Upload handler.py

handler.py

@@ -0,0 +1,70 @@
+from typing import Dict, Any, List
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+class EndpointHandler:
+    def __init__(self, path=""):
+        """Initialize the model and tokenizer.
+        
+        Args:
+            path (str): Path to the model directory. Defaults to empty string.
+        """
+        self.device = "cuda" if torch.cuda.is_available() else "cpu"
+        self.model = AutoModelForCausalLM.from_pretrained(
+            path or "merged",
+            torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+            device_map="auto"
+        )
+        self.tokenizer = AutoTokenizer.from_pretrained(path or "merged")
+
+    def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
+        """Handle inference requests.
+        
+        Args:
+            data (Dict[str, Any]): The input data containing:
+                - instruction (str): The instruction for the model
+                - input (str, optional): Additional input text
+                - max_new_tokens (int, optional): Maximum number of tokens to generate. Defaults to 512
+                - temperature (float, optional): Sampling temperature. Defaults to 0.7
+        
+        Returns:
+            Dict[str, Any]: The model's response containing:
+                - response (str): The generated text
+        """
+        # Extract parameters from the request
+        instruction = data.get("instruction", "")
+        input_text = data.get("input", "")
+        max_new_tokens = data.get("max_new_tokens", 512)
+        temperature = data.get("temperature", 0.7)
+
+        # Create prompt
+        prompt = f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.
+
+### Instruction:
+{instruction}"""
+
+        if input_text:
+            prompt += f"""
+
+### Input:
+{input_text}"""
+
+        prompt += """
+
+### Response:"""
+
+        # Generate response
+        inputs = self.tokenizer(prompt, return_tensors="pt").to(self.model.device)
+        outputs = self.model.generate(
+            **inputs,
+            max_new_tokens=max_new_tokens,
+            temperature=temperature,
+            do_sample=True,
+            pad_token_id=self.tokenizer.eos_token_id
+        )
+        
+        # Decode and extract response
+        full_response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
+        response = full_response.split("### Response:")[-1].strip()
+
+        return {"response": response}