File size: 1,806 Bytes
65520f9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
from typing import Any, Dict

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig

# from peft import PeftConfig, PeftModel


class EndpointHandler:
    def __init__(self, path=""):
        # load model and processor from path
        self.tokenizer = AutoTokenizer.from_pretrained(path)
        # try:
        # config = AutoConfig.from_pretrained(path)
        model = AutoModelForCausalLM.from_pretrained(
            path,
            # return_dict=True,
            # load_in_8bit=True,
            device_map="auto",
            torch_dtype=torch.float16,
            # trust_remote_code=True,
        )
        # model.resize_token_embeddings(len(self.tokenizer))
        # model = PeftModel.from_pretrained(model, path)
        # except Exception:
        #     model = AutoModelForCausalLM.from_pretrained(
        #         path, device_map="auto", load_in_8bit=True, torch_dtype=torch.float16, trust_remote_code=True
        #     )
        self.model = model
        self.device = "cuda" if torch.cuda.is_available() else "cpu"

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

        # preprocess
        inputs = self.tokenizer(f"User: {inputs}\n\n", return_tensors="pt")

        # pass inputs with all kwargs in data
        if parameters is not None:
            outputs = self.model.generate(**inputs.to(self.device), max_new_tokens=880, **parameters)
        else:
            outputs = self.model.generate(**inputs.to(self.device), max_new_tokens=880)

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

        return [{"generated_text": prediction}]