From 4d13333da709fbcbc1b09682fb9604ef3220332b Mon Sep 17 00:00:00 2001
From: Frank <frank@dankware>
Date: Thu, 17 Sep 2026 03:03:46 -0700
Subject: [PATCH] frank-small: wire SPEC_DRAFT_N_MAX and POLL_LEVEL knobs

Both were implicit build defaults (3 and 50); the throughput sweep needs
to move them without editing the exec line each time. Defaults preserve
current behavior exactly. Additive only.
---
 scripts/frank-small.env      | 6 +++++-
 scripts/serve-frank-small.sh | 8 +++++++-
 2 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/scripts/frank-small.env b/scripts/frank-small.env
index 0b7483b..acdcc31 100644
--- a/scripts/frank-small.env
+++ b/scripts/frank-small.env
@@ -90,8 +90,12 @@ NGL=999
 FLASH_ATTN=on
 
 # Speculative decoding: the MTP sidecar (draft context) chained with the
-# draftless ngram-mod matcher. Draft depth stays llama.cpp's default of 3.
+# draftless ngram-mod matcher. Draft depth and busy-wait level are made
+# explicit here for tuning (both were the implicit build defaults: 3 and 50;
+# the Gemma4 sidecar carries 4 nextn layers).
 SPEC_TYPE=draft-mtp,ngram-mod
+SPEC_DRAFT_N_MAX=3
+POLL_LEVEL=50
 
 # Greedy default (fleet contract); reasoning is extracted into
 # reasoning_content with an unlimited budget (the model decides).
diff --git a/scripts/serve-frank-small.sh b/scripts/serve-frank-small.sh
index a97f7e4..0a9aade 100755
--- a/scripts/serve-frank-small.sh
+++ b/scripts/serve-frank-small.sh
@@ -60,6 +60,11 @@ fi
 : "${CACHE_TYPE_V:=f16}"
 : "${BATCH_SIZE:=2048}"
 : "${UBATCH_SIZE:=512}"
+# Decode-shape knobs: the MTP drafter's per-round proposal length (build
+# default 3; the Gemma4 sidecar carries 4 nextn layers) and the engine's
+# busy-wait level (build default 50).
+: "${SPEC_DRAFT_N_MAX:=3}"
+: "${POLL_LEVEL:=50}"
 : "${NUM_THREADS:=8}"
 : "${THREAD_BATCH:=8}"
 : "${NGL:=999}"
@@ -90,7 +95,7 @@ fi
 
 SPEC_ARGS=()
 if [[ -n "${SPEC_TYPE:-}" && "$SPEC_TYPE" != "none" ]]; then
-    SPEC_ARGS=(--spec-type "$SPEC_TYPE")
+    SPEC_ARGS=(--spec-type "$SPEC_TYPE" --spec-draft-n-max "$SPEC_DRAFT_N_MAX")
 fi
 
 DRAFT_ARGS=()
@@ -122,6 +127,7 @@ exec "$LLAMA_BIN" \
     --ubatch-size "$UBATCH_SIZE" \
     --threads "$NUM_THREADS" \
     --threads-batch "$THREAD_BATCH" \
+    --poll "$POLL_LEVEL" \
     --cache-type-k "$CACHE_TYPE_K" \
     --cache-type-v "$CACHE_TYPE_V" \
     "${SPEC_ARGS[@]}" \