xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-11 00:15:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Perf][Easy] Early stop in request_block_hasher (#26112)

Browse Source 
Signed-off-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>
This commit is contained in:
Jialin Ouyang 2025-10-05 01:46:03 -07:00 committed by GitHub
parent e0986ea07b
commit 201c971e96
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/v1/core/kv_cache_utils.py
View File

@ -585,6 +585,10 @@ def get_request_block_hasher(
start_token_idx = len(request.block_hashes) * block_size start_token_idx = len(request.block_hashes) * block_size
num_tokens = request.num_tokens num_tokens = request.num_tokens
if start_token_idx + block_size > num_tokens:
# Early stop when there no new full blocks created.
return []
curr_mm_idx = 0 curr_mm_idx = 0
if start_token_idx > 0: if start_token_idx > 0:
# Set curr_mm_idx = -1 to indicate the last mm input. # Set curr_mm_idx = -1 to indicate the last mm input.