PR #14855 [WIP] Asynchronous model mover for lowvram

[WIP] Asynchronous model mover for lowvram #14855

wfjsw wants to merge 16 commits into AUTOMATIC1111:dev from wfjsw:model-mover

wfjsw changed the base branch from master to dev 2 years ago

wfjsw force pushed from 565925ec to 5295c969 2 years ago

wfjsw marked this pull request as ready for review 2 years ago

wfjsw requested a review from

AUTOMATIC1111 2 years ago

wfjsw marked this pull request as draft 2 years ago

wfjsw marked this pull request as ready for review 2 years ago

Nuullll commented on 2024-02-25

wfjsw marked this pull request as draft 2 years ago

async weight mover

c1702ea4

smart lowvram mover

4f0d5a58

cleanup

cca6102f

fix ci

f729b21b

fix ci

8278ad01

tweak

5d69b1e8

avoid infinite loop

a58ee39e

remove profiler

cf3cc4c7

remove unneeded changes

0caa7531

better impl

fffc9026

fix ci

8828c9ec

add option to revert to old behavior

1d950c77

only allow cuda for streamlined lowvram and handle sparse tensors in …

4703b95b

fix signature

572f4cdd

wfjsw force pushed from 10cf7701 to 572f4cdd 2 years ago

avoid oom on slow cards

e8df8a9f

fix impl

ed69979d

Reviewers

Nuullll

AUTOMATIC1111

Assignees

No one assigned

Labels

None yet

Milestone

No milestone

stable-diffusion-webui [WIP] Asynchronous model mover for lowvram #14855 Open

[WIP] Asynchronous model mover for lowvram #14855

stable-diffusion-webui
[WIP] Asynchronous model mover for lowvram
#14855

Open