[Bugfix] Fix multimodal configuration for Qwen3VL MOE model (#30670)

Signed-off-by: Max Hu <hyoung2991@gmail.com>

vllm/model_executor/models/qwen3_vl_moe.py

@@ -418,6 +418,11 @@ class Qwen3VLMoeForConditionalGeneration(
 
         self.config = config
         self.multimodal_config = multimodal_config
+        self.use_data_parallel = multimodal_config.mm_encoder_tp_mode == "data"
+        self.video_pruning_rate = multimodal_config.video_pruning_rate
+        self.is_multimodal_pruning_enabled = (
+            multimodal_config.is_multimodal_pruning_enabled()
+        )
 
         if not multimodal_config.get_limit_per_prompt(
             "image"