GClasses
|
The base class for the buffers that a thread needs to use (train or predict with) a neural network component.
#include <GBlock.h>
Public Member Functions | |
GContext (GRand &rand) | |
virtual | ~GContext () |
virtual void | resetState ()=0 |
Resets the state of all recurrent blocks. (This is called whenever a recurrent neural network begins with a new sequence, either for training or testing.) More... | |
Public Attributes | |
GRand & | m_rand |
|
inline |
|
inlinevirtual |
|
pure virtual |
Resets the state of all recurrent blocks. (This is called whenever a recurrent neural network begins with a new sequence, either for training or testing.)
Implemented in GClasses::GContextGRU, GClasses::GContextLSTM, GClasses::GContextRecurrent, GClasses::GContextNeuralNet, and GClasses::GContextLayer.
GRand& GClasses::GContext::m_rand |