fix llava_next regression in latest main

Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

server/text_generation_server/models/custom_modeling/flash_llama_modeling.py

@@ -424,7 +424,7 @@ def __init__(self, prefix, config, weights):
                 FlashLlamaLayer(
                     index=0,
                     prefix=(
-                        "model.layers.0" if not prefix else "{prefix}.model.layers.0"
+                        "model.layers.0" if not prefix else f"{prefix}.model.layers.0"
                     ),
                     config=config,
                     weights=weights,