Repository navigation
fix(tests): accept CUDA 13.1-13.4 in the hub tests config - #362
Merged
Merged
Conversation
The hub test environment was pinned to a single CUDA version, 13.0, which the test GPU is not currently offered on. Worker placement then finds no matching host and every hub test run fails before the worker comes up. allowedCudaVersions is an exact set with no "or newer" form, so the versions above 13.0 have to be listed explicitly. Listing the newer versions alongside the required one also widens the pool of eligible hardware for placement.
colin99d
approved these changes
Oct 21, 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
.runpod/tests.jsonpinned the hub test environment to a single CUDA version —13.0— which no host for the test GPU (NVIDIA A40) currently advertises.Placement finds no matching machine, so hub test runs fail before the worker
ever comes up.
allowedCudaVersionsis an exact set: thetests.jsonschema (gpuTypeId,gpuCount,cpuFlavor,env,allowedCudaVersions) has no minimum-CUDAfield, so "13.0 and newer" has to be written out as a list. This change sets:
13.2is the version the test GPU advertises today, so this is what actuallymakes placement succeed.
13.1,13.3and13.4are valid console selections and are offered onother hardware; keeping them costs nothing and picks up capacity as hosts
upgrade.
13.0is deliberately excluded, since the request was for versions above it.Adding it back is a one-word change if a wider set is wanted — it satisfies the
image's
cuda>=13.0requirement.Test plan
pytest tests→ 142 passed(
GET /v2/catalog/gpus?include=AVAILABILITY) that the test GPU advertises13.2, so a set containing it can place a worker(
basic_inference_test,openai_messages_test)Follow-up (not in this PR)
.runpod/hub.jsoncarries the same single-version pin,allowedCudaVersions: ["13.0"], for poolsADA_80_PRO,AMPERE_80, where13.0also isn't advertised — deploys from the published listing hit the same
placement failure. Kept out of this PR to keep the change scoped.