onnxruntime
QMoE CPU Performance Update (Up to 4x on 4-bit)
#27364
Merged

QMoE CPU Performance Update (Up to 4x on 4-bit) #27364

tianleiwu merged 12 commits into main from tlwu/20260216/qmoe_cpu_q4_perf
tianleiwu
tianleiwu Fix QMoE CPU
ed0478ad
tianleiwu prepack and cache to improve QMoE perf
2c838293
tianleiwu move cache to prepack; allow q4gemm for block_size=32
2b256017
tianleiwu Turn on q4 gemm by default if there is no accuracy loss
e9dcd411
tianleiwu refine tests
dec88677
tianleiwu tianleiwu requested a review from apsonawane apsonawane 170 days ago
tianleiwu tianleiwu requested a review from copilot-pull-request-reviewer copilot-pull-request-reviewer 170 days ago
copilot-pull-request-reviewer
copilot-pull-request-reviewer commented on 2026-02-17
tianleiwu tianleiwu added release:1.24.2
tianleiwu tianleiwu removed release:1.24.2
tianleiwu refactor CheckInputs
4af13153
tianleiwu fix build
23cdf2d9
tianleiwu robust shape storage
6d351fed
tianleiwu Merge remote-tracking branch 'origin/main' into tlwu/20260216/qmoe_cp…
100738b8
tianleiwu refactoring
0b4c783a
apsonawane
apsonawane commented on 2026-02-20
apsonawane
apsonawane commented on 2026-02-20
apsonawane
apsonawane commented on 2026-02-20
tianleiwu review feedback
abef62be
tianleiwu tianleiwu requested a review from apsonawane apsonawane 166 days ago
tianleiwu tianleiwu requested a review from copilot-pull-request-reviewer copilot-pull-request-reviewer 166 days ago
copilot-pull-request-reviewer
copilot-pull-request-reviewer commented on 2026-02-20
tianleiwu address AI feedback
11d30ce0
apsonawane
apsonawane approved these changes on 2026-02-20
tianleiwu tianleiwu enabled auto-merge (squash) 166 days ago
tianleiwu tianleiwu merged b2a6e69e into main 165 days ago
tianleiwu tianleiwu deleted the tlwu/20260216/qmoe_cpu_q4_perf branch 165 days ago
tianleiwu tianleiwu added release:1.24.3
tianleiwu tianleiwu removed release:1.24.3

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone