Bypass the softmax pytorch kernel that upscales to fp32#19865
Open
telgamal-1 wants to merge 1 commit into
Open
Bypass the softmax pytorch kernel that upscales to fp32#19865telgamal-1 wants to merge 1 commit into
telgamal-1 wants to merge 1 commit into
Conversation
Summary: Bypasses the PyTorch softmax kernel in `static_attention` that upscales activations to fp32, keeping the softmax computation in fp16. Also updates `norm.py` to handle the fp16 softmax output. Differential Revision: D106729898
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/19865
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 1 Awaiting Approval, 2 Unrelated Failures, 6 Unclassified FailuresAs of commit 4260f2a with merge base 42581f1 ( UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:
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
Bot
added
the
CLA Signed
|
Contributor
|
@telgamal-1 has exported this pull request. If you are a Meta employee, you can view the originating Diff in D106729898. |
This PR needs a
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary: Bypasses the PyTorch softmax kernel in
static_attentionthat upscales activations to fp32, keeping the softmax computation in fp16. Also updatesnorm.pyto handle the fp16 softmax output.Differential Revision: D106729898