From df41aaafe72bbc0afad4ce2187d41d832971e579 Mon Sep 17 00:00:00 2001
From: Frank <frank@dankware>
Date: Tue, 11 Aug 2026 13:13:17 -0700
Subject: [PATCH] =?UTF-8?q?Offload=20one=20more=20MoE=20expert=20to=20CPU?=
 =?UTF-8?q?=20on=20frank-default-small=20(32=E2=86=9233=20of=2040)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 scripts/frank-default-small.env | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/scripts/frank-default-small.env b/scripts/frank-default-small.env
index 099133d..13a094a 100644
--- a/scripts/frank-default-small.env
+++ b/scripts/frank-default-small.env
@@ -14,9 +14,9 @@ SPEC_TYPE=ngram-mod
 NGRAM_MOD_N_MIN=48
 NGRAM_MOD_N_MAX=64
 NGRAM_MOD_N_MATCH=24
-# Expert offload: 32 of 40 layers keep MoE weights on CPU; remaining
-# 8 layers have their expert weights resident on GPU for faster routing.
-N_CPU_MOE=32
+# Expert offload: 33 of 40 layers keep MoE weights on CPU; remaining
+# 7 layers have their expert weights resident on GPU for faster routing.
+N_CPU_MOE=33
 N_GPU_LAYERS=all
 # KV cache on GPU, K quantized to q8_0, V stays f16 (user-locked decision).
 CACHE_TYPE_K=q8_0