Repository navigation
fix(docker): install torchvision in GPU load-balancer image - #113
Merged
Merged
Conversation
This was referenced Oct 15, 2026
Dockerfile got a pinned torchvision in #101 (SLS-377), but Dockerfile-lb did not. flash build auto-excludes torchvision from tarballs, so load-balanced GPU endpoints that import torchvision fail with ModuleNotFoundError. Mirror the Dockerfile install: torchvision 0.24.1 from the cu128 index, paired with the base image torch 2.9.1, plus the build-time import check.
justinwlin
force-pushed
the
fix/lb-image-torchvision
branch
from
October 26, 2026 19:21
9d976cb to
ba8344c
Compare
This was referenced Oct 26, 2026
docs(flash): match the quickstart and tutorials to a live run on runpod-flash 1.21.0
runpod/docs#923
Open
deanq
approved these changes
Oct 26, 2026
deanq
approved these changes
Oct 27, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Installs the pinned
torchvisioninDockerfile-lb, matching what #101 did forDockerfile.Why
flash buildnever bundlestorchvision, because a bundled copy would shadow one matched to the image'storch. The queue-based GPU image has installedtorchvision==0.24.1since #101. The load-balanced GPU image never got it, so LB endpoints that importtorchvisionfail withModuleNotFoundError.Change
ARG TORCHVISION_VERSION=0.24.1, the release paired with the basetorch 2.9.1.Dockerfile.import torchvisionto the build-time verify step.Verification
Dockerfile(diffof the two blocks is empty).runpod/flash:py3.12-latesthas thetorchvisionstep in its build history, andrunpod/flash-lb:py3.12-latestdoesn't.torch==2.9.1+cu128+torchvision==0.24.1resolves cleanly against the cu128 index. PyPI's latesttorchvision0.29.1 needstorch>=2.14, which is why the pin matters.Dockerfile-lbfor Python 3.10 through 3.13, and the non-3.12 path reinstallstorchbefore this step.Related Flash PRs
runpod-flashstops getting fixes on 2026-10-29, so therunpod/flashchanges need to merge and release before then.torchvisionin the GPU load-balancer imageflash buildarchivesflash deployapikeyas a credentials fallback