public static IModelInferEngine Create(
InferenceBackend backend
)
The created engine is in unloaded state - LoadModel must be called before inference. 创建的引擎处于未加载状态 - 在推理前必须调用LoadModel。
Implementations should optimize backend-specific initialization but defer heavy operations until LoadModel is called. 实现应优化特定后端的初始化,但推迟繁重操作直到调用LoadModel。
NotSupportedException | Thrown when requesting an unsupported backend type. 当请求不受支持的后端类型时抛出。 |
InvalidOperationException | Thrown when engine initialization fails. 当引擎初始化失败时抛出。 |