pulsatrix::RlRowArgs Struct Reference

Operand pointers and dims for DeviceBackend::rl_rows (passed to kernels by value). More...

#include <device_backend.hpp>

Public Attributes
const float *	in [5] = {}
float *	out [4] = {}
int64_t	rows = 0
int64_t	cols = 0
float	scale = 0.0f
	gradient batch-mean scale (1/N or 2/N)
float	lower = 0.0f
	PPO 1 - clip_epsilon.
float	upper = 0.0f
	PPO 1 + clip_epsilon.
float	gamma = 0.0f
	discount factor
float	tau = 0.0f
	Polyak blend factor.

Detailed Description

Operand pointers and dims for DeviceBackend::rl_rows (passed to kernels by value).

Member Data Documentation

cols

int64_t pulsatrix::RlRowArgs::cols = 0

gamma

float pulsatrix::RlRowArgs::gamma = 0.0f

discount factor

in

const float* pulsatrix::RlRowArgs::in[5] = {}

lower

float pulsatrix::RlRowArgs::lower = 0.0f

PPO 1 - clip_epsilon.

out

float* pulsatrix::RlRowArgs::out[4] = {}

rows

int64_t pulsatrix::RlRowArgs::rows = 0

scale

float pulsatrix::RlRowArgs::scale = 0.0f

gradient batch-mean scale (1/N or 2/N)

tau

float pulsatrix::RlRowArgs::tau = 0.0f

Polyak blend factor.

upper

float pulsatrix::RlRowArgs::upper = 0.0f

PPO 1 + clip_epsilon.

---

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

• include/pulsatrix/device_backend.hpp