add lora name as non breaking change

Signed-off-by: Sage Ahrac <sagiahrak@gmail.com>

examples/online_serving/kv_events_subscriber.py

@@ -28,8 +28,9 @@ class BlockStored(KVCacheEvent):
     parent_block_hash: ExternalBlockHash | None
     token_ids: list[int]
     block_size: int
-    lora_name: str | None
+    lora_id: int | None
     medium: str | None
+    lora_name: str | None
 
 
 class BlockRemoved(KVCacheEvent):

tests/v1/engine/test_engine_core_client.py

@@ -635,6 +635,7 @@ def test_kv_cache_events(
             "Block size should be the same as the block size"
         )
         assert event.parent_block_hash is None, "Parent block hash should be None"
+        assert event.lora_id is None, "Lora id should be None"
         assert event.lora_name is None, "Lora name should be None"
         assert len(event.token_ids) == num_blocks * block_size, (
             "Token ids should be the same as the custom tokens"

tests/v1/kv_connector/unit/test_offloading_connector.py

@@ -527,6 +527,7 @@ def test_offloading_connector(request_runner):
     assert event.medium == "A"
     assert event.token_ids == []
     assert event.parent_block_hash is None
+    assert event.lora_id is None
     assert event.lora_name is None
     event = events[1]
     assert isinstance(event, BlockRemoved)

vllm/distributed/kv_events.py

@@ -51,8 +51,9 @@ class BlockStored(KVCacheEvent):
     parent_block_hash: ExternalBlockHash | None
     token_ids: list[int]
     block_size: int
-    lora_name: str | None
+    lora_id: int | None
     medium: str | None
+    lora_name: str | None
 
     def __hash__(self) -> int:
         return hash(

vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py

@@ -402,9 +402,10 @@ class OffloadingConnectorScheduler:
                     block_hashes=event.block_hashes,
                     parent_block_hash=None,
                     token_ids=[],
-                    lora_name=None,
+                    lora_id=None,
                     block_size=event.block_size,
                     medium=event.medium,
+                    lora_name=None,
                 )
 
 

vllm/v1/core/block_pool.py

@@ -284,10 +284,13 @@ class BlockPool:
                         num_cached_blocks * block_size : num_full_blocks * block_size
                     ],
                     block_size=block_size,
-                    lora_name=request.lora_request.name
+                    lora_id=request.lora_request.adapter_id
                     if request.lora_request
                     else None,
                     medium=MEDIUM_GPU,
+                    lora_name=request.lora_request.name
+                    if request.lora_request
+                    else None,
                 )
             )