Add FLASH_ATTN_HDIMS option to limit kernel compilation#2029
Open
Caellian wants to merge 1 commit intoOpenNMT:masterfrom
Open
Add FLASH_ATTN_HDIMS option to limit kernel compilation#2029Caellian wants to merge 1 commit intoOpenNMT:masterfrom
Caellian wants to merge 1 commit intoOpenNMT:masterfrom
Conversation
In many applications model head dimensions are known in advance and it's possible to opt-out of compiling ones that will never be used, even regardless of model choice. Signed-off-by: Tin Švagelj <tin.svagelj@live.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
In many applications model head dimensions are known in advance (even regardless of model choice) and it's possible to opt-out of compiling ones that will never be used.
In my case, I need CTranslate2 only for whisper models which means I can cut down compile times a lot by setting the
FLASH_ATTN_HDIMS="64"option. Newer LLMs also almost always use 128.Default it backwards compatible, can be explicitly set to speed up builds.