@@ -6,60 +6,57 @@ base:
     model: {}
     container: {}
     frameworks:
-      dynamo: "1.2.1"
+      dynamo: "1.6.0.dev20261010"
   dynamo:
     install: true
     source:
-      wheel: "1.2.1"
+      wheel: "1.6.0.dev20261010"
   environment:
-    DYNAMO_WHEEL_DIRS: "/srtctl-wheels"
-    # The frontend shares Grace CPU capacity with the long TP8 cold start.
+    PYTHONHASHSEED: "0"
+    WITH_NVIDIA_PEERMEM: "0"
+    EVAL_MAX_MODEL_LEN: "1048576"
     ETCD_LEASE_TTL: "7200"
-  setup_script: vllm-container-deps.sh
   slurm:
-    time_limit: "8:00:00"
+    time_limit: "4:00:00"
   health_check:
     max_attempts: 2160
     interval_seconds: 10
   resources:
     gpu_type: "gb300"
     gpus_per_node: 4
   services:
-    - name: etcd
-      type: etcd
-      placement:
-        node: infra
-    - name: nats
-      type: nats
-      placement:
-        node: infra
-      options:
-        max_payload_mb: 32
-    - name: mooncake-master
-      type: mooncake-master
-      options:
-        store_config:
-          metadata_server: "P2PHANDSHAKE"
-          global_segment_size: "150GB"
-          local_buffer_size: "4GB"
-          protocol: "rdma"
-          device_name: "@fabric.rdma-devices"
-          mode: "embedded"
-          enable_offload: false
+  - name: "etcd"
+    type: "etcd"
+    placement:
+      node: "infra"
+  - name: "nats"
+    type: "nats"
+    placement:
+      node: "infra"
+    options:
+      max_payload_mb: 32
   frontend:
-    type: dynamo
+    type: "dynamo"
     enable_multiple_frontends: false
+    env:
+      ETCD_LEASE_TTL: "120"
+      DYN_ROUTER_TEMPERATURE: "0"
+      DYN_TOKENIZER_CACHE: "1"
+      DYN_TOKENIZER_CACHE_BYTES: "8000000000"
+      DYN_TOKENIZER: "fastokens"
+      DYN_TCP_REQUEST_TIMEOUT: "30"
+      DYN_REQUEST_PLANE: "tcp"
+      DYN_FRONTEND_ENABLE_KV_EVENTS: "0"
+      DYN_LOGGING_JSONL: "false"
+      OTEL_EXPORT_ENABLED: "false"
     args:
       router-mode: "kv"
-      router-reset-states: true
-      router-temperature: 0.0
-      router-queue-threshold: 65536
+      router-session-affinity-ttl-secs: "14400"
       active-decode-blocks-threshold: "None"
       active-prefill-tokens-threshold: "None"
       active-prefill-tokens-threshold-frac: "None"
-      tokenizer: "fastokens"
   engine:
-    type: vllm
+    type: "vllm"
     connector:
   roles:
     agg:
@@ -70,6 +67,7 @@ base:
         TRANSFORMERS_CACHE: "/hf_hub_cache"
         VLLM_RPC_TIMEOUT: "600000"
         VLLM_LOG_STATS_INTERVAL: "1"
+        VLLM_V2_WARMUP_MAX_NUM_SEQS: "16"
         TILELANG_CLEANUP_TEMP_FILES: "1"
         VLLM_USE_NCCL_SYMM_MEM: "0"
         TORCH_SYMMMEM: "NVSHMEM"
@@ -81,75 +79,98 @@ base:
         VLLM_USE_RUST_FRONTEND: "1"
         VLLM_ALLREDUCE_USE_FLASHINFER: "1"
         VLLM_FLASHINFER_ALLREDUCE_BACKEND: "auto"
-        VLLM_MOONCAKE_LOAD_RECV_THREADS: "20"
         VLLM_PREFIX_CACHE_RETENTION_INTERVAL: "32768"
+        VLLM_USE_BREAKABLE_CUDAGRAPH: "0"
         VLLM_SPARSE_INDEXER_MAX_LOGITS_MB: "1024"
         VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: "1800"
         UCX_MEMTYPE_CACHE: "n"
         UCX_NET_DEVICES: "@fabric.rdma-device-ports"
         UCX_TLS: "rc,cuda_copy"
         NCCL_IB_HCA: "@fabric.rdma-devices"
         NCCL_P2P_LEVEL: "NVL"
-        MC_ENABLE_DEST_DEVICE_AFFINITY: "1"
-        MC_STORE_CLIENT_METRIC: "1"
-        MC_STORE_CLIENT_METRIC_INTERVAL: "5"
-        MC_TE_METRIC: "0"
       args:
-        served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+        served-model-name: "deepseek-ai/DeepSeek-V4-Pro-0813"
         kv-cache-dtype: "fp8"
         pipeline-parallel-size: 1
         disable-custom-all-reduce: true
-        attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV4","use_prefill_query_quantization":true,"use_fp4_indexer_cache":true}'
+        attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV4","use_prefill_query_quantization":true,"indexer_kv_dtype":"mxfp4"}'
         max-model-len: 1048576
         max-num-batched-tokens: 8192
         trust-remote-code: true
         no-enable-flashinfer-autotune: true
         block-size: 256
-        speculative-config: '{"method":"mtp","num_speculative_tokens":3}'
-        gpu-memory-utilization: 0.94
-        stream-interval: 10
+        cudagraph-metrics: true
+        compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[6,7,12,14,18,21,24,28,30,35,36,42,48,49,54,56,60,63,66,70,72,77,78,84,90,91,96,98,105,112],"mode":0}'
+        speculative-config: '{"method":"dspark","num_speculative_tokens":6,"draft_sample_method":"probabilistic"}'
+        gpu-memory-utilization: 0.9
+        stream-interval: 1
         no-disable-hybrid-kv-cache-manager: true
         tokenizer-mode: "deepseek_v4"
+        numa-bind: true
   sbatch_directives:
     mem: "0"
   srun_options:
     container-remap-root: ""
+    cpu-bind: "none"
   benchmark:
     env:
       AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
+      AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: "true"
+      AIPERF_DATASET_MMAP_CACHE_ENABLED: "false"
 
-override_tp4:
-  name: "svf-vllm-agg-gb300-tp4-mtp-agentic"
+override_tp8_c1:
+  name: "svf-vllm-agg-gb300-tp8-c1-mtp-agentic"
+  roles:
+    agg:
+      nodes: 2
+      gpus: 8
+      env:
+        DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-agg-tp8-c1-mtp-{job_id}"
+      args:
+        tensor-parallel-size: 8
+        max-num-seqs: 1
+
+# GB300 AgentX aggregate TP4 worker on one node at concurrency 1,
+# DSpark K=6 (tuning run 3184212).
+override_tp4_c1:
+  name: "svf-vllm-agg-gb300-tp4-c1-mtp-agentic"
   roles:
     agg:
       nodes: 1
       gpus: 4
       env:
-        VLLM_V2_WARMUP_MAX_NUM_SEQS: "32"
-        DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-agg-tp4-mtp-{job_id}"
+        DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-agg-tp4-c1-mtp-{job_id}"
       args:
         tensor-parallel-size: 4
-        max-num-seqs: 32
-        compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,64,68,72,76,80,84,88,92,96,100,104,108,112,116,120,124,128],"mode":0}'
-  sbatch_directives:
-    cpus-per-task: "72"
-  benchmark:
-    env:
-      # Avoid concurrent readers observing a mismatched mmap data/index pair.
-      AIPERF_DATASET_MMAP_CACHE_ENABLED: "false"
+        max-num-seqs: 1
 
-override_tp8:
-  name: "svf-vllm-agg-gb300-tp8-mtp-agentic"
+# GB300 AgentX aggregate TP4 worker on one node at concurrency 2,
+# DSpark K=6 (tuning run 3184214).
+override_tp4_c2:
+  name: "svf-vllm-agg-gb300-tp4-c2-mtp-agentic"
   roles:
     agg:
-      nodes: 2
-      gpus: 8
+      nodes: 1
+      gpus: 4
       env:
-        VLLM_V2_WARMUP_MAX_NUM_SEQS: "16"
-        DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-agg-mtp-{job_id}"
+        DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-agg-tp4-c2-mtp-{job_id}"
       args:
-        tensor-parallel-size: 8
-        max-num-seqs: 16
-        compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,64],"mode":0}'
-  sbatch_directives:
-    cpus-per-task: "144"
+        tensor-parallel-size: 4
+        max-num-seqs: 2
+
+# GB300 AgentX aggregate TP4 worker on one node at concurrency 4,
+# DSpark K=6 (tuning run 3184216).
+override_tp4_c4:
+  name: "svf-vllm-agg-gb300-tp4-c4-mtp-agentic"
+  frontend:
+    env:
+      DYN_ROUTER_QUEUE_THRESHOLD: "65536"
+  roles:
+    agg:
+      nodes: 1
+      gpus: 4
+      env:
+        DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-agg-tp4-c4-mtp-{job_id}"
+      args:
+        tensor-parallel-size: 4
+        max-num-seqs: 4