[NV]Update Kimi K2.5 NVFP4 GB200 disaggregated TRT-LLM benchmarks via Dynamo

[xinli-sw]

Opened on behalf of @richardhuo-nv

---

Note

Low Risk
CI/benchmark YAML only; no application code, but wrong recipes or concurrency rows would skew published performance results and sweep cost.

Overview
Refreshes kimik2.5-fp4-gb200-dynamo-trt in nvidia-master.yaml for disaggregated TensorRT-LLM Dynamo serving on GB200.

The container image moves from tensorrtllm-runtime:1.1.0-dev.2 to 1.3.0-dev.1-cuda13. Fixed-seq-len search spaces for 1k/1k and 8k/1k no longer use bundled allconc recipe paths; each concurrency row now points at a dedicated STP recipe from srt-slurm (sa-submission-q2-2026 / PR #89), with updated conc-list values and matching prefill/decode topology (workers, TP/EP, CONFIG_FILE).

A perf-changelog.yaml entry documents the change for sweep selection.

^{Reviewed by Cursor Bugbot for commit 29e08b9. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

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. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers.

If additional help is needed, PR authors can reach out to core maintainers over Slack.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 29e08b9. Configure here.}

[cursor]

Duplicate concurrency 1229 label

Medium Severity

Under kimik2.5-fp4-gb200-dynamo-trt ISL 8192/OSL 1024, two search-space rows use conc-list: [ 1229 ] with different recipes (ctx4dep4_gen1dep16_batch64 vs ctx6dep4_gen1dep32_batch32). The second row replaced the prior 2151 point in the same commit, so concurrency 2151 is missing and two distinct configs share one label.

Additional Locations (1)

• .github/configs/nvidia-master.yaml#L8514-L8528

 

^{Reviewed by Cursor Bugbot for commit 29e08b9. Configure here.}

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27591355916
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27591355916

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27591355916
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27591355916

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27591355916
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27591355916

[functionstackx]

overall LGTM, just. need to solve these too configs on PR validation https://github.com/SemiAnalysisAI/InferenceX/actions/runs/27591355916/job/81813641872?pr=1797