onnxruntime
QMoE: prepack int4/int8 expert weights in PrePack hook (symmetric with MatMulNBits)
#28749
Merged
Merged
QMoE: prepack int4/int8 expert weights in PrePack hook (symmetric with MatMulNBits) #28749
justinchuby
merged
13 commits into
microsoft:main
from
justinchuby:qmoe-int-prepack
QMoE: prepack int4/int8 expert weights in PrePack hook
834eae86
Add TestQMoEIntPrePackParity for the new QMoE.PrePack hook
66c7f47e
Fix import order in test_qmoe_cuda.py
516812cb
Rewrite QMoE PrePack test as a smoke test; verified passing on H200
197819fb
QMoE: address review — opt-in weights_prepacked attribute, SM guard, …
2fcb9404
QMoE: cache weight shapes during PrePack so source initializer can be…
5e1491c7
tianleiwu
requested changes
on 2026-06-02
tianleiwu
commented
on 2026-06-03
Address round-3 QMoE int-prepack review feedback
360217b4
QMoE: align PrePack test name/docs with smoke-test intent
3cbbf512
tianleiwu
commented
on 2026-06-03
tianleiwu
commented
on 2026-06-03
QMoE: pass SM75 through prepack clamp to mirror dispatch
3d4e0be1
QMoE: make weights_prepacked an optional tri-state attribute
4d898bf1
QMoE: set weights_prepacked schema default to -1 (auto)
543ab255
tianleiwu
dismissed these changes
on 2026-06-05
Merge remote-tracking branch 'origin/main' into qmoe-int-prepack
da0178c7
QMoE: address round-4 int-prepack review feedback
e9cea0d3
tianleiwu
dismissed these changes
on 2026-06-09
tianleiwu
dismissed their stale review
40 days ago
Dismissing: this review was posted with stale content from a different PR by mistake. Re-posting the correct review for PR #28749.
tianleiwu
commented
on 2026-06-09
disabled auto-merge 40 days ago
Manually disabled by user
Manually disabled by user
tianleiwu
approved these changes
on 2026-06-09
Login to write a write a comment.
Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone