[Misc] Multi-Modality Support for Loading Local image Files

FIX vllm-project#8730

Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>

vllm/multimodal/utils.py

@@ -18,13 +18,21 @@
 cached_get_tokenizer = lru_cache(get_tokenizer)
 
 
-def _load_image_from_bytes(b: bytes):
+def _load_image_from_bytes(b: bytes) -> Image.Image:
     image = Image.open(BytesIO(b))
     image.load()
     return image
 
 
-def _load_image_from_data_url(image_url: str):
+def _load_image_from_file(image_url: str) -> Image.Image:
+    # Only split once and assume the second part is the image path
+    _, image_path = image_url.split("file://", 1)
+    image = Image.open(image_path)
+    image.load()
+    return image
+
+
+def _load_image_from_data_url(image_url: str) -> Image.Image:
     # Only split once and assume the second part is the base64 encoded image
     _, image_base64 = image_url.split(",", 1)
     return load_image_from_base64(image_base64)
@@ -43,9 +51,11 @@ def fetch_image(image_url: str, *, image_mode: str = "RGB") -> Image.Image:
 
     elif image_url.startswith('data:image'):
         image = _load_image_from_data_url(image_url)
+    elif image_url.startswith('file://'):
+        image = _load_image_from_file(image_url)
     else:
         raise ValueError("Invalid 'image_url': A valid 'image_url' must start "
-                         "with either 'data:image' or 'http'.")
+                         "with either 'data:image', 'file://' or 'http'.")
 
     return image.convert(image_mode)
 
@@ -65,9 +75,11 @@ async def async_fetch_image(image_url: str,
 
     elif image_url.startswith('data:image'):
         image = _load_image_from_data_url(image_url)
+    elif image_url.startswith('file://'):
+        image = _load_image_from_file(image_url)
     else:
         raise ValueError("Invalid 'image_url': A valid 'image_url' must start "
-                         "with either 'data:image' or 'http'.")
+                         "with either 'data:image', 'file://' or 'http'.")
 
     return image.convert(image_mode)