Add DSV4 B300 Dynamo vLLM disagg

[Oseltamivir]

Summary

• Add dsv4-fp4-b300-dynamo-vllm multi-node disaggregated config.
• Add B300-adapted DSV4 Dynamo vLLM recipes derived from the B200/GB200 recipe set, mapping each worker to one full 8-GPU B300 node.
• Carry over B200 single-node vLLM DSv4 optimizations: FP4 indexer cache, FULL_AND_PIECEWISE compilation, DSv4 tool/reasoning parser flags, larger cudagraph capture, and lower GMU on Mega-MoE DP paths.
• Update the B300 NV Slurm launcher to support dynamo-vllm for dsv4/fp4.

Validation

• python utils/matrix_logic/generate_sweep_configs.py test-config --config-files .github/configs/nvidia-master.yaml --runner-config .github/configs/runners.yaml --config-keys dsv4-fp4-b300-dynamo-vllm
• python utils/matrix_logic/generate_sweep_configs.py full-sweep --config-files .github/configs/nvidia-master.yaml --runner-config .github/configs/runners.yaml --model-prefix dsv4 --framework dynamo-vllm --runner-type b300 --multi-node
• PYTHONPATH=/tmp/ix-srt-slurm-vllm/src python - <<PY ... validate_config_file(disagg-b300-*.yaml) ... PY
• python -m pytest utils/matrix_logic/ -v
• bash -n runners/launch_b300-nv.sh
• git diff --check

[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://cookbook.sglang.io/
• https://docs.vllm.ai/projects/recipes/en/latest/

If it is not, please create a PR first before we can merge your 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/sgl-cookbook

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=25591370867
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25591370867

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25591370867
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25591370867

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25591851716
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25591851716

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25591851716
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25591851716

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25591851716
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25591851716

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25594451429
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25594451429

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25594633797
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25594633797

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25595208120
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25595208120

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25604092782
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25604092782

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25608462180
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25608462180

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=25613254628
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=25613254628