Skip to content

Stop immediately specializing common constants 0/1 for plain int #128327

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
ezyang wants to merge 11 commits into from
Closed

Stop immediately specializing common constants 0/1 for plain int #128327

ezyang wants to merge 11 commits into from

Conversation

@ezyang
Copy link
Contributor

@ezyang ezyang commented Jun 10, 2024 •
edited
Loading

@ezyang
Update
7673384 
[ghstack-poisoned]
@pytorch-bot
Copy link

pytorch-bot bot commented Jun 10, 2024 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/128327

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit c3903ed with merge base 1e27af3 (image):
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

ezyang added a commit that referenced this pull request Jun 10, 2024
@ezyang
Stop immediately specializing common constants 0/1 for plain int
b59e0b2 
Fixes #128319

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

ghstack-source-id: f1ac898
Pull Request resolved: #128327
@albanD albanD removed their request for review June 10, 2024 14:57
@ezyang ezyang requested a review from anijain2305 June 10, 2024 15:30
@ezyang
Copy link
Contributor Author

ezyang commented Jun 10, 2024

oof dis CI

@ezyang
Copy link
Contributor Author

ezyang commented Jun 10, 2024

It looks like this is just grinding recompilation counts though, it seems

@ezyang
Copy link
Contributor Author

ezyang commented Jun 11, 2024

@anijain2305 will finish this off. (Animesh, feel free to open a new PR for yourself)

@ezyang
Update
23ff891 
[ghstack-poisoned]
ezyang added a commit that referenced this pull request Jun 13, 2024
@ezyang
Stop immediately specializing common constants 0/1 for plain int
cf7b027 
Fixes #128319

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

ghstack-source-id: 0e0e224
Pull Request resolved: #128327
@ezyang
Update
739afbd 
[ghstack-poisoned]
ezyang added a commit that referenced this pull request Jun 13, 2024
@ezyang
Stop immediately specializing common constants 0/1 for plain int
ff5dea7 
Fixes #128319

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

ghstack-source-id: e5c6c47
Pull Request resolved: #128327
@ezyang ezyang mentioned this pull request Jul 1, 2024
Open
@ezyang
Update
f98ef2f 
[ghstack-poisoned]
@pytorch-bot pytorch-bot bot added the release notes: linalg_frontend release notes category label Jul 1, 2024
@ezyang
Update
cc0fa3b 
[ghstack-poisoned]
@ezyang ezyang mentioned this pull request Jul 1, 2024
Closed
ezyang added a commit that referenced this pull request Jul 1, 2024
@ezyang
Stop immediately specializing common constants 0/1 for plain int
da60c74 
Fixes #128319

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

ghstack-source-id: ca9cd4e
Pull Request resolved: #128327
@ezyang
Update
c449020 
[ghstack-poisoned]
@ezyang
Copy link
Contributor Author

ezyang commented Jul 2, 2024

Do we have any place where we could document this behaviour?

I don't. This is just so incredibly inside baseball, I'm not sure what someone is going to do with this information. In practice, you figure this out by looking at the recompiles logs, I think...

@ezyang ezyang added the topic: not user facing topic category label Jul 2, 2024
@ezyang
Copy link
Contributor Author

ezyang commented Jul 2, 2024

To make rollout of this safer, I'm going to split this into bugfixes and then the actually policy change

@lezcano
Copy link
Collaborator

lezcano commented Jul 2, 2024

That being said, putting it in https://pytorch.org/docs/main/torch.compiler_dynamo_deepdive.html#are-always-specialized would be nice.

@ezyang ezyang mentioned this pull request Jul 2, 2024
Closed
@ezyang
Update
b6a7c53 
[ghstack-poisoned]
@ezyang
Copy link
Contributor Author

ezyang commented Jul 2, 2024

Done

@ezyang
Update
b4826b8 
[ghstack-poisoned]
ezyang added a commit that referenced this pull request Jul 2, 2024
@ezyang
Stop immediately specializing common constants 0/1 for plain int
d36b83b 
Fixes #128319

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

ghstack-source-id: c431efc
Pull Request resolved: #128327
@ezyang
Update
ac80fda 
[ghstack-poisoned]
@ezyang ezyang mentioned this pull request Jul 2, 2024
Closed
@ezyang
Update
348a0b1 
[ghstack-poisoned]
pytorchmergebot pushed a commit that referenced this pull request Jul 2, 2024
@ezyang @pytorchmergebot
Bug fixes for disabling 0/1 specialization on plain int (#129961)
d7680a5 
These bug fixes will be exercised in
#128327 but I separate them from
the actual policy change (which is more risky)

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

Pull Request resolved: #129961
Approved by: https://github.com/lezcano
@ezyang
Update
c3903ed 
[ghstack-poisoned]
ezyang added a commit that referenced this pull request Jul 3, 2024
@ezyang
Stop immediately specializing common constants 0/1 for plain int
f067f17 
Fixes #128319

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

ghstack-source-id: 1d47cce
Pull Request resolved: #128327
@ezyang ezyang added the ciflow/trunk Trigger trunk jobs on your pull request label Jul 3, 2024
pytorchmergebot pushed a commit that referenced this pull request Jul 3, 2024
@ezyang @pytorchmergebot
Make sparse empty constructors specialize instead of fail on symbolic…
64a04d2 
… inputs (#129983)

Exercised in #128327

Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: #129983
Approved by: https://github.com/anijain2305
@ezyang
Copy link
Contributor Author

ezyang commented Jul 3, 2024

@pytorchbot merge

@pytorchmergebot
Copy link
Collaborator

pytorchmergebot commented Jul 3, 2024

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging
Check the merge workflow status
here

@github-actions github-actions bot deleted the gh/ezyang/2804/head branch August 4, 2024 02:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@lezcano lezcano lezcano approved these changes

@antoniojkim antoniojkim Awaiting requested review from antoniojkim

@bdhirsh bdhirsh Awaiting requested review from bdhirsh

@miladm miladm Awaiting requested review from miladm

@SherlockNoMad SherlockNoMad Awaiting requested review from SherlockNoMad

@anijain2305 anijain2305 Awaiting requested review from anijain2305

@nikitaved nikitaved Awaiting requested review from nikitaved

@IvanYashchuk IvanYashchuk Awaiting requested review from IvanYashchuk

Assignees

No one assigned

Labels

ciflow/inductor ciflow/trunk Trigger trunk jobs on your pull request Merged module: dynamo module: inductor release notes: linalg_frontend release notes category topic: not user facing topic category

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@ezyang @lezcano @pytorchmergebot