llama.cpp
ggml-webgpu: improve MTP inference by using mat-vec path for small batches
#24811
Merged
Merged
ggml-webgpu: improve MTP inference by using mat-vec path for small batches #24811
yomaytk
merged
2 commits into
ggml-org:master
from
yomaytk:mat-vec-n8
ggml-webgpu: improve small batches decoding
313ac645
yomaytk
commented
on 2026-06-19
yomaytk
commented
on 2026-06-19
CISC
approved these changes
on 2026-06-19
Add barrier to the NUM_COLS loop in mul-mat-vec
9fe606fe
reeselevine
approved these changes
on 2026-06-23
Reviewers
Assignees
No one assigned
Labels
testing
ggml
WebGPU
Milestone
No milestone
Login to write a write a comment.
Login via GitHub