-
Notifications
You must be signed in to change notification settings - Fork 581
/
_schemas.py
205 lines (173 loc) · 6.83 KB
/
_schemas.py
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
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
from __future__ import annotations
import pydantic, inflection, orjson, typing as t
from ._configuration import LLMConfig
from .utils import gen_random_uuid
from ._typing_compat import Required, TypedDict, LiteralString
if t.TYPE_CHECKING:
import vllm
class MessageParam(TypedDict):
role: t.Union[t.Literal['system', 'user', 'assistant'], LiteralString]
content: str
class MessagesConverterInput(TypedDict):
add_generation_prompt: bool
messages: t.List[MessageParam]
class MetadataOutput(pydantic.BaseModel):
model_config = pydantic.ConfigDict(protected_namespaces=())
model_id: str
timeout: int
model_name: str
backend: str
configuration: str
class GenerationInputDict(TypedDict, total=False):
prompt: t.Optional[str]
prompt_token_ids: t.Optional[t.List[int]]
llm_config: Required[t.Dict[str, t.Any]]
stop: t.Optional[t.List[str]]
stop_token_ids: t.Optional[t.List[int]]
request_id: t.Optional[str]
adapter_name: t.Optional[str]
class GenerationInput(pydantic.BaseModel):
prompt: t.Optional[str] = pydantic.Field(default=None)
llm_config: LLMConfig = pydantic.Field(default_factory=dict)
prompt_token_ids: t.Optional[t.List[int]] = pydantic.Field(default=None)
stop: t.Optional[t.List[str]] = pydantic.Field(default=None)
stop_token_ids: t.Optional[t.List[int]] = pydantic.Field(default=None)
request_id: t.Optional[str] = pydantic.Field(default=None)
adapter_name: t.Optional[str] = pydantic.Field(default=None)
_class_ref: t.ClassVar[type[LLMConfig]] = pydantic.PrivateAttr()
@pydantic.field_validator('llm_config')
@classmethod
def llm_config_validator(cls, v: LLMConfig | dict[str, t.Any]) -> LLMConfig:
if isinstance(v, dict):
return cls._class_ref.model_construct_env(**v)
return v
@pydantic.field_validator('stop')
@classmethod
def stop_validator(cls, data: str | list[str] | t.Iterable[str] | None) -> list[str] | None:
if data is None:
return None
if isinstance(data, str):
return [data]
else:
return list(data)
@pydantic.model_serializer
def ser_model(self) -> dict[str, t.Any]:
flattened = self.llm_config.model_dump()
flattened.update({
'prompt': self.prompt,
'prompt_token_ids': self.prompt_token_ids,
'request_id': self.request_id,
'adapter_name': self.adapter_name,
})
if self.stop is not None:
flattened['stop'] = self.stop
if self.stop_token_ids is not None:
flattened['stop_token_ids'] = self.stop_token_ids
return flattened
def __init__(self, /, *, _internal: bool = False, **data: t.Any) -> None:
if not _internal:
raise RuntimeError(
f'Cannot instantiate GenerationInput directly. Use "{self.__class__.__qualname__}.from_dict" instead.'
)
super().__init__(**data)
@classmethod
def from_dict(cls, structured: GenerationInputDict) -> GenerationInput:
if not hasattr(cls, '_class_ref'):
raise ValueError(
'Cannot use "from_dict" from a raw GenerationInput class. Currently only supports class created from "from_config".'
)
filtered: dict[str, t.Any] = {k: v for k, v in structured.items() if v is not None}
llm_config: dict[str, t.Any] | None = filtered.pop('llm_config', None)
if llm_config is not None:
filtered['llm_config'] = cls._class_ref.model_construct_env(**llm_config)
return cls(_internal=True, **filtered)
@classmethod
def from_config(cls, llm_config: LLMConfig) -> type[GenerationInput]:
klass = pydantic.create_model(
inflection.camelize(llm_config['start_name']) + 'GenerationInput',
__base__=cls,
llm_config=(type(llm_config), llm_config),
_class_ref=(llm_config.__class__, pydantic.PrivateAttr(default=llm_config.__class__)),
)
return klass
# NOTE: parameters from vllm.RequestOutput and vllm.CompletionOutput since vllm is not available on CPU.
# OpenLLM will adapt CPU outputs to similar architecture with vLLM outputs for consistency
SampleLogprobs = t.List[t.Dict[int, float]]
PromptLogprobs = t.List[t.Optional[t.Dict[int, float]]]
FinishReason = t.Literal['length', 'stop']
class CompletionChunk(pydantic.BaseModel):
index: int
text: str
token_ids: t.List[int]
cumulative_logprob: float
logprobs: t.Optional[SampleLogprobs] = pydantic.Field(default=None)
finish_reason: t.Optional[FinishReason] = pydantic.Field(default=None)
class GenerationOutput(pydantic.BaseModel):
prompt: str
finished: bool
outputs: t.List[CompletionChunk]
prompt_token_ids: t.Optional[t.List[int]] = pydantic.Field(default=None)
prompt_logprobs: t.Optional[PromptLogprobs] = pydantic.Field(default=None)
request_id: str = pydantic.Field(default_factory=gen_random_uuid)
@staticmethod
def _preprocess_sse_message(data: str) -> str:
proc = [line[6:] for line in data.strip().split('\n') if line.startswith('data: ')]
if not proc:
return data
if len(proc) > 1:
raise ValueError('Multiple data found in SSE message.')
return proc[0]
@classmethod
def from_runner(cls, data: str) -> GenerationOutput:
data = cls._preprocess_sse_message(data)
if not data:
raise ValueError('No data found from messages.')
try:
structured = orjson.loads(data)
except orjson.JSONDecodeError as e:
raise ValueError(f'Failed to parse JSON from SSE message: {data!r}') from e
return cls.from_dict(structured)
@classmethod
def from_dict(cls, structured: dict[str, t.Any]) -> GenerationOutput:
if structured['prompt_logprobs']:
structured['prompt_logprobs'] = [
{int(k): v for k, v in it.items()} if it else None for it in structured['prompt_logprobs']
]
return cls(
prompt=structured['prompt'],
finished=structured['finished'],
prompt_token_ids=structured['prompt_token_ids'],
prompt_logprobs=structured['prompt_logprobs'],
request_id=structured['request_id'],
outputs=[
CompletionChunk(
index=it['index'],
text=it['text'],
token_ids=it['token_ids'],
cumulative_logprob=it['cumulative_logprob'],
finish_reason=it['finish_reason'],
logprobs=[{int(k): v for k, v in s.items()} for s in it['logprobs']] if it['logprobs'] else None,
)
for it in structured['outputs']
],
)
@classmethod
def from_vllm(cls, request_output: vllm.RequestOutput) -> GenerationOutput:
return cls(
prompt=request_output.prompt,
finished=request_output.finished,
request_id=request_output.request_id,
prompt_token_ids=request_output.prompt_token_ids,
prompt_logprobs=request_output.prompt_logprobs,
outputs=[
CompletionChunk(
index=it.index,
text=it.text,
token_ids=it.token_ids,
cumulative_logprob=it.cumulative_logprob,
logprobs=it.logprobs,
finish_reason=it.finish_reason,
)
for it in request_output.outputs
],
)