From d622851dc9f1d044e820a45a89ab38614d70c543 Mon Sep 17 00:00:00 2001 From: Rui Wang <45031995+wangraying@users.noreply.github.com> Date: Sun, 1 Dec 2024 09:53:43 +0800 Subject: [PATCH] [Fix] fix assertion error for chunked prefill when disabling cache (#2282) --- python/sglang/srt/managers/schedule_policy.py | 3 +++ 1 file changed, 3 insertions(+) diff --git a/python/sglang/srt/managers/schedule_policy.py b/python/sglang/srt/managers/schedule_policy.py index eeccea89b..bc35bfbd6 100644 --- a/python/sglang/srt/managers/schedule_policy.py +++ b/python/sglang/srt/managers/schedule_policy.py @@ -231,6 +231,9 @@ class PrefillAdder: break self.req_states.insert(i, (tokens_left, tokens_occupied)) + if len(self.can_run_list) != 0: + return AddReqResult.OTHER + if self.req_states is None: self.req_states = [] add_req_state(req)