From 6a7b3bc29cdd4c0bc25c87396b7567bb46c58607 Mon Sep 17 00:00:00 2001 From: Sergey-Zlobin <52936797+Sergey-Zlobin@users.noreply.github.com> Date: Thu, 29 Jan 2026 11:44:30 +0300 Subject: [PATCH] Qwen3-VL-MoE EAGLE support for vLLM-Ascend (#6327) ### What this PR does / why we need it? Qwen3-VL-MoE EAGLE support for vLLM-Ascend ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? The patch tested with Qwen3-VL-30B-A3B-Instruct model - vLLM version: v0.14.1 - vLLM main: https://github.com/vllm-project/vllm/commit/dc917cceb877dfd13f98c538c4c96158047d98bd Signed-off-by: Sergey_Zlobin --- vllm_ascend/spec_decode/eagle_proposer.py | 1 + 1 file changed, 1 insertion(+) diff --git a/vllm_ascend/spec_decode/eagle_proposer.py b/vllm_ascend/spec_decode/eagle_proposer.py index 505326c8..c23c1751 100644 --- a/vllm_ascend/spec_decode/eagle_proposer.py +++ b/vllm_ascend/spec_decode/eagle_proposer.py @@ -198,6 +198,7 @@ class EagleProposer(VllmEagleProposer): if self.get_model_name(model) in [ "Qwen2_5_VLForConditionalGeneration", "Qwen3VLForConditionalGeneration", + "Qwen3VLMoeForConditionalGeneration", ]: self.model.config.image_token_index = model.config.image_token_id elif self.get_model_name(