PR #31073 kv-cache : keep rotation constants on the GPU

kv-cache : keep rotation constants on the GPU #31073

anton-iermolenko wants to merge 3 commits into ggml-org:master from anton-iermolenko:perf/kv-rotation-resident

kv-cache : keep rotation constants on the GPU

c2108c9a

anton-iermolenko requested a review from

ggerganov 1 day ago

github-actions added testing

ggerganov assigned

ggerganov 1 day ago

tests : verify resident KV rotation buffers

a7aaba01

kv-cache : document remaining multi-device rotation work

a7931aef

Reviewers

ggerganov

Assignees

ggerganov

Labels

testing

Milestone

No milestone