[SYCL] Use batched mul_mat pathway #5591
Merged
Conversation
This file contains 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
|
@NeoZhangJianyu, @abhilash1910, @Alcpz, feedback would be appreciated |
This was referenced Feb 19, 2024
abhilash1910
approved these changes
Feb 29, 2024
There was a problem hiding this comment.
LGTM. I think we can use this until MKL adds the dtypes for batched gemm . Pinging @airMeng @ggerganov for a look when available.
@AidanBeltonS could you please rebase , should fix the android build issue. Thanks
AidanBeltonS
force-pushed
the
fix_mul_mat_pathways
branch
from
cb21f6c
to
abed262
Compare
hazelnutcloud
pushed a commit
to hazelnutcloud/llama.cpp
that referenced
this pull request
Mar 10, 2024
* Use batched mul_mat pathway * rm extra line * Explicitly state scaled data type --------- Co-authored-by: Abhilash Majumder <[email protected]>
jordankanter
pushed a commit
to jordankanter/llama.cpp
that referenced
this pull request
Mar 13, 2024
* Use batched mul_mat pathway * rm extra line * Explicitly state scaled data type --------- Co-authored-by: Abhilash Majumder <[email protected]>
hodlen
pushed a commit
to hodlen/llama.cpp
that referenced
this pull request
Apr 1, 2024
* Use batched mul_mat pathway * rm extra line * Explicitly state scaled data type --------- Co-authored-by: Abhilash Majumder <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
This PR enables using the batched mul_mat pathway when appropriate. Previously the single gemm path was being taken and it was not suitable for the type of operation causing segfaults. This PR changes things to more closely match the CUDA impl and use the batched gemm path.
This change allows a lot more tests to pass for SYCL devices. There is one limitation with this approach, we cannot use non default precision operations. As oneMKL has not open sourced the
gemm_batchfor the data types <half, half, float, float> (corresponding to <src0, src1, dst, scaling>) yet. This is something I have raised with oneMKL