pytorch/torch
Gregory Chanan d895d30876 Fix c10d build without nccl.
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/18582

Differential Revision: D14672928

Pulled By: gchanan

fbshipit-source-id: 74e9805cbaf5ebe8e3f579fe08dad72eb410b80a
2019-03-29 13:34:38 -07:00
..
_thnn Fix lstrip bug revealed by B005 lint (#18177) 2019-03-21 07:56:24 -07:00
autograd Upgrade flake8-bugbear to master, fix the new lints. (#18507) 2019-03-27 08:07:41 -07:00
backends Add ability to query if built with CUDA and MKL-DNN. (#18362) 2019-03-25 10:39:09 -07:00
contrib
csrc Fix c10d build without nccl. 2019-03-29 13:34:38 -07:00
cuda Implement reference counting for shared IPC CUDA tensors (#16854) 2019-03-25 10:24:38 -07:00
distributed Fixed a formatting issue in doc comments (#17505) 2019-03-12 09:55:29 -07:00
distributions Rename trtrs to triangular_solve (#18213) 2019-03-21 14:27:21 -07:00
for_onnx
jit Delete batch tensor (#18575) 2019-03-28 23:13:27 -07:00
legacy
lib Fix NCCL/Gloo process groups and DDP stream sync bug (#18465) 2019-03-28 15:12:40 -07:00
multiprocessing Implement reference counting for shared IPC CUDA tensors (#16854) 2019-03-25 10:24:38 -07:00
nn make InstanceNorm1d raise an error if the input is 2D (#11992) 2019-03-29 06:50:04 -07:00
onnx Resubmit PR-18512: Improved onnx export for 3 onnx ops (#18571) 2019-03-28 18:12:49 -07:00
optim ReduceLrOnPlateau: best=current -> best=copy(current) (#16364) (#16697) 2019-03-29 06:56:51 -07:00
sparse
testing
utils multiline KeyError msg python bug workaround (#18557) 2019-03-29 07:04:20 -07:00
__init__.py Introduce SobolEngine (#10505) 2019-03-26 07:53:07 -07:00
__init__.pyi.in Rename btrifact* to lu (#18435) 2019-03-29 00:34:30 -07:00
_jit_internal.py Upgrade flake8-bugbear to master, fix the new lints. (#18507) 2019-03-27 08:07:41 -07:00
_ops.py Override the resolve_library_path in FBCode (#17497) 2019-03-12 22:09:24 -07:00
_six.py Upgrade flake8-bugbear to master, fix the new lints. (#18507) 2019-03-27 08:07:41 -07:00
_storage_docs.py
_tensor_docs.py Rename btrifact* to lu (#18435) 2019-03-29 00:34:30 -07:00
_tensor_str.py
_torch_docs.py Fixed torch.arange docs (#18604) 2019-03-29 06:42:28 -07:00
_utils.py
_utils_internal.py Override the resolve_library_path in FBCode (#17497) 2019-03-12 22:09:24 -07:00
abi-check.cpp
CMakeLists.txt Add named submodule support to nn::Sequential (#17552) 2019-03-29 13:06:29 -07:00
extension.h
functional.py Rename btriunpack to lu_unpack (#18529) 2019-03-29 13:01:30 -07:00
hub.py
quasirandom.py Introduce SobolEngine (#10505) 2019-03-26 07:53:07 -07:00
random.py
README.txt
script.h
serialization.py Serialization supports pathlib.Path object for the input argument (#18562) 2019-03-28 21:01:15 -07:00
storage.py
tensor.py Rename btrifact* to lu (#18435) 2019-03-29 00:34:30 -07:00

Note [TH abstraction violation]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

TH/THC provide some hpp headers, which are proper C++ headers rather than
C headers.  These headers serve double duty as *internal implementation
detail* headers, whose contents should largely not be used by external
clients.

Ideally, we would not install these headers at all; instead, you should
use public functions (in headers like `THTensor.h`, NOT `THTensor.hpp`)
to manipulate these structs.  However, there are a few places
in torch/csrc where we violate this abstraction.  They are marked with
a pointer to this note.  Each of those sites will have to be refactored
when we refactor the guts of THTensor and related structures.