[V1] Enable prefill optimization for Gemma3n (#22628)

Signed-off-by: Yong Hoon Shin <yhshin@meta.com>

vllm/config/cache.py

@@ -145,12 +145,19 @@ class CacheConfig:
 
         self._verify_cache_dtype()
         self._verify_prefix_caching()
+        self._verify_kv_sharing_fast_prefill()
 
     def metrics_info(self):
         # convert cache_config to dict(key: str, value: str) for prometheus
         # metrics info
         return {key: str(value) for key, value in self.__dict__.items()}
 
+    def _verify_kv_sharing_fast_prefill(self) -> None:
+        if self.kv_sharing_fast_prefill and not envs.VLLM_USE_V1:
+            raise NotImplementedError(
+                "Fast prefill optimization for KV sharing is not supported "
+                "in V0 currently.")
+
     @model_validator(mode='after')
     def _verify_args(self) -> Self:
         if self.cpu_offload_gb < 0:
@@ -162,11 +169,6 @@ class CacheConfig:
                 "GPU memory utilization must be less than 1.0. Got "
                 f"{self.gpu_memory_utilization}.")
 
-        if self.kv_sharing_fast_prefill:
-            logger.warning_once(
-                "--kv-sharing-fast-prefill is currently work in progress "
-                "and not functional yet (i.e. no prefill savings)")
-
         return self
 
     def _verify_cache_dtype(self) -> None: