oa::FnMatrix::rnnCellLinear

rnnCellLinear: fused vanilla-RNN recurrent step — Linear(h, W_hh) + rnnCellPointwise. Replaces the per-timestep pair of dispatches with one kernel. inGatesI is the whole input projection [B*T, H]; inTimeOffset/inBatchStride index the current timestep's row directly, so oa::Rnn needs no per-step Slice. The hidden projection required by reverse mode is retained internally rather than exposed as an output parameter.

FnC++ public headers

Function Documentation

Matrix oa::FnMatrix::rnnCellLinear(
	const Matrix & inGatesI,
	const Matrix & inHidden,
	const Matrix & inWeightHh,
	oa::U32 inTimeOffset = 0,
	oa::U32 inBatchStride = 1,
	const Matrix & inBiasHh = Matrix{}
)

Parameters

inGatesI

const Matrix &

—

inHidden

const Matrix &

—

inWeightHh

const Matrix &

—

inTimeOffset

oa::U32

Default: 0

inBatchStride

oa::U32

Default: 1

inBiasHh

const Matrix &

Default: Matrix{}

Returns

Matrix

The declared return value.