Relax

Relax is a high-level abstraction for graph optimization and transformation in Apache TVM stack. Additionally, Apache TVM combines Relax and TensorIR together for cross-level optimization. Hence, Relax is usually working closely with TensorIR for representing and optimizing the whole IRModule

• Graph Abstraction for ML Models
  • What is Graph Abstraction?
  • Key Features of Relax
• Understand Relax Abstraction
  • End to End Model Execution
  • Key Elements of Relax
• Dataflow Pattern Language (DPL)
  • Building Patterns
  • Matching Without Rewriting
  • Rewriting Matched Patterns
  • Using DPL in Compiler Passes
  • Quick Reference
• Relax Creation
  • Create Relax programs using TVMScript
  • Create Relax programs using NNModule API
  • Create Relax programs using Block Builder API
  • Summary
• Transformation
  • Apply transformations
  • Custom Passes
  • Summary