[CPU/MLAS] Add QMoE shared layout INT2 AVX-512/AVX2 kernels and direct CPU QMoE #33785
tianleiwu
force pushed
from
e1f6c2cc
to
0b2619b0
4 days ago
tianleiwu
force pushed
from
0b2619b0
to
027a6a93
3 days ago
[CPU/MLAS] Add QMoE shared layout INT2 AVX-512/AVX2 kernels and direc…
332f2bc5
Fix review feedback for shared-layout INT2 QMoE
9de5d354
Fix INT2 QMoE review feedback and cross-platform CI failures
9e766e6d
Validate INT2 CPU CUDA parity and isolate decode profiling
9c75a759
Keep INT2 BF16 overflow protection off the decode hot path
08e6b963
Use GCC-style lambda inline attribute for clang-cl in INT2 AVX2 kernel
ffd18e6a
Address INT2 QMoE review: align FC2 bias boundary, SiLU infinity limi…
b86a8deb
tianleiwu
force pushed
from
027a6a93
to
b86a8deb
3 days ago
fix(qmoe): align INT2 tensor-core boundaries
3b49c0cf
fix(cuda): scope INT2 projection rounding to QMoE
09f67db1
Merge origin/main into shared-layout INT2 QMoE
7af5aa29
tianleiwu
dismissed their stale review
via 7af5aa29
2 days ago
tianleiwu
merged
f340e685
into main 2 days ago
tianleiwu
deleted the tlwu/20261031/qmoe-shared-layout-int2-cpu branch 2 days ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub