Technicals

These are articles that explain internal implementations of NNC. These will focus on ‘why’ and ‘how to implement’ rather than ‘how to use’ questions.

• The NNC Tensor Allocation Algorithm
  • Tensor Representation
  • Loop Representation
  • The Problem Definition
  • The Core Algorithm
  • Basic Structure
  • Candidate Selection
  • Insertion
  • Intuition
  • Loop
  • Multi-view Tensor
  • Loop with Efficient Tensor Allocation
  • Sub-Computation Graph
  • Conclusion
• NNC Static Schedule A Graph
  • Stream
  • Static Schedule
  • while and case..of
• NNC Dynamic Graph Execution
  • Naming The Variable
  • Tracing The Operation
  • Optimizations
  • Interoperability
  • Future Optimizations
  • Some Maybes
• NNC Common Neural Network Primitives
  • Model
  • Model IO
  • Fit, Evaluate, Backward, Apply Gradients
• NNC Dataframe
  • Operations on the Data
  • Iteration
  • Map
  • Reduce
  • Others
  • Use Dataframe with Addons
• NNC Micro Ops
  • Describe Micro Ops
  • Parameters
  • Simplification
  • Loop-Fusion
  • Variable Substitution
  • Automatic Differentiation