Update B200 Dsv4 configs

[wzhao18]

Note

Low Risk
Benchmark-only image and vLLM serve-flag changes; no production auth or application logic.

Overview
Updates DeepSeek-V4 FP4 B200 vLLM fixed-seq-len benchmarks to a pinned vLLM nightly image and turns on expert-parallel load balancing (EPLB) whenever DP attention (DEP) is active.

For DP_ATTENTION=true, dsv4_fp4_b200_vllm.sh now passes --enable-eplb with an NCCL communicator config alongside the existing deep_gemm_mega_moe MoE backend. perf-changelog.yaml documents this under dsv4-fp4-b200-vllm.

^{Reviewed by Cursor Bugbot for commit e1b0750. 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=26899483031
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26899483031

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26900694153
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26900694153

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26900694153
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26900694153

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26905612623
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26905612623

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26905612623
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26905612623

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26914041013
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26914041013

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=26914041013
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=26914041013

[wzhao18]

@kedarpotdar-nv @functionstackx @Oseltamivir Ready for review/merge. Thanks!

[Oseltamivir]

/reuse-sweep-run

[Oseltamivir]

lgtm

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27032104407
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27032104407