model_cache.py

# 创建一个全局模型管理器来缓存已加载的模型, 避免重复创建
class ModelCache:
    def __init__(self):
        self._detector = None
        self._detector_device = None
        self._embedder = None
        self._embedder_device = None
        self._embedder_source = None
        self._databases = {}  # 缓存已加载的数据库

    def get_detector(self, device: str):
        if self._detector is None or self._detector_device != device:
            from modules.face_detection import load_detector
            self._detector = load_detector(device)
            self._detector_device = device
        return self._detector

    def get_embedder(self, device: str, pretrained_source: str):
        if (self._embedder is None or
            self._embedder_device != device or
            self._embedder_source != pretrained_source):
            from modules.feature_extraction import load_embedder
            self._embedder = load_embedder(device, pretrained_source)
            self._embedder_device = device
            self._embedder_source = pretrained_source
        return self._embedder

    def get_database(self, path: str):
        """获取缓存的数据库实例，如果不存在则加载"""
        if path not in self._databases:
            from modules.matcher import load_db
            self._databases[path] = load_db(path)
        return self._databases[path]

    def invalidate_database(self, path: str):
        """使指定路径的数据库缓存失效"""
        if path in self._databases:
            del self._databases[path]

    def clear_database_cache(self):
        """清空所有数据库缓存"""
        self._databases.clear()

model_cache = ModelCache()