Create handler.py

handler.py

@@ -0,0 +1,20 @@
+import torch.nn.functional as F
+from torch import Tensor
+from transformers import AutoTokenizer, AutoModel, AutoProcessor
+from torch import cuda
+
+class EndpointHandler():
+    def __init__(self, path=""):
+        self.processor = AutoProcessor.from_pretrained(path)
+        self.model = AutoModel.from_pretrained(path, trust_remote_code=True)
+        self.device = "cuda" if cuda.is_available() else "cpu"
+        self.model.to(self.device)
+
+    def __call__(self, data: Dict[str, Any]) -> List[List[int]]:
+        image  = data.pop("inputs",data)
+
+        processed = self.processor(images=image, return_tensors="pt").to(self.device)
+
+        prediction = self.model(processed["pixel_values"])
+        
+        return prediction.item()