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 16, 10:45 UTC
[Misc] Fix typo in BlockSpaceManagerV1 (#8944)
juncheoll committed
Sep 30, 2024 at 00:05 UTC
1fb9c1b0bf8e65e6576ff4c45f5623d233d7194b
1 file changed
+1
-1
vllm/core/block_manager_v1.py
+1
-1
@@ -443,7 +443,7 @@ class BlockSpaceManagerV1(BlockSpaceManager):
443
# prefix tokens)
444
new_block = self.gpu_allocator.allocate(block_hash, num_hashed_tokens)
445
446
- # If the block has is None, then the block is not full.
446
+ # If the block_hash is None, then the block is not full.
447
# If the block is not full, then we expect it to have a refcount of 1.
448
if block_hash is None:
449
assert new_block.ref_count == 1