High performance self-hosted photo and video management solution.
TypeScript backup-toolfluttergoogle-photosgoogle-photos-alternativejavascriptmobile-appnestjsnodejsphoto-galleryphotosphotos-managementself-hostedsveltesveltekittypescriptvideos
feat: Add additional env variables for Machine Learning (#15326)
Tempestc5476a99b127
diff --git a/docs/docs/install/environment-variables.md b/docs/docs/install/environment-variables.mdindex 1f34b5c6d0..b4cb905a0c 100644--- a/docs/docs/install/environment-variables.md+++ b/docs/docs/install/environment-variables.md@@ -148,24 +148,26 @@ Redis (Sentinel) URL example JSON before encoding:## Machine Learning-| Variable | Description | Default | Containers |-| :-------------------------------------------------------- | :-------------------------------------------------------------------------------------------------- | :-----------------------------: | :--------------- |-| `MACHINE_LEARNING_MODEL_TTL` | Inactivity time (s) before a model is unloaded (disabled if \<= 0) | `300` | machine learning |-| `MACHINE_LEARNING_MODEL_TTL_POLL_S` | Interval (s) between checks for the model TTL (disabled if \<= 0) | `10` | machine learning |-| `MACHINE_LEARNING_CACHE_FOLDER` | Directory where models are downloaded | `/cache` | machine learning |-| `MACHINE_LEARNING_REQUEST_THREADS`<sup>\*1</sup> | Thread count of the request thread pool (disabled if \<= 0) | number of CPU cores | machine learning |-| `MACHINE_LEARNING_MODEL_INTER_OP_THREADS` | Number of parallel model operations | `1` | machine learning |-| `MACHINE_LEARNING_MODEL_INTRA_OP_THREADS` | Number of threads for each model operation | `2` | machine learning |-| `MACHINE_LEARNING_WORKERS`<sup>\*2</sup> | Number of worker processes to spawn | `1` | machine learning |-| `MACHINE_LEARNING_HTTP_KEEPALIVE_TIMEOUT_S`<sup>\*3</sup> | HTTP Keep-alive time in seconds | `2` | machine learning |-| `MACHINE_LEARNING_WORKER_TIMEOUT` | Maximum time (s) of unresponsiveness before a worker is killed | `120` (`300` if using OpenVINO) | machine learning |-| `MACHINE_LEARNING_PRELOAD__CLIP` | Name of a CLIP model to be preloaded and kept in cache | | machine learning |-| `MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION` | Name of a facial recognition model to be preloaded and kept in cache | | machine learning |-| `MACHINE_LEARNING_ANN` | Enable ARM-NN hardware acceleration if supported | `True` | machine learning |-| `MACHINE_LEARNING_ANN_FP16_TURBO` | Execute operations in FP16 precision: increasing speed, reducing precision (applies only to ARM-NN) | `False` | machine learning |-| `MACHINE_LEARNING_ANN_TUNING_LEVEL` | ARM-NN GPU tuning level (1: rapid, 2: normal, 3: exhaustive) | `2` | machine learning |-| `MACHINE_LEARNING_DEVICE_IDS`<sup>\*4</sup> | Device IDs to use in multi-GPU environments | `0` | machine learning |-| `MACHINE_LEARNING_MAX_BATCH_SIZE__FACIAL_RECOGNITION` | Set the maximum number of faces that will be processed at once by the facial recognition model | None (`1` if using OpenVINO) | machine learning |+| Variable | Description | Default | Containers |+| :---------------------------------------------------------- | :-------------------------------------------------------------------------------------------------- | :-----------------------------: | :--------------- |+| `MACHINE_LEARNING_MODEL_TTL` | Inactivity time (s) before a model is unloaded (disabled if \<= 0) | `300` | machine learning |+| `MACHINE_LEARNING_MODEL_TTL_POLL_S` | Interval (s) between checks for the model TTL (disabled if \<= 0) | `10` | machine learning |+| `MACHINE_LEARNING_CACHE_FOLDER` | Directory where models are downloaded | `/cache` | machine learning |+| `MACHINE_LEARNING_REQUEST_THREADS`<sup>\*1</sup> | Thread count of the request thread pool (disabled if \<= 0) | number of CPU cores | machine learning |+| `MACHINE_LEARNING_MODEL_INTER_OP_THREADS` | Number of parallel model operations | `1` | machine learning |+| `MACHINE_LEARNING_MODEL_INTRA_OP_THREADS` | Number of threads for each model operation | `2` | machine learning |+| `MACHINE_LEARNING_WORKERS`<sup>\*2</sup> | Number of worker processes to spawn | `1` | machine learning |+| `MACHINE_LEARNING_HTTP_KEEPALIVE_TIMEOUT_S`<sup>\*3</sup> | HTTP Keep-alive time in seconds | `2` | machine learning |+| `MACHINE_LEARNING_WORKER_TIMEOUT` | Maximum time (s) of unresponsiveness before a worker is killed | `120` (`300` if using OpenVINO) | machine learning |+| `MACHINE_LEARNING_PRELOAD__CLIP__TEXTUAL` | Name of the textual CLIP model to be preloaded and kept in cache | | machine learning |+| `MACHINE_LEARNING_PRELOAD__CLIP__VISUAL` | Name of the visual CLIP model to be preloaded and kept in cache | | machine learning |+| `MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION__RECOGNITION` | Name of the recognition portion of the facial recognition model to be preloaded and kept in cache | | machine learning |+| `MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION__DETECTION` | Name of the detection portion of the facial recognition model to be preloaded and kept in cache | | machine learning |+| `MACHINE_LEARNING_ANN` | Enable ARM-NN hardware acceleration if supported | `True` | machine learning |+| `MACHINE_LEARNING_ANN_FP16_TURBO` | Execute operations in FP16 precision: increasing speed, reducing precision (applies only to ARM-NN) | `False` | machine learning |+| `MACHINE_LEARNING_ANN_TUNING_LEVEL` | ARM-NN GPU tuning level (1: rapid, 2: normal, 3: exhaustive) | `2` | machine learning |+| `MACHINE_LEARNING_DEVICE_IDS`<sup>\*4</sup> | Device IDs to use in multi-GPU environments | `0` | machine learning |+| `MACHINE_LEARNING_MAX_BATCH_SIZE__FACIAL_RECOGNITION` | Set the maximum number of faces that will be processed at once by the facial recognition model | None (`1` if using OpenVINO) | machine learning |\*1: It is recommended to begin with this parameter when changing the concurrency levels of the machine learning service and then tune the other ones.diff --git a/machine-learning/app/config.py b/machine-learning/app/config.pyindex 92799ac692..fa7dc61d47 100644--- a/machine-learning/app/config.py+++ b/machine-learning/app/config.py@@ -14,9 +14,41 @@ from uvicorn import Serverfrom uvicorn.workers import UvicornWorker+class ClipSettings(BaseModel):+ textual: str | None = None+ visual: str | None = None+++class FacialRecognitionSettings(BaseModel):+ recognition: str | None = None+ detection: str | None = None++class PreloadModelData(BaseModel):- clip: str | None = None- facial_recognition: str | None = None+ clip: ClipSettings = ClipSettings()+ facial_recognition: FacialRecognitionSettings = FacialRecognitionSettings()++ clip_model_fallback: str | None = os.getenv("MACHINE_LEARNING_PRELOAD__CLIP", None)+ facial_recognition_model_fallback: str | None = os.getenv("MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION", None)++ def update_from_fallbacks(self) -> None:+ if self.clip_model_fallback:+ self.clip.textual = self.clip_model_fallback+ self.clip.visual = self.clip_model_fallback+ log.warning(+ "Deprecated env variable: MACHINE_LEARNING_PRELOAD__CLIP. "+ "Use MACHINE_LEARNING_PRELOAD__CLIP__TEXTUAL and "+ "MACHINE_LEARNING_PRELOAD__CLIP__VISUAL instead."+ )++ if self.facial_recognition_model_fallback:+ self.facial_recognition.recognition = self.facial_recognition_model_fallback+ self.facial_recognition.detection = self.facial_recognition_model_fallback+ log.warning(+ "Deprecated environment variable: MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION. "+ "Use MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION__RECOGNITION and "+ "MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION__DETECTION instead."+ )class MaxBatchSize(BaseModel):diff --git a/machine-learning/app/main.py b/machine-learning/app/main.pyindex 684001b875..ee8881b6c6 100644--- a/machine-learning/app/main.py+++ b/machine-learning/app/main.py@@ -76,18 +76,29 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]: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)++ if preload.clip.textual is not None:+ model = await model_cache.get(preload.clip.textual, ModelType.TEXTUAL, ModelTask.SEARCH)await load(model)- model = await model_cache.get(preload.clip, ModelType.VISUAL, ModelTask.SEARCH)+ if preload.clip.visual is not None:+ model = await model_cache.get(preload.clip.visual, 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)+ if preload.facial_recognition.detection is not None:+ model = await model_cache.get(+ preload.facial_recognition.detection,+ ModelType.DETECTION,+ ModelTask.FACIAL_RECOGNITION,+ )await load(model)- model = await model_cache.get(preload.facial_recognition, ModelType.RECOGNITION, ModelTask.FACIAL_RECOGNITION)+ if preload.facial_recognition.recognition is not None:+ model = await model_cache.get(+ preload.facial_recognition.recognition,+ ModelType.RECOGNITION,+ ModelTask.FACIAL_RECOGNITION,+ )await load(model)diff --git a/machine-learning/app/test_main.py b/machine-learning/app/test_main.pyindex e5cb63997c..5da3baded7 100644--- a/machine-learning/app/test_main.py+++ b/machine-learning/app/test_main.py@@ -700,11 +700,13 @@ class TestCache:await model_cache.get("test_model_name", ModelType.TEXTUAL, ModelTask.SEARCH)async def test_preloads_clip_models(self, monkeypatch: MonkeyPatch, mock_get_model: mock.Mock) -> None:- os.environ["MACHINE_LEARNING_PRELOAD__CLIP"] = "ViT-B-32__openai"+ os.environ["MACHINE_LEARNING_PRELOAD__CLIP__TEXTUAL"] = "ViT-B-32__openai"+ os.environ["MACHINE_LEARNING_PRELOAD__CLIP__VISUAL"] = "ViT-B-32__openai"settings = Settings()assert settings.preload is not None- assert settings.preload.clip == "ViT-B-32__openai"+ assert settings.preload.clip.textual == "ViT-B-32__openai"+ assert settings.preload.clip.visual == "ViT-B-32__openai"model_cache = ModelCache()monkeypatch.setattr("app.main.model_cache", model_cache)@@ -721,11 +723,13 @@ class TestCache:async def test_preloads_facial_recognition_models(self, monkeypatch: MonkeyPatch, mock_get_model: mock.Mock) -> None:- os.environ["MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION"] = "buffalo_s"+ os.environ["MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION__DETECTION"] = "buffalo_s"+ os.environ["MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION__RECOGNITION"] = "buffalo_s"settings = Settings()assert settings.preload is not None- assert settings.preload.facial_recognition == "buffalo_s"+ assert settings.preload.facial_recognition.detection == "buffalo_s"+ assert settings.preload.facial_recognition.recognition == "buffalo_s"model_cache = ModelCache()monkeypatch.setattr("app.main.model_cache", model_cache)@@ -740,13 +744,17 @@ class TestCache:)async def test_preloads_all_models(self, monkeypatch: MonkeyPatch, mock_get_model: mock.Mock) -> None:- os.environ["MACHINE_LEARNING_PRELOAD__CLIP"] = "ViT-B-32__openai"- os.environ["MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION"] = "buffalo_s"+ os.environ["MACHINE_LEARNING_PRELOAD__CLIP__TEXTUAL"] = "ViT-B-32__openai"+ os.environ["MACHINE_LEARNING_PRELOAD__CLIP__VISUAL"] = "ViT-B-32__openai"+ os.environ["MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION__RECOGNITION"] = "buffalo_s"+ os.environ["MACHINE_LEARNING_PRELOAD__FACIAL_RECOGNITION__DETECTION"] = "buffalo_s"settings = Settings()assert settings.preload is not None- assert settings.preload.clip == "ViT-B-32__openai"- assert settings.preload.facial_recognition == "buffalo_s"+ assert settings.preload.clip.visual == "ViT-B-32__openai"+ assert settings.preload.clip.textual == "ViT-B-32__openai"+ assert settings.preload.facial_recognition.recognition == "buffalo_s"+ assert settings.preload.facial_recognition.detection == "buffalo_s"model_cache = ModelCache()monkeypatch.setattr("app.main.model_cache", model_cache)