[v1] Hybrid Memory Allocator (#17996)

Signed-off-by: Chen Zhang <zhangch99@outlook.com>

vllm/config.py

@@ -2104,6 +2104,12 @@ class SchedulerConfig:
     default scheduler. Can be a class directly or the path to a class of form
     "mod.custom_class"."""
 
+    disable_hybrid_kv_cache_manager: bool = False
+    """If set to True, KV cache manager will allocate the same size of KV cache
+    for all attention layers even if there are multiple type of attention layers
+    like full attention and sliding window attention.
+    """
+
     def compute_hash(self) -> str:
         """
         WARNING: Whenever a new field is added to this config,
@@ -4465,6 +4471,21 @@ class VllmConfig:
         if not self.instance_id:
             self.instance_id = random_uuid()[:5]
 
+        if (envs.VLLM_USE_V1
+                and not self.scheduler_config.disable_hybrid_kv_cache_manager):
+            # logger should only print warning message for hybrid models. As we
+            # can't know whether the model is hybrid or not now, so we don't log
+            # warning message here and will log it later.
+            if not (current_platform.is_cuda() or current_platform.is_rocm()):
+                # Hybrid KV cache manager is not supported on non-GPU platforms.
+                self.disable_hybrid_kv_cache_manager = True
+            if self.kv_transfer_config is not None:
+                # Hybrid KV cache manager is not compatible with KV transfer.
+                self.disable_hybrid_kv_cache_manager = True
+            if self.kv_events_config is not None:
+                # Hybrid KV cache manager is not compatible with KV events.
+                self.disable_hybrid_kv_cache_manager = True
+
     def update_sizes_for_sequence_parallelism(self,
                                               possible_sizes: list) -> list:
         # remove the sizes that not multiple of tp_size when