GeezerVisonThe World Wide Web, restored to 800 pixels of glory. |
| Home|About|Gaming cabinet|Archives|How it works|Starter sites |
GeezerVison is showing github.com — llama-cpp-python/llama_cpp/server/app.py at main · CISC/llama-cpp-python Expand file tree/ Copy pathapp.pyMore file actions 612 lines (534 loc) · 19.9 KB / Copy pathapp.pyFile metadata and controls612 lines (534 loc) · 19.9 KB 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 from __future__ import annotations import os import json import typing import contextlib from threading import Lock from functools import partial from typing import Iterator, List, Optional, Union, Dict import llama_cpp import anyio from anyio.streams.memory import MemoryObjectSendStream from starlette.concurrency import run_in_threadpool, iterate_in_threadpool from fastapi import Depends, FastAPI, APIRouter, Request, HTTPException, status, Body from fastapi.middleware import Middleware from fastapi.middleware.cors import CORSMiddleware from fastapi.security import HTTPBearer from sse_starlette.sse import EventSourceResponse from starlette_context.plugins import RequestIdPlugin # type: ignore from starlette_context.middleware import RawContextMiddleware from llama_cpp.server.model import ( LlamaProxy, ) from llama_cpp.server.settings import ( ConfigFileSettings, Settings, ModelSettings, ServerSettings, ) from llama_cpp.server.types import ( CreateCompletionRequest, CreateEmbeddingRequest, CreateChatCompletionRequest, ModelList, TokenizeInputRequest, TokenizeInputResponse, TokenizeInputCountResponse, DetokenizeInputRequest, DetokenizeInputResponse, ) from llama_cpp.server.errors import RouteErrorHandler router = APIRouter(route_class=RouteErrorHandler) _server_settings: Optional[ServerSettings] = None def set_server_settings(server_settings: ServerSettings): global _server_settings _server_settings = server_settings def get_server_settings(): yield _server_settings _llama_proxy: Optional[LlamaProxy] = None llama_outer_lock = Lock() llama_inner_lock = Lock() def set_llama_proxy(model_settings: List[ModelSettings]): global _llama_proxy _llama_proxy = LlamaProxy(models=model_settings) def get_llama_proxy(): # NOTE: This double lock allows the currently streaming llama model to # check if any other requests are pending in the same thread and cancel # the stream if so. llama_outer_lock.acquire() release_outer_lock = True try: llama_inner_lock.acquire() try: llama_outer_lock.release() release_outer_lock = False yield _llama_proxy finally: llama_inner_lock.release() finally: if release_outer_lock: llama_outer_lock.release() _ping_message_factory: typing.Optional[typing.Callable[[], bytes]] = None def set_ping_message_factory(factory: typing.Callable[[], bytes]): global _ping_message_factory _ping_message_factory = factory def create_app( settings: Settings | None = None, server_settings: ServerSettings | None = None, model_settings: List[ModelSettings] | None = None, ): config_file = os.environ.get("CONFIG_FILE", None) if config_file is not None: if not os.path.exists(config_file): raise ValueError(f"Config file {config_file} not found!") with open(config_file, "rb") as f: # Check if yaml file if config_file.endswith(".yaml") or config_file.endswith(".yml"): import yaml config_file_settings = ConfigFileSettings.model_validate_json( json.dumps(yaml.safe_load(f)) ) else: config_file_settings = ConfigFileSettings.model_validate_json(f.read()) server_settings = ServerSettings.model_validate(config_file_settings) model_settings = config_file_settings.models if server_settings is None and model_settings is None: if settings is None: settings = Settings() server_settings = ServerSettings.model_validate(settings) model_settings = [ModelSettings.model_validate(settings)] assert ( server_settings is not None and model_settings is not None ), "server_settings and model_settings must be provided together" set_server_settings(server_settings) middleware = [Middleware(RawContextMiddleware, plugins=(RequestIdPlugin(),))] app = FastAPI( middleware=middleware, title="🦙 llama.cpp Python API", version=llama_cpp.__version__, root_path=server_settings.root_path, ) app.add_middleware( CORSMiddleware, allow_origins=["*"], allow_credentials=True, allow_methods=["*"], allow_headers=["*"], ) app.include_router(router) assert model_settings is not None set_llama_proxy(model_settings=model_settings) if server_settings.disable_ping_events: set_ping_message_factory(lambda: bytes()) return app async def get_event_publisher( request: Request, inner_send_chan: MemoryObjectSendStream[typing.Any], iterator: Iterator[typing.Any], on_complete: typing.Optional[typing.Callable[[], None]] = None, ): server_settings = next(get_server_settings()) interrupt_requests = ( server_settings.interrupt_requests if server_settings else False ) async with inner_send_chan: try: async for chunk in iterate_in_threadpool(iterator): await inner_send_chan.send(dict(data=json.dumps(chunk))) if await request.is_disconnected(): raise anyio.get_cancelled_exc_class()() if interrupt_requests and llama_outer_lock.locked(): await inner_send_chan.send(dict(data="[DONE]")) raise anyio.get_cancelled_exc_class()() await inner_send_chan.send(dict(data="[DONE]")) except anyio.get_cancelled_exc_class() as e: print("disconnected") with anyio.move_on_after(1, shield=True): print(f"Disconnected from client (via refresh/close) {request.client}") raise e finally: if on_complete: on_complete() def _logit_bias_tokens_to_input_ids( llama: llama_cpp.Llama, logit_bias: Dict[str, float], ) -> Dict[str, float]: to_bias: Dict[str, float] = {} for token, score in logit_bias.items(): token = token.encode("utf-8") for input_id in llama.tokenize(token, add_bos=False, special=True): to_bias[str(input_id)] = score return to_bias # Setup Bearer authentication scheme bearer_scheme = HTTPBearer(auto_error=False) async def authenticate( settings: Settings = Depends(get_server_settings), authorization: Optional[str] = Depends(bearer_scheme), ): # Skip API key check if it's not set in settings if settings.api_key is None: return True # check bearer credentials against the api_key if authorization and authorization.credentials == settings.api_key: # api key is valid return authorization.credentials # raise http error 401 raise HTTPException( status_code=status.HTTP_401_UNAUTHORIZED, detail="Invalid API key", ) openai_v1_tag = "OpenAI V1" @router.post( "/v1/completions", summary="Completion", dependencies=[Depends(authenticate)], response_model=Union[ llama_cpp.CreateCompletionResponse, str, ], responses={ "200": { "description": "Successful Response", "content": { "application/json": { "schema": { "anyOf": [ {"$ref": "#/components/schemas/CreateCompletionResponse"} ], "title": "Completion response, when stream=False", } }, "text/event-stream": { "schema": { "type": "string", "title": "Server Side Streaming response, when stream=True. " + "See SSE format: https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events#Event_stream_format", # noqa: E501 "example": """data: {... see CreateCompletionResponse ...} \\n\\n data: ... \\n\\n ... data: [DONE]""", } }, }, } }, tags=[openai_v1_tag], ) @router.post( "/v1/engines/copilot-codex/completions", include_in_schema=False, dependencies=[Depends(authenticate)], tags=[openai_v1_tag], ) async def create_completion( request: Request, body: CreateCompletionRequest, ) -> llama_cpp.Completion: exit_stack = contextlib.ExitStack() llama_proxy = await run_in_threadpool( lambda: exit_stack.enter_context(contextlib.contextmanager(get_llama_proxy)()) ) if llama_proxy is None: raise HTTPException( status_code=status.HTTP_503_SERVICE_UNAVAILABLE, detail="Service is not available", ) if isinstance(body.prompt, list): assert len(body.prompt) <= 1 body.prompt = body.prompt[0] if len(body.prompt) > 0 else "" llama = llama_proxy( body.model if request.url.path != "/v1/engines/copilot-codex/completions" else "copilot-codex" ) exclude = { "n", "best_of", "logit_bias_type", "user", "min_tokens", } kwargs = body.model_dump(exclude=exclude) if body.logit_bias is not None: kwargs["logit_bias"] = ( _logit_bias_tokens_to_input_ids(llama, body.logit_bias) if body.logit_bias_type == "tokens" else body.logit_bias ) if body.grammar is not None: kwargs["grammar"] = llama_cpp.LlamaGrammar.from_string(body.grammar) if body.min_tokens > 0: _min_tokens_logits_processor = llama_cpp.LogitsProcessorList( [llama_cpp.MinTokensLogitsProcessor(body.min_tokens, llama.token_eos())] ) if "logits_processor" not in kwargs: kwargs["logits_processor"] = _min_tokens_logits_processor else: kwargs["logits_processor"].extend(_min_tokens_logits_processor) iterator_or_completion: Union[ llama_cpp.CreateCompletionResponse, Iterator[llama_cpp.CreateCompletionStreamResponse], ] = await run_in_threadpool(llama, **kwargs) if isinstance(iterator_or_completion, Iterator): # EAFP: It's easier to ask for forgiveness than permission first_response = await run_in_threadpool(next, iterator_or_completion) # If no exception was raised from first_response, we can assume that # the iterator is valid and we can use it to stream the response. def iterator() -> Iterator[llama_cpp.CreateCompletionStreamResponse]: yield first_response yield from iterator_or_completion exit_stack.close() send_chan, recv_chan = anyio.create_memory_object_stream(10) return EventSourceResponse( recv_chan, data_sender_callable=partial( # type: ignore get_event_publisher, request=request, inner_send_chan=send_chan, iterator=iterator(), on_complete=exit_stack.close, ), sep="\n", ping_message_factory=_ping_message_factory, ) else: return iterator_or_completion @router.post( "/v1/embeddings", summary="Embedding", dependencies=[Depends(authenticate)], tags=[openai_v1_tag], ) async def create_embedding( request: CreateEmbeddingRequest, llama_proxy: LlamaProxy = Depends(get_llama_proxy), ): return await run_in_threadpool( llama_proxy(request.model).create_embedding, **request.model_dump(exclude={"user"}), ) @router.post( "/v1/chat/completions", summary="Chat", dependencies=[Depends(authenticate)], response_model=Union[llama_cpp.ChatCompletion, str], responses={ "200": { "description": "Successful Response", "content": { "application/json": { "schema": { "anyOf": [ { "$ref": "#/components/schemas/CreateChatCompletionResponse" } ], "title": "Completion response, when stream=False", } }, "text/event-stream": { "schema": { "type": "string", "title": "Server Side Streaming response, when stream=True" + "See SSE format: https://developer.mozilla.org/en-US/docs/Web/API/Server-sent_events/Using_server-sent_events#Event_stream_format", # noqa: E501 "example": """data: {... see CreateChatCompletionResponse ...} \\n\\n data: ... \\n\\n ... data: [DONE]""", } }, }, } }, tags=[openai_v1_tag], ) async def create_chat_completion( request: Request, body: CreateChatCompletionRequest = Body( openapi_examples={ "normal": { "summary": "Chat Completion", "value": { "model": "gpt-3.5-turbo", "messages": [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "What is the capital of France?"}, ], }, }, "json_mode": { "summary": "JSON Mode", "value": { "model": "gpt-3.5-turbo", "messages": [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Who won the world series in 2020"}, ], "response_format": {"type": "json_object"}, }, }, "tool_calling": { "summary": "Tool Calling", "value": { "model": "gpt-3.5-turbo", "messages": [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Extract Jason is 30 years old."}, ], "tools": [ { "type": "function", "function": { "name": "User", "description": "User record", "parameters": { "type": "object", "properties": { "name": {"type": "string"}, "age": {"type": "number"}, }, "required": ["name", "age"], }, }, } ], "tool_choice": { "type": "function", "function": { "name": "User", }, }, }, }, "logprobs": { "summary": "Logprobs", "value": { "model": "gpt-3.5-turbo", "messages": [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "What is the capital of France?"}, ], "logprobs": True, "top_logprobs": 10, }, }, } ), ) -> llama_cpp.ChatCompletion: # This is a workaround for an issue in FastAPI dependencies # where the dependency is cleaned up before a StreamingResponse # is complete. # https://github.com/tiangolo/fastapi/issues/11143 exit_stack = contextlib.ExitStack() llama_proxy = await run_in_threadpool( lambda: exit_stack.enter_context(contextlib.contextmanager(get_llama_proxy)()) ) if llama_proxy is None: raise HTTPException( status_code=status.HTTP_503_SERVICE_UNAVAILABLE, detail="Service is not available", ) exclude = { "n", "logit_bias_type", "user", "min_tokens", } kwargs = body.model_dump(exclude=exclude) llama = llama_proxy(body.model) if body.logit_bias is not None: kwargs["logit_bias"] = ( _logit_bias_tokens_to_input_ids(llama, body.logit_bias) if body.logit_bias_type == "tokens" else body.logit_bias ) if body.grammar is not None: kwargs["grammar"] = llama_cpp.LlamaGrammar.from_string(body.grammar) if body.min_tokens > 0: _min_tokens_logits_processor = llama_cpp.LogitsProcessorList( [llama_cpp.MinTokensLogitsProcessor(body.min_tokens, llama.token_eos())] ) if "logits_processor" not in kwargs: kwargs["logits_processor"] = _min_tokens_logits_processor else: kwargs["logits_processor"].extend(_min_tokens_logits_processor) iterator_or_completion: Union[ llama_cpp.ChatCompletion, Iterator[llama_cpp.ChatCompletionChunk] ] = await run_in_threadpool(llama.create_chat_completion, **kwargs) if isinstance(iterator_or_completion, Iterator): # EAFP: It's easier to ask for forgiveness than permission first_response = await run_in_threadpool(next, iterator_or_completion) # If no exception was raised from first_response, we can assume that # the iterator is valid and we can use it to stream the response. def iterator() -> Iterator[llama_cpp.ChatCompletionChunk]: yield first_response yield from iterator_or_completion exit_stack.close() send_chan, recv_chan = anyio.create_memory_object_stream(10) return EventSourceResponse( recv_chan, data_sender_callable=partial( # type: ignore get_event_publisher, request=request, inner_send_chan=send_chan, iterator=iterator(), on_complete=exit_stack.close, ), sep="\n", ping_message_factory=_ping_message_factory, ) else: exit_stack.close() return iterator_or_completion @router.get( "/v1/models", summary="Models", dependencies=[Depends(authenticate)], tags=[openai_v1_tag], ) async def get_models( llama_proxy: LlamaProxy = Depends(get_llama_proxy), ) -> ModelList: return { "object": "list", "data": [ { "id": model_alias, "object": "model", "owned_by": "me", "permissions": [], } for model_alias in llama_proxy ], } extras_tag = "Extras" @router.post( "/extras/tokenize", summary="Tokenize", dependencies=[Depends(authenticate)], tags=[extras_tag], ) async def tokenize( body: TokenizeInputRequest, llama_proxy: LlamaProxy = Depends(get_llama_proxy), ) -> TokenizeInputResponse: tokens = llama_proxy(body.model).tokenize(body.input.encode("utf-8"), special=True) return TokenizeInputResponse(tokens=tokens) @router.post( "/extras/tokenize/count", summary="Tokenize Count", dependencies=[Depends(authenticate)], tags=[extras_tag], ) async def count_query_tokens( body: TokenizeInputRequest, llama_proxy: LlamaProxy = Depends(get_llama_proxy), ) -> TokenizeInputCountResponse: tokens = llama_proxy(body.model).tokenize(body.input.encode("utf-8"), special=True) return TokenizeInputCountResponse(count=len(tokens)) @router.post( "/extras/detokenize", summary="Detokenize", dependencies=[Depends(authenticate)], tags=[extras_tag], ) async def detokenize( body: DetokenizeInputRequest, llama_proxy: LlamaProxy = Depends(get_llama_proxy), ) -> DetokenizeInputResponse: text = llama_proxy(body.model).detokenize(body.tokens).decode("utf-8") return DetokenizeInputResponse(text=text) |
| You are visitor 01402520 — GeezerVison does not store the pages you look up. Best viewed with Netscape Navigator 4.7 or Internet Explorer 6 at 800×600. No cookies required. |