Create handler.py

handler.py

@@ -0,0 +1,41 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+class EndpointHandler:
+    def __init__(self, model_dir):
+        # Load tokenizer and model from the provided model directory
+        self.tokenizer = AutoTokenizer.from_pretrained(model_dir, trust_remote_code=True)
+
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
+        torch_dtype = torch.float16 if self.device == "cuda" else torch.float32
+
+        self.model = AutoModelForCausalLM.from_pretrained(
+            model_dir,
+            trust_remote_code=True,
+            torch_dtype=torch_dtype,
+            device_map="auto" if self.device == "cuda" else None
+        )
+        self.model.eval()
+
+    def __call__(self, data):
+        # Extract inputs and parameters from the request payload
+        inputs = data.get("inputs", "")
+        params = data.get("parameters", {})
+
+        # Tokenize input
+        input_ids = self.tokenizer.encode(inputs, return_tensors="pt").to(self.model.device)
+
+        # Generate response
+        with torch.no_grad():
+            outputs = self.model.generate(
+                input_ids,
+                max_new_tokens=params.get("max_new_tokens", 256),
+                temperature=params.get("temperature", 0.7),
+                top_p=params.get("top_p", 0.95),
+                top_k=params.get("top_k", 50),
+                do_sample=params.get("do_sample", True)
+            )
+
+        # Decode and return
+        response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
+        return {"generated_text": response}