[Model] Allow users to control skip reading cache per request. (#28194)

Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
Signed-off-by: wang.yuqi <noooop@126.com>

vllm/v1/core/kv_cache_manager.py

@@ -185,12 +185,11 @@ class KVCacheManager:
                 - A list of blocks that are computed for the request.
                 - The number of computed tokens.
         """
-        # Prefix caching is disabled or
-        # When the request requires prompt logprobs, we skip prefix caching.
-        if not self.enable_caching or (
-            request.sampling_params is not None
-            and request.sampling_params.prompt_logprobs is not None
-        ):
+        # We skip finding the prefix cache hit when prefix caching is
+        # disabled or the request is marked as skipping kv cache read
+        # (which happens when the request requires prompt logprobs
+        # or calls a pooling model with all pooling).
+        if not self.enable_caching or request.skip_reading_prefix_cache:
             return self.empty_kv_cache_blocks, 0
 
         # NOTE: When all tokens hit the cache, we must recompute the last token