[NV] Add MiniMax-M2.5 FP8 GB200 Dynamo vLLM recipes

.github/configs/nvidia-master.yaml

@@ -10080,6 +10080,141 @@ qwen3.5-fp8-h100-sglang-agentic:
       - { tp: 8, ep: 8, offloading: none,    conc-list: [1, 2, 4, 8, 12, 14, 16] }
       - { tp: 8, ep: 8, offloading: hicache, conc-list: [12, 14, 16, 20, 24, 28, 32, 42] }
 
+minimaxm2.5-fp8-gb200-dynamo-vllm:
+  image: vllm/vllm-openai:v0.20.1
+  model: MiniMaxAI/MiniMax-M2.5
+  model-prefix: minimaxm2.5
+  runner: gb200
+  precision: fp8
+  framework: dynamo-vllm
+  multinode: true
+  disagg: true
+  scenarios:
+    fixed-seq-len:
+    - isl: 1024
+      osl: 1024
+      search-space:
+      - conc-list: [1, 4, 8, 16, 32, 64]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p1d-tp4.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [2, 32, 64, 128, 256, 512]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p2d-tp4.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [1024]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p3d-tp4ep.yaml"
+        decode:
+          num-worker: 3
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [512, 1024]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p1d-dep8.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+      - conc-list: [4096]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p4d-dep2.yaml"
+        decode:
+          num-worker: 4
+          tp: 2
+          ep: 2
+          dp-attn: true
+      - conc-list: [4096, 8192]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p3d-dep4.yaml"
+        decode:
+          num-worker: 3
+          tp: 4
+          ep: 4
+          dp-attn: true
+    - isl: 8192
+      osl: 1024
+      search-space:
+      - conc-list: [1, 4, 8, 16, 32, 64, 128]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [256, 512]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4ep.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [1024, 2048, 4096]
+        prefill:
+          num-worker: 3
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-3p2d-dep4.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 4
+          dp-attn: true
+
 
 minimaxm2.5-fp8-b200-dynamo-vllm:
   image: vllm/vllm-openai:v0.20.1

benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p1d-tp4.yaml

@@ -0,0 +1,67 @@
+name: "minimax-m2.5-vllm-disagg-gb200-1p1d-tp4"
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "1x4x8x16x32x64"

benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p2d-tp4.yaml

@@ -0,0 +1,67 @@
+name: "minimax-m2.5-vllm-disagg-gb200-1p2d-tp4"
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "2x32x64x128x256x512"

benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p3d-tp4ep.yaml

@@ -0,0 +1,72 @@
+name: "minimax-m2.5-vllm-disagg-gb200-1p3d-tp4ep"
+
+# Rate-matched tp4ep for FP8 GB200 1k/1k.
+# X_tp4ep_fp8_gb200 = 17.9k tok/s; P_per_worker = 48k; ideal X/P = 0.37; 1P:3D = 0.33 ✓
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 3
+  prefill_workers: 1
+  decode_workers: 3
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "1024"