From 794fdd39ef3dc347b0d7fc1c8765a55522b59bef Mon Sep 17 00:00:00 2001
From: hanwlax <278915462+hanwlax@users.noreply.github.com>
Date: Thu, 28 May 2026 15:26:13 +0800
Subject: [PATCH] fix: adapt dots_vlm for transformers v5 (#25829)

---
 python/sglang/srt/multimodal/processors/dots_vlm.py | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/python/sglang/srt/multimodal/processors/dots_vlm.py b/python/sglang/srt/multimodal/processors/dots_vlm.py
index bc269cdae..66fd62762 100644
--- a/python/sglang/srt/multimodal/processors/dots_vlm.py
+++ b/python/sglang/srt/multimodal/processors/dots_vlm.py
@@ -33,8 +33,16 @@ class DotsVLMImageProcessor(BaseMultimodalProcessor):
         merge_size = vision_config.spatial_merge_size
 
         self.IMAGE_FACTOR = patch_size * merge_size
-        self.MIN_PIXELS = _processor.image_processor.min_pixels
-        self.MAX_PIXELS = _processor.image_processor.max_pixels
+        self.MIN_PIXELS = getattr(
+            _processor.image_processor,
+            "min_pixels",
+            getattr(_processor.image_processor, "size", {}).get("shortest_edge"),
+        )
+        self.MAX_PIXELS = getattr(
+            _processor.image_processor,
+            "max_pixels",
+            getattr(_processor.image_processor, "size", {}).get("longest_edge"),
+        )
         self.MAX_RATIO = 200
         self.mm_tokens = MultimodalSpecialTokens(
             image_token=self.IMAGE_TOKEN,