Add DSv4-Pro FP4 GB200 SGLang disagg + MTP config (#1676)

* Add DSv4-Pro FP4 GB200 SGLang disagg + MTP config

Initial submission of the MTP-decoded variant of the DSv4-Pro FP4 disagg
GB200 SGLang config at 8k/1k. Eight prefill/decode topologies: two
low-latency (1p1d-tp8-tp8, 1p6d-dep8-tp8) and six mid-curve points
(1p1d through 6p1d-dep8-dep16). Each scenario sets
`spec-decoding: "mtp"` so the matrix turns on the MTP speculative-decode
path; chat template enabled accordingly.

Image: lmsysorg/sglang:nightly-dev-cu13-20260528-0abe6a85

* Update perf-changelog pr-link for #1676

* Use NVIDIA/srt-slurm:main for DSv4 SGLang clone (drop submission-branch pin)

* Fix GB200 DSv4-Pro FP4 model path to shared Lustre root

The dsv4/fp4 model path resolved to /mnt/numa1/models/deepseek-v4-pro/
(compute-node-local NVMe), which srtctl preflight reports as unavailable.
Point both the dynamo-sglang and dynamo-vllm branches at the shared Lustre
model root /mnt/lustre01/models/deepseek-v4-pro instead.

* Update perf-changelog.yaml with new configurations

Author: Ankur-singh

.github/configs/nvidia-master.yaml

@@ -9484,6 +9484,143 @@ qwen3.5-fp8-gb200-dynamo-sglang:
           ep: 16
           dp-attn: true
 
+
+# MTP variant of dsv4-fp4-gb200-dynamo-sglang.
+dsv4-fp4-gb200-dynamo-sglang-mtp:
+  image: lmsysorg/sglang:nightly-dev-cu13-20260528-0abe6a85
+  model: deepseek-ai/DeepSeek-V4-Pro
+  model-prefix: dsv4
+  runner: gb200
+  precision: fp4
+  framework: dynamo-sglang
+  multinode: true
+  disagg: true
+  scenarios:
+    fixed-seq-len:
+    - isl: 8192
+      osl: 1024
+      search-space:
+      # Low-latency baseline: 1p1d-tp8-tp8. 4 nodes.
+      - spec-decoding: "mtp"
+        conc-list: [1]
+        prefill:
+          num-worker: 1
+          tp: 8
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/8k1k/disagg-gb200-low-latency-1p1d-tp8-tp8-mtp.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 1
+          dp-attn: false
+      # Low-latency 1p6d-dep8-tp8: 1P (DEP=8) + 6 TP=8 decode workers. 14 nodes.
+      # Recipe runs concurrencies=32x64x128; matrix tracks the max.
+      - spec-decoding: "mtp"
+        conc-list: [128]
+        prefill:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/8k1k/disagg-gb200-low-latency-1p6d-dep8-tp8-mtp.yaml"
+        decode:
+          num-worker: 6
+          tp: 8
+          ep: 1
+          dp-attn: false
+      # Mid curve 1p1d-dep8-dep16. 6 nodes.
+      - spec-decoding: "mtp"
+        conc-list: [1024]
+        prefill:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/8k1k/disagg-gb200-mid-curve-1p1d-dep8-dep16-mtp.yaml"
+        decode:
+          num-worker: 1
+          tp: 16
+          ep: 16
+          dp-attn: true
+      # Mid curve 2p1d-dep8-dep16. 8 nodes.
+      - spec-decoding: "mtp"
+        conc-list: [2048]
+        prefill:
+          num-worker: 2
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/8k1k/disagg-gb200-mid-curve-2p1d-dep8-dep16-mtp.yaml"
+        decode:
+          num-worker: 1
+          tp: 16
+          ep: 16
+          dp-attn: true
+      # Mid curve 3p1d-dep8-dep16. 10 nodes.
+      - spec-decoding: "mtp"
+        conc-list: [3072]
+        prefill:
+          num-worker: 3
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/8k1k/disagg-gb200-mid-curve-3p1d-dep8-dep16-mtp.yaml"
+        decode:
+          num-worker: 1
+          tp: 16
+          ep: 16
+          dp-attn: true
+      # Mid curve 4p1d-dep8-dep16. 12 nodes.
+      - spec-decoding: "mtp"
+        conc-list: [6144]
+        prefill:
+          num-worker: 4
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/8k1k/disagg-gb200-mid-curve-4p1d-dep8-dep16-mtp.yaml"
+        decode:
+          num-worker: 1
+          tp: 16
+          ep: 16
+          dp-attn: true
+      # Mid curve 5p1d-dep8-dep16. 14 nodes.
+      - spec-decoding: "mtp"
+        conc-list: [8192]
+        prefill:
+          num-worker: 5
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/8k1k/disagg-gb200-mid-curve-5p1d-dep8-dep16-mtp.yaml"
+        decode:
+          num-worker: 1
+          tp: 16
+          ep: 16
+          dp-attn: true
+      # Mid curve 6p1d-dep8-dep16. 16 nodes.
+      - spec-decoding: "mtp"
+        conc-list: [16384]
+        prefill:
+          num-worker: 6
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/sglang/deepseek-v4/8k1k/disagg-gb200-mid-curve-6p1d-dep8-dep16-mtp.yaml"
+        decode:
+          num-worker: 1
+          tp: 16
+          ep: 16
+          dp-attn: true
 dsv4-fp4-b300-dynamo-vllm:
   image: vllm/vllm-openai:v0.20.1
   model: deepseek-ai/DeepSeek-V4-Pro

benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/8k1k/disagg-gb200-low-latency-1p1d-tp8-tp8-mtp.yaml

@@ -0,0 +1,124 @@
+name: "dsv4-pro-gb200-disagg-8k1k-low-latency-1p1d-tp8-tp8-mtp"
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: true
+  num_additional_frontends: 8
+
+dynamo:
+  hash: "92f5b3b8d7dd5ab9179d4b1034bd2c1c0803693e"
+  install: true
+
+model:
+  path: "deepseek-v4-pro"
+  container: "lmsysorg/sglang:nightly-dev-cu13-20260528-0abe6a85"
+  precision: "fp4"
+
+sbatch_directives:
+  cpus-per-task: "144"
+  mem: "0"
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 2
+  prefill_workers: 1
+  gpus_per_prefill: 8
+  decode_nodes: 2
+  decode_workers: 1
+  gpus_per_decode: 8
+
+backend:
+  type: sglang
+
+  prefill_environment:
+    PYTHONUNBUFFERED: "1"
+    SGLANG_RADIX_FORCE_MISS: "1"
+    SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
+    SGLANG_DEFAULT_THINKING: "1"
+    SGLANG_DSV4_REASONING_EFFORT: "max"
+    SGLANG_OPT_SWA_SPLIT_LEAF_ON_INSERT: "1"
+    SGLANG_OPT_SWA_EVICT_DROP_PAGE_MARGIN: "1"
+    NCCL_MNNVL_ENABLE: "1"
+    NCCL_CUMEM_ENABLE: "1"
+    SGLANG_MOONCAKE_CUSTOM_MEM_POOL: "True"
+    MC_FORCE_MNNVL: "1"
+    SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT: "100000"
+    SGLANG_DISAGGREGATION_WAITING_TIMEOUT: "100000"
+    SGLANG_OPT_SWA_RELEASE_LEAF_LOCK_AFTER_WINDOW: "1"
+
+  decode_environment:
+    PYTHONUNBUFFERED: "1"
+    SGLANG_RADIX_FORCE_MISS: "1"
+    SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
+    SGLANG_DEFAULT_THINKING: "1"
+    SGLANG_DSV4_REASONING_EFFORT: "max"
+    SGLANG_OPT_SWA_SPLIT_LEAF_ON_INSERT: "1"
+    SGLANG_OPT_SWA_EVICT_DROP_PAGE_MARGIN: "1"
+    NCCL_MNNVL_ENABLE: "1"
+    NCCL_CUMEM_ENABLE: "1"
+    SGLANG_MOONCAKE_CUSTOM_MEM_POOL: "True"
+    MC_FORCE_MNNVL: "1"
+    SGLANG_DISAGGREGATION_BOOTSTRAP_TIMEOUT: "100000"
+    SGLANG_DISAGGREGATION_WAITING_TIMEOUT: "100000"
+    SGLANG_OPT_SWA_RELEASE_LEAF_LOCK_AFTER_WINDOW: "1"
+
+  sglang_config:
+    prefill:
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+      model-path: "/model/"
+      trust-remote-code: true
+      tool-call-parser: deepseekv4
+
+      disaggregation-mode: "prefill"
+      disaggregation-transfer-backend: mooncake
+
+      tensor-parallel-size: 8
+      data-parallel-size: 1
+      expert-parallel-size: 1
+
+      moe-runner-backend: "flashinfer_mxfp4"
+      disable-flashinfer-autotune: true
+
+      mem-fraction-static: 0.9
+      max-running-requests: 16
+      cuda-graph-max-bs: 8
+      chunked-prefill-size: 65536
+
+    decode:
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+      model-path: "/model/"
+      trust-remote-code: true
+      tool-call-parser: deepseekv4
+
+      disaggregation-mode: "decode"
+      disaggregation-transfer-backend: mooncake
+
+      tensor-parallel-size: 8
+      data-parallel-size: 1
+      expert-parallel-size: 1
+
+      moe-runner-backend: "flashinfer_mxfp4"
+      disable-flashinfer-autotune: true
+
+      speculative-algo: "EAGLE"
+      speculative-num-steps: 3
+      speculative-eagle-topk: 1
+      speculative-num-draft-tokens: 4
+
+      mem-fraction-static: 0.9
+      max-running-requests: 8
+      cuda-graph-max-bs: 8
+      swa-full-tokens-ratio: 0.1
+      context-length: 16384
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  random_range_ratio: 0.8
+  concurrencies: "1"
+  req_rate: "inf"
+  use_chat_template: true
+  custom_tokenizer: "sa_bench_tokenizers.sglang_deepseek_v4.SGLangDeepseekV4Tokenizer"
+