In https://gerrit.wikimedia.org/r/c/machinelearning/liftwing/inference-services/+/715918 we moved the editquality docker image to a multi-stage build (see commit msg for more details) and its size went from 1.56GB to 1.12GB (-440MB). We should apply the same pattern to the other model types, and figure out if we can create a base image to re-use across them to avoid repeating configs (not a strict requirement for the task, only a nice to have in case it is possible).
Description
Description
Details
Details
Related Changes in Gerrit:
Customize query in gerrit
Revisions and Commits
Revisions and Commits
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | kevinbazira | T301409 Migrate editquality models | |||
| Resolved | • ACraze | T284689 Create migration plan for editquality models from ORES to Lift Wing | |||
| Resolved | calbon | T272874 Prepare 4 ORES English models for Lift Wing | |||
| Resolved | kevinbazira | T279004 Production images for ORES/revscoring models | |||
| Resolved | • ACraze | T290266 Move ML docker images to multi-stage build |
Event Timeline
Comment Actions
Nice job @elukey! Confirming that the 2021-09-01-140944-production version of editquality model server seems to be ~1.12GB now.
I'll try to do something similar with the articlequality model server and see how it goes
Comment Actions
Change 716629 had a related patch set uploaded (by Accraze; author: Accraze):
[machinelearning/liftwing/inference-services@main] articlequality: move blubber config to multi-stage
Comment Actions
Change 717083 had a related patch set uploaded (by Kevin Bazira; author: Kevin Bazira):
[machinelearning/liftwing/inference-services@main] refactor draftquality blubber file to reduce image sizes
Comment Actions
Thank you for proposing this approach @elukey. I have implemented it on draftquality images and the sizes have reduced.
IMAGE OLD NEW production 1.38GB 796MB test 756MB 176MB
Comment Actions
Change 717083 merged by jenkins-bot:
[machinelearning/liftwing/inference-services@main] refactor draftquality blubber file to reduce image sizes
Comment Actions
Change 719164 had a related patch set uploaded (by Kevin Bazira; author: Kevin Bazira):
[machinelearning/liftwing/inference-services@main] refactor topic to reduce image sizes
Comment Actions
Change 716629 merged by jenkins-bot:
[machinelearning/liftwing/inference-services@main] articlequality: move blubber config to multi-stage
Comment Actions
Change 719164 merged by Accraze:
[machinelearning/liftwing/inference-services@main] refactor topic to reduce image sizes
Comment Actions
Marking this as RESOLVED since all revscoring production images have been moved to multi-stage builds. Nice job everyone!
editquality: 1.56GB -> 1.12 GB
articlequality: 1.2 GB -> 800 MB
draftquality: 1.38 GB -> 796 MB
topic: 1.38 GB -> 916MB