CUDA: fix MMQ stream-k rounding if ne00 % 128 != 0#8311
Merged
JohannesGaessler merged 1 commit intoggml-org:masterfrom Jul 5, 2024
Merged
CUDA: fix MMQ stream-k rounding if ne00 % 128 != 0#8311JohannesGaessler merged 1 commit intoggml-org:masterfrom
JohannesGaessler merged 1 commit intoggml-org:masterfrom
Conversation
JohannesGaessler
added
the
Review Complexity : Medium
slaren
approved these changes
Jul 4, 2024
Nexesenex
added a commit
to Nexesenex/croco.cpp
that referenced
this pull request
Jul 5, 2024
Co-Authored-By: Johannes Gäßler <johannesg@5d6.de>
Nexesenex
added a commit
to Nexesenex/croco.cpp
that referenced
this pull request
Jul 6, 2024
Co-Authored-By: Johannes Gäßler <johannesg@5d6.de>
Nexesenex
added a commit
to Nexesenex/croco.cpp
that referenced
this pull request
Jul 11, 2024
Co-Authored-By: Johannes Gäßler <johannesg@5d6.de>
arthw
pushed a commit
to arthw/llama.cpp
that referenced
this pull request
Jul 13, 2024
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Fixes #8254 . The problem as far as I can tell is that for determining the boundaries between CUDA blocks I implicitly assumed that the ne00/ne10 dimension of the matrices is a multiple of the number of values processed by MMQ in a single iteration (128 in this case). So for deepseek where ne00 is only a multiple of 64 the results were wrong.