引導式測驗 · 中等 等級

用於英特爾硬體推理測驗的 OpenVINO

遵循 OpenVINO 轉換和編譯工作流程,同時檢查支援的運算子、目標裝置、數值變更和任務準確性。

相關引導路徑Openvino 英特爾推理
問題 1 的 8

What role does OpenVINO IR play in an inference workflow?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. What role does OpenVINO IR play in an inference workflow?
    • A format for annotating training datasets
    • A driver package that replaces the runtime
    • A runtime-oriented representation of a converted model graph
    • A loss function used during model training
  2. 使用 OpenVINO 執行環境讀取模型後會發生什麼事?
    • Reading the model automatically retrains its weights
    • The model can be compiled for an available target device
    • Reading regenerates the model class labels
    • The model is automatically uploaded to a registry
  3. 為什麼轉換成功不足以證明部署就緒?
    • Conversion produces a new training dataset
    • Every conversion changes class labels
    • A successful conversion guarantees task accuracy
    • Operators, shapes, or preprocessing may still differ or be unsupported
  4. 當應用程式可能在不同的 Intel 裝置上運行時,應檢查什麼?
    • Available devices and the target used for compilation
    • Only the source model filename
    • Only the number of classes in the training data
    • The display resolution of the device
  5. 從來源框架轉換後應比較什麼?
    • Only the size of the converted artifact
    • Only the time spent converting
    • Outputs and task metrics on representative inputs
    • Only the number of graph layers
  6. 較低的精度或量化如何影響推理?
    • Lower precision always improves accuracy
    • It may improve efficiency while changing prediction quality
    • It removes the need for preprocessing
    • It guarantees bit-identical outputs
  7. Why benchmark on the actual target device?
    • Architecture alone determines performance on every device
    • Conversion time is the same as inference latency
    • CPU and accelerator execution behave identically
    • Hardware and runtime versions affect supported operations and speed
  8. Which statement about OpenVINO is accurate?
    • OpenVINO can support inference deployment but does not prove model correctness
    • It replaces the training framework for every model workflow
    • It supports every trained model without compatibility checks
    • It automatically certifies production accuracy