Add MiniMax-M3-MXFP8 GB300 disaggregated Dynamo-vLLM EAGLE3 MTP - #2478
Add MiniMax-M3-MXFP8 GB300 disaggregated Dynamo-vLLM EAGLE3 MTP#2478hshrivastava-droid wants to merge 3 commits into
Conversation
|
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 |
hshrivastava-droid
added
full-sweep-enabled
full-sweep-fail-fast
and removed
full-sweep-enabled
labels
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
|
|
||
| - config-keys: | ||
| - minimaxm3-fp8-gb300-dynamo-vllm-mtp | ||
| scenario-type: |
There was a problem hiding this comment.
🟡 The new perf-changelog.yaml entry for minimaxm3-fp8-gb300-dynamo-vllm-mtp has pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXX — a leftover placeholder that should be 2478 (this PR's number). Every other entry in the file substitutes the real PR number, so this one breaks the audit-trail link.
Extended reasoning...
The newly-added perf-changelog.yaml entry for minimaxm3-fp8-gb300-dynamo-vllm-mtp ends with:
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXXXXX is a literal, unsubstituted placeholder rather than an actual PR number. Every other entry in this file — including the immediately preceding ones for PRs 2360, 2213, 2421, and 2420 — has the real PR number filled in at this field, e.g. pull/2420, pull/2421. This PR is #2478 (per the PR metadata), so the correct value here is pull/2478.
The likely origin is that AGENTS.md/the recipe-authoring template shows pull/XXX as a generic placeholder to be filled in with the actual PR number once it's known (which is typically only after the PR is opened and assigned a number). It looks like this substitution step was simply missed when finalizing this entry.
Nothing in the benchmark execution path or CI reads or parses pr-link at runtime — it's purely a human-facing changelog field used to trace which PR introduced a given perf-changelog entry. So this doesn't cause any functional failure, build break, or benchmark regression. Its only effect is that anyone auditing perf-changelog.yaml later (e.g., trying to find "what PR added the MiniMax-M3 MTP GB300 entry") will follow a dead link (.../pull/XXX) instead of landing on the actual PR page.
Proof:
- Open
perf-changelog.yamland find the entry forconfig-keys: [minimaxm3-fp8-gb300-dynamo-vllm-mtp](line ~5379). - Its
pr-linkreadshttps://github.com/SemiAnalysisAI/InferenceX/pull/XXX. - Navigating to that URL 404s / resolves to nothing meaningful, since
XXXis not a valid PR number. - Compare with any adjacent entry, e.g. the one for
qwen3.5-fp4-b300-agentx-mtpa few lines above, whosepr-linkishttps://github.com/SemiAnalysisAI/InferenceX/pull/2421— a working link to the actual PR. - Since this diff is itself PR Add MiniMax-M3-MXFP8 GB300 disaggregated Dynamo-vLLM EAGLE3 MTP #2478, the fix is simply to replace
XXXwith2478so the link resolves correctly.
The fix is a one-line change: replace XXX with 2478 in the pr-link field. This is a documentation/reference-only defect with no runtime impact, so it should not block merge, but it's a trivial fix the author should make before merging to keep the changelog's audit trail accurate.
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30840867368 |
…UTLASS MSA decode
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30842732639 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
6 similar comments
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
1 similar comment
|
see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=30848195936 |
|
closing asM3 8k1k is dperpecated in favor of m3 agentx https://github.com/SemiAnalysisAI/InferenceX/blob/main/MODELS.md |
2 participants
No description provided.