add API for re-functionalization an aten op; other functionalization fixes by bdhirsh · Pull Request #79420 · pytorch/pytorch

bdhirsh · 2022-06-13T15:37:56Z

I moved out the changes to FunctionalTensorWrapper.h from the LTC <> functionalization PR into a separate PR here, so dealing with XLA failures will be a bit easier.

Specifically, the LTC PR will make a few operators like pixel_shuffle that are functional, but decompose into view ops, require re-functionalization once they hit the XLA backend. This PR exposes a helper utility to do that through functionalize_aten_op. This PR also contains the changes to:

fix detach() for FunctionalTensorWrapper
fix some undefined tensor handling cases

I have an XLA patch here to do the re-functionalizing: pytorch/xla#3646

Stack from ghstack (oldest at bottom):

…fixes [ghstack-poisoned]

facebook-github-bot · 2022-06-13T15:38:02Z

🔗 Helpful links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/79420
📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓Need help or want to give feedback on the CI? Visit our office hours
↩️ [fb-only] Re-run with SSH instructions

❌ 2 New Failures

As of commit a9a2430 (more details on the Dr. CI page):

Expand to see more

2/2 failures introduced in this PR

🕵️ 2 new failures recognized by patterns

The following CI failures do not appear to be due to upstream breakages

pull / win-vs2019-cuda11.6-py3 / build (1/2)

Step: "Build" (full log | diagnosis details | 🔁 rerun)

2022-06-22T06:32:05.3559534Z FAILED: bin/torch_cpu.dll lib/torch_cpu.lib

2022-06-22T06:31:49.5755404Z cub.cu
2022-06-22T06:31:49.5755910Z [5594/6323] Building CUDA object caffe2\CMakeFiles\torch_cuda_cu.dir\__\aten\src\ATen\native\cuda\DistributionLogNormalKernel.cu.obj
2022-06-22T06:31:49.5756335Z DistributionLogNormalKernel.cu
2022-06-22T06:31:57.0485914Z [5595/6323] Building CUDA object caffe2\CMakeFiles\torch_cuda_cu.dir\__\aten\src\ATen\native\cuda\DistributionUniform.cu.obj
2022-06-22T06:31:57.0486356Z DistributionUniform.cu
2022-06-22T06:31:59.6497073Z [5596/6323] Building CUDA object caffe2\CMakeFiles\torch_cuda_cu.dir\__\aten\src\ATen\native\cuda\EmbeddingBackwardKernel.cu.obj
2022-06-22T06:31:59.6497512Z EmbeddingBackwardKernel.cu
2022-06-22T06:32:00.4615771Z [5597/6323] Building CUDA object caffe2\CMakeFiles\torch_cuda_cu.dir\__\aten\src\ATen\native\cuda\Embedding.cu.obj
2022-06-22T06:32:00.4616106Z Embedding.cu
2022-06-22T06:32:05.3559024Z [5598/6323] Linking CXX shared library bin\torch_cpu.dll
2022-06-22T06:32:05.3559534Z FAILED: bin/torch_cpu.dll lib/torch_cpu.lib 
2022-06-22T06:32:05.3561390Z cmd.exe /C "cd . && C:\Jenkins\Miniconda3\Library\bin\cmake.exe -E vs_link_dll --intdir=caffe2\CMakeFiles\torch_cpu.dir --rc=C:\PROGRA~2\WI3CF2~1\10\bin\100190~1.0\x64\rc.exe --mt=C:\PROGRA~2\WI3CF2~1\10\bin\100190~1.0\x64\mt.exe --manifests  -- C:\PROGRA~2\MICROS~2\2019\BUILDT~1\VC\Tools\MSVC\1428~1.293\bin\Hostx64\x64\link.exe /nologo @CMakeFiles\torch_cpu.rsp  /out:bin\torch_cpu.dll /implib:lib\torch_cpu.lib /pdb:bin\torch_cpu.pdb /dll /version:0.0 /machine:x64 /ignore:4049 /ignore:4217 /ignore:4099 /INCREMENTAL:NO /NODEFAULTLIB:vcomp  -WHOLEARCHIVE:C:/actions-runner/_work/pytorch/pytorch/build/lib/caffe2_protos.lib -WHOLEARCHIVE:C:/actions-runner/_work/pytorch/pytorch/build/lib/onnx.lib  && cd ."
2022-06-22T06:32:05.3563694Z LINK: command "C:\PROGRA~2\MICROS~2\2019\BUILDT~1\VC\Tools\MSVC\1428~1.293\bin\Hostx64\x64\link.exe /nologo @CMakeFiles\torch_cpu.rsp /out:bin\torch_cpu.dll /implib:lib\torch_cpu.lib /pdb:bin\torch_cpu.pdb /dll /version:0.0 /machine:x64 /ignore:4049 /ignore:4217 /ignore:4099 /INCREMENTAL:NO /NODEFAULTLIB:vcomp -WHOLEARCHIVE:C:/actions-runner/_work/pytorch/pytorch/build/lib/caffe2_protos.lib -WHOLEARCHIVE:C:/actions-runner/_work/pytorch/pytorch/build/lib/onnx.lib /MANIFEST /MANIFESTFILE:bin\torch_cpu.dll.manifest" failed (exit code 1120) with the following output:
2022-06-22T06:32:05.3564564Z    Creating library lib\torch_cpu.lib and object lib\torch_cpu.exp
2022-06-22T06:32:05.3565485Z FunctionalTensorWrapper.cpp.obj : error LNK2019: unresolved external symbol "__declspec(dllimport) public: __cdecl at::functionalization::_RemoveExcludeDispatchKeyGuard::~_RemoveExcludeDispatchKeyGuard(void)" (__imp_??1_RemoveExcludeDispatchKeyGuard@functionalization@at@@QEAA@XZ) referenced in function "public: __cdecl at::functionalization::ReenableFunctionalize::~ReenableFunctionalize(void)" (??1ReenableFunctionalize@functionalization@at@@QEAA@XZ)
2022-06-22T06:32:05.3566316Z bin\torch_cpu.dll : fatal error LNK1120: 1 unresolved externals
2022-06-22T06:32:05.4100685Z [5599/6323] Building CUDA object caffe2\CMakeFiles\torch_cuda_cu.dir\__\aten\src\ATen\native\cuda\DistributionBernoulli.cu.obj
2022-06-22T06:32:05.4101071Z DistributionBernoulli.cu
2022-06-22T06:32:06.5446701Z [5600/6323] Building CUDA object caffe2\CMakeFiles\torch_cuda_cu.dir\__\aten\src\ATen\native\cuda\DistributionRandomKernel.cu.obj
2022-06-22T06:32:06.5447211Z DistributionRandomKernel.cu
2022-06-22T06:32:09.6694112Z [5601/6323] Building CUDA object caffe2\CMakeFiles\torch_cuda_cu.dir\__\aten\src\ATen\native\cuda\ForeachBinaryOpList.cu.obj

pull / win-vs2019-cpu-py3 / build (2/2)

Step: "Build" (full log | diagnosis details | 🔁 rerun)

2022-06-22T06:32:24.1339740Z FAILED: bin/torch_cpu.dll lib/torch_cpu.lib

2022-06-22T06:29:24.5937706Z [5767/5849] Building CXX object caffe2\torch\CMakeFiles\torch_python.dir\csrc\Layout.cpp.obj
2022-06-22T06:29:24.5986128Z [5768/5849] Building CXX object caffe2\torch\CMakeFiles\torch_python.dir\csrc\QScheme.cpp.obj
2022-06-22T06:29:25.0694138Z [5769/5849] Building CXX object caffe2\torch\CMakeFiles\torch_python.dir\csrc\MemoryFormat.cpp.obj
2022-06-22T06:29:25.2282097Z [5770/5849] Building CXX object test_jit\CMakeFiles\jitbackend_test.dir\test_backend_lib.cpp.obj
2022-06-22T06:29:25.6435049Z [5771/5849] Building CXX object caffe2\torch\CMakeFiles\torch_python.dir\csrc\Dtype.cpp.obj
2022-06-22T06:29:25.8737219Z [5772/5849] Building CXX object caffe2\torch\CMakeFiles\torch_python.dir\csrc\DynamicTypes.cpp.obj
2022-06-22T06:29:25.8981089Z [5773/5849] Building CXX object caffe2\torch\CMakeFiles\torch_python.dir\csrc\Generator.cpp.obj
2022-06-22T06:29:26.1163955Z [5774/5849] Building CXX object caffe2\torch\CMakeFiles\torch_python.dir\csrc\Exceptions.cpp.obj
2022-06-22T06:29:26.3268386Z [5775/5849] Building CXX object test_jit\CMakeFiles\torchbind_test.dir\test_custom_class_registrations.cpp.obj
2022-06-22T06:32:24.1339410Z [5776/5849] Linking CXX shared library bin\torch_cpu.dll
2022-06-22T06:32:24.1339740Z FAILED: bin/torch_cpu.dll lib/torch_cpu.lib 
2022-06-22T06:32:24.1340803Z cmd.exe /C "cd . && C:\Jenkins\Miniconda3\Library\bin\cmake.exe -E vs_link_dll --intdir=caffe2\CMakeFiles\torch_cpu.dir --rc=C:\PROGRA~2\WI3CF2~1\10\bin\100190~1.0\x64\rc.exe --mt=C:\PROGRA~2\WI3CF2~1\10\bin\100190~1.0\x64\mt.exe --manifests  -- C:\PROGRA~2\MICROS~2\2019\BUILDT~1\VC\Tools\MSVC\1428~1.293\bin\Hostx64\x64\link.exe /nologo @CMakeFiles\torch_cpu.rsp  /out:bin\torch_cpu.dll /implib:lib\torch_cpu.lib /pdb:bin\torch_cpu.pdb /dll /version:0.0 /machine:x64 /ignore:4049 /ignore:4217 /ignore:4099 /INCREMENTAL:NO /NODEFAULTLIB:vcomp  -WHOLEARCHIVE:C:/actions-runner/_work/pytorch/pytorch/build/lib/caffe2_protos.lib -WHOLEARCHIVE:C:/actions-runner/_work/pytorch/pytorch/build/lib/onnx.lib  && cd ."
2022-06-22T06:32:24.1342522Z LINK: command "C:\PROGRA~2\MICROS~2\2019\BUILDT~1\VC\Tools\MSVC\1428~1.293\bin\Hostx64\x64\link.exe /nologo @CMakeFiles\torch_cpu.rsp /out:bin\torch_cpu.dll /implib:lib\torch_cpu.lib /pdb:bin\torch_cpu.pdb /dll /version:0.0 /machine:x64 /ignore:4049 /ignore:4217 /ignore:4099 /INCREMENTAL:NO /NODEFAULTLIB:vcomp -WHOLEARCHIVE:C:/actions-runner/_work/pytorch/pytorch/build/lib/caffe2_protos.lib -WHOLEARCHIVE:C:/actions-runner/_work/pytorch/pytorch/build/lib/onnx.lib /MANIFEST /MANIFESTFILE:bin\torch_cpu.dll.manifest" failed (exit code 1120) with the following output:
2022-06-22T06:32:24.1343352Z    Creating library lib\torch_cpu.lib and object lib\torch_cpu.exp
2022-06-22T06:32:24.1344449Z FunctionalTensorWrapper.cpp.obj : error LNK2019: unresolved external symbol "__declspec(dllimport) public: __cdecl at::functionalization::_RemoveExcludeDispatchKeyGuard::~_RemoveExcludeDispatchKeyGuard(void)" (__imp_??1_RemoveExcludeDispatchKeyGuard@functionalization@at@@QEAA@XZ) referenced in function "public: __cdecl at::functionalization::ReenableFunctionalize::~ReenableFunctionalize(void)" (??1ReenableFunctionalize@functionalization@at@@QEAA@XZ)
2022-06-22T06:32:24.1345274Z bin\torch_cpu.dll : fatal error LNK1120: 1 unresolved externals
2022-06-22T06:32:24.1345544Z ninja: build stopped: subcommand failed.
2022-06-22T06:32:24.1528171Z -- Building version 1.13.0a0+gita9a2430
2022-06-22T06:32:24.1529433Z cmake -GNinja -DBUILD_ENVIRONMENT=win-vs2019-cpu-py3 -DBUILD_PYTHON=True -DBUILD_TEST=True -DBUILD_TYPE=release -DBUILD_WHEEL=1 -DCMAKE_BUILD_TYPE=Release -DCMAKE_GENERATOR=Ninja -DCMAKE_INCLUDE_PATH=C:\actions-runner\_work\pytorch\pytorch\build\win_tmp\mkl\include -DCMAKE_INSTALL_PREFIX=C:\actions-runner\_work\pytorch\pytorch\torch -DCMAKE_PREFIX_PATH=C:\Jenkins\Miniconda3\Lib\site-packages -DNUMPY_INCLUDE_DIR=C:\Jenkins\Miniconda3\lib\site-packages\numpy\core\include -DPYTHON_EXECUTABLE=C:\Jenkins\Miniconda3\python.exe -DPYTHON_INCLUDE_DIR=C:\Jenkins\Miniconda3\Include -DPYTHON_LIBRARY=C:\Jenkins\Miniconda3/libs/python38.lib -DTORCH_BUILD_VERSION=1.13.0a0+gita9a2430 -DUSE_CUDA=0 -DUSE_NUMPY=True C:\actions-runner\_work\pytorch\pytorch
2022-06-22T06:32:24.1530724Z cmake --build . --target install --config Release -- -j 8
2022-06-22T06:32:24.2116699Z

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

ezyang

I've already reviewed this right

bdhirsh · 2022-06-13T19:02:47Z

I've already reviewed this right

yup!

…nalization fixes" I moved out the changes to `FunctionalTensorWrapper.h` from the LTC <> functionalization PR into a separate PR here, so dealing with XLA failures will be a bit easier. Specifically, the LTC PR will make a few operators like `pixel_shuffle` that are functional, but decompose into view ops, require re-functionalization once they hit the XLA backend. This PR exposes a helper utility to do that through `functionalize_aten_op`. This PR also contains the changes to: - fix `detach()` for `FunctionalTensorWrapper` - fix some undefined tensor handling cases I have an XLA patch here to do the re-functionalizing: pytorch/xla#3646 [ghstack-poisoned]

pytorchmergebot · 2022-06-22T06:08:27Z

Rebased gh/bdhirsh/199/orig onto refs/remotes/origin/master because #75527 was rebased, please pull locally before adding more changes (for example, via ghstack checkout https://github.com/pytorch/pytorch/pull/79420)

github-actions · 2022-08-21T09:35:15Z

Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as Stale.
Feel free to remove the Stale label if you feel this was a mistake.
If you are unable to remove the Stale label please contact a maintainer in order to do so.
If you want the bot to never mark this PR stale again, add the no-stale label.
Stale pull requests will automatically be closed after 30 days of inactivity.

bdhirsh · 2022-08-23T17:13:17Z

closing, a version of this was already landed

add API for re-functionalization an aten op; other functionalization …

062d076

…fixes [ghstack-poisoned]

bdhirsh mentioned this pull request Jun 13, 2022

add a new alias key for functional to non functional decompositions #76358

Closed

facebook-github-bot added the cla signed label Jun 13, 2022

bdhirsh mentioned this pull request Jun 13, 2022

integrate functionalization <> LTC torchscript backend #75527

Closed

bdhirsh requested a review from ezyang June 13, 2022 16:48

bdhirsh mentioned this pull request Jun 13, 2022

run opted out decompositions pytorch/xla#3646

Merged

ezyang approved these changes Jun 13, 2022

View reviewed changes

bdhirsh and others added 10 commits June 13, 2022 12:52

github-actions bot added the Stale label Aug 21, 2022

bdhirsh closed this Aug 23, 2022

facebook-github-bot deleted the gh/bdhirsh/249/head branch September 23, 2022 14:20

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add API for re-functionalization an aten op; other functionalization fixes#79420

add API for re-functionalization an aten op; other functionalization fixes#79420
bdhirsh wants to merge 11 commits intogh/bdhirsh/249/basefrom
gh/bdhirsh/249/head

bdhirsh commented Jun 13, 2022 •

edited by pytorchmergebot

Loading

Uh oh!

facebook-github-bot commented Jun 13, 2022 •

edited

Loading

🕵️ 2 new failures recognized by patterns

pull / win-vs2019-cuda11.6-py3 / build (1/2)

pull / win-vs2019-cpu-py3 / build (2/2)

Uh oh!

ezyang left a comment

Uh oh!

bdhirsh commented Jun 13, 2022

Uh oh!

pytorchmergebot commented Jun 22, 2022

Uh oh!

github-actions bot commented Aug 21, 2022

Uh oh!

bdhirsh commented Aug 23, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

bdhirsh commented Jun 13, 2022 • edited by pytorchmergebot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

facebook-github-bot commented Jun 13, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful links

❌ 2 New Failures

🕵️ 2 new failures recognized by patterns

pull / win-vs2019-cuda11.6-py3 / build (1/2)

pull / win-vs2019-cpu-py3 / build (2/2)

Uh oh!

ezyang left a comment

Choose a reason for hiding this comment

Uh oh!

bdhirsh commented Jun 13, 2022

Uh oh!

pytorchmergebot commented Jun 22, 2022

Uh oh!

github-actions bot commented Aug 21, 2022

Uh oh!

bdhirsh commented Aug 23, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

bdhirsh commented Jun 13, 2022 •

edited by pytorchmergebot

Loading

facebook-github-bot commented Jun 13, 2022 •

edited

Loading