resolve conflict in data parallel model
* master: (201 commits)
  Add cost inference to ConvGradient and WeightedSum operators (pytorch#10744)
  Move collapse dims into a single place (pytorch#11272)
  Fix some more warnings (pytorch#11257)
  Fix the batchnorm onnx exporting when affine=False
  Improve error message to include return types too (pytorch#11245)
  Check doxygen output in travis (pytorch#11124)
  Accept more numpy scalars as doubles (pytorch#9659)
  Fixed log message (pytorch#10874)
  Fix to distribution.__repr__ with lazy attributes (pytorch#11263)
  Add import export step to end to end tests
  Add complex hooks for out of tree complex implementation. (pytorch#11216)
  Unify opt flag for cmake codegen (pytorch#11227)
  nomnigraph - fix memory error in NN subgraph matchOp (pytorch#11127)
  Port PackedSequences functions to C++ (pytorch#11224)
  Treat numerical differences as warnings instead of errors when tracing (pytorch#11246)
  add a Float16UniformFill (pytorch#11123)
  Implement torch.tensordot (pytorch#10025)
  keep net type info when generating model complete net (pytorch#11032)
  Get rid of some uses of type() (pytorch#11215)
  Reorganize methods in Type, add CPUTypeDefault/CUDATypeDefault (pytorch#11205)
  ...

Summary:
`__repr__` currently fails for distributions with lazy attributes in PyTorch master, throwing a `KeyError`. This fixes the issue.

**Additionally:**
 - Added `logits` to `arg_constraints` for distributions that accept either `probs` or `logits`. This is both to have `__repr__` display the `logits` param when available, and to be able to do validation checks (e.g. NaN checks) when the logit parametrization is used. fritzo, alicanb - I think there were reasons why we had not done so in the first place, but I am unable to recall now. It passes all the tests, but let me know if there is something that I am missing at the moment.
 - There are certain distributions, e.g. `OneHotCategorical` which won't show any parameters because it uses a `categorical` instance under the hood and neither `logits` / `probs` in `arg_constraints` are present in the instance's `__dict__`. This isn't addressed in this PR.

cc. vishwakftw, fritzo, nadavbh12, apaszke
Pull Request resolved: pytorch#11263

Differential Revision: D9654959

Pulled By: apaszke

fbshipit-source-id: 16f5b20243fe8e2c13e9c528050d4df0b8ea6e45