8000 [Inductor] Construct subgraph with benchmarking args not example_inputs (#153667) by PaulZhang12 · Pull Request #153752 · pytorch/pytorch · GitHub
[go: up one dir, main page]
Skip to content

[Inductor] Construct subgraph with benchmarking args not example_inputs (#153667) #153752

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
Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[Inductor] Construct subgraph with benchmarking args not example_inputs (#153667) #153752

wants to merge 1 commit into from
+121 −43

Conversation

PaulZhang12
Copy link
Contributor
@PaulZhang12 PaulZhang12 commented May 16, 2025 •
edited by pytorch-bot bot
Loading

Summary:

If the inputs to a subgraph has FlexibleLayout, the subgraph does not currently freeze the layouts here. Therefore, the example_inputs generated might not be consistent in layout with the args based in for benchmarking

Test Plan:
`
M, N, K = (4, 128, 14240)
import torch
torch.set_default_device("cuda")

torch.compil
8000
e(mode='max-autotune-no-cudagraphs')
def foo(x, y):
    return (x + 1) @ y

inps = [torch.rand([M, K], device='cuda', dtype=torch.bfloat16), torch.rand([K, N], device='cuda', dtype=torch.bfloat16)]
foo(*inps)

`

To produce FlexibleLayout with the stride change. Added to test_subgraph_choice

Differential Revision: D74900879

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

@PaulZhang12 @facebook-github-bot
[Inductor] Construct subgraph with benchmarking args not example_inpu…
de91865 
…ts (pytorch#153667)

Summary:

If the inputs to a subgraph has FlexibleLayout, the subgraph does not currently freeze the layouts here. Therefore, the `example_inputs` generated might not be consistent in layout with the `args` based in for benchmarking

Test Plan:
`
  M, N, K = (4, 128, 14240)
    import torch
    torch.set_default_device("cuda")

    torch.compile(mode='max-autotune-no-cudagraphs')
    def foo(x, y):
        return (x + 1) @ y

    inps = [torch.rand([M, K], device='cuda', dtype=torch.bfloat16), torch.rand([K, N], device='cuda', dtype=torch.bfloat16)]
    foo(*inps)
`

To produce FlexibleLayout with the stride change. Added to `test_subgraph_choice`

Differential Revision: D74900879
@pytorch-bot PyTorch Bot
Copy link
pytorch-bot bot commented May 16, 2025 •
edited
Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/153752

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

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

❌ 1 New Failure, 10 Unrelated Failures

As of commit de91865 with merge base cf226cb (image):

NEW FAILURE - The following job has failed:

BROKEN TRUNK - The following jobs failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

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

@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D74900879

@PaulZhang12 PaulZhang12 added the topic: not user facing topic category label May 16, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
ciflow/inductor fb-exported module: inductor topic: not user facing topic category
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@PaulZhang12 @facebook-github-bot
0