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.
|
For the operand dimension that is not directly a loop dimension, a relation equations between them(operand dimension) and the loop dimension is required. More...
Public Member Functions | |
def | __init__ (self, LayerDim dim_1, LayerDim dim_2, LayerDim dim_3, int coef_2, int coef_3) |
![]() | |
def | __init__ (self, Any data) |
int | __len__ (self) |
Iterator[Any] | __iter__ (self) |
def | __getitem__ (self, Any key) |
bool | __contains__ (self, Any key) |
def | __str__ (self) |
def | __repr__ (self) |
Any | __jsonrepr__ (self) |
def | __eq__ (self, object other) |
def | __hash__ (self) |
Static Public Member Functions | |
tuple[PrLoop, LoopList, PrScalingFactors] | extract_pr_loop_info (list["LayerDimRelation"] relations) |
Public Attributes | |
dim_1 | |
dim_2 | |
dim_3 | |
coef_2 | |
coef_3 | |
data | |
![]() | |
data | |
For the operand dimension that is not directly a loop dimension, a relation equations between them(operand dimension) and the loop dimension is required.
e.g. dim1 = coef2 * dim2 + coef3 * dim3
|
static |
coef_2 |
coef_3 |
data |
dim_1 |
dim_2 |
dim_3 |