ci: GPU smoke test (vast.ai) for released backends#41
Merged
Conversation
On-demand (workflow_dispatch) + nightly: rents the cheapest matching GPU, downloads the released archive, quantizes a tiny f16 model to NVFP4 with the shipped llama-quantize, serves it with -ngl 99, asserts a coherent answer, runs llama-bench and asserts the GPU backend actually did the work -- a silent CPU fallback fails the run. Result is posted as a non-required commit status gpu-smoke/<backend> on the released commit. Deliberately NOT a required PR check: spot GPU rental is slow, nondeterministic and costs money. Hardening baked in from manual runs: X11 client libs + LunarG loader for the NVIDIA Vulkan ICD in headless containers, nohup+poll against vast hosts dropping long ssh sessions, no grep|head under pipefail, box is destroyed in an always() step.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds
backend-smoke.yml+.github/scripts/gpu-smoke/{rent,remote-smoke}.sh.What a run does (per backend, matrix over
linux-x64-vulkan/linux-x64-cuda-13.3):gpu_query(default RTX 5090, ~$0.30/h), sequential offers with retries, box destroyed inalways().release_tag(defaultdev-latest).llama-quantize.llama-server -ngl 99→/health→ chat completion → asserts the answer contains "Paris" (temperature 0).llama-bench→ asserts the GPU backend actually did the work (loaded Vulkan/CUDA backend+ NVIDIA device in the log) — a silent CPU fallback fails the run.gpu-smoke/<backend>on the commit the release resolves to → visible on the dev→master promotion PR.Triggers:
workflow_dispatch(tag/backends/gpu filter as inputs) + nightly cron againstdev-latest(~$1/night). Deliberately not a required check — spot rental is slow, nondeterministic and billed.Hardening from the manual 5090 runs: X11 libs + LunarG loader for the NVIDIA Vulkan ICD in headless containers; nohup+poll because vast hosts drop long ssh sessions; no
grep | headunder pipefail.Secrets used:
VAST_API_KEY,VAST_SSH_KEY(already configured).Note: the Run workflow button appears once this lands on the default branch (master) via the dev→master promotion; until then it is dispatchable with
gh workflow run backend-smoke.yml --ref dev. The cuda-13.3 leg needs #40 merged sodev-latestactually contains the cuda asset.