From b7a9d0f232dde05bad32afcf23a0fd4cad36586b Mon Sep 17 00:00:00 2001
From: OMRailgun <yxz1149901132@163.com>
Date: Mon, 30 Mar 2026 21:51:59 +0800
Subject: [PATCH 1/2] Update config.py

update bailingMMAPI
---
 vlmeval/config.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/vlmeval/config.py b/vlmeval/config.py
index 05af2ff76..c1d877a17 100644
--- a/vlmeval/config.py
+++ b/vlmeval/config.py
@@ -645,6 +645,9 @@
     "BailingMM-Pro-0120": partial(
         api.bailingMMAPI, model="BailingMM-Pro-0120", temperature=0, retry=10
     ),
+    "Ming-flash-omni-2.0": partial(
+        api.bailingMMAPI, model="Ming-flash-omni-2.0", temperature=0, retry=10
+    ),
     # BlueLM-2.5
     "BlueLM-2.5-3B": partial(api.BlueLM_API, model="BlueLM-2.5-3B", temperature=0, retry=3),
     # JiuTian-VL

From 761d313c069863edb54b8f24e69a5366ae503e20 Mon Sep 17 00:00:00 2001
From: OMRailgun <yxz1149901132@163.com>
Date: Mon, 13 Apr 2026 17:59:03 +0800
Subject: [PATCH 2/2] Update bailingmm.py

---
 vlmeval/api/bailingmm.py | 13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)

diff --git a/vlmeval/api/bailingmm.py b/vlmeval/api/bailingmm.py
index fdbcf0400..2d266baa3 100644
--- a/vlmeval/api/bailingmm.py
+++ b/vlmeval/api/bailingmm.py
@@ -42,20 +42,18 @@ def image_to_base64(self, image_path):
 
     def prepare_inputs(self, inputs):
         msgs = cp.deepcopy(inputs)
-        content = []
+        messages = [{"role": "user","content": []}]
         for i, msg in enumerate(msgs):
             if msg['type'] == 'text':
-                pass
+                messages[-1]["content"].append({"type": "text","text": msg["value"]})
             else:
                 try:
                     image_data = self.image_to_base64(msg['value'])
+                    messages[-1]["content"].append({"type": "image_url","image_url": {"url": f"data:image;base64,{image_data}"}})
                 except Exception as e:
                     if self.verbose:
                         logger.error(e)
-                    image_data = ''
-                msg['value'] = image_data
-            content.append(msg)
-        return content
+        return messages
 
     def generate_inner(self, inputs, **kwargs) -> str:
         assert isinstance(inputs, str) or isinstance(inputs, list)
@@ -63,11 +61,10 @@ def generate_inner(self, inputs, **kwargs) -> str:
         inputs = [inputs] if isinstance(inputs, str) else inputs
 
         messages = self.prepare_inputs(inputs)
-
         service_url = "https://bailingchat.alipay.com/api/proxy/eval/antgmm/completions"
 
         payload = {
-            "structInput": json.dumps([{"role": "user", "content": messages}]),
+            "structInput": json.dumps(messages, ensure_ascii=False),
             "sk": self.key,
             "model": self.model,
             "timeout": 180000