Repository navigation
ci(buildkite): run SM90 tests on one whole H100 instead of H200 MIG - #128
Closed
jinzhen-lin wants to merge 3 commits into
Closed
jinzhen-lin wants to merge 3 commits into
jinzhen-lin wants to merge 3 commits into
Conversation
Move the H200 base and sampled tests from the MIG queues (h200_18gb, h200_35gb) to the whole-GPU H200 queue. Follow vLLM ci-infra's whole-GPU Docker preset: request all GPUs, and drop the MIG-only device passthrough, /dev/nvidiactl mount, allocator setting and agent image pre-pull variable. Co-authored-by: Claude Code (Claude Opus 5.5) <noreply@anthropic.com> Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
The whole-GPU H200 queue hands a job all eight GPUs. Use the single-GPU H100 Kubernetes pool instead, as vLLM does for its humming tests: the scheduler assigns one whole GPU. H100 and H200 share the SM90 heuristics. Co-authored-by: Claude Code (Claude Opus 5.5) <noreply@anthropic.com> Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
jinzhen-lin
changed the title
Set HUMMING_COMPILER=nvcc on the H100 base and sampled tests to try nvcc instead of the default NVRTC there. Other devices keep NVRTC. Co-authored-by: Claude Code (Claude Opus 5.5) <noreply@anthropic.com> Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Move the SM90 base and sampled tests from the H200 MIG queues to one whole H100, following vLLM ci-infra's single-GPU H100 Kubernetes preset (
h100_plugin_template), which vLLM also uses for its humming tests.h200_18gb->mithril-h100-pool.h200_35gb->mithril-h100-pool.nvidia.com/gpu: 1, replacing the Docker plugin and its MIG device passthrough.The whole-GPU
H200queue was not used because it hands a job all eight GPUs; vLLM only uses it for 8-GPU jobs. H100 and H200 share humming's SM90 heuristics.Not verified locally: only that the YAML parses and both steps resolve to the new queue and pod spec.
🤖 Generated with Claude Code