handler.py

from typing import Any, Dict

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

from peft import PeftConfig, PeftModel
from transformers import pipeline


class EndpointHandler:
    def __init__(self, path=""):
        # load model and processor from path
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
        config = PeftConfig.from_pretrained(path)
        model = AutoModelForCausalLM.from_pretrained(
            config.base_model_name_or_path,
            return_dict=True,
            torch_dtype=torch.float16,
            trust_remote_code=True,
        )
        self.tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path, trust_remote_code=True)
        model = PeftModel.from_pretrained(model, path)
        self.model = model
        self.model.to(torch.float16)
        self.model.to(self.device)
        self.model = self.model.merge_and_unload()
        self.model.eval()
        self.pipeline = pipeline('text-generation', 
                                model = self.model, 
                                tokenizer=self.tokenizer,
                                device=self.device,
                                torch_dtype=torch.float16)


    def __call__(self, data: Dict[str, Any]) -> Dict[str, str]:
        # process input
        inputs = data.pop("inputs", data)
        parameters = data.pop("parameters", None)

        # pass inputs with all kwargs in data
        if parameters is not None:
            outputs = self.pipeline(**inputs, **parameters)
        else:
            outputs = self.pipeline(**inputs)

        return outputs