[B300][vLLM] Add MiniMax-M2.5 FP4 disagg Dynamo configs

.github/configs/nvidia-master.yaml

@@ -10532,3 +10532,216 @@ minimaxm2.5-fp4-b200-dynamo-vllm:
           tp: 4
           ep: 4
           dp-attn: true
+
+minimaxm2.5-fp4-b300-dynamo-vllm:
+  image: vllm/vllm-openai:v0.20.1
+  model: nvidia/MiniMax-M2.5-NVFP4
+  model-prefix: minimaxm2.5
+  runner: b300
+  precision: fp4
+  framework: dynamo-vllm
+  multinode: true
+  disagg: true
+  scenarios:
+    fixed-seq-len:
+    - isl: 1024
+      osl: 1024
+      search-space:
+      - conc-list: [4, 16]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [4]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp8-1p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 1
+          dp-attn: false
+      - conc-list: [8, 16]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4-1p2d.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [32, 64, 128]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [64, 128, 256, 1024]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/tp4ep-1p3d.yaml"
+        decode:
+          num-worker: 3
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [4096]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-1p2d.yaml"
+        decode:
+          num-worker: 2
+          tp: 2
+          ep: 2
+          dp-attn: true
+      - conc-list: [2048, 4096]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d.yaml"
+        decode:
+          num-worker: 3
+          tp: 2
+          ep: 2
+          dp-attn: true
+      - conc-list: [6144, 8192]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep2-2p3d-c6144.yaml"
+        decode:
+          num-worker: 3
+          tp: 2
+          ep: 2
+          dp-attn: true
+      - conc-list: [1024, 1536, 2048, 4096]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/1k1k/dep8-2p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+    - isl: 8192
+      osl: 1024
+      search-space:
+      - conc-list: [2, 4, 8, 16]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4-1p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [4]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp8-1p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 1
+          dp-attn: false
+      - conc-list: [32, 128]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-1p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [64, 128, 256, 512]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/tp4ep-2p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [384, 512]
+        prefill:
+          num-worker: 4
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep4-4p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: true
+      - conc-list: [384]
+        prefill:
+          num-worker: 4
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5/8k1k/dep8-4p1d.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true

benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-b300/1k1k/dep2-1p2d.yaml

@@ -0,0 +1,72 @@
+name: "minimax-m2.5-vllm-disagg-b300-decode-2xdep2"
+
+model:
+  path: "minimax-m2.5-nvfp4"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp4"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 2
+  gpus_per_prefill: 1
+  gpus_per_decode: 2
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+  allow_prefill_decode_colocation: true
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+    UCX_TLS: "cuda_copy,rc"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+    UCX_TLS: "cuda_copy,rc"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      max-model-len: 2048
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+      stream-interval: 128
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      no-enable-prefix-caching: true
+      max-model-len: 2048
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+      max-num-seqs: 864
+      gpu-memory-utilization: 0.90
+      stream-interval: 128
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "4096"
+  random_range_ratio: 0.8

benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-b300/1k1k/dep2-2p3d-c6144.yaml

@@ -0,0 +1,72 @@
+name: "minimax-m2.5-vllm-disagg-b300-decode-2p3xdep2-c6144"
+
+model:
+  path: "minimax-m2.5-nvfp4"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp4"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 2
+  decode_workers: 3
+  gpus_per_prefill: 1
+  gpus_per_decode: 2
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+  allow_prefill_decode_colocation: true
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+    UCX_TLS: "cuda_copy,rc"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+    UCX_TLS: "cuda_copy,rc"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      max-model-len: 2048
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+      stream-interval: 128
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      no-enable-prefix-caching: true
+      max-model-len: 2048
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+      max-num-seqs: 864
+      gpu-memory-utilization: 0.90
+      stream-interval: 128
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "6144x8192"
+  random_range_ratio: 0.8