handler.py

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

class EndpointHandler:
    def __init__(self, path=""):
        # Load model and tokenizer
        self.tokenizer = AutoTokenizer.from_pretrained(path)
        self.model = AutoModelForCausalLM.from_pretrained(
            path,
            torch_dtype=torch.float16,
            device_map="auto"
        )
        
    def __call__(self, data):
        # Parse the input data
        inputs = data.pop("inputs", data)
        parameters = data.pop("parameters", {})
        
        # Set default parameters if not provided
        max_new_tokens = parameters.get("max_new_tokens", 256)
        temperature = parameters.get("temperature", 0.7)
        do_sample = parameters.get("do_sample", True)
        
        # Tokenize and generate
        input_ids = self.tokenizer(inputs, return_tensors="pt").input_ids.to(self.model.device)
        
        with torch.no_grad():
            outputs = self.model.generate(
                input_ids,
                max_new_tokens=max_new_tokens,
                do_sample=do_sample,
                temperature=temperature,
                **parameters
            )
            
        response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
        
        return {"generated_text": response}