Learning Roadmap

Basic

• InferenceSession/ORTModule
• Graph/Node/NodeArg
• Onnx/Op/Schema/Kernel
• ORTValue/Tensor
• GraphTransformer
• Per-op Gradient Building
• Performance Investigation

Intermediate

• ExecutionProvider
• IOBinding/dlpack
• PyBind
• Gradient Graph Building
• CUDA Programming

Advanced

• Execution Engine

  • SessionState
  • ExecutionFrame

• Memory

• Distributed Training

Domain Experts

• Performance optimization for CUDA kernels
• Hyper-parameter tuning