PR #35186 [draft] Fake turboquant

[draft] Fake turboquant #35186

isanghao wants to merge 6 commits into openvinotoolkit:master from isanghao:fake_tbq

intermediate stage of kv_copy

83133e3e

turboquant accuracy test passes

b4c02597

option for fake turboquant

12c72d4c

Slight accuracy improvement with turboquant:

a7750e69

bug fix for accuracy

d4bc3104

enable turboquant only when head size is multiply of 128

6ca2cff0

isanghao added do_not_review

isanghao added do_not_merge

github-actions added category: GPU

Reviewers

No reviews

Assignees

No one assigned

Labels

category: GPU do_not_review do_not_merge

Milestone

No milestone