From 7be9ffcd9f5c1a93705caac7d556794dd2608099 Mon Sep 17 00:00:00 2001 From: Roger Wang Date: Thu, 25 Sep 2025 03:16:45 -0700 Subject: [PATCH] [Misc] Fix Qwen3-VL `video_grid_thw` typing (#25646) Signed-off-by: Roger Wang --- vllm/model_executor/models/qwen3_vl.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/model_executor/models/qwen3_vl.py b/vllm/model_executor/models/qwen3_vl.py index d4f1547fd8e50..4ea93d3fdf417 100644 --- a/vllm/model_executor/models/qwen3_vl.py +++ b/vllm/model_executor/models/qwen3_vl.py @@ -1249,7 +1249,7 @@ class Qwen3VLForConditionalGeneration(nn.Module, SupportsMultiModal, rope_type="rope_3d") else: video_embeds = self.visual(pixel_values_videos, - grid_thw=grid_thw) + grid_thw=grid_thw_list) # Split concatenated embeddings for each video item. # Using prod on grid_thw_list instead of grid_thw.prod avoids CUDA sync