From 6871aad03dc8cb5408bcdf0323799bde5aa56262 Mon Sep 17 00:00:00 2001 From: chen <103103266+ckl117@users.noreply.github.com> Date: Fri, 7 Nov 2025 19:43:25 +0800 Subject: [PATCH] [BugFix] fix token_processor zmq (#4827) * fix_token_processor_zmq * check pooling_model's token_ids is None * revert --- fastdeploy/output/token_processor.py | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/fastdeploy/output/token_processor.py b/fastdeploy/output/token_processor.py index f14d10d94..e6efcc33e 100644 --- a/fastdeploy/output/token_processor.py +++ b/fastdeploy/output/token_processor.py @@ -241,6 +241,11 @@ class TokenProcessor: task_id = task.request_id token_ids = stream_data.tokens # numpy.array + if token_ids is not None and token_ids[-1] <= 0: + if envs.ENABLE_V1_KVCACHE_SCHEDULER: + if task_id in self.resource_manager.to_be_rescheduled_request_id_set: + self.resource_manager.reschedule_preempt_task(task_id) + continue current_time = time.time() if self.tokens_counter[task_id] == 0: