dlprim::solvers::Adam Class Reference

Inheritance diagram for dlprim::solvers::Adam:

Public Member Functions
	Adam (Context &ctx)
void	init (Net &n, ExecutionContext const &q)
	Prepare solver - takes all parameters that need to be trained and prepares buffers.
void	zero_grad (Net &n, ExecutionContext const &e)
	zero all gradients before accumulating them for next batch
void	apply (Net &n, ExecutionContext const &e)
	apply solver updates
Public Member Functions inherited from dlprim::solvers::SolverBase
void	step (Net &n, ExecutionContext const &e)
	shortcut for single training step zero_grad, forward, backward, apply

Public Attributes
float	lr = 0.001
float	beta1 = 0.9
float	beta2 = 0.999
float	eps = 1e-8
float	weight_decay = 0.0005

---

The documentation for this class was generated from the following file:

• include/dlprim/solvers/adam.hpp