Kenning

• Introduction
• Kenning
  • Kenning installation
  • Kenning structure
  • Kenning usage
  • Example use case of Kenning - optimizing a classifier
  • Using Kenning as a library in Python scripts
  • Adding new implementations
• Deep Learning deployment stack
  • From training to deployment
  • Dataset preparation
  • Model preparation and training
  • Model optimization
  • Model compilation and deployment
• Defining optimization pipelines in Kenning
  • JSON specification
  • Model evaluation using its native framework
  • Optimizing and running a model on a single device
  • Compiling a model and running it remotely
• Using Kenning via command line arguments
  • Command-line arguments for classes
  • Model training
  • In-framework inference performance measurements
  • ONNX conversion
  • Testing inference on target hardware
  • Running inference
  • Generating performance reports
• Kenning measurements
  • Performance metrics
• ONNX support in deep learning frameworks
  • ONNX support grid in deep learning frameworks
  • ONNX conversion support grid
• Sample autogenerated report
  • Pet Dataset classification using TVM-compiled TensorFlow model
  • Inference performance metrics for build.local-cpu-tvm-tensorflow-classification.json
  • Inference quality metrics for build.local-cpu-tvm-tensorflow-classification.json
• Developing Kenning blocks
  • Model metadata
  • Implementing a new Kenning component
• Kenning API
  • Deployment API overview
  • Dataset
  • ModelWrapper
  • Optimizer
  • Runtime
  • RuntimeProtocol
  • Measurements
  • ONNXConversion
  • DataProvider
  • OutputCollector