auto-round
perf: vectorize Qwen4-Exp QSA indexer to speed up full-attention blocks
#2579

Merged

perf: vectorize Qwen4-Exp QSA indexer to speed up full-attention blocks #2579

lvliang-intel merged 7 commits into main from lvl/improve_quantize_efficency

vectorize Qwen4-Exp QSA indexer to speed up full-attention blocks

cfab4009

lvliang-intel requested a review from

copilot-pull-request-reviewer 9 days ago

Merge branch 'main' into lvl/improve_quantize_efficency

4a829966

copilot-pull-request-reviewer commented on 2026-11-04

fix: address review on Qwen4-Exp QSA indexer (mask dtype, non-prefix …

9c31f210

Merge branch 'lvl/improve_quantize_efficency' of https://github.com/i…

c6eb1029

Merge branch 'main' into lvl/improve_quantize_efficency

a412331e

Merge branch 'main' into lvl/improve_quantize_efficency

090a8917

Merge branch 'main' into lvl/improve_quantize_efficency

4c96fbee

lvliang-intel requested a review from

xin3he 7 days ago

lvliang-intel requested a review from

n1ck-guo 7 days ago

chensuyue added this to the 0.17.0 milestone 3 days ago

n1ck-guo approved these changes on 2026-11-11

lvliang-intel merged b1f816bb into main 2 days ago

lvliang-intel deleted the lvl/improve_quantize_efficency branch 2 days ago

Reviewers

n1ck-guo

xin3he

Assignees

No one assigned

Labels

None yet

Milestone

0.17.0