Fix libtorch static builds that regressed #1056
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Just today I wondered about this. Thank you! |
|
@seemethere @ezyang can you please help review or send to the right person ? |
|
@malfet can you take a look at this? |
malfet
approved these changes
Jun 10, 2022
malfet
previously requested changes
Jun 10, 2022
|
Also, we probably want to keep debug symbol information, but only for shared library builds (as static ones indeed can not be separated into library and debug symbols) |
libtorch-static used to ship libtorch.a until 1.0.1 and has regressed since forcing people to build from source. This fixes the build commands and add BUILD_SHARED_LIBS=OFF It turns off USE_TENSORPIPE until pytorch/tensorpipe#449 is fixed. Tested on my system to verify a libtorch.a is produced but can't test the CI
|
Friendly ping to review. Thanks thanks |
ezyang
approved these changes
Jun 14, 2022
|
thanks @ezyang |
|
Reverting as it broke all static builds in nightlies, see https://hud.pytorch.org/pytorch/pytorch/commit/a4fa62064e1863d108c655f541bce4d806cf1823 |
malfet
added a commit
that referenced
this pull request
Jun 16, 2022
|
here's the error (string interpolation in ninja?) |
atalman
added a commit
to atalman/builder
that referenced
this pull request
Jun 17, 2022
This reverts commit 17bfd84.
|
Any ideas on what went wrong here ? The same command works ok on couple systems i tried. Is the build.ninja file available somewhere ? |
|
I have isolated the breakage to USE_MKL / USE_MKLDNN generating wrong Ninja rules like I have filed pytorch/pytorch#80012 to fix the issue in PyTorch but will land a fix for static builds with MKL disabled for now until it is fixed. |
powderluv
added a commit
to powderluv/builder
that referenced
this pull request
Jun 22, 2022
Prior attempt to land was reverted due to a failure with MKLDNN pytorch#1056 Disable MKLDNN in static builds until it is fixed. It is tracked in pytorch/pytorch#80012 TEST: With and without MKLDNN to recreate the last failure and test that it builds without MKLDNN.
malfet
pushed a commit
that referenced
this pull request
Jun 22, 2022
Prior attempt to land was reverted due to a failure with MKLDNN #1056 Disable MKLDNN in static builds until it is fixed. It is tracked in pytorch/pytorch#80012 TEST: With and without MKLDNN to recreate the last failure and test that it builds without MKLDNN.
powderluv
added a commit
to nod-ai/libtorch
that referenced
this pull request
Jun 28, 2022
Prior attempt to land was reverted due to a failure with MKLDNN pytorch#1056 Disable MKLDNN in static builds until it is fixed. It is tracked in pytorch/pytorch#80012 TEST: With and without MKLDNN to recreate the last failure and test that it builds without MKLDNN.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
libtorch-static used to ship libtorch.a until 1.0.1 and has
regressed since forcing people to build from source.
This fixes the build commands and add BUILD_SHARED_LIBS=OFF
It also removes build it as a debug build, and turns off
USE_TENSORPIPE until pytorch/tensorpipe#449
is fixed.
Tested on my system to verify a libtorch.a is produced but can't
test the CI
Fixes: pytorch/pytorch#70898