onnxruntime
Implement new experimental lookup-based matrix multiplication method(TMAC)
#26695
Merged
Overview
Commits
52
Changes
View On GitHub

Implement new experimental lookup-based matrix multiplication method(TMAC) #26695

jambayk merged 52 commits into main from vraspar/lut-gemm
vraspar
liqunfu init code structure for matmul 2 bits
5484560d
liqunfu add and pass q4dq tests for q2bit - rename file and test name later
8c1cfe11
liqunfu some fixes
f6f22e30
liqunfu add apis to neon and other avxs
3e1a9514
liqunfu fix neon build
01300610
liqunfu disable 2bit test
b4aad013
liqunfu 2 bit quantize to support model builder
ff531cbe
carzh Merge remote-tracking branch 'msft/main' into carzh/bitnet-reverse-la…
6849ea2c
carzh fix compile errors
e85431e5
carzh resolve build failure update
96427403
HectorSVC 2 bits check
892222a4
fixed bug causing int8 tests to fail
07b7f3fc
carzh Merge remote-tracking branch 'origin/main' into carzh/bitnet-reverse-…
5fb2edd1
carzh lintrunner
493ebd15
carzh prepack wip -- not prepacking b data because dispatch to check for ml…
b4b143fc
carzh fixed dispatch issue, added acc level 4 tests, and now running into a…
534b8e6d
deep sigh
70d65889
builds somehow
ad2572b4
update
b312815d
udpate
bfeac34b
vraspar Implement Pre Packing of qweight for tmac
a5de1080
vraspar Implement Pre packing for Scales and zero points
7ff8218e
vraspar Transform zero points before interleaving
6d8e8ece
vraspar Initial implementation of tmac kernel config
5d19daf0
vraspar Move pre packing scales and zp code to qlutgemm and use tmac_params
c6000562
update
5cf99e6b
bug fixes
f9a9b47d
vraspar Fix bug in scale unpacking
5687e5e6
vraspar Fix issues with TMAC GEMM kernels and remove hard coded variables
6f084186
vraspar Fix bug in LUT table generation
6191aadb
vraspar Fix casting issue
f2de7764
vraspar add session option and clean up
9ef6d75f
vraspar Refactor QNBit GEMM Implementation for AVX2
59c00550
vraspar Refactor dispatch
457cfa37
vraspar Add test cases
bdb29823
vraspar rewrite test_sqlutgemm.cpp
289e53e0
vraspar Add more robust checking before using LUT kernels
fabae089
vraspar Merge remote-tracking branch 'origin/main' into vraspar/lut-gemm
5d8a6ee3
github-actions
github-actions commented on 2025-12-16
github-advanced-security
github-advanced-security commented on 2025-12-16
vraspar revert graph_transform_test.cc
b1fcda14
vraspar Clean up: revert unchanged files
3eb22b08
vraspar Apply linting and clean up
f61c3d8b
vraspar Add headers, update binding, and general clean up + linting
bebcb647
vraspar Fix zero point test cases
6a2e8221
jambayk
jambayk commented on 2026-01-02
jambayk
jambayk commented on 2026-01-02
vraspar Refactor ComputeBPackedLUT to remove unused parameters and simplify f…
a19b2f6e
vraspar Merge remote-tracking branch 'origin/main' into vraspar/lut-gemm
26678b21
vraspar Fix compiler warnings
e5f80cbe
vraspar Improve error handling in TMACComputeGemm_avx2 for batch size and sca…
b518ce9e
vraspar vraspar marked this pull request as ready for review 17 days ago
vraspar vraspar requested a review from edgchen1 edgchen1 17 days ago
edgchen1
edgchen1 commented on 2026-01-06
vraspar Apply feedback and use PrePacking
f94e51ef
vraspar vraspar added release:1.24.0
vraspar update platform.cpp
7b708ad8
vraspar use MLAS_THROW_EX for qlutgemm.cpp
58e93ec2
hariharans29
hariharans29 commented on 2026-01-11
hariharans29
hariharans29 commented on 2026-01-11
hariharans29
hariharans29 commented on 2026-01-11
hariharans29
hariharans29 commented on 2026-01-11
hariharans29
hariharans29 commented on 2026-01-11
hariharans29
hariharans29 commented on 2026-01-11
Add LUT GEMM 2-bit tests and fix Python quantization reference implem…
469cde7d
vraspar vraspar requested a review from hariharans29 hariharans29 6 days ago
vraspar vraspar requested a review from edgchen1 edgchen1 6 days ago
vraspar Merge remote-tracking branch 'origin/main' into vraspar/lut-gemm
48fd9829
hariharans29
hariharans29 commented on 2026-01-15
hariharans29
hariharans29 commented on 2026-01-15
hariharans29
hariharans29 approved these changes on 2026-01-15
hariharans29
hariharans29 commented on 2026-01-15
jambayk jambayk merged 8e050d16 into main 4 days ago
jambayk jambayk deleted the vraspar/lut-gemm branch 4 days ago
hariharans29
hariharans29 commented on 2026-01-16
hariharans29
hariharans29 commented on 2026-01-16
hariharans29
hariharans29 commented on 2026-01-16

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
release:1.24.0
Milestone
No milestone
Terms Of Use Privacy Policy Refund Policy Impressum