Update handler.py

handler.py

@@ -1,12 +1,17 @@
 from typing import Dict, List, Any
-from transformers import pipeline
+from transformers import pipeline, BitsAndBytesConfig
 from PIL import Image    
 import requests
+import torch
 
 class EndpointHandler():
     def __init__(self, path=""):
-        print("init")
-        self.pipeline = pipeline("image-to-text", model=path)
+        quantization_config = BitsAndBytesConfig(
+        load_in_4bit=True,
+        bnb_4bit_compute_dtype=torch.float16
+        )
+        pipe = pipeline("image-to-text", path="", model_kwargs={"quantization_config": quantization_config})
+
 
     def __call__(self, data: Dict[str, Any]):
         totalarr = []