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 15, 06:00 UTC
[Core] add remove_seq_from_computed_blocks_tracker to BlockSpaceManager (#19686)
Signed-off-by: 刘全 <quan.liu2@dbappsecurity.com.cn> Co-authored-by: 刘全 <quan.liu2@dbappsecurity.com.cn>
quanliu committed
Jun 17, 2025 at 12:40 UTC
5c76b9cdafb2abe205d675497f3ddb5603a07170
2 files changed
+7
vllm/core/interfaces.py
+4
@@ -133,3 +133,7 @@ class BlockSpaceManager(ABC):
133
@abstractmethod
134
def get_num_cached_tokens(self, seq: Sequence) -> int:
135
pass
136
+
137
+ @abstractmethod
138
+ def remove_seq_from_computed_blocks_tracker(self, seq: Sequence) -> None:
139
+ pass
\ No newline at end of file
vllm/core/placeholder_block_space_manager.py
+3
@@ -98,3 +98,6 @@ class PlaceholderBlockSpaceManager(BlockSpaceManager):
98
99
def get_num_cached_tokens(self, seq: Sequence) -> int:
100
return 0
101
+
102
+ def remove_seq_from_computed_blocks_tracker(self, seq: Sequence) -> None:
103
+ return