Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
119 commits
Select commit Hold shift + click to select a range
2231d8e
Making the Sampler::reset to accept a new size
jaliyae Nov 12, 2018
aedc16b
add default nullopt to existing samplers' reset method
jaliyae Nov 12, 2018
9b5d23a
Merge remote-tracking branch 'upstream/master' into jaliyae/samplers
jaliyae Nov 13, 2018
b15f676
fix review comments
jaliyae Nov 13, 2018
60963c2
Add "axis" and "axis_w" arguments in FC to support customized axix to…
gujinghui Nov 21, 2018
34db39d
Add onnxifi support to SparseLengthsWeightedSum (#14210)
Nov 21, 2018
d9cdcc9
Add list inequality operator (#14129)
Nov 22, 2018
85d3fcc
Removed redundant allreduce options in DDP (#14208)
teng-li Nov 22, 2018
d55ba77
Remove extra include
Nov 22, 2018
b149456
alias analysis (#14018)
suo Nov 22, 2018
b26f82b
Robust NCCL barrier improvement to cover all devices combinations (#1…
teng-li Nov 22, 2018
341b485
Updating submodules
Nov 22, 2018
992e275
Updating submodules
Nov 22, 2018
8c4910b
Cuda version comparison with CUDA_VERSION_STRING (#14302)
achalshah20 Nov 22, 2018
fb8c3d6
removing quantization utility functions moved to fbgemm (#14301)
jspark1105 Nov 22, 2018
3918e22
Updating submodules
Nov 22, 2018
e91c8e2
ignore generated caffe2 docs and virtualenvs
JoelMarcey Nov 22, 2018
4160c13
Move intrusive_ptr to c10/util
smessmer Nov 22, 2018
0e93a03
Fix include paths for intrusive_ptr (#13692)
smessmer Nov 22, 2018
08b77d3
Use ADL to find toString (#14021)
smessmer Nov 22, 2018
7fc34a4
Convert gumbel_softmax, lp pooling weak functions and modules (#14232)
wanchaol Nov 22, 2018
f79fb58
Make sure we bind input/output of Onnxifi op positionally (#14214)
Nov 22, 2018
d55b25a
Remove individual "using c10:xxx" statements (#13168)
smessmer Nov 22, 2018
1e05f4b
Updating submodules
Nov 23, 2018
c36156e
Option to preserve bitwise accuracy of gradient checkpointed vs non-c…
definitelynotmcarilli Nov 23, 2018
7557a99
Allow dataloader to accept a custom memory pinning function (#14171)
definitelynotmcarilli Nov 23, 2018
8e3240d
Some minor fixes for Windows build script (#14218)
peterjc123 Nov 23, 2018
33d091f
shape analysis fix (#14325)
suo Nov 23, 2018
6e1e203
quote NUMPY_INCLUDE_DIR (#14341)
shkit Nov 24, 2018
60e7d04
Add Recency Weighted into SparseLookup (#14291)
olittle Nov 24, 2018
1ef9490
Enable native wrappers for the remainder of nn functions.
gchanan Nov 26, 2018
a603689
Batch more matrix multiplies (#13456)
apaszke Nov 26, 2018
a13fd7e
Allow torch.utils.cpp_extension.load to load shared libraries that ar…
goldsborough Nov 26, 2018
b30c803
allow concatenating "hybrid" (sparse/dense) tensors along their dense…
Nov 26, 2018
fa73037
Add proper from_blob overloads (#13982)
goldsborough Nov 26, 2018
90ed2f5
minimize code compiled with avx2 and header includes from them (#14313)
jspark1105 Nov 26, 2018
a10a993
Fix -Wreturn-std-move (#14113)
goldsborough Nov 26, 2018
0651b59
Updating submodules
Nov 26, 2018
80ba65e
remove unnecessary zero_point argument from constructors (#14323)
jspark1105 Nov 26, 2018
2fa3c83
fix tensor advanced indexing with assignment (#14311)
suo Nov 26, 2018
0d1f382
Removing Caffe2-specific conda infra
pjh5 Nov 26, 2018
8cc5d54
Updating submodules
Nov 26, 2018
b13f91d
Allow graph fuser to move chunks past multiple nodes. (#14055)
zou3519 Nov 26, 2018
735cd06
FeedTensor returns a Tensor (#14196)
jerryzh168 Nov 26, 2018
bb7fb7e
remove CAFFE2_API from IdWrapper (#14044)
anderspapitto Nov 26, 2018
5918de8
Revert D13166669: [pytorch][PR] Allow dataloader to accept a custom m…
Nov 26, 2018
b18063b
Fix caffe2 => onnx exporter for ConvTranspose (#14143)
Nov 26, 2018
afb2c0c
changing some rpath stuff (#14304)
pjh5 Nov 26, 2018
cef23a4
More JIT type hierarchy refinement (#14127)
wanchaol Nov 27, 2018
1ca0ec7
fix typo in `torch.sum` documentation (#14250)
Nov 27, 2018
6f3002a
Fixed c10d test (#14389)
teng-li Nov 27, 2018
f639249
Fix dataloader iterator test (#14045)
goldsborough Nov 27, 2018
2d6f039
Fixed file init_method write/read race (#14388)
teng-li Nov 27, 2018
9e1805d
Switch Int8ChannelShuffle operator to QNNPACK (#14362)
Nov 27, 2018
1255801
backward for sparse.addmm(D, S, D, alpha, beta) -> D (#13345)
weiyangfb Nov 27, 2018
507cb16
Delete OPENMP_STUB translation. (#14286)
ezyang Nov 27, 2018
3423900
Remove unused executors, part 1 (#14117)
Nov 27, 2018
7953b32
Remove unused executors, part 2 (#14115)
Nov 27, 2018
f019a2d
Remove unused executors, part 3 (#14199)
Nov 27, 2018
a0ef8af
Updating submodules
Nov 27, 2018
3ff7071
Implement NaN-propagating max/min on Vec256.
Nov 27, 2018
388258f
Add hash functions for Stream, CUDAStream; fix Device hash function (…
ezyang Nov 27, 2018
c5cc1e3
Delete legacy THCStream (long live THCStream). (#14246)
ezyang Nov 27, 2018
c19af59
Use integer math to compute output size of pooling operations (#14405)
f0k Nov 27, 2018
49fe678
Add variable_factories.h to cppdocs (#14381)
goldsborough Nov 27, 2018
a3cfab2
per-group and per-channel quantization (#14340)
jspark1105 Nov 27, 2018
50bc9dc
fix doc for sparse.addmm (#14403)
weiyangfb Nov 27, 2018
936c2bb
Use new style barrier support in c10d/gloo (#14294)
pietern Nov 27, 2018
5865561
Remove algorithm caching in ProcessGroupGloo (#14295)
pietern Nov 27, 2018
52f5022
Refer to all work with ProcessGroup prefix (#14296)
pietern Nov 27, 2018
03864b7
Add option structs and timeout field (#14297)
pietern Nov 27, 2018
9598d38
Make ProcessGroup::Work::wait() throw (#14298)
pietern Nov 27, 2018
2cc35c1
Barrier synchronizes with prior work before completing (#14386)
pietern Nov 27, 2018
66c8bbf
Add boolean dispatch for function overloading (#14081)
Nov 27, 2018
bd62948
Updating submodules
Nov 27, 2018
8408dff
Add Type support to the fuser, fuse more (#14336)
t-vi Nov 27, 2018
e22cc7c
Print default values and introduce ir view classes (#14176)
zdevito Nov 27, 2018
5fefb29
Tensor construction: combine Resize+mutable_data - 4/4 (#13856)
jerryzh168 Nov 27, 2018
ffbc390
Fixed torch.multiprocessing.spawn for not being able to spawn like da…
teng-li Nov 27, 2018
3fca4bd
Trace in-place ops (#14254)
suo Nov 27, 2018
50e9c56
Move Scalar and ScalarType to c10/core
smessmer Nov 27, 2018
44e21cf
Fix include paths for Scalar.h and ScalarType.h
smessmer Nov 27, 2018
8b1ca28
Move ScalarTypeUtils.h to c10
smessmer Nov 27, 2018
0b10f14
Move UniqueVoidPtr to c10
smessmer Nov 27, 2018
3a71d5e
Move Allocator.h to c10
smessmer Nov 27, 2018
507ed90
Fix include paths for Allocator.h
smessmer Nov 27, 2018
3d4d09f
Move Storage and StorageImpl to c10
smessmer Nov 27, 2018
6c2e816
Fix include paths for Storage.h and StorageImpl.h
smessmer Nov 27, 2018
ed10ef9
Move TensorTypeId to c10/core
smessmer Nov 27, 2018
3257ac1
Fix include paths for TensorTypeId.h and TensorTypeIdRegistration.h
smessmer Nov 27, 2018
f9c27d6
Remove fake dependencies from TensorImpl to caffe2 (#14141)
smessmer Nov 27, 2018
1b80644
Revert D13192228: [pytorch][PR] [jit] Add boolean dispatch for functi…
Nov 27, 2018
1b93cb7
Document device_guard in native_functions.yaml (#14235)
ezyang Nov 27, 2018
b7856a3
Add scaffolding for HIP backend in ATen/core. (#14285)
ezyang Nov 27, 2018
e392d42
Allowing TaskGroups to carry remote nets (#14342)
heslami Nov 27, 2018
ba6c49c
Add test of ONNX_ATEN (#14259)
zrphercule Nov 27, 2018
5c84145
condition blob in while_op test changes data type (#14279)
jerryzh168 Nov 27, 2018
ae1b376
Windows local build: restore original working dir after activating VC…
Nov 27, 2018
0199d59
Resubmit: Set the correct engine name for position weighted pooling w…
chocjy Nov 27, 2018
006505b
Speed-up "advanced" indexing operations (#13420)
colesbury Nov 27, 2018
1620161
when BUILD_CAFFE2_OPS is OFF, torch-python needs a direct dep on nccl…
anderspapitto Nov 27, 2018
75bac5a
Eliminate necessity of HIPify on AccumulateType.h (#14412)
ezyang Nov 28, 2018
226a01e
Handling of pretty-printing methods (#14378)
zdevito Nov 28, 2018
23d111c
Fix clang tidy errors
wanchaol Nov 28, 2018
6f2307b
Allow building libraries with setuptools that dont have abi suffix (#…
goldsborough Nov 28, 2018
82175f3
Move Affine grid to C++ (#14392)
Nov 28, 2018
23f901a
fix enable_cpu_fuser
zdevito Nov 28, 2018
d75f751
Add boolean dispatch for function overloading (#14425)
Nov 28, 2018
662f66e
Add poisson_nll_loss to script
Nov 28, 2018
b08a186
roll along multiple dimensions
nairbv Nov 28, 2018
a0def0b
check for invalid ranges in torch.arange
nairbv Nov 28, 2018
4cdcbbf
Use nn module tests in test_jit (#14238)
Nov 28, 2018
7d07fcd
Fixed SyncParam/QueueReduction/SyncReduction test for 2+ GPUs (#14452)
teng-li Nov 28, 2018
3d98810
Revert D13192230: [pytorch][PR] [jit] Use nn module tests in test_jit
Nov 28, 2018
a38ed02
PT1 Stable Release Distributed Documentation (#14444)
teng-li Nov 28, 2018
e1275bb
merging with master
jaliyae Nov 28, 2018
1a34ee6
resolve review comments
jaliyae Nov 28, 2018
e3b7bfd
fix review comments
jaliyae Nov 28, 2018
627d83e
clang tidy fixes
jaliyae Nov 28, 2018
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
6 changes: 6 additions & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -192,6 +192,9 @@ MANIFEST-*
# generated version file
caffe2/version.py

# generated caffe2 docs
docs/caffe2/

# setup.py intermediates
.eggs
caffe2.egg-info
Expand Down Expand Up @@ -221,3 +224,6 @@ TAGS
# Files generated when a patch is rejected
*.orig
*.rej

# virtual environments
venv/
15 changes: 1 addition & 14 deletions .jenkins/caffe2/build.sh
Original file line number Diff line number Diff line change
Expand Up @@ -92,7 +92,7 @@ fi


###############################################################################
# Use special scripts for Android, conda, and setup builds
# Use special scripts for Android and setup builds
###############################################################################
if [[ "${BUILD_ENVIRONMENT}" == *-android* ]]; then
export ANDROID_NDK=/opt/ndk
Expand All @@ -102,19 +102,6 @@ if [[ "${BUILD_ENVIRONMENT}" == *-android* ]]; then
CMAKE_ARGS+=("-DUSE_ZSTD=ON")
"${ROOT_DIR}/scripts/build_android.sh" ${CMAKE_ARGS[*]} "$@"
exit 0
elif [[ "${BUILD_ENVIRONMENT}" == conda* ]]; then
"${ROOT_DIR}/scripts/build_anaconda.sh" --skip-tests --install-locally "$@"
report_compile_cache_stats

# This build will be tested against onnx tests, which needs onnx installed.
# At this point the visible protbuf installation will be in conda, since one
# of Caffe2's dependencies uses conda, so the correct protobuf include
# headers are those in conda as well
# This path comes from install_anaconda.sh which installs Anaconda into the
# docker image
PROTOBUF_INCDIR=/opt/conda/include pip install -b /tmp/pip_install_onnx "file://${ROOT_DIR}/third_party/onnx#egg=onnx"
report_compile_cache_stats
exit 0
fi


Expand Down
24 changes: 3 additions & 21 deletions .jenkins/caffe2/test.sh
Original file line number Diff line number Diff line change
Expand Up @@ -15,14 +15,6 @@ fi
# The prefix must mirror the setting from build.sh
INSTALL_PREFIX="/usr/local/caffe2"

# Anaconda builds have a special install prefix and python
if [[ "$BUILD_ENVIRONMENT" == conda* ]]; then
# This path comes from install_anaconda.sh which installs Anaconda into the
# docker image
PYTHON="/opt/conda/bin/python"
INSTALL_PREFIX="/opt/conda/"
fi

# Add the site-packages in the caffe2 install prefix to the PYTHONPATH
SITE_DIR=$($PYTHON -c "from distutils import sysconfig; print(sysconfig.get_python_lib(prefix=''))")
INSTALL_SITE_DIR="${INSTALL_PREFIX}/${SITE_DIR}"
Expand All @@ -34,11 +26,9 @@ if [[ "${BUILD_ENVIRONMENT}" == *-android* ]]; then
fi

# Set PYTHONPATH and LD_LIBRARY_PATH so that python can find the installed
# Caffe2. This shouldn't be done on Anaconda, as Anaconda should handle this.
if [[ "$BUILD_ENVIRONMENT" != conda* ]]; then
export PYTHONPATH="${PYTHONPATH}:$INSTALL_SITE_DIR"
export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:${INSTALL_PREFIX}/lib"
fi
# Caffe2.
export PYTHONPATH="${PYTHONPATH}:$INSTALL_SITE_DIR"
export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:${INSTALL_PREFIX}/lib"

cd "$ROOT_DIR"

Expand Down Expand Up @@ -97,13 +87,6 @@ if [[ "$BUILD_ENVIRONMENT" == *-cuda* ]]; then
EXTRA_TESTS+=("$CAFFE2_PYPATH/contrib/nccl")
fi

conda_ignore_test=()
if [[ $BUILD_ENVIRONMENT == conda* ]]; then
# These tests both assume Caffe2 was built with leveldb, which is not the case
conda_ignore_test+=("--ignore $CAFFE2_PYPATH/python/dataio_test.py")
conda_ignore_test+=("--ignore $CAFFE2_PYPATH/python/operator_test/checkpoint_test.py")
fi

rocm_ignore_test=()
if [[ $BUILD_ENVIRONMENT == *-rocm* ]]; then
# Currently these tests are failing on ROCM platform:
Expand All @@ -130,7 +113,6 @@ pip install --user pytest-sugar
--ignore "$CAFFE2_PYPATH/python/operator_test/matmul_op_test.py" \
--ignore "$CAFFE2_PYPATH/python/operator_test/pack_ops_test.py" \
--ignore "$CAFFE2_PYPATH/python/mkl/mkl_sbn_speed_test.py" \
${conda_ignore_test[@]} \
${rocm_ignore_test[@]} \
"$CAFFE2_PYPATH/python" \
"${EXTRA_TESTS[@]}"
Expand Down
2 changes: 2 additions & 0 deletions .jenkins/pytorch/win-build.sh
Original file line number Diff line number Diff line change
Expand Up @@ -99,8 +99,10 @@ if "%REBUILD%"=="" (
:: Install ninja
if "%REBUILD%"=="" ( pip install ninja )

set WORKING_DIR=%CD%
call "C:\\Program Files (x86)\\Microsoft Visual Studio\\2017\\Community\\VC\\Auxiliary\\Build\\vcvarsall.bat" x64
call "C:\\Program Files (x86)\\Microsoft Visual Studio\\2017\\Community\\VC\\Auxiliary\\Build\\vcvarsall.bat" x86_amd64
cd %WORKING_DIR%

git submodule update --init --recursive

Expand Down
2 changes: 2 additions & 0 deletions .jenkins/pytorch/win-test.sh
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,9 @@ if NOT "%BUILD_ENVIRONMENT%"=="" (
)
pip install ninja future hypothesis

set WORKING_DIR=%CD%
call "C:\\Program Files (x86)\\Microsoft Visual Studio\\2017\\Community\\VC\\Auxiliary\\Build\\vcvarsall.bat" x86_amd64
cd %WORKING_DIR%

set PATH=C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v9.0\\bin;C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v9.0\\libnvvp;%PATH%
set CUDA_PATH=C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v9.0
Expand Down
5 changes: 4 additions & 1 deletion .travis.yml
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,10 @@ matrix:
install: pip install mypy mypy-extensions
script: mypy @mypy-files.txt
- env: CPP_DOC_CHECK
install: sudo apt-get install -y doxygen
python: "3.6"
install:
- sudo apt-get install -y doxygen
- pip install -r requirements.txt
script: cd docs/cpp/source && ./check-doxygen.sh
- env: CLANG_TIDY
python: "3.6"
Expand Down
2 changes: 1 addition & 1 deletion CONTRIBUTING.md
Original file line number Diff line number Diff line change
Expand Up @@ -412,7 +412,7 @@ Here are a few well known pitfalls and workarounds:
* The idiom `static_assert(f() == f())` to test if `f` is constexpr
does not work; you'll get "error C2131: expression did not evaluate
to a constant". Don't use these asserts on Windows.
(Example: `aten/src/ATen/core/intrusive_ptr.h`)
(Example: `c10/util/intrusive_ptr.h`)

* (NVCC) Code you access inside a `static_assert` will eagerly be
evaluated as if it were device code, and so you might get an error
Expand Down
2 changes: 1 addition & 1 deletion aten/src/ATen/ATen.h
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
#include "ATen/core/Generator.h"
#include "ATen/core/Layout.h"
#include "ATen/core/Scalar.h"
#include "ATen/core/Storage.h"
#include "c10/core/Storage.h"
#include "ATen/core/TensorMethods.h"
#include "ATen/core/TensorOptions.h"
#include "c10/util/Exception.h"
7 changes: 5 additions & 2 deletions aten/src/ATen/AccumulateType.h
Original file line number Diff line number Diff line change
Expand Up @@ -6,17 +6,20 @@
// Example:
// using accscalar_t = acc_type<scalar_t, true>;

#ifdef __CUDACC__
#if defined(__CUDACC__)
#include <cuda.h>
#include <cuda_fp16.h>
#elif defined(__HIPCC__)
#include <hip/hip_runtime.h>
#include <hip/hip_fp16.h>
#endif

namespace at {

template <typename T, bool is_cuda>
struct AccumulateType { };

#ifdef __CUDACC__
#if defined(__CUDACC__) || defined(__HIPCC__)
template <> struct AccumulateType<half, true> { using type = float; };
#endif
template <> struct AccumulateType<Half, true> { using type = float; };
Expand Down
2 changes: 1 addition & 1 deletion aten/src/ATen/Allocator.h
Original file line number Diff line number Diff line change
@@ -1,2 +1,2 @@
#pragma once
#include <ATen/core/Allocator.h>
#include <c10/core/Allocator.h>
2 changes: 1 addition & 1 deletion aten/src/ATen/DeviceGuard.h
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@

#include <c10/DeviceGuard.h>
#include <ATen/core/Tensor.h>
#include <ATen/core/ScalarType.h> // TensorList whyyyyy
#include <c10/core/ScalarType.h> // TensorList whyyyyy

namespace at {

Expand Down
2 changes: 1 addition & 1 deletion aten/src/ATen/InferSize.h
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
#pragma once

#include <ATen/ScalarType.h>
#include <c10/core/ScalarType.h>
#include <c10/util/Optional.h>
#include <sstream>
#include <vector>
Expand Down
11 changes: 6 additions & 5 deletions aten/src/ATen/ScalarOps.h
Original file line number Diff line number Diff line change
@@ -1,18 +1,19 @@
#pragma once

#include "ATen/core/Scalar.h"
#include <c10/core/Scalar.h>
#include "ATen/Tensor.h"

namespace at {
// This is in the c10 namespace because we use ADL to find the functions in it.
namespace c10 {

// FIXME: this should be (and was) Scalar::toTensor, but there is currently no way
// to implement this without going through Derived Types (which are not part of core).
inline Tensor scalar_to_tensor(Scalar s) {
inline at::Tensor scalar_to_tensor(Scalar s) {
if (s.isFloatingPoint()) {
return CPU(kDouble).scalarTensor(s);
return at::CPU(kDouble).scalarTensor(s);
} else {
AT_ASSERT(s.isIntegral());
return CPU(kLong).scalarTensor(s);
return at::CPU(kLong).scalarTensor(s);
}
}

Expand Down
2 changes: 1 addition & 1 deletion aten/src/ATen/ScalarType.h
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
#pragma once
#include <ATen/core/ATenGeneral.h> // for BC reasons
#include <ATen/core/Backend.h>
#include <ATen/core/ScalarType.h>
#include <c10/core/ScalarType.h>
2 changes: 1 addition & 1 deletion aten/src/ATen/Storage.h
Original file line number Diff line number Diff line change
@@ -1,2 +1,2 @@
#pragma once
#include <ATen/core/Storage.h>
#include <c10/core/Storage.h>
2 changes: 1 addition & 1 deletion aten/src/ATen/TensorOperators.h
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
#pragma once

#include "ATen/core/Scalar.h"
#include <c10/core/Scalar.h>
#include "ATen/Tensor.h"
#include "ATen/Type.h"

Expand Down
6 changes: 3 additions & 3 deletions aten/src/ATen/Utils.h
Original file line number Diff line number Diff line change
@@ -1,13 +1,13 @@
#pragma once

#include "ATen/core/ATenGeneral.h"
#include "ATen/core/StorageImpl.h"
#include <c10/core/StorageImpl.h>
#include "ATen/core/UndefinedTensorImpl.h"

#include <ATen/core/ScalarType.h>
#include <c10/core/ScalarType.h>
#include "ATen/Formatting.h"
#include <c10/util/ArrayRef.h>
#include "c10/util/Exception.h"
#include <c10/util/Exception.h>

#include <algorithm>
#include <sstream>
Expand Down
Loading