Use fabi-version=11 to ensure compatibility between gcc7 and gcc9 binaries by atalman · Pull Request #81058 · pytorch/pytorch

atalman · 2022-07-07T20:28:52Z

Test using cuda 11.3 manywheel binary:

import torch
print(torch.__version__)
print(torch._C._PYBIND11_BUILD_ABI)

Output

1.13.0.dev20220707+cu113
_cxxabi1011

Functorch test torch : 1.13.0.dev20220707+cu113, functorch with cu102

import torch
print(torch.__version__)
print(torch._C._PYBIND11_BUILD_ABI)
from functorch import vmap
x = torch.randn(2, 3, 5)
vmap(lambda x: x, out_dims=3)(x)

Output

1.13.0.dev20220707+cu113
_cxxabi1011
/home/atalman/temp/testc1.py:5: UserWarning: Failed to initialize NumPy: No module named 'numpy' (Triggered internally at ../torch/csrc/utils/tensor_numpy.cpp:73.)
  x = torch.randn(2, 3, 5)
Traceback (most recent call last):
  File "/home/atalman/temp/testc1.py", line 6, in <module>
    vmap(lambda x: x, out_dims=3)(x)
  File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 361, in wrapped
    return _flat_vmap(
  File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 488, in _flat_vmap
    return _unwrap_batched(batched_outputs, out_dims, vmap_level, batch_size, func)
  File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 165, in _unwrap_batched
    flat_outputs = [
  File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 166, in <listcomp>
    _remove_batch_dim(batched_output, vmap_level, batch_size, out_dim)
IndexError: Dimension out of range (expected to be in range of [-3, 2], but got 3)

Related Builder PR: pytorch/builder#1083

Test PR: #81232

facebook-github-bot · 2022-07-07T20:29:03Z

🔗 Helpful links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/81058
📄 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

✅ No Failures (0 Pending)

As of commit b03cacd (more details on the Dr. CI page):

Expand to see more

💚 💚 Looks good so far! There are no failures yet. 💚 💚

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.

malfet

This is a wrong location to put this constraint, as it has nothing to do with generic PyTorch builds, but specific to how we are packaging it for release

CMakeLists.txt

malfet

You need to add

if(${GLIBCXX_USE_CXX11_ABI} EQUAL 0)
 string(APPEND CMAKE_CXX_FLAGS " -fabi-version=11")
endif()

to

pytorch/caffe2/CMakeLists.txt

Line 1227 in 28776c4

message(STATUS "Determined _GLIBCXX_USE_CXX11_ABI=${GLIBCXX_USE_CXX11_ABI}")

atalman · 2022-07-11T21:59:01Z

You need to add
if(${GLIBCXX_USE_CXX11_ABI} EQUAL 0)
 string(APPEND CMAKE_CXX_FLAGS " -fabi-version=11")
endif()
to

pytorch/caffe2/CMakeLists.txt

Line 1227 in 28776c4

message(STATUS "Determined _GLIBCXX_USE_CXX11_ABI=${GLIBCXX_USE_CXX11_ABI}")

in 28776c4

As discussed will not be doing this change in this PR, but refactor in pytorch/builder#1084

atalman · 2022-07-12T13:10:39Z

@pytorchbot merge

pytorchmergebot · 2022-07-12T13:12:02Z

@pytorchbot successfully started a merge job. Check the current status here

pytorchmergebot · 2022-07-12T13:12:09Z

Merge failed due to Refusing to merge as mandatory check(s) pull failed for rule superuser
Raised by https://github.com/pytorch/pytorch/actions/runs/2656739238

atalman · 2022-07-12T14:20:42Z

@pytorchbot merge

pytorchmergebot · 2022-07-12T14:21:59Z

@pytorchbot successfully started a merge job. Check the current status here

atalman · 2022-07-12T17:55:07Z

@pytorchbot merge

pytorchmergebot · 2022-07-12T17:56:27Z

@pytorchbot successfully started a merge job. Check the current status here

github-actions · 2022-07-12T17:57:09Z

Hey @atalman.
You've committed this PR, but it does not have both a 'release notes: ...' and 'topics: ...' label. Please add one of each to the PR. The 'release notes: ...' label should represent the part of PyTorch that this PR changes (fx, autograd, distributed, etc) and the 'topics: ...' label should represent the kind of PR it is (not user facing, new feature, bug fix, perf improvement, etc). The list of valid labels can be found here for the 'release notes: ...' and here for the 'topics: ...'.
For changes that are 'topic: not user facing' there is no need for a release notes label.

JackCaoG · 2022-07-13T17:15:39Z

@malfet can we revert this one? It breaks both ci and internal build for pytorch/xla. We are trying to figure out how to build with this option. @yeounoh FYI

…aries (#81058) (#81058) Summary: Fixes: #80489 Test using cuda 11.3 manywheel binary: ``` import torch print(torch.__version__) print(torch._C._PYBIND11 (d55b25a633b7e2e6122becf6dbdf0528df6e8b13)_BUILD_ABI) ```` Output ``` 1.13.0.dev20220707+cu113 _cxxabi1011 ``` Functorch test torch : 1.13.0.dev20220707+cu113, functorch with cu102 ``` import torch print(torch.__version__) print(torch._C._PYBIND11 (d55b25a633b7e2e6122becf6dbdf0528df6e8b13)_BUILD_ABI) from functorch import vmap x = torch.randn(2, 3, 5) vmap(lambda x: x, out_dims=3)(x) ``` Output ``` 1.13.0.dev20220707+cu113 _cxxabi1011 /home/atalman/temp/testc1.py:5: UserWarning: Failed to initialize NumPy: No module named 'numpy' (Triggered internally at ../torch/csrc/utils/tensor_numpy.cpp:73.) x = torch.randn(2, 3, 5) Traceback (most recent call last): File "/home/atalman/temp/testc1.py", line 6, in <module> vmap(lambda x: x, out_dims=3)(x) File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 361, in wrapped return _flat_vmap( File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 488, in _flat_vmap return _unwrap_batched(batched_outputs, out_dims, vmap_level, batch_size, func) File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 165, in _unwrap_batched flat_outputs = [ File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 166, in <listcomp> _remove_batch_dim(batched_output, vmap_level, batch_size, out_dim) IndexError: Dimension out of range (expected to be in range of [-3, 2], but got 3) ``` Related Builder PR: pytorch/builder#1083 Test PR: #81232 Pull Request resolved: #81058 Approved by: https://github.com/zou3519, https://github.com/malfet Test Plan: contbuild & OSS CI, see https://hud.pytorch.org/commit/pytorch/pytorch/d552ba3b4f53da9b6a5f6e0463111e43b367ef8a Reviewed By: DanilBaibak Differential Revision: D37813240 Pulled By: atalman fbshipit-source-id: 94d94e777b0e9d5da106173c06117b3019ba71c4

JackCaoG · 2022-07-13T17:56:29Z

OK CI is actually OK, it is our internal build failed. I will try to look into it this afternoon.

JackCaoG · 2022-07-14T01:05:51Z

this seems to break our setup when using gcc8. I run into errors like

Building wheel torch-1.13.0a0+gitc657c3d
-- Building version 1.13.0a0+gitc657c3d
cmake -DBUILD_PYTHON=True -DBUILD_TEST=True -DCMAKE_BUILD_TYPE=Debug -DCMAKE_INSTALL_PREFIX=/pytorch/torch -DCMAKE_PREFIX_PATH=/root/anaconda3/envs/pytorch/lib/python3.7/site-packages -DGLIBCXX_USE_CXX11_ABI=0 -DNUMPY_INCLUDE_DIR=/root/anaconda3/envs/pytorch/lib/python3.7/site-packages/numpy/core/include -DPYTHON_EXECUTABLE=/root/anaconda3/envs/pytorch/bin/python -DPYTHON_INCLUDE_DIR=/root/anaconda3/envs/pytorch/include/python3.7m -DPYTHON_LIBRARY=/root/anaconda3/envs/pytorch/lib/libpython3.7m.so.1.0 -DTORCH_BUILD_VERSION=1.13.0a0+gitc657c3d -DUSE_CUDA=0 -DUSE_NUMPY=True /pytorch
-- The CXX compiler identification is Clang 8.0.1
-- The C compiler identification is Clang 8.0.1
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/clang++-8 - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/clang-8 - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Not forcing any particular BLAS to be found
-- Could not find ccache. Consider installing ccache to speed up compilation.
-- Performing Test COMPILER_WORKS
-- Performing Test COMPILER_WORKS - Success
-- Performing Test SUPPORT_GLIBCXX_USE_C99
-- Performing Test SUPPORT_GLIBCXX_USE_C99 - Failed
CMake Error at cmake/MiscCheck.cmake:63 (message):
  The C++ compiler does not support required functions.  This is very likely
  due to a known bug in GCC 5 (and maybe other versions) on Ubuntu 17.10 and
  newer.  For more information, see:
  https://github.com/pytorch/pytorch/issues/5229
Call Stack (most recent call first):
  CMakeLists.txt:695 (include)

(pytorch) root@803de85a4f7e:/# gcc --version
gcc (Debian 8.3.0-6) 8.3.0

any insight?

JackCaoG · 2022-07-14T03:22:43Z

I guess pytorch/xla would just always enable GLIBCXX_USE_CXX11_ABI which seems to solve the issue. We have been doing this for the release wheel and ci anyway

…aries (pytorch#81058) (pytorch#81058) Summary: Fixes: pytorch#80489 Test using cuda 11.3 manywheel binary: ``` import torch print(torch.__version__) print(torch._C._PYBIND11 (pytorch@d55b25a633b7e2e6122becf6dbdf0528df6e8b13)_BUILD_ABI) ```` Output ``` 1.13.0.dev20220707+cu113 _cxxabi1011 ``` Functorch test torch : 1.13.0.dev20220707+cu113, functorch with cu102 ``` import torch print(torch.__version__) print(torch._C._PYBIND11 (pytorch@d55b25a633b7e2e6122becf6dbdf0528df6e8b13)_BUILD_ABI) from functorch import vmap x = torch.randn(2, 3, 5) vmap(lambda x: x, out_dims=3)(x) ``` Output ``` 1.13.0.dev20220707+cu113 _cxxabi1011 /home/atalman/temp/testc1.py:5: UserWarning: Failed to initialize NumPy: No module named 'numpy' (Triggered internally at ../torch/csrc/utils/tensor_numpy.cpp:73.) x = torch.randn(2, 3, 5) Traceback (most recent call last): File "/home/atalman/temp/testc1.py", line 6, in <module> vmap(lambda x: x, out_dims=3)(x) File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 361, in wrapped return _flat_vmap( File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 488, in _flat_vmap return _unwrap_batched(batched_outputs, out_dims, vmap_level, batch_size, func) File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 165, in _unwrap_batched flat_outputs = [ File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 166, in <listcomp> _remove_batch_dim(batched_output, vmap_level, batch_size, out_dim) IndexError: Dimension out of range (expected to be in range of [-3, 2], but got 3) ``` Related Builder PR: pytorch/builder#1083 Test PR: pytorch#81232 Pull Request resolved: pytorch#81058 Approved by: https://github.com/zou3519, https://github.com/malfet Test Plan: contbuild & OSS CI, see https://hud.pytorch.org/commit/pytorch/pytorch/d552ba3b4f53da9b6a5f6e0463111e43b367ef8a Reviewed By: DanilBaibak Differential Revision: D37813240 Pulled By: atalman fbshipit-source-id: 94d94e777b0e9d5da106173c06117b3019ba71c4

…aries (#81058) (#81058) (#81884) Summary: Fixes: #80489 Test using cuda 11.3 manywheel binary: ``` import torch print(torch.__version__) print(torch._C._PYBIND11 (d55b25a633b7e2e6122becf6dbdf0528df6e8b13)_BUILD_ABI) ```` Output ``` 1.13.0.dev20220707+cu113 _cxxabi1011 ``` Functorch test torch : 1.13.0.dev20220707+cu113, functorch with cu102 ``` import torch print(torch.__version__) print(torch._C._PYBIND11 (d55b25a633b7e2e6122becf6dbdf0528df6e8b13)_BUILD_ABI) from functorch import vmap x = torch.randn(2, 3, 5) vmap(lambda x: x, out_dims=3)(x) ``` Output ``` 1.13.0.dev20220707+cu113 _cxxabi1011 /home/atalman/temp/testc1.py:5: UserWarning: Failed to initialize NumPy: No module named 'numpy' (Triggered internally at ../torch/csrc/utils/tensor_numpy.cpp:73.) x = torch.randn(2, 3, 5) Traceback (most recent call last): File "/home/atalman/temp/testc1.py", line 6, in <module> vmap(lambda x: x, out_dims=3)(x) File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 361, in wrapped return _flat_vmap( File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 488, in _flat_vmap return _unwrap_batched(batched_outputs, out_dims, vmap_level, batch_size, func) File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 165, in _unwrap_batched flat_outputs = [ File "/home/atalman/conda/lib/python3.9/site-packages/functorch/_src/vmap.py", line 166, in <listcomp> _remove_batch_dim(batched_output, vmap_level, batch_size, out_dim) IndexError: Dimension out of range (expected to be in range of [-3, 2], but got 3) ``` Related Builder PR: pytorch/builder#1083 Test PR: #81232 Pull Request resolved: #81058 Approved by: https://github.com/zou3519, https://github.com/malfet Test Plan: contbuild & OSS CI, see https://hud.pytorch.org/commit/pytorch/pytorch/d552ba3b4f53da9b6a5f6e0463111e43b367ef8a Reviewed By: DanilBaibak Differential Revision: D37813240 Pulled By: atalman fbshipit-source-id: 94d94e777b0e9d5da106173c06117b3019ba71c4

#4137) * Added pytorch patch file to revert pytorch/pytorch#81058 * updated the diff file.

pytorch/pytorch#81058

facebook-github-bot added the cla signed label Jul 7, 2022

atalman added ciflow/binaries_wheel Trigger binary build and upload jobs for wheel on the PR ciflow/binaries_conda labels Jul 7, 2022

atalman changed the title ~~[WIP] Testing use fabi-version=11~~ Use fabi-version=11 to ensure compatibility between gcc7 and gcc9 binaries Jul 8, 2022

zou3519 approved these changes Jul 8, 2022

View reviewed changes

malfet requested changes Jul 8, 2022

View reviewed changes

malfet reviewed Jul 8, 2022

View reviewed changes

CMakeLists.txt Outdated Show resolved Hide resolved

atalman mentioned this pull request Jul 11, 2022

Use _GLIBCXX_USE_CXX11_ABI flag in conda builds, Add _PYBIND11_BUILD_ABI test for compatibility gcc7 vs gcc9, pytorch/builder#1083

Merged

malfet approved these changes Jul 11, 2022

View reviewed changes

CMakeLists.txt Outdated Show resolved Hide resolved

malfet reviewed Jul 11, 2022

View reviewed changes

atalman mentioned this pull request Jul 11, 2022

Refactor code around usage of GLIBCXX_USE_CXX11_ABI flag pytorch/builder#1084

Open

atalman added 5 commits July 12, 2022 07:03

Testing use fabi-version=11

76bbe2d

Temporary setting only Linux flag

4298fa6

Move fabi option to already exiting if case

60499ba

Move compatibility flag under GLIBCXX_USE_CXX11_ABI condition

82a5664

Move fabi flag to use case where GLIBCXX_USE_CXX11_ABI=0

f88c7bb

atalman force-pushed the testing_fabi_11 branch from 774d9f1 to f88c7bb Compare July 12, 2022 14:05

Fix wording

b03cacd

pytorchmergebot added the Merged label Jul 12, 2022

pytorchmergebot closed this in d552ba3 Jul 12, 2022

JackCaoG mentioned this pull request Jul 14, 2022

Enable cxx abi by default pytorch/xla#3714

Merged

atalman mentioned this pull request Jul 21, 2022

Use fabi-version=11 to ensure compatibility between gcc7 and gcc9 bin… #81884

Merged

vanbasten23 added a commit to pytorch/xla that referenced this pull request Oct 27, 2022

Added pytorch patch file to revert pytorch/pytorch#81058

b085a91

vanbasten23 mentioned this pull request Oct 27, 2022

Added pytorch patch file to revert https://github.com/pytorch/pytorch… pytorch/xla#4137

Merged

vanbasten23 added a commit to pytorch/xla that referenced this pull request Oct 28, 2022

Added pytorch patch file to revert https://github.com/pytorch/pytorch… (

dffb915

#4137) * Added pytorch patch file to revert pytorch/pytorch#81058 * updated the diff file.

vanbasten23 mentioned this pull request Nov 1, 2022

PyTorch/XLA should disable cxx_abi during release. pytorch/xla#4146

Closed

steventk-g added a commit to pytorch/xla that referenced this pull request Feb 20, 2023

Added pytorch patch file to revert PR #81058

86ddbe4

pytorch/pytorch#81058

steventk-g added a commit to pytorch/xla that referenced this pull request Feb 20, 2023

Add pytorch patch file to revert pytorch/pytorch#81058

768d7e1

vanbasten23 mentioned this pull request Jul 21, 2023

Disable cxx_abi when building PyTorch/XLA in r2.0. pytorch/xla#5332

Merged

Conversation

atalman commented Jul 7, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

facebook-github-bot commented Jul 7, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful links

✅ No Failures (0 Pending)

Uh oh!

malfet left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

malfet left a comment

Choose a reason for hiding this comment

Uh oh!

atalman commented Jul 11, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

atalman commented Jul 12, 2022

Uh oh!

pytorchmergebot commented Jul 12, 2022

Uh oh!

pytorchmergebot commented Jul 12, 2022

Uh oh!

atalman commented Jul 12, 2022

Uh oh!

pytorchmergebot commented Jul 12, 2022

Uh oh!

atalman commented Jul 12, 2022

Uh oh!

pytorchmergebot commented Jul 12, 2022

Uh oh!

github-actions bot commented Jul 12, 2022

Uh oh!

JackCaoG commented Jul 13, 2022

Uh oh!

JackCaoG commented Jul 13, 2022

Uh oh!

JackCaoG commented Jul 14, 2022

Uh oh!

JackCaoG commented Jul 14, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

atalman commented Jul 7, 2022 •

edited

Loading

facebook-github-bot commented Jul 7, 2022 •

edited

Loading

atalman commented Jul 11, 2022 •

edited

Loading