[JIT] Cache compilation of free functions #30503
Closed
Conversation
This file contains hidden or 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
facebook-github-bot
added
the
oncall: jit
suo
previously requested changes
Nov 27, 2019
There was a problem hiding this comment.
Good change, but we need one additional thing to make it safe I think. Compiling a redefined function will lead to weird behavior:
def foo():
# something
fn1 = torch.jit.script(foo)
def foo():
# something else
fn2 = torch.jit.script(foo)
In the current implementation, fn2 will be the same as fn1. I think the fix is simple; you can use the fn pyobject ptr as the key into a compilation cache maintained from the python side.
|
Can you also add a test to verify that we the corner case right? |
Good point, and yea change sounds good. We could consider doing something like this for all objects which are script-convertible. This would also fix: #30421 The change for modules is a little more complicated because you have to check that attributes haven't changed. |
|
actually hold off on review i need to fix something edit: fixed |
suo
approved these changes
Nov 28, 2019
eellison
pushed a commit
that referenced
this pull request
Dec 2, 2019
Run Constant Propagation upon compilation only on ops with non-aliasing inputs and outputs. This speeds up the first run of `torchvision.models.resnet18` by over 50% and speeds up compilation by about 25% (although the effects didn't seem additive with with #30503, so I'm going to land this PR first and then see if caching still has a sizable impact). Running constant prop only with non-aliasing types does a lot of graph cleanup by removing constant ifs and a bunch of other smaller ops. It also avoids all the jitter problems we had when we tried running full constant prop previously. Bc it is idempotent it doesn't jitter, and it doesn't jitter graphs constructed from tracing because tracing doesn't emit any ops that only involve non-aliasing inputs. Full constant prop isn't idempotent because what ops are run depends on the state of mutation in alias db, which will often change upon successive iterations of constant propagation, and bc it affects graphs constructed from tracing. [ghstack-poisoned]
eellison
pushed a commit
that referenced
this pull request
Dec 2, 2019
Run Constant Propagation upon compilation only on ops with non-aliasing inputs and outputs. This speeds up the first run of `torchvision.models.resnet18` by over 50% and speeds up compilation by about 25% (although the effects didn't seem additive with with #30503, so I'm going to land this PR first and then see if caching still has a sizable impact). Running constant prop only with non-aliasing types does a lot of graph cleanup by removing constant ifs and a bunch of other smaller ops. It also avoids all the jitter problems we had when we tried running full constant prop previously. Bc it is idempotent it doesn't jitter, and it doesn't jitter graphs constructed from tracing because tracing doesn't emit any ops that only involve non-aliasing inputs. Full constant prop isn't idempotent because what ops are run depends on the state of mutation in alias db, which will often change upon successive iterations of constant propagation, and bc it affects graphs constructed from tracing. [ghstack-poisoned]
added 2 commits
December 3, 2019 12:59
There was a problem hiding this comment.
@eellison has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
added 2 commits
December 3, 2019 18:30
There was a problem hiding this comment.
@eellison is landing this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
eellison
pushed a commit
that referenced
this pull request
Dec 4, 2019
Run Constant Propagation upon compilation only on ops with non-aliasing inputs and outputs. This speeds up the first run of `torchvision.models.resnet18` by over 50% and speeds up compilation by about 25% (although the effects didn't seem additive with with #30503, so I'm going to land this PR first and then see if caching still has a sizable impact). Running constant prop only with non-aliasing types does a lot of graph cleanup by removing constant ifs and a bunch of other smaller ops. It also avoids all the jitter problems we had when we tried running full constant prop previously. Bc it is idempotent it doesn't jitter, and it doesn't jitter graphs constructed from tracing because tracing doesn't emit any ops that only involve non-aliasing inputs. Full constant prop isn't idempotent because what ops are run depends on the state of mutation in alias db, which will often change upon successive iterations of constant propagation, and bc it affects graphs constructed from tracing. Edit: if we were okay with running constant propagation on graphs constructed from tracing (potentially making them hard to debug), an alternative would be to run constant propagation until the graph reaches a fixed point. [ghstack-poisoned]
eellison
pushed a commit
that referenced
this pull request
Dec 4, 2019
Run Constant Propagation upon compilation only on ops with non-aliasing inputs and outputs. This speeds up the first run of `torchvision.models.resnet18` by over 50% and speeds up compilation by about 25% (although the effects didn't seem additive with with #30503, so I'm going to land this PR first and then see if caching still has a sizable impact). Running constant prop only with non-aliasing types does a lot of graph cleanup by removing constant ifs and a bunch of other smaller ops. It also avoids all the jitter problems we had when we tried running full constant prop previously. Bc it is idempotent it doesn't jitter, and it doesn't jitter graphs constructed from tracing because tracing doesn't emit any ops that only involve non-aliasing inputs. Full constant prop isn't idempotent because what ops are run depends on the state of mutation in alias db, which will often change upon successive iterations of constant propagation, and bc it affects graphs constructed from tracing. Edit: if we were okay with running constant propagation on graphs constructed from tracing (potentially making them hard to debug), an alternative would be to run constant propagation until the graph reaches a fixed point. [ghstack-poisoned]
eellison
pushed a commit
that referenced
this pull request
Dec 5, 2019
Run Constant Propagation upon compilation only on ops with non-aliasing inputs and outputs. This speeds up the first run of `torchvision.models.resnet18` by over 50% and speeds up compilation by about 25% (although the effects didn't seem additive with with #30503, so I'm going to land this PR first and then see if caching still has a sizable impact). Running constant prop only with non-aliasing types does a lot of graph cleanup by removing constant ifs and a bunch of other smaller ops. It also avoids all the jitter problems we had when we tried running full constant prop previously. Bc it is idempotent it doesn't jitter, and it doesn't jitter graphs constructed from tracing because tracing doesn't emit any ops that only involve non-aliasing inputs. Full constant prop isn't idempotent because what ops are run depends on the state of mutation in alias db, which will often change upon successive iterations of constant propagation, and bc it affects graphs constructed from tracing. Edit: if we were okay with running constant propagation on graphs constructed from tracing (potentially making them hard to debug), an alternative would be to run constant propagation until the graph reaches a fixed point. [ghstack-poisoned]
facebook-github-bot
pushed a commit
that referenced
this pull request
Dec 9, 2019
Summary: Pull Request resolved: #30544 Run Constant Propagation upon compilation only on ops with non-aliasing inputs and outputs. This speeds up the first run of `torchvision.models.resnet18` by over 50% and speeds up compilation by about 25% (although the effects didn't seem additive with with #30503, so I'm going to land this PR first and then see if caching still has a sizable impact). Running constant prop only with non-aliasing types does a lot of graph cleanup by removing constant ifs and a bunch of other smaller ops. It also avoids all the jitter problems we had when we tried running full constant prop previously. Bc it is idempotent it doesn't jitter, and it doesn't jitter graphs constructed from tracing because tracing doesn't emit any ops that only involve non-aliasing inputs. Full constant prop isn't idempotent because what ops are run depends on the state of mutation in alias db, which will often change upon successive iterations of constant propagation, and bc it affects graphs constructed from tracing. Edit: if we were okay with running constant propagation on graphs constructed from tracing (potentially making them hard to debug), an alternative would be to run constant propagation until the graph reaches a fixed point. Test Plan: Imported from OSS Differential Revision: D18833607 Pulled By: eellison fbshipit-source-id: 92a0adb4882d67ed5a0db5c279f5e122aeeba54a
wuhuikx
pushed a commit
to wuhuikx/pytorch
that referenced
this pull request
Jan 30, 2020
Summary: We don't have to recompile free functions if we've already compiled them. Improved compilation of resnet18 by 27%. Pull Request resolved: pytorch#30503 Differential Revision: D18796501 Pulled By: eellison fbshipit-source-id: 2dee0fc5fcf9adc5b92213f8cb813730d71b376f
wuhuikx
pushed a commit
to wuhuikx/pytorch
that referenced
this pull request
Jan 30, 2020
Summary: Pull Request resolved: pytorch#30544 Run Constant Propagation upon compilation only on ops with non-aliasing inputs and outputs. This speeds up the first run of `torchvision.models.resnet18` by over 50% and speeds up compilation by about 25% (although the effects didn't seem additive with with pytorch#30503, so I'm going to land this PR first and then see if caching still has a sizable impact). Running constant prop only with non-aliasing types does a lot of graph cleanup by removing constant ifs and a bunch of other smaller ops. It also avoids all the jitter problems we had when we tried running full constant prop previously. Bc it is idempotent it doesn't jitter, and it doesn't jitter graphs constructed from tracing because tracing doesn't emit any ops that only involve non-aliasing inputs. Full constant prop isn't idempotent because what ops are run depends on the state of mutation in alias db, which will often change upon successive iterations of constant propagation, and bc it affects graphs constructed from tracing. Edit: if we were okay with running constant propagation on graphs constructed from tracing (potentially making them hard to debug), an alternative would be to run constant propagation until the graph reaches a fixed point. Test Plan: Imported from OSS Differential Revision: D18833607 Pulled By: eellison fbshipit-source-id: 92a0adb4882d67ed5a0db5c279f5e122aeeba54a
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
We don't have to recompile free functions if we've already compiled them.
Improved compilation of resnet18 by 27%.