[model] support ERNIE-4.5-VL Models (#9521)

src/llamafactory/data/mm_plugin.py

@@ -465,6 +465,38 @@ class BasePlugin(MMPluginMixin):
         self._validate_input(processor, images, videos, audios)
         return self._get_mm_inputs(images, videos, audios, processor)
 
+@dataclass
+class ErnieVLPlugin(BasePlugin):
+    @override
+    def process_messages(
+        self,
+        messages: list[dict[str, str]],
+        images: list["ImageInput"],
+        videos: list["VideoInput"],
+        audios: list["AudioInput"],
+        processor: Optional["MMProcessor"],
+    ) -> list[dict[str, str]]:
+        self._validate_input(processor, images, videos, audios)
+        self._validate_messages(messages, images, videos, audios)
+        messages = deepcopy(messages)
+        image_idx, video_idx = 0, 0
+        for message in messages:
+            content = message["content"]
+            image_token = self.image_token or "<|image@placeholder|>"
+            video_token = self.video_token or "<|video@placeholder|>"
+            while IMAGE_PLACEHOLDER in content:
+                image_idx += 1
+                content = content.replace(
+                    IMAGE_PLACEHOLDER, f"Picture {image_idx}:<|IMAGE_START|>{image_token}<|IMAGE_END|>", 1
+                )
+            while VIDEO_PLACEHOLDER in content:
+                video_idx += 1
+                content = content.replace(
+                    VIDEO_PLACEHOLDER, f"Video {video_idx}:<|VIDEO_START|>{video_token}<|VIDEO_END|>", 1
+                )
+            message["content"] = content
+        return messages
+
 
 @dataclass
 class Gemma3Plugin(BasePlugin):
@@ -2039,6 +2071,7 @@ class VideoLlavaPlugin(BasePlugin):
 
 PLUGINS = {
     "base": BasePlugin,
+    "ernie_vl": ErnieVLPlugin,
     "gemma3": Gemma3Plugin,
     "glm4v": GLM4VPlugin,
     "gemma3n": Gemma3nPlugin,