#include <Synchronizer.h>
Public Member Functions | |
Synchronizer (Synchronizer_Helper &) | |
virtual | ~Synchronizer () |
void | synchronize () |
The synchronization strategy. | |
void | set_all_iters_done () |
bool | is_all_iters_done () |
Check if the Execution_Strategy has finished. |
The synchronization strategy that uses the Synchronizer_Helper to perform the concrete synchronization steps. It provides slots, so to say, for synchronization and calls the Synchronizer_Helper to perform its duties for that slot.
This greatly reduces the burden on the model writer
So whenever a model needs to be scaled by providing a distributed inferencing solution, all the model writer has to do is write a Synchronizer_Helper.
This is a default synchronization strategy provided and users can implement their own strategy by extending this
The main aim is to run the inferencing locally while keeping the model in sync globally through the use of this Synchronizer and a Distributed Map
Synchronizer::Synchronizer | ( | Synchronizer_Helper & | sync_helper | ) |
Synchronizer::~Synchronizer | ( | ) | [virtual] |
bool Synchronizer::is_all_iters_done | ( | ) |
Check if the Execution_Strategy has finished.
void Synchronizer::set_all_iters_done | ( | ) |
Set that all iterations of the Execution_Strategy have been done
void Synchronizer::synchronize | ( | ) |
The synchronization strategy.