[AgentX] MiniMax-M3.1-Flash B200/GB200/B300 SGLang: add conc 1 low-latency configs / [AgentX] MiniMax-M3.1-Flash B200/GB200/B300 SGLang：新增 conc 1 低延迟配置

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/minimaxm31flash/sglang/b200-fp4-mtp/agentic-low-latency.yaml

@@ -0,0 +1,64 @@
+base:
+  schema: 2
+  name: minimaxm31flash-fp4-b200-sglang-agentic-low-latency
+  resources:
+    gpu_type: b200
+    gpus_per_node: 8
+  frontend:
+    type: sglang
+    enable_multiple_frontends: false
+  observability:
+    enabled: false
+    tachometer:
+      enabled: false
+  engine: sglang
+  # Cold loads of the 250 GB checkpoint plus JIT MSA kernel builds and graph capture outlast the default.
+  health_check:
+    max_attempts: 1440
+    interval_seconds: 5
+  roles:
+    agg:
+      nodes: 1
+      workers: 1
+      gpus: 4
+      # The MSA-V1 MonoKernel serves batch-size-1 DSpark steps; larger batches take
+      # the standard path.
+      args:
+        served-model-name: MiniMaxAI/MiniMax-M3.1-Flash
+        trust-remote-code: true
+        reasoning-parser: minimax-m3
+        tool-call-parser: minimax-m3
+        tensor-parallel-size: 4
+        moe-runner-backend: flashinfer_trtllm_routed
+        page-size: 128
+        speculative-algorithm: DSPARK
+        speculative-use-block-verification: true
+        speculative-num-steps: 1
+        speculative-eagle-topk: 1
+        speculative-num-draft-tokens: 8
+        speculative-dspark-block-size: 7
+        monokernel-backend: msa
+        disable-attn-tp-gather: true
+        max-running-requests: 8
+        mem-fraction-static: 0.8
+        watchdog-timeout: 3600
+        enable-metrics: true
+      env:
+        PYTHONNOUSERSITE: '1'
+        PYTHONUNBUFFERED: '1'
+        SGLANG_TIMEOUT_KEEP_ALIVE: '1800'
+        FMHA_SM100_ALLOW_JIT: '1'
+        ALLOW_JIT: '1'
+        # Canonical MegaAttention placement; it needs the GPUs to itself.
+        SGLANG_FORCE_COARSE_WAR_BARRIER: '1'
+  benchmark:
+    env:
+      # Warmup can leave bytes unacknowledged past AIPerf's 30 s default.
+      AIPERF_HTTP_TCP_USER_TIMEOUT: '900000'
+      AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'sglang:'
+
+override_tp4_c1:
+  benchmark:
+    env:
+      CONC: '1'
+      KV_OFFLOADING: none

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/minimaxm31flash/sglang/b300-fp4-mtp/agentic-low-latency.yaml

@@ -0,0 +1,64 @@
+base:
+  schema: 2
+  name: minimaxm31flash-fp4-b300-sglang-agentic-low-latency
+  resources:
+    gpu_type: b300
+    gpus_per_node: 8
+  frontend:
+    type: sglang
+    enable_multiple_frontends: false
+  observability:
+    enabled: false
+    tachometer:
+      enabled: false
+  engine: sglang
+  # Cold loads of the 250 GB checkpoint plus JIT MSA kernel builds and graph capture outlast the default.
+  health_check:
+    max_attempts: 1440
+    interval_seconds: 5
+  roles:
+    agg:
+      nodes: 1
+      workers: 1
+      gpus: 4
+      # The MSA-V1 MonoKernel serves batch-size-1 DSpark steps; larger batches take
+      # the standard path.
+      args:
+        served-model-name: MiniMaxAI/MiniMax-M3.1-Flash
+        trust-remote-code: true
+        reasoning-parser: minimax-m3
+        tool-call-parser: minimax-m3
+        tensor-parallel-size: 4
+        moe-runner-backend: flashinfer_trtllm_routed
+        page-size: 128
+        speculative-algorithm: DSPARK
+        speculative-use-block-verification: true
+        speculative-num-steps: 1
+        speculative-eagle-topk: 1
+        speculative-num-draft-tokens: 8
+        speculative-dspark-block-size: 7
+        monokernel-backend: msa
+        disable-attn-tp-gather: true
+        max-running-requests: 8
+        mem-fraction-static: 0.8
+        watchdog-timeout: 3600
+        enable-metrics: true
+      env:
+        PYTHONNOUSERSITE: '1'
+        PYTHONUNBUFFERED: '1'
+        SGLANG_TIMEOUT_KEEP_ALIVE: '1800'
+        FMHA_SM100_ALLOW_JIT: '1'
+        ALLOW_JIT: '1'
+        # Canonical MegaAttention placement; it needs the GPUs to itself.
+        SGLANG_FORCE_COARSE_WAR_BARRIER: '1'
+  benchmark:
+    env:
+      # Warmup can leave bytes unacknowledged past AIPerf's 30 s default.
+      AIPERF_HTTP_TCP_USER_TIMEOUT: '900000'
+      AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'sglang:'
+
+override_tp4_c1:
+  benchmark:
+    env:
+      CONC: '1'
+      KV_OFFLOADING: none

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/minimaxm31flash/sglang/gb200-fp4-mtp/agentic-low-latency.yaml

@@ -0,0 +1,64 @@
+base:
+  schema: 2
+  name: minimaxm31flash-fp4-gb200-sglang-agentic-low-latency
+  resources:
+    gpu_type: gb200
+    gpus_per_node: 4
+  frontend:
+    type: sglang
+    enable_multiple_frontends: false
+  observability:
+    enabled: false
+    tachometer:
+      enabled: false
+  engine: sglang
+  # Cold loads of the 250 GB checkpoint plus JIT MSA kernel builds and graph capture outlast the default.
+  health_check:
+    max_attempts: 1440
+    interval_seconds: 5
+  roles:
+    agg:
+      nodes: 1
+      workers: 1
+      gpus: 4
+      # The MSA-V1 MonoKernel serves batch-size-1 DSpark steps; larger batches take
+      # the standard path.
+      args:
+        served-model-name: MiniMaxAI/MiniMax-M3.1-Flash
+        trust-remote-code: true
+        reasoning-parser: minimax-m3
+        tool-call-parser: minimax-m3
+        tensor-parallel-size: 4
+        moe-runner-backend: flashinfer_trtllm_routed
+        page-size: 128
+        speculative-algorithm: DSPARK
+        speculative-use-block-verification: true
+        speculative-num-steps: 1
+        speculative-eagle-topk: 1
+        speculative-num-draft-tokens: 8
+        speculative-dspark-block-size: 7
+        monokernel-backend: msa
+        disable-attn-tp-gather: true
+        max-running-requests: 8
+        mem-fraction-static: 0.8
+        watchdog-timeout: 3600
+        enable-metrics: true
+      env:
+        PYTHONNOUSERSITE: '1'
+        PYTHONUNBUFFERED: '1'
+        SGLANG_TIMEOUT_KEEP_ALIVE: '1800'
+        FMHA_SM100_ALLOW_JIT: '1'
+        ALLOW_JIT: '1'
+        # Canonical MegaAttention placement; it needs the GPUs to itself.
+        SGLANG_FORCE_COARSE_WAR_BARRIER: '1'
+  benchmark:
+    env:
+      # Warmup can leave bytes unacknowledged past AIPerf's 30 s default.
+      AIPERF_HTTP_TCP_USER_TIMEOUT: '900000'
+      AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'sglang:'
+
+override_tp4_c1:
+  benchmark:
+    env:
+      CONC: '1'
+      KV_OFFLOADING: none

inferencex-e2e/configs/nvidia-master.yaml

@@ -5633,6 +5633,8 @@ minimaxm31flash-fp4-b200-sglang-agentic-dspark:
     - search-space:
       # Balanced: TP4.
       - { tp: 4, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 16, 32], srt-recipe: agentic.yaml }
+      # Low-Latency: MSA-V1 MonoKernel at batch size 1.
+      - { tp: 4, spec-decoding: mtp, kv-offloading: none, conc-list: [1], srt-recipe: agentic-low-latency.yaml }
     - dram-utilization: 0.80
       search-space:
       - { tp: 4, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [32, 40, 48, 56, 64], srt-recipe: agentic.yaml }
@@ -5825,6 +5827,8 @@ minimaxm31flash-fp4-gb200-sglang-agentic-dspark:
     - search-space:
       # Balanced: TP4.
       - { tp: 4, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 16, 32], srt-recipe: agentic.yaml }
+      # Low-Latency: MSA-V1 MonoKernel at batch size 1.
+      - { tp: 4, spec-decoding: mtp, kv-offloading: none, conc-list: [1], srt-recipe: agentic-low-latency.yaml }
     - dram-utilization: 0.80
       search-space:
       - { tp: 4, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [32, 40, 48, 56, 64], srt-recipe: agentic.yaml }
@@ -9681,6 +9685,8 @@ minimaxm31flash-fp4-b300-sglang-agentic-dspark:
     - search-space:
       # Balanced: TP4.
       - { tp: 4, spec-decoding: mtp, kv-offloading: none, conc-list: [1, 4, 8, 16, 32], srt-recipe: agentic.yaml }
+      # Low-Latency: MSA-V1 MonoKernel at batch size 1.
+      - { tp: 4, spec-decoding: mtp, kv-offloading: none, conc-list: [1], srt-recipe: agentic-low-latency.yaml }
     - dram-utilization: 0.80
       search-space:
       - { tp: 4, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [32, 64, 80, 96], srt-recipe: agentic.yaml }

inferencex-e2e/perf-changelog.yaml

@@ -11489,3 +11489,13 @@
     - "Update the MI355X DeepSeek-V4-Pro-0813 SGLang MoRI disaggregated AgentX serving image from lmsysorg/sglang-rocm:v0.5.22-rocm720-mi35x-20261019 to lmsysorg/sglang-rocm:v0.5.22-rocm10-mi35x-20261026, the latest published MI35x ROCm 10 nightly (digest sha256:edabad83728ab04e3eae39f315eb16261eeb64dbdb051112bd22fc88560b60c4)."
     - "Preserve all eight concurrency points (1, 16, 32, 48, 192, 256, 384, 512), prefill/decode topology, DSpark settings, UMBP offload, MoRI transfer, and benchmark client image. The workload binder supplies the serving image to every selected recipe variant."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4571
+
+- config-keys:
+    - minimaxm31flash-fp4-b200-sglang-agentic-dspark
+    - minimaxm31flash-fp4-gb200-sglang-agentic-dspark
+    - minimaxm31flash-fp4-b300-sglang-agentic-dspark
+  description:
+    - "Add a Low-Latency TP4 point at conc 1 with the MSA-V1 MonoKernel to the existing curve, following the GB300 recipe."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4509
+  scenario-type: [agentic-coding]
+  append-only: true