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.
|
This class captures multi-dimensional operational array size. More...
Public Member Functions | |
def | __init__ (self, OperationalUnit operational_unit, dict[OADimension, int] dimension_sizes) |
def | __jsonrepr__ (self) |
bool | __eq__ (self, Any other) |
![]() | |
def | __init__ (self, dict[OADimension, int] dimension_sizes) |
Public Attributes | |
total_unit_count | |
total_area | |
unit | |
dimension_sizes | |
![]() | |
dimension_sizes | |
This class captures multi-dimensional operational array size.
def __init__ | ( | self, | |
OperationalUnit | operational_unit, | ||
dict[OADimension, int] | dimension_sizes | ||
) |
operational_unit | an OperationalUnit object including precision and single operation energy, later we can add idle energy also (e.g. for situations that one or two of the input operands is zero). |
dimensions | define the name and size of each multiplier array dimensions, e.g. {'D1': 3, 'D2': 5}. |
bool __eq__ | ( | self, | |
Any | other | ||
) |
def __jsonrepr__ | ( | self | ) |
dimension_sizes |
total_area |
total_unit_count |
unit |