diff --git a/benchmarks/single_node/agentic/dsv4_fp4_b300_vllm_mtp.sh b/benchmarks/single_node/agentic/dsv4_fp4_b300_vllm_mtp.sh index 50e429b2fe..c208e5fb5d 100755 --- a/benchmarks/single_node/agentic/dsv4_fp4_b300_vllm_mtp.sh +++ b/benchmarks/single_node/agentic/dsv4_fp4_b300_vllm_mtp.sh @@ -93,6 +93,13 @@ if [ "$DP_ATTENTION" = "true" ]; then agentic_pip_install --quiet "vllm-router==$VLLM_ROUTER_VERSION" fi +# AIPerf automatically scrapes the public endpoint's /metrics URL. That is the +# vLLM engine for pure TP, but the native router for DP-attention. Explicitly +# add the engine endpoint so every topology captures vLLM metrics; AIPerf +# deduplicates it against the automatic endpoint in pure-TP runs. +export AIPERF_SERVER_METRICS_URLS="http://localhost:${VLLM_BACKEND_PORT}/metrics" +export AIPERF_REQUIRED_SERVER_METRIC_PREFIX="vllm:" + # Match the environment used by v4pro-b300.yaml. export VLLM_USE_V2_MODEL_RUNNER=1 export VLLM_ENGINE_READY_TIMEOUT_S=3600 diff --git a/perf-changelog.yaml b/perf-changelog.yaml index 7f25314145..f8e593b9dd 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -5526,6 +5526,15 @@ - "Enable SGLang backend metrics on every aggregate, prefill, and decode engine, and fail before publishing a partial trace artifact if required sglang: metrics are absent." - "Record the recipes' active HiCache host-DRAM tier and, for disaggregated points, Dynamo router commit 5a638087 in nvidia-master metadata so generated artifacts no longer report kv_offloading=none, allocated_cpu_dram_gb=0, or a null router." pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2477 + +- config-keys: + - dsv4-fp4-b300-vllm-agentic-mtp + scenario-type: + - agentic-coding + description: + - "Refresh submission with up to date AgentX harness." + - "Capture the vLLM backend Prometheus endpoint for every topology and require vllm: metrics before publishing AgentX results." + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2517 - config-keys: - kimik3-fp4-h200-vllm-agentic