update handler

handler.py

@@ -1,33 +1,27 @@
-import torch
+from typing import Dict, List, Any
 from PIL import Image
-import io
-from transformers import CLIPProcessor, CLIPModel
+from io import BytesIO
+from transformers import pipeline
+import base64
 
-class EndpointHandler:
-    def __init__(self, model_dir):
-        # Initialize the model and processor with a model directory if necessary
-        self.model = CLIPModel.from_pretrained(model_dir)
-        self.processor = CLIPProcessor.from_pretrained(model_dir)
 
-    def preprocess(self, data):
-        # Preprocess the image and text
-        image_data = data['image']
-        text_data = data['text']
-        
-        # Convert image bytes to a PIL Image
-        image = Image.open(io.BytesIO(image_data))
-        
-        # Use the processor to prepare inputs
-        inputs = self.processor(text=[text_data], images=image, return_tensors="pt", padding=True)
-        return inputs
+class EndpointHandler():
+    def __init__(self, path=""):
+        self.pipeline=pipeline("zero-shot-image-classification",model=path)
+    
+    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
+        """
+       data args:
+            images (:obj:`string`)
+            candiates (:obj:`list`)
+      Return:
+            A :obj:`list`:. The list contains items that are dicts should be liked {"label": "XXX", "score": 0.82}
+        """
+        inputs = data.pop("inputs", data)
 
-    def inference(self, inputs):
-        # Run the model
-        outputs = self.model(**inputs)
-        return outputs
+        # decode base64 image to PIL
+        image = Image.open(BytesIO(base64.b64decode(inputs['image'])))
 
-    def postprocess(self, outputs):
-        # Extract and process the output logits
-        logits_per_image = outputs.logits_per_image  # this is the image-text similarity score
-        probs = logits_per_image.softmax(dim=1).tolist()[0]
-        return probs
+        # run prediction one image wit provided candiates
+        prediction = self.pipeline(images=[image], candidate_labels=inputs["candiates"])
+        return prediction[0]