[NV]Add Kimi K2.5 NVFP4 GB300 disaggregated TRT-LLM benchmarks via Dynamo

[xinli-sw]

opened on behalf of @richardhuo-nv

---

Note

Low Risk
Benchmark and launcher configuration only; no application logic, auth, or data-path changes beyond new model aliases for CI runs.

Overview
Adds GB300 benchmark coverage for Kimi K2.5 NVFP4 using disaggregated TensorRT-LLM via Dynamo, keyed as kimik2.5-fp4-gb300-dynamo-trt in nvidia-master.yaml.

The new block mirrors the existing GB200 Kimi TRT setup but targets runner: gb300, uses image tensorrtllm-runtime:1.3.0-dev.1-cuda13, and wires 25 STP search-space points (14 at ISL1K/OSL1K, 11 at ISL8K/OSL1K) to srt-slurm recipe YAMLs under gb300Nvfp4 with varying concurrency, prefill worker counts, and decode TP/EP topologies. perf-changelog.yaml documents the new config key and recipe source.

runners/launch_gb300-nv.sh is updated so kimik2.5 + fp4 resolves MODEL_PATH to /scratch/models/Kimi-K2.5-NVFP4 and the matching SRT_SLURM_MODEL_PREFIX, and the unsupported-model message lists this combo.

^{Reviewed by Cursor Bugbot for commit 947db0a. 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.

[functionstackx]

waiting on kimi nvfp4 gb300 vllm PR first

would like prefer if @SemiAnalysisAI/core quick whip up "best effort/good faith attempt" at an vllm kimi nvfp4 gb300 config rn and then merge then first in so then we can merge in trtllm?

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27663808752
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27663808752