A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 8, 08:45 UTC
[KVConnector] Clean up redundant code in KV connectors (#34147)
Signed-off-by: Martin Hickey <martin.hickey@ie.ibm.com>
Martin Hickey committed
Feb 13, 2026 at 08:14 UTC
47e9b63e1afeb074b0fa584e0169e27d517b4e7b
4 files changed
+2
-11
vllm/distributed/kv_transfer/kv_connector/v1/__init__.py
+1
-1
@@ -6,7 +6,7 @@ from vllm.distributed.kv_transfer.kv_connector.v1.base import (
6
SupportsHMA,
7
supports_hma,
8
)
9
-from vllm.distributed.kv_transfer.kv_connector.v1.decode_bench_connector import ( # noqa E:501
9
+from vllm.distributed.kv_transfer.kv_connector.v1.decode_bench_connector import ( # noqa: E501
10
DecodeBenchConnector,
11
)
12
vllm/distributed/kv_transfer/kv_connector/v1/example_connector.py
-8
@@ -145,7 +145,6 @@ class ExampleConnector(KVConnectorBase_V1):
145
num_pages * page_size, -1
146
)
147
dst_kv_cache_layer[slot_mapping, ...] = src_kv_cache
148
- dst_kv_cache_layer.reshape(dst_kv_cache_layer_shape)
148
else:
149
num_pages = dst_kv_cache_layer_shape[1]
150
page_size = dst_kv_cache_layer_shape[2]
@@ -153,18 +152,11 @@ class ExampleConnector(KVConnectorBase_V1):
152
2, num_pages * page_size, -1
153
)
154
dst_kv_cache_layer[:, slot_mapping, ...] = src_kv_cache
156
- dst_kv_cache_layer.reshape(dst_kv_cache_layer_shape)
155
156
# Get the metadata
157
metadata: KVConnectorMetadata = self._get_connector_metadata()
158
assert isinstance(metadata, ExampleConnectorMetadata)
159
162
- if metadata is None:
163
- logger.warning(
164
- "In connector.start_load_kv, but the connector metadata is None"
165
- )
166
- return
167
-
160
attn_metadata = forward_context.attn_metadata
161
if attn_metadata is None:
162
logger.warning("In connector.start_load_kv, but the attn_metadata is None")
vllm/distributed/kv_transfer/kv_connector/v1/lmcache_mp_connector.py
-1
@@ -36,7 +36,6 @@ except ImportError:
36
)
37
38
if TYPE_CHECKING:
39
- from vllm.config import VllmConfig
39
from vllm.distributed.kv_events import KVCacheEvent
40
from vllm.distributed.kv_transfer.kv_connector.v1.metrics import (
41
KVConnectorPromMetrics,
vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py
+1
-1
@@ -85,7 +85,7 @@ class OffloadingConnectorStats(KVConnectorStats):
85
for transfer_type, ops_list in self.data.items():
86
assert isinstance(ops_list, list)
87
total_bytes = 0
88
- total_time = 0
88
+ total_time = 0.0
89
for op in ops_list:
90
assert isinstance(op, dict)
91
total_bytes += op["op_size"]