From 2f1928354903ae0c6edfe76cc90081eb513ead2c Mon Sep 17 00:00:00 2001
From: youkaichao <youkaichao@gmail.com>
Date: Sat, 6 Apr 2024 19:14:06 -0700
Subject: [PATCH] [Core] latency optimization (#3890)

---
 vllm/core/block_manager_v1.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/vllm/core/block_manager_v1.py b/vllm/core/block_manager_v1.py
index b2aaeb33c5299..e7e3b4dc1e9b4 100644
--- a/vllm/core/block_manager_v1.py
+++ b/vllm/core/block_manager_v1.py
@@ -328,7 +328,7 @@ class BlockSpaceManagerV1(BlockSpaceManager):
         self,
         seq: Sequence,
     ) -> bool:
-        token_ids_len = len(seq.data.get_token_ids())
+        token_ids_len = seq.data.get_len()
         return token_ids_len > 0 and token_ids_len % seq.block_size == 0
 
     def _maybe_promote_last_block(