Skip to content

Support extra ops modes for LLM Models#18670

Open
Jiseong-oh wants to merge 2 commits intomainfrom
extra_ops_modes
Open

Support extra ops modes for LLM Models#18670
Jiseong-oh wants to merge 2 commits intomainfrom
extra_ops_modes

Conversation

@Jiseong-oh
Copy link
Copy Markdown
Collaborator

@Jiseong-oh Jiseong-oh commented Apr 2, 2026

Summary

Add Extra Ops for Exynos Backend

  • Cos, GroupNorm, Index, Log, Pow, Rms_Norm, Sigmoid, sing, Splite_with_sizes_copy, sum_int_list, tanh and Topk
  • All ops are verfied
  • These ops will be used for supporting LLM models

Support Perf mode

  • It can be supported Perf mode with experimental
  • This mode MUST BE used on device farm before testing on the phone

cc @SS-JIA @digantdesai @kimishpatel

@pytorch-bot
Copy link
Copy Markdown

pytorch-bot bot commented Apr 2, 2026

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18670

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

❌ 2 New Failures, 2 Unrelated Failures

As of commit ab5de0d with merge base 94d70b9 (image):

NEW FAILURES - The following jobs have 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.

@meta-cla meta-cla bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Apr 2, 2026
@Jiseong-oh Jiseong-oh added partner: samsung For backend delegation, kernels, demo, etc. from the 3rd-party partner, Samsung module: samsung CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. and removed CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. labels Apr 2, 2026
@github-actions
Copy link
Copy Markdown

github-actions bot commented Apr 2, 2026

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

Jiseong-oh and others added 2 commits April 2, 2026 12:06
The previously supported single ops are being migrated to the current
dev branch, as follows.
cos, group_norm, index, log, pow, rms_norm, sigmoid, sin,
split_with_sizes_copy, sum_int_list, tanh and topk.

Co-authored-by: xz.linghu <xz.linghu@samsung.com>
Signed-off-by: jiseong.oh <jiseong.oh@samsung.com>
LiteCore provide a new api named "graphgen_set_perf_mode".
This commit invokes this api to set performance mode.

Co-authored-by: Jingya Zhang <jingya.zhang@samsung.com>
Signed-off-by: jiseong.oh <jiseong.oh@samsung.com>
@Jiseong-oh Jiseong-oh changed the title Extra ops modes Support extra ops modes for LLM Models Apr 3, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: samsung partner: samsung For backend delegation, kernels, demo, etc. from the 3rd-party partner, Samsung

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant