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, 16:00 UTC
[V1][Minor] Optimize get_cached_block (#16135)
Woosuk Kwon committed
Apr 6, 2025 at 13:48 UTC
b10e51989551cd80dd74079429ccf91f0807bd92
1 file changed
+5
-5
vllm/v1/core/block_pool.py
+5
-5
@@ -67,11 +67,11 @@ class BlockPool:
67
Returns:
68
The cached block if it exists, or None.
69
"""
70
- if block_hash in self.cached_block_hash_to_block:
71
- first_block_id = list(
72
- self.cached_block_hash_to_block[block_hash].keys())[0]
73
- return self.cached_block_hash_to_block[block_hash][first_block_id]
74
- return None
70
+ cached_blocks = self.cached_block_hash_to_block.get(block_hash)
71
+ if not cached_blocks:
72
+ return None
73
+ first_block_id = next(iter(cached_blocks))
74
+ return cached_blocks[first_block_id]
75
76
def cache_full_blocks(
77
self,