Why can't we pip install?

I’m curious, why can tensorflow have

$ pip install tensorflow
$ pip install tensorflow-gpu

but for PyTorch we have to use this unconvential pip matrix which greatly increases the complexity of creating an open-source tool for other users based on PyTorch. We have to jump hoops with the Pipfiles, the setup.py and getting the instructions all together for installation which causes a higher barrier for entry.

It seems in the case of cuda 9, PyTorch can be installed as pip install torch
https://pytorch.org/get-started/locally/#cuda-90-1

Right, which is the case for one of 12 30ish possible installations.

Hi, I have one question when I use anaconda virtual environment . I tried to install pytorch 1.0.0 and tensorflow under the same environment, but when I have successfully installed pytorch, tensorflow can not be established correctly, pip shows that no tensorflow version can be found. I am confused about it , if pytorch 1.0.0 can not stay with tensorflow?

@Lausanne I believe that is a separate, unrelated issue. My reference to Tensorflow is merely an example of what seems to be possible but is not with PyTorch.