fix double-free kv cache for requests that have already finished and been freed during preemption (#18694)
This commit is contained in:
@@ -76,6 +76,7 @@ class TestPrefillAdder(CustomTestCase):
|
||||
req.output_ids = [0] * output_len
|
||||
req.sampling_params = SimpleNamespace(max_new_tokens=max_new_tokens)
|
||||
req.time_stats = SimpleNamespace(wait_queue_entry_time=wait_time)
|
||||
req.finished.return_value = False
|
||||
return req
|
||||
|
||||
def create_adder(self, running_batch):
|
||||
|
||||
Reference in New Issue
Block a user