File size: 1,178 Bytes
4b5a3ab
 
 
 
 
 
 
 
 
 
 
 
 
b287914
 
4b5a3ab
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
from typing import Dict, List, Any
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
import torch

class EndpointHandler:
    def __init__(self, path=""):
        # load model and processor from path
        self.model =  AutoModelForSeq2SeqLM.from_pretrained(path, device_map="auto", load_in_8bit=True)
        self.tokenizer = AutoTokenizer.from_pretrained(path)

    def __call__(self, data: Dict[str, Any]) -> Dict[str, str]:
        """
        Args:
            data (:dict:):
                The payload with the text prompt and generation parameters.
        """
        # process input
        inputs = data.pop("inputs", data)
        parameters = data.pop("parameters", None)

        # preprocess
        input_ids = self.tokenizer(inputs, return_tensors="pt").input_ids

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

        # postprocess the prediction
        prediction = self.tokenizer.decode(outputs[0], skip_special_tokens=True)

        return [{"generated_text": prediction}]