config: update GB300 DSV41flash vLLM image and retune TP4/DEP2

Move the GB300 DeepSeek-V4.1-Flash vLLM AgentX recipe to
nightly-dev-arm64-cu130-ac9126e58aa7 and enable FlashInfer autotuning.
Run TP4 at CONC 1-16 and replace TP2 with DEP2 (TP1 x DP2 + EP2,
MegaMoE) at CONC 8-192 behind a consistent-hash vLLM Router. Append the
perf-changelog entry and document the GB300 settings.

将 GB300 DeepSeek-V4.1-Flash vLLM AgentX 配方更新到
nightly-dev-arm64-cu130-ac9126e58aa7，并开启 FlashInfer autotune。
TP4 覆盖 CONC 1-16；用 DEP2（TP1 x DP2 + EP2，MegaMoE）替代 TP2，覆盖
CONC 8-192，前置一致性哈希 vLLM Router。追加 perf-changelog 条目，并补充
GB300 配置文档。

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Author: Juntian777

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv41flash/vllm/gb300-fp4-mtp/agentic.yaml

@@ -6,7 +6,7 @@ base:
   name: dsv41flash-fp4-gb300-vllm-agentic
   model:
     path: hf:deepseek-ai/DeepSeek-V4.1-Flash
-    container: vllm/vllm-openai:nightly-ac68c3087215e0a4f3cdfa218508c6aada57235d
+    container: vllm/vllm-openai:nightly-dev-arm64-cu130-ac9126e58aa7
     precision: fp4
   resources:
     gpu_type: gb300
@@ -37,17 +37,21 @@ base:
         enable-auto-tool-choice: true
         reasoning-parser: deepseek_v41
         engram-config: '{"cpu_offload":true}'
-        # FlashInfer sparse attention with MXFP4 indexer KV and sparse logits.
+        kernel-config: '{"enable_flashinfer_autotune":true}'
         attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV41","indexer_kv_dtype":"mxfp4","indexer_sparse_logits":true}'
         kv-cache-dtype: fp8
         # Five-token DSpark with probabilistic drafting. Throughput runs replace
         # block rejection with the golden acceptance length.
         speculative-config: '{"method":"dspark","num_speculative_tokens":5,"draft_sample_method":"probabilistic","rejection_sample_method":"block","enable_adaptive_verification":false}'
         max-model-len: 1048576
-        max-num-seqs: 256
+        # Piecewise graph capture sizes are multiples of the six-token DSpark
+        # verification block, denser for small batches.
+        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
+        max-cudagraph-capture-size: 8190
+        max-num-batched-tokens: 8192
         disable-uvicorn-access-log: true
       env:
-        VLLM_DISABLED_KERNELS: FlashInferCutedslMxfp8LinearKernel
+        VLLM_USE_V2_MODEL_RUNNER: '1'
         VLLM_USE_RUST_FRONTEND: '1'
         PYTHONUNBUFFERED: '1'
         VLLM_ENGINE_READY_TIMEOUT_S: '7200'
@@ -58,216 +62,66 @@ base:
       MODEL: deepseek-ai/DeepSeek-V4.1-Flash
       AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'vllm:'
 
-# One variant per point. Piecewise graph capture sizes are multiples of the
-# six-token DSpark verification block, denser for small batches, and each
-# batched-token limit matches the largest captured graph: CONC <= 4 and TP2
-# CONC 128 capture up to 2046 tokens (the latter at 0.97 memory utilization),
-# every other point up to 8190.
-override_tp4_c1:
-  roles:
-    agg:
-      gpus: 4
-      args:
-        tensor-parallel-size: 4
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046]}'
-        max-cudagraph-capture-size: 2046
-        max-num-batched-tokens: 2048
-  benchmark:
-    env:
-      CONC: '1'
-
-override_tp4_c2:
-  roles:
-    agg:
-      gpus: 4
-      args:
-        tensor-parallel-size: 4
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046]}'
-        max-cudagraph-capture-size: 2046
-        max-num-batched-tokens: 2048
-  benchmark:
-    env:
-      CONC: '2'
-
-override_tp4_c4:
-  roles:
-    agg:
-      gpus: 4
-      args:
-        tensor-parallel-size: 4
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046]}'
-        max-cudagraph-capture-size: 2046
-        max-num-batched-tokens: 2048
-  benchmark:
-    env:
-      CONC: '4'
-
-override_tp4_c8:
-  roles:
-    agg:
-      gpus: 4
-      args:
-        tensor-parallel-size: 4
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '8'
-
-override_tp4_c16:
-  roles:
-    agg:
-      gpus: 4
-      args:
-        tensor-parallel-size: 4
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '16'
-
-override_tp4_c32:
-  roles:
-    agg:
-      gpus: 4
-      args:
-        tensor-parallel-size: 4
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '32'
-
-override_tp4_c64:
-  roles:
-    agg:
-      gpus: 4
-      args:
-        tensor-parallel-size: 4
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '64'
+# Native zip expansion creates one variant per concurrency.
 
-override_tp4_c128:
+# TP4 covers low concurrency with THP Engram tables.
+zip_override_tp4:
   roles:
     agg:
       gpus: 4
       args:
         tensor-parallel-size: 4
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '128'
-
-override_tp2_c2:
-  roles:
-    agg:
-      gpus: 2
-      args:
-        tensor-parallel-size: 2
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046]}'
-        max-cudagraph-capture-size: 2046
-        max-num-batched-tokens: 2048
-  benchmark:
-    env:
-      CONC: '2'
-
-override_tp2_c4:
-  roles:
-    agg:
-      gpus: 2
-      args:
-        tensor-parallel-size: 2
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046]}'
-        max-cudagraph-capture-size: 2046
-        max-num-batched-tokens: 2048
-  benchmark:
-    env:
-      CONC: '4'
-
-override_tp2_c8:
-  roles:
-    agg:
-      gpus: 2
-      args:
-        tensor-parallel-size: 2
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '8'
-
-override_tp2_c16:
-  roles:
-    agg:
-      gpus: 2
-      args:
-        tensor-parallel-size: 2
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '16'
-
-override_tp2_c32:
-  roles:
-    agg:
-      gpus: 2
-      args:
-        tensor-parallel-size: 2
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '32'
-
-override_tp2_c64:
-  roles:
-    agg:
-      gpus: 2
-      args:
-        tensor-parallel-size: 2
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
-        max-cudagraph-capture-size: 8190
-        max-num-batched-tokens: 8192
-  benchmark:
-    env:
-      CONC: '64'
-
-override_tp2_c128:
-  roles:
-    agg:
-      gpus: 2
-      args:
-        tensor-parallel-size: 2
-        compilation-config: '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046]}'
-        max-cudagraph-capture-size: 2046
-        max-num-batched-tokens: 2048
-        gpu-memory-utilization: 0.97
-  benchmark:
-    env:
-      CONC: '128'
-
-override_tp2_c1:
+        engram-config: '{"cpu_offload":true,"use_thp":true}'
+        max-num-seqs: [8, 8, 16, 32]
+        compilation-config:
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092]}'
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092]}'
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190],"decoder_replay_cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2048,2304,2560,2816,3072,3328,3584,3840,4096]}'
+        max-cudagraph-capture-size: [4092, 4092, 8190, 8190]
+        max-num-batched-tokens: [4096, 4096, 8192, 8192]
+  benchmark:
+    env:
+      CONC: ['1', '4', '8', '16']
+
+# DEP2: TP1 x DP2 + EP2 with MegaMoE behind a consistent-hash router;
+# MegaAttention at CONC 128 and above.
+zip_override_dep2:
+  frontend:
+    type: vllm-router
+    args:
+      policy: consistent_hash
+      prometheus-host: 127.0.0.1
+      prometheus-port: 18000
+      request-timeout-secs: 14400
+      disable-retries: true
+  setup_script: vllm-router-0.1.14.sh
   roles:
     agg:
       gpus: 2
       args:
-        tensor-parallel-size: 2
-        max-cudagraph-capture-size: 64
-        max-num-batched-tokens: 4096
-        max-num-seqs: 16
-  benchmark:
-    env:
-      CONC: '1'
+        tensor-parallel-size: 1
+        data-parallel-size: 2
+        enable-expert-parallel: true
+        kernel-config: '{"enable_flashinfer_autotune":true,"moe_backend":"deep_gemm_mega_moe"}'
+        engram-config: '{"cpu_offload":true,"embedding_across_dp":true}'
+        max-num-seqs: [16, 16, 32, 64, 128, 192]
+        attention-config:
+          - '{"backend":"FLASHINFER_MLA_SPARSE_DSV41","indexer_kv_dtype":"mxfp4","indexer_sparse_logits":true}'
+          - '{"backend":"FLASHINFER_MLA_SPARSE_DSV41","indexer_kv_dtype":"mxfp4","indexer_sparse_logits":true}'
+          - '{"backend":"FLASHINFER_MLA_SPARSE_DSV41","indexer_kv_dtype":"mxfp4","indexer_sparse_logits":true}'
+          - '{"backend":"FLASHINFER_MLA_SPARSE_DSV41","indexer_kv_dtype":"mxfp4","indexer_sparse_logits":true}'
+          - '{"backend":"FLASHMLA_MEGA_ATTN_DSV41","indexer_kv_dtype":"mxfp4","indexer_sparse_logits":true}'
+          - '{"backend":"FLASHMLA_MEGA_ATTN_DSV41","indexer_kv_dtype":"mxfp4","indexer_sparse_logits":true}'
+        compilation-config:
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190]}'
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190],"decoder_replay_cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2048,2304,2560,2816,3072,3328,3584,3840,4096]}'
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190],"decoder_replay_cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2048,2304,2560,2816,3072,3328,3584,3840,4096]}'
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092,6144,8190],"decoder_replay_cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2048,2304,2560,2816,3072,3328,3584,3840,4096]}'
+          - '{"mode":"VLLM_COMPILE","cudagraph_mode":"FULL_AND_PIECEWISE","cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2046,3072,4092],"decoder_replay_cudagraph_capture_sizes":[6,12,18,24,30,36,48,60,72,96,120,144,192,240,288,384,480,576,768,1020,1536,2048]}'
+        max-cudagraph-capture-size: [8190, 8190, 8190, 8190, 8190, 4092]
+        max-num-batched-tokens: [8192, 8192, 8192, 8192, 16384, 16384]
+  benchmark:
+    env:
+      CONC: ['8', '16', '32', '64', '128', '192']

inferencex-e2e/configs/nvidia-master.yaml

@@ -8534,7 +8534,7 @@ dsv41flash-fp4-b200-sglang-agentic-dspark:
       - { tp: 2, ep: 2, kv-offloading: none, spec-decoding: mtp, conc-list: [1, 2, 4, 8], srt-recipe: benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/b200-fp4-mtp/agentic.yaml }
 
 dsv41flash-fp4-gb300-vllm-agentic-dspark:
-  image: vllm/vllm-openai:nightly-ac68c3087215e0a4f3cdfa218508c6aada57235d
+  image: vllm/vllm-openai:nightly-dev-arm64-cu130-ac9126e58aa7
   model: deepseek-ai/DeepSeek-V4.1-Flash
   model-prefix: dsv41flash
   runner: cluster:gb300-nv
@@ -8546,13 +8546,11 @@ dsv41flash-fp4-gb300-vllm-agentic-dspark:
     - dram-utilization: 0.80
       search-space:
       # Engram weights use UVA DRAM; the KV cache stays GPU-resident.
-      - { tp: 4, kv-offloading: none, spec-decoding: mtp, conc-list: [1, 2, 4, 8, 16, 32, 64, 128], srt-recipe: benchmarks/single_node/srt-slurm-recipes/dsv41flash/vllm/gb300-fp4-mtp/agentic.yaml }
-      # TP2 halves the GPU count per replica. Weights rise to ~175 GiB on each
-      # 277 GiB GPU with the Engram tables in pinned host DRAM, so the arm
-      # takes the same caps as the B200 TP2 arm: batched tokens 4096 (the
-      # indexer's logits buffer is 32 GiB at the upstream 16384) and graph
-      # capture stopped at 512, leaving ~36 GiB of KV per GPU.
-      - { tp: 2, kv-offloading: none, spec-decoding: mtp, conc-list: [1, 2, 4, 8, 16, 32, 64, 128], srt-recipe: benchmarks/single_node/srt-slurm-recipes/dsv41flash/vllm/gb300-fp4-mtp/agentic.yaml }
+      # TP4 covers low concurrency with FlashInfer attention and THP Engram tables.
+      - { tp: 4, kv-offloading: none, spec-decoding: mtp, conc-list: [1, 4, 8, 16], srt-recipe: benchmarks/single_node/srt-slurm-recipes/dsv41flash/vllm/gb300-fp4-mtp/agentic.yaml }
+      # DEP2 is native TP1 x DP2 + EP2 with MegaMoE behind a consistent-hash vLLM
+      # Router; FlashInfer attention up to conc 64, MegaAttention above.
+      - { tp: 2, ep: 2, dp-attn: true, kv-offloading: none, spec-decoding: mtp, conc-list: [8, 16, 32, 64, 128, 192], router: { name: vllm-router, version: "0.1.14" }, srt-recipe: benchmarks/single_node/srt-slurm-recipes/dsv41flash/vllm/gb300-fp4-mtp/agentic.yaml }
 
 # H200 AgentX arm for DeepSeek-V4.1-Flash. Upstream marks h200 verified and says
 # the GB200 NVL4 TP4 layout becomes TP8 on 8-GPU nodes, so this is TP8.

inferencex-e2e/docs/configuration-procedures.md

@@ -328,7 +328,7 @@ EAGLE3 K3, golden AL 2.78 and indexer CP are unchanged.
 
 The GB200 DSpark recipe uses a minimum CUDA graph capture size of 64 tokens to cover concurrent AgentX subagents. This raises c1/c2/c4 from 8/16/32 to 64; c8 and above retain their existing sizes. The full trace, AL 3.51, and Engram UVA settings are preserved; low-concurrency tail latency improvements require CI confirmation.
 The B200 DSpark recipe uses the same minimum capture size and preserves the same workload settings.
-The GB300 DSpark recipe uses the same minimum capture size and preserves the same workload settings.
+The GB300 DSpark recipe sets explicit capture sizes per point, described below.
 The H200 DSpark recipe uses the same minimum capture size and preserves the same workload settings.
 
 The B300 DSpark recipe sets explicit capture sizes per point, described below.
@@ -342,7 +342,7 @@ weights determine the recipe's `precision: fp4` label.
 
 The GPU-specific entry points share the text-only serving behavior, `deepseek_v41` tokenizer and
 parsers, 1M context, and the shared AgentX trace replay, power, metrics, and eval
-helpers. The TP4 concurrency range is 1–128. The shared script sizes graph capture
+helpers. Unless noted below, the TP4 concurrency range is 1–128. The shared script sizes graph capture
 for the six-token DSpark verification block. The srt-slurm single-node path mounts the checkout at `/infmax-workspace`, so
 AgentX runtime directories are not created under `/workspace`. Cluster model paths and persistent caches are reused.
 The recipe probes the serving port on the compute node and selects an available
@@ -355,6 +355,11 @@ behind a consistent-hash vLLM Router, switching to MegaAttention at 128 and abov
 `FULL_AND_PIECEWISE` CUDA graphs sized in multiples of the six-token verification block.
 Other SKUs continue to use the shared script.
 
+The GB300 entry uses `vllm/vllm-openai:nightly-dev-arm64-cu130-ac9126e58aa7` with FlashInfer
+autotuning. TP4 covers concurrency 1–16; DEP2 (TP1 x DP2 + EP2, DeepGEMM MegaMoE) covers 8–192
+behind a consistent-hash vLLM Router, switching to MegaAttention at 128 and above. All points use
+`FULL_AND_PIECEWISE` CUDA graphs sized in multiples of the six-token verification block.
+
 The GB300 launcher allows 7200 seconds for engine readiness. In [run 34504969146](https://github.com/SemiAnalysisAI/InferenceX/actions/runs/34504969146), the Rust frontend exhausted its 3600-second deadline while the engine was still capturing graphs; model loading alone took 18–23 minutes. This extends startup time without changing the benchmark duration or decoding settings.
 
 GPU sweep and eval evidence is required before calling any recipe validated.