try something else

handler.py

@@ -12,12 +12,12 @@ class EndpointHandler():
         self.processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
 
 
-    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
-        inputs = data.get("inputs")
-        text = inputs.get("text")
-        imageData = inputs.get("image")
-        image = Image.open(BytesIO(base64.b64decode(imageData)))
-        inputs = self.processor(text=text, images=image, return_tensors="pt", padding=True)
-        outputs = self.model(**inputs)
-        embeddings = outputs.image_embeds.detach().numpy().flatten().tolist()
+    def __call__(self, data):
+        images = list(map(file_to_image, [data]))
+        inputs = self.processor(images=images, return_tensors="jax", padding=True) # converts the images into model-acceptable inputs and applies padding
+        emb = self.model.get_image_features(**inputs)
         return { "embeddings": embeddings }
+
+def file_to_image(file):
+  image = Image.open(BytesIO(file)).convert("RGB")
+  return image