onnxruntime
Save much memory at model loading time by converting weights to OrtValues early
#26345

Merged

Save much memory at model loading time by converting weights to OrtValues early #26345

yuslepukhin merged 8 commits into main from yuslepukhin/convert_weights_early

Revert "Properly remove in-memory references (#25652)"

13d2c375

yuslepukhin requested a review from

skottmckay 326 days ago

yuslepukhin requested a review from

fs-eire 326 days ago

yuslepukhin requested a review from

adrianlizarraga 326 days ago

yuslepukhin requested a review from

copilot-pull-request-reviewer 326 days ago

copilot-pull-request-reviewer commented on 2025-10-17

yuslepukhin marked this pull request as ready for review 326 days ago

Debug and disable invalid test

17fb9a5d

Move

032b5a32

Do not allocate CPU initialziers on Arena and save some space

a15b6a55

Adjust tests for Reserve calls

7e3c38ec

Merge branch 'main' into yuslepukhin/convert_weights_early

1e741dd6

skottmckay commented on 2025-10-25

Address review comments

18127ac8

tianleiwu commented on 2025-10-27

Address comments

2ae70f8b

tianleiwu approved these changes on 2025-10-30

yuslepukhin merged b8609880 into main 313 days ago

yuslepukhin deleted the yuslepukhin/convert_weights_early branch 313 days ago

Reviewers

tianleiwu

skottmckay

fs-eire

adrianlizarraga

Assignees

No one assigned

Labels

None yet

Milestone

No milestone