From 45889a6185992b30f09d5df73db333e43f11ba61 Mon Sep 17 00:00:00 2001 From: Jade Zheng Date: Fri, 12 Dec 2025 22:36:23 +0800 Subject: [PATCH] [Bugfix] Pass vllm_config to kv_connector_no_forward in NPUModelRunner (#4970) ### What this PR does / why we need it? The newest version crashes in PD separation scenarios because the function is missing the `vllm_config` parameter. - vLLM version: v0.12.0 - vLLM main: https://github.com/vllm-project/vllm/commit/ad32e3e19ccf0526cb6744a5fed09a138a5fb2f9 Signed-off-by: Jade Zheng --- vllm_ascend/worker/model_runner_v1.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/vllm_ascend/worker/model_runner_v1.py b/vllm_ascend/worker/model_runner_v1.py index c6c881eb..20f9badf 100644 --- a/vllm_ascend/worker/model_runner_v1.py +++ b/vllm_ascend/worker/model_runner_v1.py @@ -1444,7 +1444,8 @@ class NPUModelRunner(GPUModelRunner): ) # Return empty ModelRunnerOuptut if there's no work to do. return EMPTY_MODEL_RUNNER_OUTPUT - return self.kv_connector_no_forward(scheduler_output) + return self.kv_connector_no_forward(scheduler_output, + self.vllm_config) if self.dynamic_eplb: self.eplb_updator.forward_before()