* upstream/master: (26 commits)
  cudnn 7 upgrade with spatialBN fix (pytorch#11291)
  Ignore FuseGraph Call on Windows (pytorch#11015)
  defer resolution of mkl to a cmake wrapper library (pytorch#11298)
  Cleanup dependency of distributed flags (pytorch#11221)
  Move minimal wrapdim functionality to core, remove THTensor include i… (pytorch#11283)
  Change includes from ATen/Storage.h to ATen/core/Storage.h (pytorch#11217)
  Fix scalar tensor assert in fusion compiler (pytorch#10952)
  Add dead code elimination pass (pytorch#10101)
  Distributed Data Parallel CPU module for C10D (pytorch#11168)
  Back out "[pt1][tensor] Add strides to caffe2::Tensor"
  Fix conv gradient conversion (pytorch#11312)
  Bag of clang tidy fixes for torch/csrc/ and torch/csrc/autograd (pytorch#11050)
  Sparse tensor printing; add NotImplemented autograd fn (pytorch#10181)
  Add convertToCaffe2Proto to python API
  fix doc for functional.dropout* (pytorch#10417)
  typo fix Tranpose2D -> Transpose2D (pytorch#11281)
  Remove THFinalizer
  Forward declarations of needed curand functions (pytorch#10911)
  nomnigraph - simplify core graph API and test (pytorch#11256)
  Small fixes to cppdocs for sync script (pytorch#11300)
  ...

Summary:
I'm setting up an automatic sync job for cppdocs and need two fixes to the cpp docs config:

1. Right now the cppdocs use the `torch` package to figure out the version. For C++ docs all I really need from the built package are the generated Tensor.h and Functions.h files. I can actually generate those directly via `aten/src/ATen/gen.py`, so I can skip building PyTorch altogether and save 10 minutes in the sync job! For this I need to avoid using the torch package in the docs.
2. Internal proxy issues prevent using the git link for sphinx_rtd_theme. We can just use the pip package for the cppdocs (not for the normal PyTorch docs)

soumith ezyang
Pull Request resolved: pytorch#11300

Differential Revision: D9667193

Pulled By: goldsborough

fbshipit-source-id: 5567e0b3d3bdce03f5856babdb4ff76bcee91846