karylab_agents/vllm
1
0
Fork 0
forked from Karylab-cklius/vllm
Code Pull Requests 2 Actions 1 Packages Activity

[Frontend]Flatten beam-search beams with itertools.chain instead of sum (#48829)

Browse Source 
Signed-off-by: Wang Xingda <wangxingda1993@126.com>
Co-authored-by: 王兴达 <wangxingda@360itdeMacBook-Pro.local>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Wang Xingda
2026-07-17 23:09:14 +01:00
committed by GitHub
co-authored by 王兴达 mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent c9be3a8aa1
commit fae543015c
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/entrypoints/generate/beam_search/offline.py
+3 -1
View File
@@ -207,7 +207,9 @@ class BeamSearchOfflineMixin(OfflineInferenceMixin):
Returns True if all beams are exhausted and search should stop.
"""
all_beams: list[BeamSearchSequence] = list(
sum((instance.beams for instance in instances_batch), [])
itertools.chain.from_iterable(
instance.beams for instance in instances_batch
)
)
pos = [0] + list(
itertools.accumulate(len(instance.beams) for instance in instances_batch)