PR #24384 [CPU] Add 8bit support to matmulnbits quantizer

[CPU] Add 8bit support to matmulnbits quantizer #24384

jiafatom merged 14 commits into main from fajin/matmulnbit8bit_quantizer

fajin-corp requested a review 256 days ago

fajin-corp assigned

fajin-corp 256 days ago

fajin-corp assigned

liqunfu 256 days ago

liqunfu commented on 2025-04-10

jiafatom commented on 2025-04-10

liqunfu dismissed these changes on 2025-04-11

init

16a3a8e2

finished quantizeAndTranspose for 8b and 2b

d61cfcd7

finished dequantize for 2b, 4b, 8b

04000081

changed interface

819a3c80

fixed q4 ut

f9fff2a7

fixed 4bit ut

9d9bb0ef

debugging q8

ad90bac4

fixed q8 ut

26b27632

finished ut

978dab16

updating quantizer

b90938c0

added q8 quantizer

a5a1784b

Add todo

8929e694

fix linting

700c1d55

fajin-corp dismissed their stale review via 700c1d55 256 days ago

fajin-corp force pushed from 1be4c8eb to 700c1d55 256 days ago

add todo

5125c527

jiafatom approved these changes on 2025-04-14

liqunfu approved these changes on 2025-04-14

jiafatom merged 9a993c37 into main 253 days ago

jiafatom deleted the fajin/matmulnbit8bit_quantizer branch 253 days ago

Reviewers

jiafatom

liqunfu

Assignees

liqunfu

fajin-corp

Labels

None yet

Milestone

No milestone

onnxruntime [CPU] Add 8bit support to matmulnbits quantizer #24384 Merged

[CPU] Add 8bit support to matmulnbits quantizer #24384

onnxruntime
[CPU] Add 8bit support to matmulnbits quantizer
#24384

Merged