xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-21 06:46:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

[KV offload] Offloading connector async scheduling support (#27648)

Browse Source 
Signed-off-by: KevinCheung2259 <2651309292@qq.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
This commit is contained in:
Yue Zhang 2025-11-02 05:08:56 +08:00 committed by GitHub
parent 1e88fb751b
commit 685c99ee77
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py
View File

@ -274,8 +274,8 @@ class OffloadingConnectorScheduler:
if num_new_blocks <= 0: if num_new_blocks <= 0:
continue continue
num_gpu_blocks = num_blocks * self.block_size_factor # NOTE: In async scheduling, placeholders may temporarily make
assert len(req.block_hashes) >= num_gpu_blocks # len(req.block_hashes) < num_blocks * self.block_size_factor.
new_block_hashes = self._get_block_hashes( new_block_hashes = self._get_block_hashes(
req, start_idx=start_block_idx, end_idx=num_blocks req, start_idx=start_block_idx, end_idx=num_blocks