fix(launch): request explicit Slurm memory for single-node jobs and fit host pools to measured DRAM

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv4/sglang/b200-fp4-mtp/agentic.yaml

@@ -160,7 +160,7 @@ override_tp8_hicache_c8:
     env:
       CONC: '8'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
 
 override_tp8_hicache_c10:
   roles:
@@ -181,7 +181,7 @@ override_tp8_hicache_c10:
     env:
       CONC: '10'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
 
 override_tp8_hicache_c16:
   roles:
@@ -202,7 +202,7 @@ override_tp8_hicache_c16:
     env:
       CONC: '16'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
 
 override_dep8_hicache_c64:
   frontend:
@@ -248,7 +248,7 @@ override_dep8_hicache_c64:
     env:
       CONC: '64'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
       AIPERF_HTTP_X_SMG_ROUTING_KEY_FROM_CORRELATION_ID: 'true'
       AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: 'true'
 
@@ -296,7 +296,7 @@ override_dep8_hicache_c96:
     env:
       CONC: '96'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
       AIPERF_HTTP_X_SMG_ROUTING_KEY_FROM_CORRELATION_ID: 'true'
       AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: 'true'
 
@@ -344,7 +344,7 @@ override_dep8_hicache_c128:
     env:
       CONC: '128'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
       AIPERF_HTTP_X_SMG_ROUTING_KEY_FROM_CORRELATION_ID: 'true'
       AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: 'true'
 
@@ -393,6 +393,6 @@ override_dep8_hicache_c160:
     env:
       CONC: '160'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
       AIPERF_HTTP_X_SMG_ROUTING_KEY_FROM_CORRELATION_ID: 'true'
       AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: 'true'

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv4/vllm/b200-fp4-mtp/agentic.yaml

@@ -184,15 +184,15 @@ override_dep8_c32:
         max-num-batched-tokens: 8248
         max-num-seqs: 8
         compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[7,14,21,28,35,42,49,56],"mode":0}'
-        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":243375000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
+        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":241500000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
       env:
         PYTORCH_ALLOC_CONF: 'expandable_segments:True'
         PYTHONHASHSEED: '42'
   benchmark:
     env:
       CONC: '32'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1947'
+      TOTAL_CPU_DRAM_GB: '1932'
       AIPERF_HTTP_X_SESSION_ID_FROM_CORRELATION_ID: '1'
 
 override_dep8_c64:
@@ -220,15 +220,15 @@ override_dep8_c64:
         max-num-batched-tokens: 8304
         max-num-seqs: 16
         compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[7,14,21,28,35,42,49,56,63,70,77,84,91,98,105,112],"mode":0}'
-        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":243375000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
+        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":241500000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
       env:
         PYTORCH_ALLOC_CONF: 'expandable_segments:True'
         PYTHONHASHSEED: '42'
   benchmark:
     env:
       CONC: '64'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1947'
+      TOTAL_CPU_DRAM_GB: '1932'
       AIPERF_HTTP_X_SESSION_ID_FROM_CORRELATION_ID: '1'
 
 override_dep8_c96:
@@ -256,15 +256,15 @@ override_dep8_c96:
         max-num-batched-tokens: 8360
         max-num-seqs: 24
         compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[7,14,21,28,35,42,49,56,63,70,77,84,91,98,105,112,119,126,133,140,147,154,161,168],"mode":0}'
-        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":243375000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
+        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":241500000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
       env:
         PYTORCH_ALLOC_CONF: 'expandable_segments:True'
         PYTHONHASHSEED: '42'
   benchmark:
     env:
       CONC: '96'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1947'
+      TOTAL_CPU_DRAM_GB: '1932'
       AIPERF_HTTP_X_SESSION_ID_FROM_CORRELATION_ID: '1'
 
 override_dep8_c128:
@@ -292,15 +292,15 @@ override_dep8_c128:
         max-num-batched-tokens: 8416
         max-num-seqs: 32
         compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[7,14,21,28,35,42,49,56,63,70,77,84,91,98,105,112,119,126,133,140,147,154,161,168,175,182,189,196,203,210,217,224],"mode":0}'
-        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":243375000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
+        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":241500000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
       env:
         PYTORCH_ALLOC_CONF: 'expandable_segments:True'
         PYTHONHASHSEED: '42'
   benchmark:
     env:
       CONC: '128'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1947'
+      TOTAL_CPU_DRAM_GB: '1932'
       AIPERF_HTTP_X_SESSION_ID_FROM_CORRELATION_ID: '1'
 
 override_dep8_c160:
@@ -328,15 +328,15 @@ override_dep8_c160:
         max-num-batched-tokens: 8472
         max-num-seqs: 40
         compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[7,14,21,28,35,42,49,56,63,70,77,84,91,98,105,112,119,126,133,140,147,154,161,168,175,182,189,196,203,210,217,224,231,238,245,252,259,266,273,280],"mode":0}'
-        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":243375000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
+        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":241500000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
       env:
         PYTORCH_ALLOC_CONF: 'expandable_segments:True'
         PYTHONHASHSEED: '42'
   benchmark:
     env:
       CONC: '160'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1947'
+      TOTAL_CPU_DRAM_GB: '1932'
       AIPERF_HTTP_X_SESSION_ID_FROM_CORRELATION_ID: '1'
 
 override_dep8_c192:
@@ -364,13 +364,13 @@ override_dep8_c192:
         max-num-batched-tokens: 8528
         max-num-seqs: 48
         compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[7,14,21,28,35,42,49,56,63,70,77,84,91,98,105,112,119,126,133,140,147,154,161,168,175,182,189,196,203,210,217,224,231,238,245,252,259,266,273,280,287,294,301,308,315,322,329,336],"mode":0}'
-        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":243375000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
+        kv-transfer-config: '{"kv_connector":"SimpleCPUOffloadConnector","kv_role":"kv_both","kv_connector_extra_config":{"cpu_bytes_to_use_per_rank":241500000000,"enable_cross_layers_blocks":"true","lazy_offload":false}}'
       env:
         PYTORCH_ALLOC_CONF: 'expandable_segments:True'
         PYTHONHASHSEED: '42'
   benchmark:
     env:
       CONC: '192'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1947'
+      TOTAL_CPU_DRAM_GB: '1932'
       AIPERF_HTTP_X_SESSION_ID_FROM_CORRELATION_ID: '1'

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/glm5.2/sglang/b200-fp4-mtp/agentic.yaml

@@ -56,12 +56,12 @@ base:
       AIPERF_HTTP_TCP_USER_TIMEOUT: '900000'
       AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'sglang:'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
 
 # One variant per point. Admission is 2x CONC and the graph batch matches it,
 # capped at 64. MLA ranks hold complete per-token KV, so ratio 0.75 gives too
 # few host slots past C8; C12/C16 pin a 169 GB/rank pool (~1,662 GB across TP8
-# with the DSA indexer, within the 1,731 GB budget).
+# with the DSA indexer, within the 1,717 GB budget).
 override_c1:
   roles:
     agg:

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/glm5.2/sglang/b200-fp8-mtp/agentic.yaml

@@ -56,12 +56,12 @@ base:
       AIPERF_HTTP_TCP_USER_TIMEOUT: '900000'
       AIPERF_REQUIRED_SERVER_METRIC_PREFIX: 'sglang:'
       KV_OFFLOADING: dram
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
 
 # One variant per point. Admission is 2x CONC and the graph batch matches it,
 # capped at 64. MLA ranks hold complete per-token KV, so ratio 0.75 gives too
 # few host slots past C8; C12/C16 pin a 169 GB/rank pool (~1,662 GB across TP8
-# with the DSA indexer, within the 1,731 GB budget).
+# with the DSA indexer, within the 1,717 GB budget).
 override_c1:
   roles:
     agg:

inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/minimaxm3/trtllm/b200-fp4-mtp/agentic.yaml

@@ -94,8 +94,8 @@ base:
       KV_OFFLOADING: dram
 
 # One variant per point. The batch cap is CONC; graphs capture every batch up
-# to 20 and a thinned ladder above. The host pool is 250 GiB at TP4 and
-# 200 GiB at TP8.
+# to 20 and a thinned ladder above. The per-rank host pool is TOTAL_CPU_DRAM_GB / 4
+# at TP4 and 200 GiB at TP8.
 override_tp4_c5:
   roles:
     agg:
@@ -111,11 +111,11 @@ override_tp4_c5:
             - 4
             - 5
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '5'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp4_c10:
   roles:
@@ -137,11 +137,11 @@ override_tp4_c10:
             - 9
             - 10
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '10'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp4_c15:
   roles:
@@ -168,11 +168,11 @@ override_tp4_c15:
             - 14
             - 15
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '15'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp4_c20:
   roles:
@@ -204,11 +204,11 @@ override_tp4_c20:
             - 19
             - 20
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '20'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp4_c25:
   roles:
@@ -240,11 +240,11 @@ override_tp4_c25:
             - 23
             - 25
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '25'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp4_c30:
   roles:
@@ -276,11 +276,11 @@ override_tp4_c30:
             - 27
             - 30
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '30'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp4_c35:
   roles:
@@ -311,11 +311,11 @@ override_tp4_c35:
             - 32
             - 35
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '35'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp4_c40:
   roles:
@@ -347,11 +347,11 @@ override_tp4_c40:
             - 38
             - 40
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '40'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp4_c45:
   roles:
@@ -381,11 +381,11 @@ override_tp4_c45:
             - 42
             - 45
         kv_cache_config:
-          host_cache_size: 268435456000
+          host_cache_size: 268250000000
   benchmark:
     env:
       CONC: '45'
-      TOTAL_CPU_DRAM_GB: '865'
+      TOTAL_CPU_DRAM_GB: '1073'
 
 override_tp8_c1:
   roles:
@@ -402,7 +402,7 @@ override_tp8_c1:
   benchmark:
     env:
       CONC: '1'
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'
 
 override_tp8_c5:
   roles:
@@ -423,4 +423,4 @@ override_tp8_c5:
   benchmark:
     env:
       CONC: '5'
-      TOTAL_CPU_DRAM_GB: '1731'
+      TOTAL_CPU_DRAM_GB: '1717'