openvino
[CPU] Add separate quantization of key and value to SDPA
#35366

Merged

[CPU] Add separate quantization of key and value to SDPA #35366

EgorDuplensky merged 2 commits into openvinotoolkit:master from EgorDuplensky:turboquant-pr1-non-batched-kv-cache-quantize

[CPU] Add separate quantization of key and value to SDPA

c42556b1

EgorDuplensky requested a review 67 days ago

github-actions added category: CPU

github-actions added category: build

EgorDuplensky assigned

maxnick 66 days ago

maxnick added this to the 2026.2 milestone 66 days ago

maxnick commented on 2026-04-16

Reuse attn_memcpy kernel

b95d0ed9

maxnick approved these changes on 2026-04-17

EgorDuplensky merged b47503cb into master 65 days ago

EgorDuplensky deleted the turboquant-pr1-non-batched-kv-cache-quantize branch 65 days ago

Reviewers

maxnick

Assignees

maxnick

Labels

category: CPU category: build

Milestone

2026.2