4x cells: KV pin 10 -> 16 GiB (NVFP4 4x has the headroom; 2x stays 10)

recipes/glm53-flash-nvfp4-r18-4x-dcp4/recipe.toml

@@ -102,7 +102,7 @@ args = [
     "--gpu-memory-utilization", "0.90",
     # Pinned 10 GiB KV (≈1.16M tokens) — GMU sizing has no headroom at
     # 92.72 GiB weights/node; pinning also keeps warmup off the swap path
-    "--kv-cache-memory-bytes", "10737418240",
+    "--kv-cache-memory-bytes", "17179869184",  # 16 GiB: 4x NVFP4 has the headroom (47.1 GiB weights/node vs 108 usable); 2x stays 10
     "--decode-context-parallel-size", "4",
     "--cp-kv-cache-interleave-size", "4",
     # No --kv-cache-dtype: let the r12 fork pick its default (the explicit

recipes/glm53-flash-nvfp4-r18-4x-nvfp4kv/recipe.toml

@@ -102,7 +102,7 @@ args = [
     "--gpu-memory-utilization", "0.90",
     # Pinned 10 GiB KV (≈1.16M tokens) — GMU sizing has no headroom at
     # 92.72 GiB weights/node; pinning also keeps warmup off the swap path
-    "--kv-cache-memory-bytes", "10737418240",
+    "--kv-cache-memory-bytes", "17179869184",  # 16 GiB: 4x NVFP4 has the headroom (47.1 GiB weights/node vs 108 usable); 2x stays 10
     "--kv-cache-dtype", "nvfp4_ds_mla",
     # No --kv-cache-dtype: let the r12 fork pick its default (the explicit
     # fp8 / fp8_ds_mla both failed auto-selection on this build).

recipes/glm53-flash-nvfp4-r18-4x/recipe.toml

@@ -102,7 +102,7 @@ args = [
     "--gpu-memory-utilization", "0.90",
     # Pinned 10 GiB KV (≈1.16M tokens) — GMU sizing has no headroom at
     # 92.72 GiB weights/node; pinning also keeps warmup off the swap path
-    "--kv-cache-memory-bytes", "10737418240",
+    "--kv-cache-memory-bytes", "17179869184",  # 16 GiB: 4x NVFP4 has the headroom (47.1 GiB weights/node vs 108 usable); 2x stays 10
     # No --kv-cache-dtype: let the r12 fork pick its default (the explicit
     # fp8 / fp8_ds_mla both failed auto-selection on this build).
     # The GLM5Next model wires the B12X sparse-MLA backend (sm_120/sm_121)