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.
|
Class that keeps yields only the cost model evaluation that has minimal latency of all cost model evaluations generated by it's substages created by list_of_callables. More...
Public Member Functions | |
def | __init__ (self, list[StageCallable] list_of_callables, *bool reduce_minimal_keep_others=False, **Any kwargs) |
Initialize the compare stage. More... | |
def | run (self) |
Run the compare stage by comparing a new cost model output with the current best found result. More... | |
![]() | |
def | __init__ (self, list["StageCallable"] list_of_callables, **Any kwargs) |
def | __iter__ (self) |
bool | is_leaf (self) |
Public Attributes | |
keep_others | |
![]() | |
kwargs | |
list_of_callables | |
Class that keeps yields only the cost model evaluation that has minimal latency of all cost model evaluations generated by it's substages created by list_of_callables.
def __init__ | ( | self, | |
list[StageCallable] | list_of_callables, | ||
*bool | reduce_minimal_keep_others = False , |
||
**Any | kwargs | ||
) |
Initialize the compare stage.
def run | ( | self | ) |
Run the compare stage by comparing a new cost model output with the current best found result.
Reimplemented from Stage.
keep_others |