File size: 2,563 Bytes
6e6f8a3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
from typing import List, Dict
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

class EndpointHandler:
    def __init__(self, path: str):
        # Load model and tokenizer
        self.tokenizer = AutoTokenizer.from_pretrained(path)
        self.model = AutoModelForCausalLM.from_pretrained(
            path,
            torch_dtype=torch.float32,  # Use float32 for CPU
            device_map="auto"
        )
        
        # Set up generation parameters
        self.default_params = {
            "max_length": 1000,
            "temperature": 0.7,
            "top_p": 0.7,
            "top_k": 50,
            "repetition_penalty": 1.0,
            "do_sample": True,
            "pad_token_id": self.tokenizer.pad_token_id,
            "eos_token_id": self.tokenizer.eos_token_id
        }

    def __call__(self, data: Dict):
        """
        Args:
            data: Dictionary with "inputs" and optional "parameters"
        Returns:
            Generated text
        """
        # Extract messages from input
        messages = data.get("inputs", {}).get("messages", [])
        if not messages:
            return {"error": "No messages provided"}

        # Format input text
        input_text = ""
        for msg in messages:
            role = msg.get("role", "")
            content = msg.get("content", "")
            input_text += f"{role}: {content}\n"

        # Get generation parameters
        params = {**self.default_params}
        if "parameters" in data:
            params.update(data["parameters"])

        # Tokenize input
        inputs = self.tokenizer(
            input_text,
            return_tensors="pt",
            padding=True,
            truncation=True,
            max_length=512
        )

        # Generate response
        with torch.no_grad():
            outputs = self.model.generate(
                inputs["input_ids"],
                attention_mask=inputs["attention_mask"],
                **params
            )

        # Decode response
        generated_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)

        return [{"generated_text": generated_text}]

    def preprocess(self, request):
        """
        Prepare request for inference
        """
        if request.content_type != "application/json":
            raise ValueError("Content type must be application/json")
        
        data = request.json
        return data

    def postprocess(self, data):
        """
        Post-process model output
        """
        return data