Fix race condition in async lora unload (#9084)

2025-08-11 22:59:29 -07:00
parent 4093d460ce
commit 5ded39cab2
4 changed files with 35 additions and 19 deletions
--- a/python/sglang/srt/utils.py
+++ b/python/sglang/srt/utils.py
@@ -2960,7 +2960,7 @@ class ConcurrentCounter:
        This suspends the calling coroutine without blocking the thread, allowing
        other tasks to run while waiting. When the counter becomes zero, the coroutine resumes.
        """
-        self.wait_for(lambda count: count == 0)
+        await self.wait_for(lambda count: count == 0)


@lru_cache(maxsize=1)