chore: 更新文档, 修正代码格式
This commit is contained in:
@@ -380,7 +380,7 @@ class InferenceScheduler:
|
||||
self._running = False
|
||||
if hasattr(self, "_loop_thread"):
|
||||
self._loop_thread.join(timeout=1.0)
|
||||
|
||||
|
||||
# Clear KV cache to free GPU memory
|
||||
if self.kv_cache is not None:
|
||||
k_cache, v_cache = self.kv_cache
|
||||
@@ -388,10 +388,10 @@ class InferenceScheduler:
|
||||
k_cache.detach()
|
||||
if v_cache is not None:
|
||||
v_cache.detach()
|
||||
|
||||
|
||||
# Clear seq mask
|
||||
self.seq_mask.detach()
|
||||
|
||||
|
||||
# Clear task lists
|
||||
self.waiting_queue.clear()
|
||||
self.active_tasks.clear()
|
||||
|
||||
Reference in New Issue
Block a user