflow-gpu, it turns out that I did't have the CUDA. So I wonder if the Nividia driver includs the "CUDA" and why the 'nvidia-smi' told me "CUDA Version：10.1" but I did't have one.And How do I confirm that CUDA is installed.
Then I try to install the pytorch on the Pytorch's website. The course told me the best way to install is using miniconda, because miniconda can install the cuda and cudnn automatically. but I'm used to using the pip,and I wonder if the pip can install the CUDA and cuDnn automatically ,just like miniconda？ Then I start to download with pip,but I find it is so slowly. it will take me 20h.I used to replace the source of the PIP to install the numpy and so on, but I can't change THE COMMAND，So I have no idea about how to accelerate it.
When I install the pytorch with the pip command above, it contains the CUDA unexpectly! I don't understand why!! But when I use 'nvcc -V' in CMD, It doesn't work!
Finnaly , if I install the pytorch-gpu with conda or pip , when I install tensorflow, should I install "CUDA" and "cuDnn" again?
I am searching for a long time on net. But no use. Please help or try to give some ideas how to achieve this.
