[Feat] Add modalities for vision server when handling pixel values for llava (#1346)

2024-09-09 17:07:34 +08:00
parent 8e6bdf851c
commit 662ecd9368
11 changed files with 40 additions and 2 deletions
--- a/python/sglang/srt/openai_api/protocol.py
+++ b/python/sglang/srt/openai_api/protocol.py
@@ -213,6 +213,7 @@ class ChatCompletionMessageContentImageURL(BaseModel):
 class ChatCompletionMessageContentImagePart(BaseModel):
    type: Literal["image_url"]
    image_url: ChatCompletionMessageContentImageURL
+    modalities: Optional[Literal["image", "multi-images", "video"]] = "image"


 ChatCompletionMessageContentPart = Union[