refactor(srt): compose main's new GB300 and B300 AgentX recipes from fragments

#3816, #3840 and #3744 added full recipes: the Qwen3.8 GB300 and Qwen3.5
GB300 DEP16 recipes carry model blocks, the client command, benchmark MODEL,
TOTAL_CPU_DRAM_GB, KV_OFFLOADING and the client paths the launcher hands
multi-node points, and both new recipes and the DSv4.1-Flash B300 one set the
identity repo and image the binder writes. Like the other AgentX fragments
they drop them, and WEKA_LOADER_OVERRIDE, which names both models' default
corpus.

The Qwen3.8 HiCache pool was the 225 GB matrix allowance less 1 GB of
alignment headroom; it now takes '@dram.per-gpu-gb' and the entry's
dram-utilization drops from 1.0 to 0.995, so the pool stays 224 GB and the
bound TOTAL_CPU_DRAM_GB is 224. Each point otherwise binds to main's
selected variant, with the Qwen3.5 model alias bound as its Hub id like the
other GB300 Qwen3.5 AgentX points.

Author: adibarra

inferencex-e2e/benchmarks/multi_node/srt-slurm-recipes/qwen3.5/sglang/gb300-fp8/agentx/disagg-dep16-mtp6-variants.yaml

@@ -1,9 +1,5 @@
 base:
   schema: 2
-  model:
-    path: qwen3.5-fp8
-    container: lmsysorg/sglang:nightly-dev-cu13-20260913-fa663e72
-    precision: fp8
   dynamo:
     install: true
     source:
@@ -227,16 +223,7 @@ base:
     mem: '0'
     container-remap-root: ''
   benchmark:
-    type: custom
-    command: bash /infmax-workspace/benchmarks/srt_agentic.sh
     env:
-      INFMAX_CONTAINER_WORKSPACE: /infmax-workspace
-      RESULT_DIR: /logs/agentic
-      PORT: '8000'
-      IS_MULTINODE: 'true'
-      WEKA_LOADER_OVERRIDE: semianalysis_cc_traces_weka_062126_256k
-      AIPERF_DATASET_MMAP_CACHE_DIR: /aiperf_mmap_cache
-      HF_HUB_CACHE: /hf_hub_cache
       AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: 'true'
       AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: '0'
       AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'sglang:'

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv41flash/sglang/b300-fp4-mtp/agentic.yaml

@@ -3,10 +3,8 @@ base:
   name: dsv41flash-fp4-b300-sglang-agentic
   identity:
     model:
-      repo: deepseek-ai/DeepSeek-V4.1-Flash
       revision: dba1be0a40aa45a94ad051997016db3960a90277
-    container:
-      image: lmsysorg/sglang:nightly-dev-cu13-20261004-295a53c9@sha256:7cfd4dd8e0e902d2a3c0560565963c4e172091e1586aaed15ff5a0a131394d34
+    container: {}
     frameworks:
       sglang: 0.0.0.dev1+g295a53c90
       sglang-router: 0.3.2

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.8next/sglang/gb300-fp4-mtp/agentic.yaml

@@ -1,19 +1,12 @@
 # Stock ARM64 SGLang, one serving GPU per point. No source/package overlay.
-# Size the host cache against the public GB300 profile's per-worker DRAM allowance.
-# The matrix floors the allowance to 225 GB; H224 leaves pool-alignment headroom.
+# The HiCache host pool takes the point's per-GPU DRAM budget.
 base:
   schema: 2
   name: qwen3.8next-fp4-gb300-sglang-agentic-stock-v0521
   identity:
     model:
-      repo: nvidia/Qwen3.8-Flash-Next-NVFP4
       revision: fc694b54fb0174e0913e6adf86691ef85a4ead47
-    container:
-      image: lmsysorg/sglang:v0.5.21-cu130@sha256:1f176b3a1dd1205d2a462ab80ea9fc17769a32e53cb6f3c690efa2dc2c384977
-  model:
-    path: hf:nvidia/Qwen3.8-Flash-Next-NVFP4
-    container: lmsysorg/sglang:v0.5.21-cu130@sha256:1f176b3a1dd1205d2a462ab80ea9fc17769a32e53cb6f3c690efa2dc2c384977
-    precision: fp4
+    container: {}
   resources:
     gpu_type: gb300
     gpus_per_node: 4
@@ -82,7 +75,7 @@ base:
         enable-hierarchical-cache: true
         # One combined packed target/draft KV + Mamba host-pool budget.
         # Verify actual allocations, not whole-node Slurm memory reservation.
-        hicache-size: 224
+        hicache-size: '@dram.per-gpu-gb'
         hicache-io-backend: kernel
         hicache-mem-layout: page_first
         hicache-write-policy: write_back
@@ -104,15 +97,8 @@ base:
     mem: '0'
     container-remap-root: ''
   benchmark:
-    type: custom
-    command: bash /infmax-workspace/benchmarks/srt_agentic.sh
     env:
-      MODEL: nvidia/Qwen3.8-Flash-Next-NVFP4
       AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'sglang:'
-      WEKA_LOADER_OVERRIDE: semianalysis_cc_traces_weka_062126_256k
-      # Native single-node binding verifies these against the generated point.
-      TOTAL_CPU_DRAM_GB: '225'
-      KV_OFFLOADING: dram
 
 # Only client concurrency changes; serving/admission/graph/cache settings stay fixed.
 override_c1:

inferencex-e2e/configs/nvidia-master.yaml

@@ -9174,9 +9174,8 @@ qwen3.8next-fp4-gb300-sglang-agentic-mtp:
   srt-recipe-dir: qwen3.8next/sglang/gb300-fp4-mtp
   scenarios:
     agentic-coding:
-    # Use the GB300 profile's proportional per-worker host DRAM allowance.
-    # Generated allowance: 225 GB. Recipe H224 reserves alignment headroom.
-    - dram-utilization: 1.0
+    # Sizes the 224 GB HiCache host pool the KV, Mamba and draft caches share.
+    - dram-utilization: 0.995
       search-space:
       - tp: 1
         ep: 1