added debugging

handler.py

@@ -1,5 +1,6 @@
 from io import BytesIO
 import base64
+import traceback
 
 from PIL import Image
 import torch
@@ -14,25 +15,29 @@ class EndpointHandler():
         self.processor = CLIPProcessor.from_pretrained("rbanfield/clip-vit-large-patch14")
 
     def __call__(self, data):
+        try:
 
-        text_input = None
-        if isinstance(data, dict):
-            inputs = data.pop("inputs", None)
-            text_input = inputs.get('text',None)
-            image_data = BytesIO(base64.b64decode(inputs['image'])) if 'image' in inputs else None
-        else:
-            # assuming its an image sent via binary
-            image_data = BytesIO(data)
+            text_input = None
+            if isinstance(data, dict):
+                print('data is a dict: ', data)
+                inputs = data.pop("inputs", None)
+                text_input = inputs.get('text',None)
+                image_data = BytesIO(base64.b64decode(inputs['image'])) if 'image' in inputs else None
+            else:
+                # assuming its an image sent via binary
+                image_data = BytesIO(data)
 
 
-        if text_input:
-            processor = self.processor(text=text_input, return_tensors="pt", padding=True).to(device)
-            with torch.no_grad():
-                return {'embeddings':self.text_model(**processor).pooler_output.tolist()[0]}
-        elif image_data:
-            image = Image.open(image_data)
-            processor = self.processor(images=image, return_tensors="pt").to(device)
-            with torch.no_grad():
-                return {'embeddings':self.image_model(**processor).image_embeds.tolist()[0]}
-        else:
-            return {'embeddings':None}
+            if text_input:
+                processor = self.processor(text=text_input, return_tensors="pt", padding=True).to(device)
+                with torch.no_grad():
+                    return {'embeddings':self.text_model(**processor).pooler_output.tolist()[0]}
+            elif image_data:
+                image = Image.open(image_data)
+                processor = self.processor(images=image, return_tensors="pt").to(device)
+                with torch.no_grad():
+                    return {'embeddings':self.image_model(**processor).image_embeds.tolist()[0]}
+            else:
+                return {'embeddings':None}
+        except Exception:
+            return {'Error':traceback.format_exc()}