cuda : fix multi-seq, quantized FA#14820
Closed
ggerganov wants to merge 1 commit intogg/tests-fa-non-contfrom
Closed
cuda : fix multi-seq, quantized FA#14820ggerganov wants to merge 1 commit intogg/tests-fa-non-contfrom
ggerganov wants to merge 1 commit intogg/tests-fa-non-contfrom
Conversation
23 tasks
Collaborator
|
I have something WIP to fix this by adding non-contiguous support to the dequantization kernels. There is still a bug somewhere, I'll try to make a PR this evening.
This is more an issue with kernel launch overhead because you're launching one kernel per sequence + each kernel will have poor hardware utilization. |
Member
Author
|
Replaced by #14822 |
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
target #14756
Relax the requirement for contiguously allocated K/V buffers in the quantized case.
I am not 100% this is the most optimal solution in terms of memory usage, but at least the results are OK now.