EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Bugfix] Pass vllm_config to kv_connector_no_forward in NPUModelRunner (#4970)

Browse Source 
### 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:
ad32e3e19c

Signed-off-by: Jade Zheng <zheng.shoujian@outlook.com>
This commit is contained in:
Jade Zheng
2025-12-12 22:36:23 +08:00
committed by GitHub
parent fa367e3b1a
commit 45889a6185
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm_ascend/worker/model_runner_v1.py
View File

@@ -1444,7 +1444,8 @@ class NPUModelRunner(GPUModelRunner):
) )
# Return empty ModelRunnerOuptut if there's no work to do. # Return empty ModelRunnerOuptut if there's no work to do.
return EMPTY_MODEL_RUNNER_OUTPUT 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: if self.dynamic_eplb:
self.eplb_updator.forward_before() self.eplb_updator.forward_before()