minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[lora] Exclude finished requests from running_loras (#27597)

Browse Source
This commit is contained in:
Erik Wijmans
2026-06-08 23:05:43 -07:00
committed by GitHub
parent 9a3e845fc1
commit dff695c76b
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/managers/scheduler.py
+3 -1
View File
@@ -2641,7 +2641,9 @@ class Scheduler(
self.chunked_req = adder.add_chunked_req(self.chunked_req) self.chunked_req = adder.add_chunked_req(self.chunked_req)
if self.enable_lora: if self.enable_lora:
running_loras = {req.lora_id for req in self.running_batch.reqs} running_loras = {
req.lora_id for req in self.running_batch.reqs if not req.finished()
}
# Account for LoRAs that are already loaded in the adder, such as chunked requests # Account for LoRAs that are already loaded in the adder, such as chunked requests
running_loras.update(req.lora_id for req in adder.can_run_list) running_loras.update(req.lora_id for req in adder.can_run_list)