Skip to content
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

[AOTI] Handle empty input args #114682

Closed
wants to merge 5 commits into from
Closed

[AOTI] Handle empty input args #114682

wants to merge 5 commits into from

Conversation

desertfire
Copy link
Contributor

@desertfire desertfire commented Nov 28, 2023 •
edited by pytorch-bot bot

Stack from ghstack (oldest at bottom):

Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler

@desertfire
[AOTI] Handle empty input args
5823076 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

[ghstack-poisoned]
@pytorch-bot PyTorch Bot
Copy link

pytorch-bot bot commented Nov 28, 2023 •
edited

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/114682

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

✅ You can merge normally! (1 Unrelated Failure)

As of commit 32208a3 with merge base 5cfda9b (image):

FLAKY - The following job failed but was likely due to flakiness present on trunk:

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

desertfire added a commit that referenced this pull request Nov 28, 2023
@desertfire
[AOTI] Handle empty input args
cf2c1da 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

ghstack-source-id: c625ba743a81e4021587c636b7da498b6dff3b63
Pull Request resolved: #114682
@desertfire
Copy link
Contributor Author

@pytorchbot rebase

@pytorchmergebot
Copy link
Collaborator

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

@desertfire
Update on "[AOTI] Handle empty input args"
71c77d5 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler

[ghstack-poisoned]
@pytorchmergebot
Copy link
Collaborator

Successfully rebased gh/desertfire/274/orig onto refs/remotes/origin/viable/strict, please pull locally before adding more changes (for example, via ghstack checkout https://github.com/pytorch/pytorch/pull/114682)

pytorchmergebot pushed a commit that referenced this pull request Dec 1, 2023
@desertfire
[AOTI] Handle empty input args
87d877a 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

ghstack-source-id: 9852d79694ea8185b45e71ae43e858ba7ac8bfe4
Pull Request resolved: #114682
@chenyang78
Copy link
Contributor

BTW, the failing test looks real. Please take a look. Thanks.

@desertfire
Copy link
Contributor Author

BTW, the failing test looks real. Please take a look. Thanks.

@eellison , test_empty_cpu_tensor fails because now we correctly recognize there are two devices in the graph, cuda and cpu, and thus this checking fails,

pytorch/torch/_inductor/compile_fx.py

Line 360 in 453d509

(set(compiled_graph.device_types) == {"cuda"}, "non-cuda device in graph"),
. I saw your #114144, which looks like a very special case. Was it meant to solve some specific problem?

@eellison
Copy link
Contributor

eellison commented Dec 5, 2023 •
edited

@desertfire this is a real case that we observed for internal workloads. if there is an empty, no-data cpu output tensor, we shouldn't skip cudagraphs bc of it

@desertfire
Update on "[AOTI] Handle empty input args"
1d55229 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler

[ghstack-poisoned]
desertfire added a commit that referenced this pull request Dec 5, 2023
@desertfire
[AOTI] Handle empty input args
133edf9 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

ghstack-source-id: 7aacfdc42f769246af55a6ab6659890eac108d7d
Pull Request resolved: #114682
@desertfire
Update on "[AOTI] Handle empty input args"
866bcc0 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler

[ghstack-poisoned]
desertfire added a commit that referenced this pull request Dec 5, 2023
@desertfire
[AOTI] Handle empty input args
304b865 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

ghstack-source-id: 769ed59ec8208e049c319f14809e066ad37ce1c3
Pull Request resolved: #114682
@desertfire
Update on "[AOTI] Handle empty input args"
32208a3 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler

[ghstack-poisoned]
desertfire added a commit that referenced this pull request Dec 5, 2023
@desertfire
[AOTI] Handle empty input args
8f8c082 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

ghstack-source-id: 6fcbc38ca0c4b323299eb415bbf3c81f5f76ce8d
Pull Request resolved: #114682
@desertfire desertfire added the topic: not user facing topic category label Dec 5, 2023
@desertfire
Copy link
Contributor Author

@pytorchbot merge

@pytorch-bot pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Dec 5, 2023
@pytorchmergebot
Copy link
Collaborator

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

@facebook-github-bot facebook-github-bot deleted the gh/desertfire/274/head branch December 8, 2023 15:29
hyperfraise pushed a commit to hyperfraise/pytorch that referenced this pull request Dec 21, 2023
@desertfire @hyperfraise
[AOTI] Handle empty input args (pytorch#114682)
d3faaed 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

Pull Request resolved: pytorch#114682
Approved by: https://github.com/chenyang78
hyperfraise pushed a commit to hyperfraise/pytorch that referenced this pull request Dec 21, 2023
@desertfire @hyperfraise
[AOTI] Handle empty input args (pytorch#114682)
ce2d45d 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

Pull Request resolved: pytorch#114682
Approved by: https://github.com/chenyang78
ZhiweiYan-96 pushed a commit to ZhiweiYan-96/pytorch that referenced this pull request Dec 22, 2023
@desertfire @ZhiweiYan-96
[AOTI] Handle empty input args (pytorch#114682)
6fa2329 
Summary: When the model takes no inputs, AOTInductor relies on checking weights to figure out which device to compile the model into. Currently recording buffer device type happens too late, and this PR fixes that.

Pull Request resolved: pytorch#114682
Approved by: https://github.com/chenyang78
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@chenyang78 chenyang78 chenyang78 approved these changes

@muchulee8 muchulee8 Awaiting requested review from muchulee8

Assignees
No one assigned
Labels
ciflow/inductor ciflow/trunk Trigger trunk jobs on your pull request Merged module: inductor topic: not user facing topic category
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet
4 participants
@desertfire @pytorchmergebot @chenyang78 @eellison