From a86ef76c155330060c79f7333688c848bd989427 Mon Sep 17 00:00:00 2001 From: Sage Ahrac Date: Tue, 16 Dec 2025 20:33:55 +0200 Subject: [PATCH] add lora name as non breaking change Signed-off-by: Sage Ahrac --- examples/online_serving/kv_events_subscriber.py | 3 ++- tests/v1/engine/test_engine_core_client.py | 1 + tests/v1/kv_connector/unit/test_offloading_connector.py | 1 + vllm/distributed/kv_events.py | 3 ++- .../kv_transfer/kv_connector/v1/offloading_connector.py | 3 ++- vllm/v1/core/block_pool.py | 5 ++++- 6 files changed, 12 insertions(+), 4 deletions(-) diff --git a/examples/online_serving/kv_events_subscriber.py b/examples/online_serving/kv_events_subscriber.py index e98ce6c6c9008..d30510634f47d 100644 --- a/examples/online_serving/kv_events_subscriber.py +++ b/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): diff --git a/tests/v1/engine/test_engine_core_client.py b/tests/v1/engine/test_engine_core_client.py index 09cdc64aba3c1..adf71f60f473c 100644 --- a/tests/v1/engine/test_engine_core_client.py +++ b/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" diff --git a/tests/v1/kv_connector/unit/test_offloading_connector.py b/tests/v1/kv_connector/unit/test_offloading_connector.py index 86d05172e159b..e1ea0b2980565 100644 --- a/tests/v1/kv_connector/unit/test_offloading_connector.py +++ b/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) diff --git a/vllm/distributed/kv_events.py b/vllm/distributed/kv_events.py index 144f3ea2b028e..8137bdb758fe6 100644 --- a/vllm/distributed/kv_events.py +++ b/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( diff --git a/vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py b/vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py index 2243073d61579..10aaa972f3280 100644 --- a/vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py +++ b/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, ) diff --git a/vllm/v1/core/block_pool.py b/vllm/v1/core/block_pool.py index bcd872c2f29a2..33d0b795b3115 100644 --- a/vllm/v1/core/block_pool.py +++ b/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, ) )