perf(agentx): refresh dsv4-fp4-b300-vllm-agentic-mtp harness - #2517
Conversation
704adc8 to
6ef1b6b
Compare
|
Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase For PR verification, add the PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. See GitHub's docs on re-running failed jobs 感谢你的贡献！请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人，你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板，包括保留英文语句 如需进行 PR 验证，请为此 PR 添加 PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档 |
|
Claude finished @cquil11's task in 1m 37s —— View job PR Review
LGTM - no blocking issues found Verified:
Note: I could not run |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
LGTM — this is a changelog-only addition with no config or harness code changes.
Extended reasoning...
Overview
The diff touches only perf-changelog.yaml, appending a single new entry for dsv4-fp4-b300-vllm-agentic-mtp referencing PR #2517. No config, script, or harness files are modified in this diff.
Security risks
None. This is a static YAML changelog entry with no executable code, no secrets, and no external inputs.
Level of scrutiny
Low. Changelog-only additions are low-risk and mechanical. I verified the entry uses the documented schema (config-keys, scenario-type, description, pr-link) consistent with other entries and covered by utils/test_process_changelog.py, and confirmed dsv4-fp4-b300-vllm-agentic-mtp exists as a config key in configs/nvidia-master.yaml.
Other factors
The bug hunting system found no issues. One candidate concern was raised and ruled out: that the changelog claims a harness refresh without an accompanying config/harness change — but since this PR's diff is scoped to the changelog file alone, that's expected for this PR and not a defect in the diff itself.
…00-vllm-agentic-mtp
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31150499711 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31152051032 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31153425491 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=31192604550 |
|
/stage-results 31192604550 |
|
@cquil11 staged run 31192604550: https://inferencemax-app-git-staging-semianalysisai.vercel.app/inference?i_dates=2026-08-07~r31192604550 This run remains available across future @cquil11 已将运行 31192604550 发布到预发布环境：https://inferencemax-app-git-staging-semianalysisai.vercel.app/inference?i_dates=2026-08-07~r31192604550 后续的 |
|
/reuse-sweep-run 31192604550 |
# Conflicts: # utils/aiperf
1 participant
Summary
dsv4-fp4-b300-vllm-agentic-mtp.vllm:metrics before publishing results. This preserves automatic direct-server collection for pure TP and adds engine metrics alongside router metrics for DP-attention.Validation
bash -n benchmarks/single_node/agentic/dsv4_fp4_b300_vllm_mtp.shpython3 utils/validate_perf_changelog.py --changelog-file perf-changelog.yaml --base-ref origin/main --head-ref HEADpython3 -m pytest utils/test_process_changelog.py utils/matrix_logic/test_generate_sweep_configs.py -q(116 passed)--server-metrics.git diff --check