auto-round
perf: avoid unused SVD reconstruction and select the CUDA driver in SVDQuant
#2356
Merged

perf: avoid unused SVD reconstruction and select the CUDA driver in SVDQuant #2356

chensuyue merged 7 commits into main from wangchang/svdquant-svd-factors
changwangss
changwangss perf: avoid unused SVD reconstruction in SVDQuant
ef46bad8
changwangss changwangss marked this pull request as ready for review 27 days ago
changwangss perf: reuse the selected SVD driver during quantization
f59dae70
AutoRoundBot
azure-pipelines
changwangss fix: initialize the SVD driver once on the quantization device
bc5527b5
changwangss refactor: keep SVD driver initialization private to the transform
d9a6122a
changwangss refactor: pass SVD driver and device explicitly
44e30b8f
changwangss refactor: reuse the configured device for SVDQuant
da9cde2a
changwangss refactor: run SVDQuant on the weight device
4de495e3
changwangss changwangss changed the title perf: avoid unused SVD reconstruction in SVDQuant perf: avoid unused SVD reconstruction and reuse the CUDA driver in SVDQuant 27 days ago
changwangss changwangss requested a review from yiliu30 yiliu30 27 days ago
changwangss changwangss changed the title perf: avoid unused SVD reconstruction and reuse the CUDA driver in SVDQuant perf: avoid unused SVD reconstruction and select the CUDA driver in SVDQuant 27 days ago
AutoRoundBot
azure-pipelines
chensuyue
yiliu30
yiliu30 approved these changes on 2026-09-16
chensuyue chensuyue merged 903ba0ce into main 25 days ago
chensuyue chensuyue deleted the wangchang/svdquant-svd-factors branch 25 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone