1
0
mirror of https://github.com/immich-app/immich.git synced 2024-11-27 09:21:05 +02:00
immich/machine-learning/app/main.py

231 lines
7.8 KiB
Python
Raw Normal View History

import asyncio
import gc
import os
2023-12-14 21:51:24 +02:00
import signal
import threading
import time
from concurrent.futures import ThreadPoolExecutor
from contextlib import asynccontextmanager
from functools import partial
feat(server): separate face clustering job (#5598) * separate facial clustering job * update api * fixed some tests * invert clustering * hdbscan * update api * remove commented code * wip dbscan * cleanup removed cluster endpoint remove commented code * fixes updated tests minor fixes and formatting fixed queuing refinements * scale search range based on library size * defer non-core faces * optimizations removed unused query option * assign faces individually for correctness fixed unit tests remove unused method * don't select face embedding update sql linting fixed ml typing * updated job mock * paginate people query * select face embeddings because typeorm * fix setting face detection concurrency * update sql formatting linting * simplify logic remove unused imports * more specific delete signature * more accurate typing for face stubs * add migration formatting * chore: better typing * don't select embedding by default remove unused import * updated sql * use normal try/catch * stricter concurrency typing and enforcement * update api * update job concurrency panel to show disabled queues formatting * check jobId in queueAll fix tests * remove outdated comment * better facial recognition icon * wording wording formatting * fixed tests * fix * formatting & sql * try to fix sql check * more detailed description * update sql * formatting * wording * update `minFaces` description --------- Co-authored-by: Jason Rasmussen <jrasm91@gmail.com> Co-authored-by: Alex Tran <alex.tran1502@gmail.com>
2024-01-18 07:08:48 +02:00
from typing import Any, AsyncGenerator, Callable, Iterator
from zipfile import BadZipFile
import orjson
from fastapi import Depends, FastAPI, File, Form, HTTPException
from fastapi.responses import ORJSONResponse, PlainTextResponse
from onnxruntime.capi.onnxruntime_pybind11_state import InvalidProtobuf, NoSuchFile
from PIL.Image import Image
from pydantic import ValidationError
from starlette.formparsers import MultiPartParser
from app.models import get_model_deps
from app.models.base import InferenceModel
from app.models.transforms import decode_pil
from .config import PreloadModelData, log, settings
from .models.cache import ModelCache
from .schemas import (
InferenceEntries,
InferenceEntry,
InferenceResponse,
ModelFormat,
ModelIdentity,
ModelTask,
ModelType,
PipelineRequest,
T,
)
MultiPartParser.max_file_size = 2**26 # spools to disk if payload is 64 MiB or larger
2023-05-17 19:07:17 +02:00
model_cache = ModelCache(revalidate=settings.model_ttl > 0)
2023-12-14 21:51:24 +02:00
thread_pool: ThreadPoolExecutor | None = None
lock = threading.Lock()
active_requests = 0
last_called: float | None = None
2023-04-26 12:39:24 +02:00
2023-12-14 21:51:24 +02:00
@asynccontextmanager
async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
2023-12-14 21:51:24 +02:00
global thread_pool
log.info(
(
"Created in-memory cache with unloading "
f"{f'after {settings.model_ttl}s of inactivity' if settings.model_ttl > 0 else 'disabled'}."
)
)
try:
if settings.request_threads > 0:
# asyncio is a huge bottleneck for performance, so we use a thread pool to run blocking code
thread_pool = ThreadPoolExecutor(settings.request_threads) if settings.request_threads > 0 else None
log.info(f"Initialized request thread pool with {settings.request_threads} threads.")
if settings.model_ttl > 0 and settings.model_ttl_poll_s > 0:
asyncio.ensure_future(idle_shutdown_task())
if settings.preload is not None:
await preload_models(settings.preload)
yield
finally:
log.handlers.clear()
for model in model_cache.cache._cache.values():
del model
if thread_pool is not None:
thread_pool.shutdown()
gc.collect()
2023-12-14 21:51:24 +02:00
async def preload_models(preload: PreloadModelData) -> None:
log.info(f"Preloading models: {preload}")
if preload.clip is not None:
model = await model_cache.get(preload.clip, ModelType.TEXTUAL, ModelTask.SEARCH)
await load(model)
model = await model_cache.get(preload.clip, ModelType.VISUAL, ModelTask.SEARCH)
await load(model)
if preload.facial_recognition is not None:
model = await model_cache.get(preload.facial_recognition, ModelType.DETECTION, ModelTask.FACIAL_RECOGNITION)
await load(model)
model = await model_cache.get(preload.facial_recognition, ModelType.RECOGNITION, ModelTask.FACIAL_RECOGNITION)
await load(model)
2023-12-14 21:51:24 +02:00
def update_state() -> Iterator[None]:
global active_requests, last_called
active_requests += 1
last_called = time.time()
try:
yield
finally:
active_requests -= 1
def get_entries(entries: str = Form()) -> InferenceEntries:
try:
request: PipelineRequest = orjson.loads(entries)
without_deps: list[InferenceEntry] = []
with_deps: list[InferenceEntry] = []
for task, types in request.items():
for type, entry in types.items():
parsed: InferenceEntry = {
"name": entry["modelName"],
"task": task,
"type": type,
"options": entry.get("options", {}),
}
dep = get_model_deps(parsed["name"], type, task)
(with_deps if dep else without_deps).append(parsed)
return without_deps, with_deps
except (orjson.JSONDecodeError, ValidationError, KeyError, AttributeError) as e:
log.error(f"Invalid request format: {e}")
raise HTTPException(422, "Invalid request format.")
app = FastAPI(lifespan=lifespan)
@app.get("/")
async def root() -> ORJSONResponse:
return ORJSONResponse({"message": "Immich ML"})
2023-04-26 12:39:24 +02:00
@app.get("/ping")
def ping() -> PlainTextResponse:
return PlainTextResponse("pong")
2023-12-14 21:51:24 +02:00
@app.post("/predict", dependencies=[Depends(update_state)])
async def predict(
entries: InferenceEntries = Depends(get_entries),
image: bytes | None = File(default=None),
text: str | None = Form(default=None),
) -> Any:
if image is not None:
inputs: Image | str = await run(lambda: decode_pil(image))
elif text is not None:
inputs = text
else:
raise HTTPException(400, "Either image or text must be provided")
response = await run_inference(inputs, entries)
return ORJSONResponse(response)
async def run_inference(payload: Image | str, entries: InferenceEntries) -> InferenceResponse:
outputs: dict[ModelIdentity, Any] = {}
response: InferenceResponse = {}
async def _run_inference(entry: InferenceEntry) -> None:
model = await model_cache.get(entry["name"], entry["type"], entry["task"], ttl=settings.model_ttl)
inputs = [payload]
for dep in model.depends:
try:
inputs.append(outputs[dep])
except KeyError:
message = f"Task {entry['task']} of type {entry['type']} depends on output of {dep}"
raise HTTPException(400, message)
model = await load(model)
output = await run(model.predict, *inputs, **entry["options"])
outputs[model.identity] = output
response[entry["task"]] = output
without_deps, with_deps = entries
await asyncio.gather(*[_run_inference(entry) for entry in without_deps])
if with_deps:
await asyncio.gather(*[_run_inference(entry) for entry in with_deps])
if isinstance(payload, Image):
response["imageHeight"], response["imageWidth"] = payload.height, payload.width
return response
async def run(func: Callable[..., T], *args: Any, **kwargs: Any) -> T:
2023-12-14 21:51:24 +02:00
if thread_pool is None:
return func(*args, **kwargs)
partial_func = partial(func, *args, **kwargs)
return await asyncio.get_running_loop().run_in_executor(thread_pool, partial_func)
async def load(model: InferenceModel) -> InferenceModel:
if model.loaded:
return model
def _load(model: InferenceModel) -> InferenceModel:
2024-06-20 20:13:18 +02:00
if model.load_attempts > 1:
raise HTTPException(500, f"Failed to load model '{model.model_name}'")
2023-12-14 21:51:24 +02:00
with lock:
try:
model.load()
except FileNotFoundError as e:
if model.model_format == ModelFormat.ONNX:
raise e
log.exception(e)
log.warning(
f"{model.model_format.upper()} is available, but model '{model.model_name}' does not support it."
)
model.model_format = ModelFormat.ONNX
model.load()
return model
try:
2024-06-20 20:13:18 +02:00
return await run(_load, model)
except (OSError, InvalidProtobuf, BadZipFile, NoSuchFile):
2024-06-20 20:13:18 +02:00
log.warning(f"Failed to load {model.model_type.replace('_', ' ')} model '{model.model_name}'. Clearing cache.")
model.clear_cache()
2024-06-20 20:13:18 +02:00
return await run(_load, model)
async def idle_shutdown_task() -> None:
while True:
log.debug("Checking for inactivity...")
2023-12-14 21:51:24 +02:00
if (
last_called is not None
and not active_requests
and not lock.locked()
and time.time() - last_called > settings.model_ttl
):
log.info("Shutting down due to inactivity.")
2023-12-14 21:51:24 +02:00
os.kill(os.getpid(), signal.SIGINT)
break
await asyncio.sleep(settings.model_ttl_poll_s)