refactor(srt): size the DSv4-Pro ATOM AgentX LMCache MP L1 from the DRAM budget

Main's C256 and C384 points of dsv4-fp4-mi355x-atom-agentic-mtp start an
LMCache MP server with a literal 1000 GB L1 and set TOTAL_CPU_DRAM_GB 1019
(dram-utilization 0.34). The L1 now takes '@dram.total-gb' and the binder
writes TOTAL_CPU_DRAM_GB. dram-utilization 0.3334 makes the budget 1000 GB,
so the bound L1 stays 1000 GB.

ATOM's lmcache.mp.max_pinned_state_bytes keeps its literal 1000000000000:
srtctl needs the extra-config JSON brace-doubled, and the binder resolves no
reference inside it.

Author: adibarra

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv4/atom/mi355x-fp4-mtp/agentic.yaml

@@ -224,7 +224,9 @@ override_tp8_c128_dep:
 
 # DRAM variants: ATOM's lmcache_mp tier backed by srtctl's lmcache-server
 # service (one LMCache server per node on port 8750, run from the LMCache build in
-# the ATOM image) within the 1019 GB budget (dram-utilization 0.34).
+# the ATOM image) whose L1 is the point's DRAM budget (1000 GB at dram-utilization
+# 0.3334). max_pinned_state_bytes repeats it in bytes: the binder resolves no
+# reference inside the brace-doubled JSON.
 override_tp8_c256_dep_lmcache:
   services:
     - name: lmcache
@@ -238,7 +240,7 @@ override_tp8_c256_dep_lmcache:
         - --supported-transfer-mode
         - lmcache_driven
         - --l1-size-gb
-        - '1000'
+        - '@dram.total-gb'
         - --l1-use-lazy
         - --l1-read-ttl-seconds
         - '900'
@@ -280,7 +282,6 @@ override_tp8_c256_dep_lmcache:
     env:
       CONC: '256'
       KV_OFFLOADING: 'dram'
-      TOTAL_CPU_DRAM_GB: '1019'
       AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: '1'
       AGENTIC_WARMUP_GRACE_PERIOD: '3600'
 
@@ -297,7 +298,7 @@ override_tp8_c384_dep_lmcache:
         - --supported-transfer-mode
         - lmcache_driven
         - --l1-size-gb
-        - '1000'
+        - '@dram.total-gb'
         - --l1-use-lazy
         - --l1-read-ttl-seconds
         - '900'
@@ -339,6 +340,5 @@ override_tp8_c384_dep_lmcache:
     env:
       CONC: '384'
       KV_OFFLOADING: 'dram'
-      TOTAL_CPU_DRAM_GB: '1019'
       AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: '1'
       AGENTIC_WARMUP_GRACE_PERIOD: '3600'

inferencex-e2e/configs/amd-master.yaml

@@ -966,8 +966,8 @@ dsv4-fp4-mi355x-vllm-agentic-mtp:
 # DeepSeek-V4-Pro-0813 FP4 AgentX on one MI355X node using ATOM DSpark K6.
 # Both throughput bands use thinking_on golden AL 3.77; eval uses real
 # acceptance. Keep the historical config key for changelog compatibility.
-# C256 and C384 offload KV to an LMCache MP server whose 1000 GB L1 fits the
-# 1019 GB budget of dram-utilization 0.34.
+# C256 and C384 offload KV to an LMCache MP server whose L1 is the 1000 GB
+# budget of dram-utilization 0.3334.
 dsv4-fp4-mi355x-atom-agentic-mtp:
   image: rocm/atom-dev:nightly_202610071700
   model: deepseek-ai/DeepSeek-V4-Pro-0813
@@ -979,7 +979,7 @@ dsv4-fp4-mi355x-atom-agentic-mtp:
   srt-recipe-dir: dsv4/atom/mi355x-fp4-mtp
   scenarios:
     agentic-coding:
-    - dram-utilization: 0.34
+    - dram-utilization: 0.3334
       search-space:
       - { tp: 8, ep: 1, dp-attn: false, kv-offloading: none, spec-decoding: draft_model, conc-list: [1, 2, 4, 8, 16], srt-recipe: agentic.yaml }
       - { tp: 8, ep: 8, dp-attn: true, kv-offloading: none, spec-decoding: draft_model, conc-list: [48, 64, 96, 128], srt-recipe: agentic.yaml }