Removing custom RMSNorm support #9966

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

ThomasJannaud wants to merge 1 commit into pytorch:main from ThomasJannaud:export-D72484196

+0 −4

Contributor

ThomasJannaud commented Apr 8, 2025 •

edited

Loading

Summary:
In #9844, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

We are pretty much reverting this diff, in favor of the existing aten.rms_norm

Differential Revision: D72484196

ThomasJannaud requested a review from tarun292 as a code owner

April 8, 2025 18:26

pytorch-bot bot commented Apr 8, 2025 •

edited

Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/9966

📄 Preview Python docs built from this PR

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

❌ 1 New Failure, 9 Cancelled Jobs

As of commit 60fce8e with merge base 20c7047 ():

NEW FAILURE - The following job has failed:

pull / unittest-arm / linux-job (gh)
RuntimeError: Command docker exec -t f59f104aedcbcd0f365c67c20820020b6f510293ad0aa69fd87267f94c2e4956 /exec failed with exit code 1

CANCELLED JOBS - The following jobs were cancelled. Please retry:

pull / test-binary-size-linux / linux-job (gh)
##[error]The operation was canceled.
pull / test-llama-runner-linux (fp32, xnnpack+custom+quantize_kv, linux.2xlarge, executorch-ubuntu-22.04... / linux-job (gh)
pull / test-models-linux (add_mul, xnnpack-quantization-delegation, linux.2xlarge) / linux-job (gh)
##[error]The operation was canceled.
pull / test-models-linux (ic3, portable, linux.2xlarge) / linux-job (gh)
pull / test-models-linux (ic3, xnnpack-quantization-delegation, linux.2xlarge) / linux-job (gh)
##[error]The operation was canceled.
pull / test-models-linux (linear, xnnpack-quantization-delegation, linux.2xlarge) / linux-job (gh)
pull / test-models-linux (mobilebert, portable, linux.2xlarge) / linux-job (gh)
pull / test-models-linux (mv2, xnnpack-quantization-delegation, linux.2xlarge) / linux-job (gh)
pull / test-models-linux (resnet50, xnnpack-quantization-delegation, linux.2xlarge) / linux-job (gh)
##[error]The operation was canceled.

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

facebook-github-bot added the CLA Signed label

Contributor

facebook-github-bot commented Apr 8, 2025

This pull request was exported from Phabricator. Differential Revision: D72484196

facebook-github-bot added the fb-exported label

Contributor Author

ThomasJannaud commented Apr 8, 2025

@pytorchbot label "topic: not user facing"

pytorch-bot bot added the topic: not user facing label

skrtskrtfb approved these changes

View reviewed changes

ThomasJannaud added a commit to ThomasJannaud/executorch that referenced this pull request


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

e7b2cd1

…rch#9966)

Summary:

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from 99ba87d to e7b2cd1 Compare

April 8, 2025 20:10

Contributor

facebook-github-bot commented Apr 8, 2025

This pull request was exported from Phabricator. Differential Revision: D72484196

ThomasJannaud added a commit to ThomasJannaud/executorch that referenced this pull request


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

5b53f8c

…rch#9966)

Summary:

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from e7b2cd1 to 5b53f8c Compare

April 8, 2025 20:43

Contributor

facebook-github-bot commented Apr 8, 2025

This pull request was exported from Phabricator. Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from 5b53f8c to 03b33c8 Compare

April 8, 2025 23:08

ThomasJannaud added a commit to ThomasJannaud/executorch that referenced this pull request


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

03b33c8

…rch#9966)

Summary:

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Differential Revision: D72484196

Contributor

facebook-github-bot commented Apr 8, 2025

This pull request was exported from Phabricator. Differential Revision: D72484196

ThomasJannaud added a commit to ThomasJannaud/executorch that referenced this pull request


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

b69c5f9

…rch#9966)

Summary:

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from 03b33c8 to b69c5f9 Compare

April 8, 2025 23:31

Contributor

facebook-github-bot commented Apr 8, 2025

This pull request was exported from Phabricator. Differential Revision: D72484196

ThomasJannaud added a commit to ThomasJannaud/executorch that referenced this pull request


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

1cbf851

…rch#9966)

Summary:
Pull Request resolved: pytorch#9966

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from b69c5f9 to 1cbf851 Compare

April 8, 2025 23:35

ThomasJannaud added a commit to ThomasJannaud/executorch that referenced this pull request


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

5db1fae

…rch#9966)

Summary:

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from 1cbf851 to 5db1fae Compare

April 9, 2025 17:02

Contributor

facebook-github-bot commented Apr 9, 2025

This pull request was exported from Phabricator. Differential Revision: D72484196

ThomasJannaud added a commit to ThomasJannaud/executorch that referenced this pull request


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

59c391b

…rch#9966)

Summary:
Pull Request resolved: pytorch#9966

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from 5db1fae to 59c391b Compare

April 9, 2025 17:08

ThomasJannaud added a commit to ThomasJannaud/executorch that referenced this pull request


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

a12dbc0

…rch#9966)

Summary:

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from 59c391b to a12dbc0 Compare

April 9, 2025 17:11

Contributor

facebook-github-bot commented Apr 9, 2025

This pull request was exported from Phabricator. Differential Revision: D72484196

ThomasJannaud changed the title ~~Adding RMSNorm support to arbitrary x and normalized_dim shapes~~ Removing custom RMSNorm support


          Adding RMSNorm support to arbitrary x and normalized_dim shapes (pyto…

60fce8e

…rch#9966)

Summary:

In D72014553, we were adding initial support for RMS norm for an input in 3 or 4 dimensions, and a weight of dimension 1 (same size as x[:-1])

In this diff, we allow for:
- input of arbitrary shape
- shape broadcasting of w (w must have dim <= 1)

Reviewed By: Vysarat

Differential Revision: D72484196

ThomasJannaud force-pushed the export-D72484196 branch from a12dbc0 to 60fce8e Compare

April 9, 2025 21:02

Contributor

facebook-github-bot commented Apr 9, 2025

This pull request was exported from Phabricator. Differential Revision: D72484196

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed fb-exported topic: not user facing