[Juntian777]

Overview

Document the DeepSeek-V4.1-Flash B300 aggregate AgentX config from SemiAnalysisAI/InferenceX#4332 (run 39351570949). The recipe already covers B200 aggregate and B300/GB300 Dynamo disaggregated serving; B300 aggregate is missing. No open PR covers it.

Claims

Guide text only, in recipes/models/deepseek-ai/DeepSeek-V4.1-Flash.yaml. The command builder output does not change.

• New subsection Aggregate serving on B300 under ## AgentX tuning on NVIDIA GPUs, with the image, checkpoint and a link to the full InferenceX config.
• One table gives per-point --max-num-seqs, --max-num-batched-tokens, --prefill-schedule-interval and attention backend for TP4 (c1, c8, c16), DEP2 (c8–c64) and DEP2 with KV offload (c128, c192).
• TP4: THP Engram tables, decode_backend=latency_specialized, and capture size 5 for the one-request DSpark draft graph.
• DEP2: MegaMoE, Engram embedding across DP, consistent-hash vLLM Router; at c128 and c192, SimpleCPUOffloadConnector with lazy offload, 50 GiB per DP rank.

Validation

• All values come from the InferenceX recipe at [Kernel] Support Fp8 Checkpoints (Dynamic + Static) #4332 head ad96b211. All 9 benchmark points and both accuracy evals (TP4 c16, DEP2 offload c192) passed in the run above.
• node scripts/validate-source.mjs --source vllm --dir <this checkout>/recipes in vllm-project/recipes passes, and pre-commit passes.
• No GPU run for this PR.

Details

decode_backend=latency_specialized for DSv4.1 comes from #61868 and #61869, which are not yet on main; the guide says so and names the image (vllm/vllm-openai:cu134-nightly-x86_64-dev-20261019-mk, public on Docker Hub). Every other flag and config key exists on main. The per-point settings are too specific for the general command builder, so this follows #62561 and stays in the guide.

AI assistance was used (Claude Code).

---

Pull Request Checklist

• I used vLLM's /pr-checklist skill. (Mandatory for agents, optional for humans).

• AI assistance was used during the creation of this PR.

• Design Fit: Minimizes impact on core components, reuses existing functionality, and justifies added complexity.

• Testing and Validation: Validates the change and ensures any added tests are meaningful and reliable, with CI coverage or documented CI resource constraints and validation performed outside CI.

• Code Quality and Style: Keeps code and comments clear and concise, and updates relevant documentation and examples.

• Pull Request Contents: Includes a brief summary and relevant links, supports claims with evidence, explains root causes and implementation trade-offs, and follows the contributing guide.

🤖 Generated with Claude Code

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[ivanium]

/ci run

[github-actions]

✅ Triggered Buildkite CI #96122 for commit 7e0cb86f2e69.