PR #1597 Leverage mmap to offloading the tensors

Leverage mmap to offloading the tensors #1597

JohannesGaessler merged 3 commits into ggml-org:master from howard0su:cuda_load

JohannesGaessler commented on 2023-05-26

github-actions commented on 2023-05-27

howard0su changed the title ~~Leverage mmap for CUDA loading~~ Leverage mmap to offloading the tensors 2 years ago

howard0su force pushed 2 years ago

howard0su requested a review from

JohannesGaessler 2 years ago

JohannesGaessler commented on 2023-05-30

howard0su force pushed 2 years ago

JohannesGaessler commented on 2023-06-10

Rebase to latest

921d87ca

howard0su force pushed to 921d87ca 2 years ago

Show progress

34ca572e

howard0su requested a review from

JohannesGaessler 2 years ago

howard0su requested a review from

0cc4m 2 years ago

howard0su commented on 2023-06-11

JohannesGaessler commented on 2023-06-11

Add assert to make sure we only allocate temp buffer for non-CPU back…

61726bd9

JohannesGaessler approved these changes on 2023-06-12

JohannesGaessler merged 58970a4c into master 2 years ago

Reviewers

JohannesGaessler

0cc4m

github-actions

Assignees

No one assigned

Labels

None yet

Milestone

No milestone