Reduce frank-default-large context for KV headroom

docs/LLM/deployment.md

@@ -145,8 +145,9 @@ The active service on Dankclaw is the charlie12345/ROCmFPX llama.cpp runtime
 against system ROCm 7.2.4 / amdclang 22) with a **strict Qwen MTP** draft
 (`--spec-type draft-mtp --spec-mtp-strict-qwen`) running on ROCm0 with all
 layers (`--spec-draft-ngl all`) — no separate draft file, because the MTP head
-is embedded in the main GGUF. K/V are f16, full 262144 context (1 slot,
-concurrent requests queue, the server never preempts a running generation),
+is embedded in the main GGUF. K/V are f16, 196608 context (64K below the
+native 262144 maximum; 1 slot, concurrent requests queue, the server never
+preempts a running generation),
 `-t 16 -tb 32`, batch 512 / ubatch 256 (ubatch is swappable; sweep it on 8k
 context with `scripts/bench-ubatch-rocmi4.sh`), 16 context checkpoints, 8 GiB
 prompt cache, and reasoning in deepseek format. The unit file is

docs/LLM/eemin-rocmi4-mtp.md

@@ -112,9 +112,10 @@ path's greedy decisions; it does not remove the W4A4 activation approximation.
 
 ### Serving settings
 
-- 262144 context (native max), 1 slot, f16 KV, context checkpoints, prompt
-  cache. Concurrent requests queue; the server never preempts a running
-  generation.
+- 196608 context (64K below the native 262144 maximum), 1 slot, f16 KV,
+  context checkpoints, prompt cache. The lower limit leaves headroom for the
+  model plus KV cache in the host's unified-memory budget. Concurrent requests
+  queue; the server never preempts a running generation.
 - `-t 16 -tb 32`, batch 512 / ubatch 256 (the qualified W4A4 profile).
 - `--temp 0.7 --top-p 0.95`, reasoning on (`deepseek` puts `<think>` into
   `message.reasoning_content`).

scripts/frank-default-large.env

@@ -36,11 +36,11 @@ ALIAS=frank-default-large
 BIND_ADDR=0.0.0.0
 LISTEN_PORT=7770
 
-# Serving: full 262144 context, 1 slot. f16 KV, context checkpoints, prompt
-# cache, strict-MTP embedded draft on ROCm0. Unified memory; rocm-smi reports
-# the 64 GiB budget. Concurrent requests queue; the server never preempts a
-# running generation.
-CTX_SIZE=262144
+# Serving: 196608 context (64K below the model's 262144 maximum), 1 slot. f16
+# KV, context checkpoints, prompt cache, strict-MTP embedded draft on ROCm0.
+# This leaves sufficient unified-memory headroom for the model plus KV cache.
+# Concurrent requests queue; the server never preempts a running generation.
+CTX_SIZE=196608
 PARALLEL=1
 # Batch/ubatch follow the qualified ROCmI4 W4A4 profile (see the bench in
 # scripts/bench-ubatch-rocmi4.sh for the ubatch sweep on 8k contexts).

scripts/frank-default-large.service

@@ -1,5 +1,5 @@
 [Unit]
-Description=CIRU llama.cpp Qwen3.8-27B ROCmFP4 STRIX MTP + Sharp template (dankclaw, frank-default-large)
+Description=ROCmFPX llama.cpp Qwen3.8-27B ROCmI4 embedded MTP + Sharp template (dankclaw, frank-default-large)
 After=network-online.target
 Wants=network-online.target
 

tests/frank_large_config.bats

@@ -39,6 +39,11 @@ TEMPLATE_FILE="$REPO_ROOT/scripts/chat-template.jinja"
     grep -q '^MTP_BACKEND_SAMPLING=on$' "$ENV_FILE"
 }
 
+@test "frank-default-large fits the model and KV cache in unified memory" {
+    grep -q '^CTX_SIZE=196608$' "$ENV_FILE"
+    grep -q ': "${CTX_SIZE:=262144}"' "$SERVE_FILE"
+}
+
 @test "frank-default-large exposes a swappable ubatch" {
     grep -q '^UBATCH_SIZE=256$' "$ENV_FILE"
     grep -q ': "${UBATCH_SIZE:=256}"' "$SERVE_FILE"