refactor: multimodal data (#4754)

2025-04-01 00:57:51 +08:00
parent c7457191a0
commit 5cb552b1d4
36 changed files with 989 additions and 1138 deletions
--- a/test/srt/test_vision_openai_server.py
+++ b/test/srt/test_vision_openai_server.py
@@ -155,9 +155,7 @@ class TestOpenAIVisionServer(CustomTestCase):
                    "content": [
                        {
                            "type": "image_url",
-                            "image_url": {
-                                "url": "https://raw.githubusercontent.com/sgl-project/sglang/main/test/lang/example_image.png"
-                            },
+                            "image_url": {"url": IMAGE_MAN_IRONING_URL},
                            "modalities": "multi-images",
                        },
                        {
@@ -399,14 +397,14 @@ class TestOpenAIVisionServer(CustomTestCase):
            {
                "role": "user",
                "content": [
-                    {
-                        "type": "text",
-                        "text": prompt,
-                    },
                    {
                        "type": "audio_url",
                        "audio_url": {"url": f"{audio_file_name}"},
                    },
+                    {
+                        "type": "text",
+                        "text": prompt,
+                    },
                ],
            }
        ]