From 75a84c81272a06a160bcc4620f05c718a3c5f12d Mon Sep 17 00:00:00 2001
From: WeiQing Chen <40507679+david6666666@users.noreply.github.com>
Date: Tue, 9 Dec 2025 20:27:37 +0800
Subject: [PATCH] [Bugfix] fix issue251, qwen3 omni does not support chunked
 prefill now (#256)

Signed-off-by: David Chen <530634352@qq.com>
---
 vllm_omni/model_executor/stage_configs/qwen3_omni_moe.yaml | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/vllm_omni/model_executor/stage_configs/qwen3_omni_moe.yaml b/vllm_omni/model_executor/stage_configs/qwen3_omni_moe.yaml
index e73ad0205..6480b7012 100644
--- a/vllm_omni/model_executor/stage_configs/qwen3_omni_moe.yaml
+++ b/vllm_omni/model_executor/stage_configs/qwen3_omni_moe.yaml
@@ -20,6 +20,7 @@ stage_args:
       engine_output_type: latent  # Output hidden states for talker
       distributed_executor_backend: "mp"
       enable_prefix_caching: false
+      max_num_batched_tokens: 32768
       hf_config_name: thinker_config
       tensor_parallel_size: 2
     final_output: true
@@ -49,6 +50,7 @@ stage_args:
        engine_output_type: latent  # Output codec codes for code2wav
       #  tensor_parallel_size: 2
        enable_prefix_caching: false
+       max_num_batched_tokens: 32768
        distributed_executor_backend: "mp"
        hf_config_name: talker_config
     engine_input_source: [0]