frank-default-small: pull expert blocks 28,29 back to GPU

Fused MoE (fused_moe / fused_up_gate / fused_mmad) is on by default in
ik_llama - no flag needed; only -no-fmoe disables it. Slide two more
expert blocks (28,29) from CPU onto the RTX 3090 (~1.26 GiB) for the
fused-MoE run, leaving ~1.6 GiB headroom. Update override regex, test,
and KB (including a note that the headroom is now tighter than the
2-3 GiB ideal).

docs/LLM/ikllama-ornith-dankserver.md

@@ -149,13 +149,13 @@ cache RAM vs f16.
 The expert-offload pattern from upstream llama.cpp carries over: the
 `--override-tensor` regex pins the sparse expert (`ffn_*_exps`) tensors to
 the CPU buft, leaving every full-attention block plus blocks 35-40 on CUDA.
-On 2026-08-26 we brought expert blocks **30, 32, 33, 34 back to CUDA**
-(~2.5 GiB) to use the ~5.4 GiB GPU headroom for lower latency; the rest of
-that CPU range stays on CPU.
+On 2026-08-26 we brought expert blocks **30, 32, 33, 34** back to CUDA
+(~2.5 GiB), then **28 and 29** (~1.26 GiB) for the fused-MoE run, to use the
+GPU headroom for lower latency; the rest of that CPU range stays on CPU.
 
 - `OVERRIDE_EXPERT` — the layer-bounded regex listing only the blocks whose
   experts stay on CPU: `(0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|
-  24|25|26|28|29)`. Blocks 30,32,33,34 (and the always-on full-attention +
+  24|25|26)`. Blocks 28,29,30,32,33,34 (and the always-on full-attention +
   35-40 set) are left off the list so their experts load to CUDA.
 - `OVERRIDE_BUFT=CPU`.
 - `N_GPU_LAYERS=999` — ik_llama `-ngl` / `--gpu-layers` only accepts an
@@ -163,9 +163,16 @@ that CPU range stays on CPU.
   layers as possible", while the expert override still pins the listed exps
   to CPU.
 
+> **Fused MoE** (`fused_moe`, `fused_up_gate`, `fused_mmad`) is **on by
+> default** in ik_llama — there is no positive flag; only `-no-fmoe`
+> disables it. `only_active_experts` is also on by default. Confirmed in the
+> boot log for the running service.
+
 To move more experts back to the GPU later, drop more block numbers from
 `OVERRIDE_EXPERT` (each block is ~630 MiB; verify `nvidia-smi` headroom
-first and keep ~2-3 GiB margin for KV/compute).
+first and keep ~2-3 GiB margin for KV/compute). After 28,29 were added the
+free headroom was ~1.6 GiB, which is tighter than ideal — add more only
+with a headroom check.
 
 ### Qwen-Sharp chat template via --jinja
 

scripts/frank-default-small.env

@@ -66,9 +66,10 @@ SPEC_TYPE=none
 # CPU buft name on this host.
 #
 # 2026-08-26: brought back expert blocks 30,32,33,34 to CUDA (~2.5 GiB) to
-# use the ~5.4 GiB GPU headroom for lower latency; the remaining larger-CPU
+# use the ~5.4 GiB GPU headroom for lower latency; then 28,29 (~1.26 GiB)
+# for the fused-MoE run. Left ~1.6 GiB headroom. The remaining larger-CPU
 # range below those stays on CPU.
-OVERRIDE_EXPERT='blk\.(0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29)\.ffn_(up|down|gate|gate_up)_(ch|)exps'
+OVERRIDE_EXPERT='blk\.(0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26)\.ffn_(up|down|gate|gate_up)_(ch|)exps'
 OVERRIDE_BUFT=CPU
 # ik_llama --gpu-layers/-ngl does NOT accept the literal "all"; use 999 to
 # mean "load as many layers as possible". Expert exps are pinned to CPU by

tests/frank_small_config.bats

@@ -22,7 +22,7 @@ TEMPLATE_FILE="$REPO_ROOT/scripts/chat-template.jinja"
 
 @test "frank-default-small keeps full-attention and final blocks on CUDA" {
     source "$ENV_FILE"
-    [[ "$OVERRIDE_EXPERT" == 'blk\.(0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26|28|29)\.ffn_(up|down|gate|gate_up)_(ch|)exps' ]]
+    [[ "$OVERRIDE_EXPERT" == 'blk\.(0|1|2|4|5|6|8|9|10|12|13|14|16|17|18|20|21|22|24|25|26)\.ffn_(up|down|gate|gate_up)_(ch|)exps' ]]
     [[ "$OVERRIDE_EXPERT" != *'|3|'* ]]
     [[ "$OVERRIDE_EXPERT" != *'|7|'* ]]
     [[ "$OVERRIDE_EXPERT" != *'|11|'* ]]