protocol.py 24.8 KB
Newer Older
1
2
# Adapted from
# https://github.com/lm-sys/FastChat/blob/168ccc29d3f7edc50823016105c024fe2282732a/fastchat/protocol/openai_api_protocol.py
Zhuohan Li's avatar
Zhuohan Li committed
3
import time
4
from typing import Any, Dict, List, Literal, Optional, Union
Zhuohan Li's avatar
Zhuohan Li committed
5

6
import torch
7
from pydantic import BaseModel, ConfigDict, Field, model_validator
8
from transformers import PreTrainedTokenizer
9
from typing_extensions import Annotated
Zhuohan Li's avatar
Zhuohan Li committed
10

11
from vllm.entrypoints.chat_utils import ChatCompletionMessageParam
12
from vllm.entrypoints.openai.logits_processors import get_logits_processors
13
from vllm.pooling_params import PoolingParams
14
from vllm.sampling_params import LogitsProcessor, SamplingParams
15
from vllm.utils import random_uuid
16

Zhuohan Li's avatar
Zhuohan Li committed
17

18
19
20
21
22
23
class OpenAIBaseModel(BaseModel):
    # OpenAI API does not allow extra fields
    model_config = ConfigDict(extra="forbid")


class ErrorResponse(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
24
25
26
27
    object: str = "error"
    message: str
    type: str
    param: Optional[str] = None
28
    code: int
Zhuohan Li's avatar
Zhuohan Li committed
29
30


31
class ModelPermission(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
32
33
34
35
36
37
38
39
40
41
42
    id: str = Field(default_factory=lambda: f"modelperm-{random_uuid()}")
    object: str = "model_permission"
    created: int = Field(default_factory=lambda: int(time.time()))
    allow_create_engine: bool = False
    allow_sampling: bool = True
    allow_logprobs: bool = True
    allow_search_indices: bool = False
    allow_view: bool = True
    allow_fine_tuning: bool = False
    organization: str = "*"
    group: Optional[str] = None
43
    is_blocking: bool = False
Zhuohan Li's avatar
Zhuohan Li committed
44
45


46
class ModelCard(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
47
48
49
    id: str
    object: str = "model"
    created: int = Field(default_factory=lambda: int(time.time()))
Woosuk Kwon's avatar
Woosuk Kwon committed
50
    owned_by: str = "vllm"
Zhuohan Li's avatar
Zhuohan Li committed
51
52
    root: Optional[str] = None
    parent: Optional[str] = None
53
    max_model_len: Optional[int] = None
Zhuohan Li's avatar
Zhuohan Li committed
54
55
56
    permission: List[ModelPermission] = Field(default_factory=list)


57
class ModelList(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
58
59
60
61
    object: str = "list"
    data: List[ModelCard] = Field(default_factory=list)


62
class UsageInfo(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
63
64
65
66
67
    prompt_tokens: int = 0
    total_tokens: int = 0
    completion_tokens: Optional[int] = 0


68
class ResponseFormat(OpenAIBaseModel):
69
    # type must be "json_object" or "text"
70
    type: Literal["text", "json_object"]
71
72


73
class StreamOptions(OpenAIBaseModel):
74
75
    include_usage: Optional[bool] = True
    continuous_usage_stats: Optional[bool] = True
76
77


78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
class FunctionDefinition(OpenAIBaseModel):
    name: str
    description: Optional[str] = None
    parameters: Optional[Dict[str, Any]] = None


class ChatCompletionToolsParam(OpenAIBaseModel):
    type: Literal["function"] = "function"
    function: FunctionDefinition


class ChatCompletionNamedFunction(OpenAIBaseModel):
    name: str


class ChatCompletionNamedToolChoiceParam(OpenAIBaseModel):
    function: ChatCompletionNamedFunction
    type: Literal["function"] = "function"


98
class ChatCompletionRequest(OpenAIBaseModel):
99
100
    # Ordered by official OpenAI API documentation
    # https://platform.openai.com/docs/api-reference/chat/create
101
    messages: List[ChatCompletionMessageParam]
102
103
104
105
    model: str
    frequency_penalty: Optional[float] = 0.0
    logit_bias: Optional[Dict[str, float]] = None
    logprobs: Optional[bool] = False
106
    top_logprobs: Optional[int] = 0
107
    max_tokens: Optional[int] = None
108
109
110
    n: Optional[int] = 1
    presence_penalty: Optional[float] = 0.0
    response_format: Optional[ResponseFormat] = None
111
112
113
    seed: Optional[int] = Field(None,
                                ge=torch.iinfo(torch.long).min,
                                le=torch.iinfo(torch.long).max)
114
    stop: Optional[Union[str, List[str]]] = Field(default_factory=list)
Zhuohan Li's avatar
Zhuohan Li committed
115
    stream: Optional[bool] = False
116
    stream_options: Optional[StreamOptions] = None
117
118
    temperature: Optional[float] = 0.7
    top_p: Optional[float] = 1.0
119
120
121
    tools: Optional[List[ChatCompletionToolsParam]] = None
    tool_choice: Optional[Union[Literal["none"],
                                ChatCompletionNamedToolChoiceParam]] = "none"
Zhuohan Li's avatar
Zhuohan Li committed
122
    user: Optional[str] = None
123
124

    # doc: begin-chat-completion-sampling-params
125
    best_of: Optional[int] = None
126
127
128
129
130
131
    use_beam_search: bool = False
    top_k: int = -1
    min_p: float = 0.0
    repetition_penalty: float = 1.0
    length_penalty: float = 1.0
    early_stopping: bool = False
132
    stop_token_ids: Optional[List[int]] = Field(default_factory=list)
133
134
135
136
137
138
    include_stop_str_in_output: bool = False
    ignore_eos: bool = False
    min_tokens: int = 0
    skip_special_tokens: bool = True
    spaces_between_special_tokens: bool = True
    truncate_prompt_tokens: Optional[Annotated[int, Field(ge=1)]] = None
139
140
141
    # doc: end-chat-completion-sampling-params

    # doc: begin-chat-completion-extra-params
142
    echo: bool = Field(
143
144
145
146
147
        default=False,
        description=(
            "If true, the new message will be prepended with the last message "
            "if they belong to the same role."),
    )
148
    add_generation_prompt: bool = Field(
149
150
151
152
153
154
        default=True,
        description=
        ("If true, the generation prompt will be added to the chat template. "
         "This is a parameter used by chat template in tokenizer config of the "
         "model."),
    )
155
    add_special_tokens: bool = Field(
156
157
158
159
160
        default=False,
        description=(
            "If true, special tokens (e.g. BOS) will be added to the prompt "
            "on top of what is added by the chat template. "
            "For most models, the chat template takes care of adding the "
161
            "special tokens so this should be set to false (as is the "
162
163
            "default)."),
    )
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
    documents: Optional[List[Dict[str, str]]] = Field(
        default=None,
        description=
        ("A list of dicts representing documents that will be accessible to "
         "the model if it is performing RAG (retrieval-augmented generation)."
         " If the template does not support RAG, this argument will have no "
         "effect. We recommend that each document should be a dict containing "
         "\"title\" and \"text\" keys."),
    )
    chat_template: Optional[str] = Field(
        default=None,
        description=(
            "A Jinja template to use for this conversion. "
            "If this is not passed, the model's default chat template will be "
            "used instead."),
    )
    chat_template_kwargs: Optional[Dict[str, Any]] = Field(
        default=None,
        description=("Additional kwargs to pass to the template renderer. "
                     "Will be accessible by the chat template."),
    )
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
    guided_json: Optional[Union[str, dict, BaseModel]] = Field(
        default=None,
        description=("If specified, the output will follow the JSON schema."),
    )
    guided_regex: Optional[str] = Field(
        default=None,
        description=(
            "If specified, the output will follow the regex pattern."),
    )
    guided_choice: Optional[List[str]] = Field(
        default=None,
        description=(
            "If specified, the output will be exactly one of the choices."),
    )
    guided_grammar: Optional[str] = Field(
        default=None,
        description=(
            "If specified, the output will follow the context free grammar."),
    )
204
205
206
207
208
209
    guided_decoding_backend: Optional[str] = Field(
        default=None,
        description=(
            "If specified, will override the default guided decoding backend "
            "of the server for this specific request. If set, must be either "
            "'outlines' / 'lm-format-enforcer'"))
210
211
212
213
214
    guided_whitespace_pattern: Optional[str] = Field(
        default=None,
        description=(
            "If specified, will override the default whitespace pattern "
            "for guided json decoding."))
215
216

    # doc: end-chat-completion-extra-params
Zhuohan Li's avatar
Zhuohan Li committed
217

218
219
220
221
222
223
224
    def to_sampling_params(
            self, tokenizer: PreTrainedTokenizer,
            guided_decode_logits_processor: Optional[LogitsProcessor],
            default_max_tokens: int) -> SamplingParams:
        max_tokens = self.max_tokens
        if max_tokens is None:
            max_tokens = default_max_tokens
225

226
        # We now allow logprobs being true without top_logrobs.
227
228
229
230
231
        logits_processors = get_logits_processors(
            logit_bias=self.logit_bias,
            allowed_token_ids=None,
            tokenizer=tokenizer,
        )
232
233
        if guided_decode_logits_processor:
            logits_processors.append(guided_decode_logits_processor)
234

235
236
        return SamplingParams(
            n=self.n,
237
            best_of=self.best_of,
238
239
240
241
242
            presence_penalty=self.presence_penalty,
            frequency_penalty=self.frequency_penalty,
            repetition_penalty=self.repetition_penalty,
            temperature=self.temperature,
            top_p=self.top_p,
243
            top_k=self.top_k,
244
            min_p=self.min_p,
Nick Hill's avatar
Nick Hill committed
245
            seed=self.seed,
246
247
            stop=self.stop,
            stop_token_ids=self.stop_token_ids,
248
249
            logprobs=self.top_logprobs if self.logprobs else None,
            prompt_logprobs=self.top_logprobs if self.echo else None,
250
            ignore_eos=self.ignore_eos,
251
            max_tokens=max_tokens,
252
            min_tokens=self.min_tokens,
253
            use_beam_search=self.use_beam_search,
254
            early_stopping=self.early_stopping,
255
256
            skip_special_tokens=self.skip_special_tokens,
            spaces_between_special_tokens=self.spaces_between_special_tokens,
257
258
            include_stop_str_in_output=self.include_stop_str_in_output,
            length_penalty=self.length_penalty,
259
            logits_processors=logits_processors,
260
            truncate_prompt_tokens=self.truncate_prompt_tokens,
261
262
        )

263
264
265
266
267
268
269
270
271
    @model_validator(mode='before')
    @classmethod
    def validate_stream_options(cls, values):
        if (values.get('stream_options') is not None
                and not values.get('stream')):
            raise ValueError(
                "stream_options can only be set if stream is true")
        return values

272
273
274
275
276
277
278
279
    @model_validator(mode="before")
    @classmethod
    def check_guided_decoding_count(cls, data):
        guide_count = sum([
            "guided_json" in data and data["guided_json"] is not None,
            "guided_regex" in data and data["guided_regex"] is not None,
            "guided_choice" in data and data["guided_choice"] is not None
        ])
280
        # you can only use one kind of guided decoding
281
282
283
284
        if guide_count > 1:
            raise ValueError(
                "You can only use one kind of guided decoding "
                "('guided_json', 'guided_regex' or 'guided_choice').")
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
        # you can only either use guided decoding or tools, not both
        if guide_count > 1 and "tool_choice" in data and data[
                "tool_choice"] != "none":
            raise ValueError(
                "You can only either use guided decoding or tools, not both.")
        return data

    @model_validator(mode="before")
    @classmethod
    def check_tool_choice(cls, data):
        if "tool_choice" in data and data["tool_choice"] != "none":
            if not isinstance(data["tool_choice"], dict):
                raise ValueError("Currently only named tools are supported.")
            if "tools" not in data or data["tools"] is None:
                raise ValueError(
                    "When using `tool_choice`, `tools` must be set.")
301
302
        return data

303
304
305
306
307
308
309
310
    @model_validator(mode="before")
    @classmethod
    def check_logprobs(cls, data):
        if "top_logprobs" in data and data["top_logprobs"] is not None:
            if "logprobs" not in data or data["logprobs"] is False:
                raise ValueError(
                    "when using `top_logprobs`, `logprobs` must be set to true."
                )
311
            elif data["top_logprobs"] < 0:
312
                raise ValueError(
313
                    "`top_logprobs` must be a value a positive value.")
314
315
        return data

Zhuohan Li's avatar
Zhuohan Li committed
316

317
class CompletionRequest(OpenAIBaseModel):
318
319
    # Ordered by official OpenAI API documentation
    # https://platform.openai.com/docs/api-reference/completions/create
Zhuohan Li's avatar
Zhuohan Li committed
320
    model: str
321
    prompt: Union[List[int], List[List[int]], str, List[str]]
322
    best_of: Optional[int] = None
Zhuohan Li's avatar
Zhuohan Li committed
323
324
325
    echo: Optional[bool] = False
    frequency_penalty: Optional[float] = 0.0
    logit_bias: Optional[Dict[str, float]] = None
326
327
    logprobs: Optional[int] = None
    max_tokens: Optional[int] = 16
328
    n: int = 1
329
    presence_penalty: Optional[float] = 0.0
330
331
332
    seed: Optional[int] = Field(None,
                                ge=torch.iinfo(torch.long).min,
                                le=torch.iinfo(torch.long).max)
333
334
    stop: Optional[Union[str, List[str]]] = Field(default_factory=list)
    stream: Optional[bool] = False
335
    stream_options: Optional[StreamOptions] = None
336
337
338
    suffix: Optional[str] = None
    temperature: Optional[float] = 1.0
    top_p: Optional[float] = 1.0
Zhuohan Li's avatar
Zhuohan Li committed
339
    user: Optional[str] = None
340
341

    # doc: begin-completion-sampling-params
342
343
344
345
346
347
    use_beam_search: bool = False
    top_k: int = -1
    min_p: float = 0.0
    repetition_penalty: float = 1.0
    length_penalty: float = 1.0
    early_stopping: bool = False
348
    stop_token_ids: Optional[List[int]] = Field(default_factory=list)
349
350
351
352
353
    include_stop_str_in_output: bool = False
    ignore_eos: bool = False
    min_tokens: int = 0
    skip_special_tokens: bool = True
    spaces_between_special_tokens: bool = True
354
    truncate_prompt_tokens: Optional[Annotated[int, Field(ge=1)]] = None
355
    allowed_token_ids: Optional[List[int]] = None
356
357
358
    # doc: end-completion-sampling-params

    # doc: begin-completion-extra-params
359
360
    add_special_tokens: bool = Field(
        default=True,
361
        description=(
362
363
            "If true (the default), special tokens (e.g. BOS) will be added to "
            "the prompt."),
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
    )
    response_format: Optional[ResponseFormat] = Field(
        default=None,
        description=
        ("Similar to chat completion, this parameter specifies the format of "
         "output. Only {'type': 'json_object'} or {'type': 'text' } is "
         "supported."),
    )
    guided_json: Optional[Union[str, dict, BaseModel]] = Field(
        default=None,
        description=("If specified, the output will follow the JSON schema."),
    )
    guided_regex: Optional[str] = Field(
        default=None,
        description=(
            "If specified, the output will follow the regex pattern."),
    )
    guided_choice: Optional[List[str]] = Field(
        default=None,
        description=(
            "If specified, the output will be exactly one of the choices."),
    )
    guided_grammar: Optional[str] = Field(
        default=None,
        description=(
            "If specified, the output will follow the context free grammar."),
    )
391
392
393
394
395
396
    guided_decoding_backend: Optional[str] = Field(
        default=None,
        description=(
            "If specified, will override the default guided decoding backend "
            "of the server for this specific request. If set, must be one of "
            "'outlines' / 'lm-format-enforcer'"))
397
398
399
400
401
    guided_whitespace_pattern: Optional[str] = Field(
        default=None,
        description=(
            "If specified, will override the default whitespace pattern "
            "for guided json decoding."))
402
403

    # doc: end-completion-extra-params
Zhuohan Li's avatar
Zhuohan Li committed
404

405
406
407
408
409
410
411
412
    def to_sampling_params(
            self, tokenizer: PreTrainedTokenizer,
            guided_decode_logits_processor: Optional[LogitsProcessor],
            default_max_tokens: int) -> SamplingParams:
        max_tokens = self.max_tokens
        if max_tokens is None:
            max_tokens = default_max_tokens

413
414
        echo_without_generation = self.echo and self.max_tokens == 0

415
416
417
418
419
        logits_processors = get_logits_processors(
            logit_bias=self.logit_bias,
            allowed_token_ids=self.allowed_token_ids,
            tokenizer=tokenizer,
        )
420
421
        if guided_decode_logits_processor:
            logits_processors.append(guided_decode_logits_processor)
422

423
424
425
426
427
428
429
430
431
432
        return SamplingParams(
            n=self.n,
            best_of=self.best_of,
            presence_penalty=self.presence_penalty,
            frequency_penalty=self.frequency_penalty,
            repetition_penalty=self.repetition_penalty,
            temperature=self.temperature,
            top_p=self.top_p,
            top_k=self.top_k,
            min_p=self.min_p,
Nick Hill's avatar
Nick Hill committed
433
            seed=self.seed,
434
435
            stop=self.stop,
            stop_token_ids=self.stop_token_ids,
436
            logprobs=self.logprobs,
437
            ignore_eos=self.ignore_eos,
438
            max_tokens=max_tokens if not echo_without_generation else 1,
439
            min_tokens=self.min_tokens,
440
            use_beam_search=self.use_beam_search,
441
            early_stopping=self.early_stopping,
442
443
            prompt_logprobs=self.logprobs if self.echo else None,
            skip_special_tokens=self.skip_special_tokens,
444
            spaces_between_special_tokens=self.spaces_between_special_tokens,
445
446
            include_stop_str_in_output=self.include_stop_str_in_output,
            length_penalty=self.length_penalty,
447
            logits_processors=logits_processors,
448
            truncate_prompt_tokens=self.truncate_prompt_tokens,
449
450
        )

451
452
453
454
455
456
457
458
459
460
461
462
463
464
    @model_validator(mode="before")
    @classmethod
    def check_guided_decoding_count(cls, data):
        guide_count = sum([
            "guided_json" in data and data["guided_json"] is not None,
            "guided_regex" in data and data["guided_regex"] is not None,
            "guided_choice" in data and data["guided_choice"] is not None
        ])
        if guide_count > 1:
            raise ValueError(
                "You can only use one kind of guided decoding "
                "('guided_json', 'guided_regex' or 'guided_choice').")
        return data

465
466
467
468
    @model_validator(mode="before")
    @classmethod
    def check_logprobs(cls, data):
        if "logprobs" in data and data[
469
470
                "logprobs"] is not None and not data["logprobs"] >= 0:
            raise ValueError("if passed, `logprobs` must be a positive value.")
471
472
        return data

473
474
475
476
477
    @model_validator(mode="before")
    @classmethod
    def validate_stream_options(cls, data):
        if data.get("stream_options") and not data.get("stream"):
            raise ValueError(
478
                "Stream options can only be defined when stream is true.")
479
480
        return data

Zhuohan Li's avatar
Zhuohan Li committed
481

482
class EmbeddingRequest(OpenAIBaseModel):
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
    # Ordered by official OpenAI API documentation
    # https://platform.openai.com/docs/api-reference/embeddings
    model: str
    input: Union[List[int], List[List[int]], str, List[str]]
    encoding_format: Optional[str] = Field('float', pattern='^(float|base64)$')
    dimensions: Optional[int] = None
    user: Optional[str] = None

    # doc: begin-embedding-pooling-params
    additional_data: Optional[Any] = None

    # doc: end-embedding-pooling-params

    def to_pooling_params(self):
        return PoolingParams(additional_data=self.additional_data)


500
class CompletionLogProbs(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
501
502
503
    text_offset: List[int] = Field(default_factory=list)
    token_logprobs: List[Optional[float]] = Field(default_factory=list)
    tokens: List[str] = Field(default_factory=list)
504
505
    top_logprobs: List[Optional[Dict[str,
                                     float]]] = Field(default_factory=list)
Zhuohan Li's avatar
Zhuohan Li committed
506
507


508
class CompletionResponseChoice(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
509
510
    index: int
    text: str
511
    logprobs: Optional[CompletionLogProbs] = None
512
513
    finish_reason: Optional[str] = None
    stop_reason: Optional[Union[int, str]] = Field(
514
515
516
517
518
519
        default=None,
        description=(
            "The stop string or token id that caused the completion "
            "to stop, None if the completion finished for some other reason "
            "including encountering the EOS token"),
    )
Zhuohan Li's avatar
Zhuohan Li committed
520
521


522
class CompletionResponse(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
523
524
525
526
527
528
529
530
    id: str = Field(default_factory=lambda: f"cmpl-{random_uuid()}")
    object: str = "text_completion"
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    choices: List[CompletionResponseChoice]
    usage: UsageInfo


531
class CompletionResponseStreamChoice(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
532
533
    index: int
    text: str
534
    logprobs: Optional[CompletionLogProbs] = None
535
536
    finish_reason: Optional[str] = None
    stop_reason: Optional[Union[int, str]] = Field(
537
538
539
540
541
542
        default=None,
        description=(
            "The stop string or token id that caused the completion "
            "to stop, None if the completion finished for some other reason "
            "including encountering the EOS token"),
    )
Zhuohan Li's avatar
Zhuohan Li committed
543
544


545
class CompletionStreamResponse(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
546
547
548
549
550
    id: str = Field(default_factory=lambda: f"cmpl-{random_uuid()}")
    object: str = "text_completion"
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    choices: List[CompletionResponseStreamChoice]
551
    usage: Optional[UsageInfo] = Field(default=None)
552
553


554
class EmbeddingResponseData(OpenAIBaseModel):
555
556
    index: int
    object: str = "embedding"
557
    embedding: Union[List[float], str]
558
559


560
class EmbeddingResponse(OpenAIBaseModel):
561
562
563
564
565
566
567
568
    id: str = Field(default_factory=lambda: f"cmpl-{random_uuid()}")
    object: str = "list"
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    data: List[EmbeddingResponseData]
    usage: UsageInfo


569
570
571
572
573
574
575
576
577
578
579
class FunctionCall(OpenAIBaseModel):
    name: str
    arguments: str


class ToolCall(OpenAIBaseModel):
    id: str = Field(default_factory=lambda: f"chatcmpl-tool-{random_uuid()}")
    type: Literal["function"] = "function"
    function: FunctionCall


580
class ChatMessage(OpenAIBaseModel):
581
582
    role: str
    content: str
583
    tool_calls: List[ToolCall] = Field(default_factory=list)
584
585


586
587
588
589
590
591
592
593
594
595
596
597
598
599
class ChatCompletionLogProb(OpenAIBaseModel):
    token: str
    logprob: float = -9999.0
    bytes: Optional[List[int]] = None


class ChatCompletionLogProbsContent(ChatCompletionLogProb):
    top_logprobs: List[ChatCompletionLogProb] = Field(default_factory=list)


class ChatCompletionLogProbs(OpenAIBaseModel):
    content: Optional[List[ChatCompletionLogProbsContent]] = None


600
class ChatCompletionResponseChoice(OpenAIBaseModel):
601
602
    index: int
    message: ChatMessage
603
    logprobs: Optional[ChatCompletionLogProbs] = None
604
    finish_reason: Optional[str] = None
605
    stop_reason: Optional[Union[int, str]] = None
606
607


608
class ChatCompletionResponse(OpenAIBaseModel):
609
    id: str = Field(default_factory=lambda: f"chatcmpl-{random_uuid()}")
610
    object: Literal["chat.completion"] = "chat.completion"
611
612
613
614
615
616
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    choices: List[ChatCompletionResponseChoice]
    usage: UsageInfo


617
class DeltaMessage(OpenAIBaseModel):
618
619
    role: Optional[str] = None
    content: Optional[str] = None
620
    tool_calls: List[ToolCall] = Field(default_factory=list)
621
622


623
class ChatCompletionResponseStreamChoice(OpenAIBaseModel):
624
625
    index: int
    delta: DeltaMessage
626
    logprobs: Optional[ChatCompletionLogProbs] = None
627
    finish_reason: Optional[str] = None
628
    stop_reason: Optional[Union[int, str]] = None
629
630


631
class ChatCompletionStreamResponse(OpenAIBaseModel):
632
    id: str = Field(default_factory=lambda: f"chatcmpl-{random_uuid()}")
633
    object: Literal["chat.completion.chunk"] = "chat.completion.chunk"
634
635
636
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    choices: List[ChatCompletionResponseStreamChoice]
637
    usage: Optional[UsageInfo] = Field(default=None)
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658


class BatchRequestInput(OpenAIBaseModel):
    """
    The per-line object of the batch input file.

    NOTE: Currently only the `/v1/chat/completions` endpoint is supported.
    """

    # A developer-provided per-request id that will be used to match outputs to
    # inputs. Must be unique for each request in a batch.
    custom_id: str

    # The HTTP method to be used for the request. Currently only POST is
    # supported.
    method: str

    # The OpenAI API relative URL to be used for the request. Currently
    # /v1/chat/completions is supported.
    url: str

659
660
    # The parameters of the request.
    body: ChatCompletionRequest
661
662


663
664
665
666
667
668
669
670
class BatchResponseData(OpenAIBaseModel):
    # HTTP status code of the response.
    status_code: int = 200

    # An unique identifier for the API request.
    request_id: str

    # The body of the response.
671
    body: Optional[ChatCompletionResponse] = None
672
673


674
675
676
677
678
679
680
681
682
683
684
class BatchRequestOutput(OpenAIBaseModel):
    """
    The per-line object of the batch output and error files
    """

    id: str

    # A developer-provided per-request id that will be used to match outputs to
    # inputs.
    custom_id: str

685
    response: Optional[BatchResponseData]
686
687
688
689

    # For requests that failed with a non-HTTP error, this will contain more
    # information on the cause of the failure.
    error: Optional[Any]
690
691


692
693
694
695
696
697
698
699
700
701
702
class TokenizeCompletionRequest(OpenAIBaseModel):
    model: str
    prompt: str

    add_special_tokens: bool = Field(default=True)


class TokenizeChatRequest(OpenAIBaseModel):
    model: str
    messages: List[ChatCompletionMessageParam]

703
704
    add_generation_prompt: bool = Field(default=True)
    add_special_tokens: bool = Field(default=False)
705
706
707


TokenizeRequest = Union[TokenizeCompletionRequest, TokenizeChatRequest]
708
709
710
711
712


class TokenizeResponse(OpenAIBaseModel):
    count: int
    max_model_len: int
713
    tokens: List[int]
714
715
716
717
718
719
720
721
722


class DetokenizeRequest(OpenAIBaseModel):
    model: str
    tokens: List[int]


class DetokenizeResponse(OpenAIBaseModel):
    prompt: str