ZigZag - Deep Learning Hardware Design Space Exploration
This repository presents the novel version of our tried-and-tested hardware Architecture-Mapping Design Space Exploration (DSE) Framework for Deep Learning (DL) accelerators. ZigZag bridges the gap between algorithmic DL decisions and their acceleration cost on specialized accelerators through a fast and accurate hardware cost estimation.
Classes | Variables
zigzag.datatypes Namespace Reference

Classes

class  OperandABC
 Abstract Base Class for all dimension- and operand-like classes. More...
 
class  LayerOperand
 Operand from the layer definition, e.g. More...
 
class  MemoryOperand
 Operand from the memory definition, e.g. More...
 
class  LayerDim
 (for-loop) dimension of a workload layer (e.g. More...
 
class  OADimension
 Operational Array Dimension. More...
 
class  Constants
 Store constant objects used throughout ZigZag (instead of hardcoding them) More...
 

Variables

 ArrayType = np.ndarray[Any, Any]
 

Variable Documentation

◆ ArrayType

ArrayType = np.ndarray[Any, Any]