llama.cpp
sycl-exp : dequant q4 k improvements
#7972
Merged

sycl-exp : dequant q4 k improvements #7972

AidanBeltonS
Remove double lines
4a481556
Single load for half2
cb3fb420
Store scales in local mem
604ef6bf
Vectorize q load
a235b7c5
AidanBeltonS AidanBeltonS requested a review from joeatodd joeatodd 1 year ago
github-actions github-actions added SYCL
mofosyne mofosyne added Review Complexity : Medium
joeatodd
joeatodd approved these changes on 2024-06-18
joeatodd joeatodd merged 0e4699e6 into codeplay/sycl-main 1 year ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone