oa::FnMatrix::softmaxScaledMasked

softmaxScaledMasked: fused transformer attention score normalisation. Computes: out = softmax(scores * scale + mask) over the last dimension.

Function Documentation

Matrix oa::FnMatrix::softmaxScaledMasked( const Matrix & inScores, const Matrix & inMask, oa::F32 inScale )
softmaxScaledMasked: fused transformer attention score normalisation. Computes: out = softmax(scores * scale + mask) over the last dimension.

Parameters

inScores
const Matrix &

inMask
const Matrix &

inScale
oa::F32

Returns

Matrix

The declared return value.