fix(agentx): finish per-lane warmup migration

Author: cquil11

.gitmodules

@@ -1,4 +1,3 @@
 [submodule "utils/aiperf"]
 	path = utils/aiperf
 	url = https://github.com/SemiAnalysisAI/aiperf.git
-	branch = main

benchmarks/multi_node/amd_utils/server_sglang.sh

@@ -986,7 +986,7 @@ if [ "$NODE_RANK" -eq 0 ]; then
                       DECODE_NUM_WORKERS DECODE_TP DECODE_EP DECODE_DP_ATTN DECODE_HARDWARE \
                       KV_OFFLOADING KV_OFFLOAD_BACKEND KV_OFFLOAD_BACKEND_METADATA TOTAL_CPU_DRAM_GB KV_P2P_TRANSFER \
                       WEKA_LOADER_OVERRIDE AIPERF_FAILED_REQUEST_THRESHOLD \
-                      AIPERF_AGENTIC_CACHE_WARMUP_DURATION AIPERF_UNSAFE_OVERRIDE \
+                      AIPERF_WARMUP_REQUESTS_PER_LANE AIPERF_EXPERIMENTAL_FAST AIPERF_UNSAFE_OVERRIDE \
                       AIPERF_TRAJECTORY_START_MIN_RATIO AIPERF_TRAJECTORY_START_MAX_RATIO \
                       AIPERF_DATASET_WEKA_LIVE_ASSISTANT_RESPONSES ROUTER_PORT TQDM_MININTERVAL; do
                 if [[ -n "${!_v+x}" ]]; then

benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/agg-gb300-tp4-mtp-agentic.yaml

@@ -141,7 +141,6 @@ benchmark:
     PORT: "8000"
     IS_MULTINODE: "true"
     AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
-    AIPERF_AGENTIC_CACHE_WARMUP_DURATION: "600"
     AIPERF_DATASET_MMAP_CACHE_DIR: "/aiperf_mmap_cache"
     HF_HUB_CACHE: "/hf_hub_cache"
     WEKA_LOADER_OVERRIDE: "semianalysis_cc_traces_weka_062126"

benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/agg-gb300-tp8-mtp-agentic.yaml

@@ -144,7 +144,6 @@ benchmark:
     # the zero decode-worker count instead of duplicating TP into P and D.
     IS_MULTINODE: "true"
     AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
-    AIPERF_AGENTIC_CACHE_WARMUP_DURATION: "600"
     AIPERF_DATASET_MMAP_CACHE_DIR: "/aiperf_mmap_cache"
     HF_HUB_CACHE: "/hf_hub_cache"
     WEKA_LOADER_OVERRIDE: "semianalysis_cc_traces_weka_062126"

benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep4-dep8-c128-mtp-agentic.yaml

@@ -198,7 +198,6 @@ benchmark:
     IS_MULTINODE: "true"
     AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
     AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: "true"
-    AIPERF_AGENTIC_CACHE_WARMUP_DURATION: "600"
     AIPERF_DATASET_MMAP_CACHE_DIR: "/aiperf_mmap_cache"
     HF_HUB_CACHE: "/hf_hub_cache"
     WEKA_LOADER_OVERRIDE: "semianalysis_cc_traces_weka_062126"

benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep8-dep8-c384-mtp-agentic.yaml

@@ -199,7 +199,6 @@ benchmark:
     IS_MULTINODE: "true"
     AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
     AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: "true"
-    AIPERF_AGENTIC_CACHE_WARMUP_DURATION: "600"
     AIPERF_DATASET_MMAP_CACHE_DIR: "/aiperf_mmap_cache"
     HF_HUB_CACHE: "/hf_hub_cache"
     WEKA_LOADER_OVERRIDE: "semianalysis_cc_traces_weka_062126"

perf-changelog.yaml

@@ -5234,7 +5234,7 @@
 - config-keys:
     - kimik3-fp4-b300-vllm-agentic
   description:
-    - "Pin shared AIPerf to c57ad643 and replace duration-based warmup with 10 deterministic requests per lane."
+    - "Pin shared AIPerf to agentx-v1.0.0 (c57ad643) and replace duration-based warmup with 10 deterministic requests per lane."
     - "Use one warmup request per lane with the 20-minute agentx-fast profile, and emit TTFT, ITL, and throughput metrics every 30 seconds."
     - "Trigger an agentx-fast validation sweep on the existing Kimi K3 B300 AgentX configuration."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2415