Add support for returning LSE from FlexAttention (and also differentiating through it) #133159
Conversation
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
…ating through it) [ghstack-poisoned]
Chillee
requested review from
albanD,
jbschlosser and
mikaylagawarecki
as code owners
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/133159
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 46c331c with merge base e890d88 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
… differentiating through it)" This PR changes the "contract" of `flex_attention_hop` to return LSE in base 2. However, we undo that and return LSE in base e from the `flex_attention` frontend. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
… differentiating through it)" This PR changes the "contract" of `flex_attention_hop` to return LSE in base 2. However, we undo that and return LSE in base e from the `flex_attention` frontend. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
… differentiating through it)" This PR changes the "contract" of `flex_attention_hop` to return LSE in base 2. However, we undo that and return LSE in base e from the `flex_attention` frontend. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
yanboliang
approved these changes
Aug 10, 2024
| kernel_options, | ||
| ) | ||
| def sdpa_hop(q, k, v, score_mod): | ||
| return flex_attention(q, k, v, score_mod, return_lse=True) |
There was a problem hiding this comment.
Finally fixed these annoying tests, lol!
… differentiating through it)" This PR changes the "contract" of `flex_attention_hop` to return LSE in base 2. However, we undo that and return LSE in base e from the `flex_attention` frontend. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
Chillee
added
ciflow/trunk
|
@pytorchbot merge |
Merge startedYour 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 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Stack from ghstack (oldest at bottom):
This PR changes the "contract" of
flex_attention_hopto return LSE in base 2. However, we undo that and return LSE in base e from theflex_attentionfrontend.cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang