io_struct.py 2.99 KB
Newer Older
Lianmin Zheng's avatar
Lianmin Zheng committed
1
2
3
4
5
6
7
8
9
import uuid
from dataclasses import dataclass
from typing import Dict, List, Optional, Union

from sglang.srt.sampling_params import SamplingParams


@dataclass
class GenerateReqInput:
10
    # The input prompt
Lianmin Zheng's avatar
Lianmin Zheng committed
11
    text: Union[List[str], str]
12
    # The image input
Lianmin Zheng's avatar
Lianmin Zheng committed
13
    image_data: Optional[Union[List[str], str]] = None
14
    # The sampling_params
Lianmin Zheng's avatar
Lianmin Zheng committed
15
    sampling_params: Union[List[Dict], Dict] = None
16
    # The request id
Lianmin Zheng's avatar
Lianmin Zheng committed
17
    rid: Optional[Union[List[str], str]] = None
18
    # Whether return logprobs of the prompts
19
    return_logprob: Optional[Union[List[bool], bool]] = None
20
    # The start location of the prompt for return_logprob
21
    logprob_start_len: Optional[Union[List[int], int]] = None
22
    # Whether to stream output
Lianmin Zheng's avatar
Lianmin Zheng committed
23
24
25
26
27
28
29
30
31
32
    stream: bool = False

    def post_init(self):
        is_single = isinstance(self.text, str)

        if is_single:
            if self.sampling_params is None:
                self.sampling_params = {}
            if self.rid is None:
                self.rid = uuid.uuid4().hex
33
34
35
36
            if self.return_logprob is None:
                self.return_logprob = False
            if self.logprob_start_len is None:
                self.logprob_start_len = 0
Lianmin Zheng's avatar
Lianmin Zheng committed
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
        else:
            num = len(self.text)

            if self.image_data is None:
                self.image_data = [None] * num
            elif not isinstance(self.image_data, list):
                self.image_data = [self.image_data] * num

            if self.sampling_params is None:
                self.sampling_params = [{}] * num
            elif not isinstance(self.sampling_params, list):
                self.sampling_params = [self.sampling_params] * num

            if self.rid is None:
                self.rid = [uuid.uuid4().hex for _ in range(num)]
            else:
                assert isinstance(self.rid, list)

55
56
57
58
            if self.return_logprob is None:
                self.return_logprob = [False] * num
            elif not isinstance(self.return_logprob, list):
                self.return_logprob = [self.return_logprob] * num
Lianmin Zheng's avatar
Lianmin Zheng committed
59

60
61
62
63
            if self.logprob_start_len is None:
                self.logprob_start_len = [0] * num
            elif not isinstance(self.logprob_start_len, list):
                self.logprob_start_len = [self.logprob_start_len] * num
Lianmin Zheng's avatar
Lianmin Zheng committed
64
65
66
67
68


@dataclass
class TokenizedGenerateReqInput:
    rid: str
Liangsheng Yin's avatar
Liangsheng Yin committed
69
    input_text: str
Lianmin Zheng's avatar
Lianmin Zheng committed
70
71
72
    input_ids: List[int]
    pixel_values: List[float]
    image_hash: int
shiyi.c_98's avatar
shiyi.c_98 committed
73
    image_size: List[int]
Lianmin Zheng's avatar
Lianmin Zheng committed
74
    sampling_params: SamplingParams
75
76
    return_logprob: bool
    logprob_start_len: int
Lianmin Zheng's avatar
Lianmin Zheng committed
77
78
79
80
81
82
83
    stream: bool


@dataclass
class BatchTokenIDOut:
    rids: List[str]
    output_tokens: List[List[int]]
Liangsheng Yin's avatar
Liangsheng Yin committed
84
    output_and_jump_forward_strs: List[str]
Lianmin Zheng's avatar
Lianmin Zheng committed
85
86
87
88
89
90
91
92
93
94
95
96
    hit_stop_str: List[Optional[str]]
    skip_special_tokens: List[bool]
    meta_info: List[Dict]
    finished: List[bool]


@dataclass
class BatchStrOut:
    rids: List[str]
    output_str: List[str]
    meta_info: List[Dict]
    finished: List[bool]
Liangsheng Yin's avatar
Liangsheng Yin committed
97
98
99
100
101


@dataclass
class FlushCacheReq:
    pass
Cody Yu's avatar
Cody Yu committed
102
103
104
105

@dataclass
class DetokenizeReqInput:
    input_ids: List[int]