frozenleaves/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-10-20 14:53:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

[V1][Minor] Avoid frequently creating ConstantList (#12653)

Browse Source 
A small optimization to avoid creating a new `ConstantList` every time `request.kv_block_hashes` is used.

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon
2025-02-01 23:43:20 -08:00
committed by GitHub
parent e497f33491
commit abfcdcdf27
1 changed files with 2 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/v1/request.py
View File

@ -64,6 +64,7 @@ class Request:
# Cache the computed kv block hashes of the request to avoid
# recomputing.
self._kv_block_hashes: List[BlockHashType] = []
self.kv_block_hashes = ConstantList(self._kv_block_hashes)
# Read-only views
# Prevent directly appending to the these lists since
@ -121,13 +122,9 @@ class Request:
num_tokens = self.mm_positions[input_id]["length"]
return num_tokens
@property
def kv_block_hashes(self) -> ConstantList["BlockHashType"]:
# Prevent directly appending to the kv_block_hashes.
return ConstantList(self._kv_block_hashes)
def set_kv_block_hashes(self, value: List["BlockHashType"]) -> None:
self._kv_block_hashes = value
self.kv_block_hashes = ConstantList(self._kv_block_hashes)
def append_kv_block_hashes(self, block_hash: "BlockHashType") -> None:
self._kv_block_hashes.append(block_hash)