Reduce small model context and rebalance GPU layers

scripts/serve-frank-small.sh

@@ -37,7 +37,7 @@ fi
 : "${ALIAS:=frank-default-small}"
 : "${BIND_ADDR:=0.0.0.0}"
 : "${LISTEN_PORT:=7770}"
-: "${CTX_SIZE:=786432}"
+: "${CTX_SIZE:=491520}"
 : "${PARALLEL:=3}"
 : "${SPEC_TYPE:=draft-mtp}"
 : "${SPEC_DRAFT_N_MAX:=4}"
@@ -49,7 +49,7 @@ fi
 # Expert offload: --override-tensor regex pinned to the CPU buft. Same pattern
 # the internal --cpu-moe / --n-cpu-moe use; explicit here so we only move the
 # MoE expert weights and leave every other tensor on the GPU.
-: "${OVERRIDE_EXPERT:=\.ffn_(up|down|gate|gate_up)_(ch|)exps}"
+: "${OVERRIDE_EXPERT:=blk\.(0|1|2|3|4|5|6|7|8|9|10|11|12|13|14|15|16|17|18|19|20|21|22|23|24|25|26|27|28|29|30|31|32)\.ffn_(up|down|gate|gate_up)_(ch|)exps}"
 : "${OVERRIDE_BUFT:=CPU}"
 : "${N_GPU_LAYERS:=all}"
 : "${BATCH_SIZE:=4096}"