[NV]Add GLM-5 NVFP4 GB300 disagg-non-mtp TRT-LLM benchmarks via Dynamo

[xinli-sw]

Note

Low Risk
Benchmark and CI runner configuration only; no changes to inference serving or application logic.

Overview
Adds glm5-fp4-gb300-dynamo-trt to nvidia-master.yaml: multinode, disaggregated STP (non-MTP) runs on GB300 using the Dynamo TensorRT-LLM runtime image, with 22 fixed-seq-len points (13 at 1K/1K, 9 at 8K/1K). Each point pairs prefill/decode worker topology with an external CONFIG_FILE recipe from NVIDIA srt-slurm (sa-submission-q2-2026).

launch_gb300-nv.sh now handles glm5 + fp4 + dynamo-trt before the generic GLM-5 FP4 branch, setting SERVED_MODEL_NAME, /scratch/models/GLM-5-NVFP4, and SRT_SLURM_MODEL_PREFIX=nvidia/GLM-5-NVFP4.

perf-changelog.yaml documents the new config key and container/recipe source.

^{Reviewed by Cursor Bugbot for commit 8aaff9f. 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.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27621749102
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27621749102

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27622347964
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27622347964

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27622347964
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27622347964

[functionstackx]

lgtm! thank you for the contribution!

[xinli-sw]

/reuse-sweep-run

[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 8aaff9f. Configure here.}

[cursor]

Changelog deletion breaks sweep

High Severity

This PR’s perf-changelog.yaml diff removes a nested config-keys line while restructuring the dsr1-fp8-gb300-dynamo-trt entry. utils/process_changelog.py rejects any non-empty deleted lines in the changelog diff, so Run Sweep fails when it compares the PR to origin/main.

 

^{Reviewed by Cursor Bugbot for commit 8aaff9f. Configure here.}