Scaled dot-product attention

gh-eq-5fbfc39da368b3afsource-comment-equation

Attention(Q,K,V) = softmax(QK^T/√d_k)V

Read the equation

Query-key dot products are divided by √dₖ, row-wise softmax converts them into weights that sum to one, and those weights average the value vectors. The code implements these stages with maximum subtraction for numerical stability. Compatible nonempty query/key widths and matching key/value counts are needed for the standard interpretation.

This exact expression also occurs in another recovered revision. Its mathematical annotation is shared; the source references below preserve this occurrence.

Terms & definitions

Term	Meaning in this source
Q	Query matrix
K	Key matrix
V	Value matrix
d_k	Query/key feature width

How it works

Read the definition from its right-hand expression to the quantity on the left. Its surrounding source records the intended input, state, or parameter context. Where the source leaves an operation unspecified, this page preserves that gap.

Assumptions & limits

Source recovery establishes that the expression appears in this record. Its mathematical validity, implementation behavior, and interpretation must be assessed separately. See the linked source and research notes for the checks available for this version.

Original context

Read the surrounding source passage

// Package transformer implements transformer-based generators (76-85)
// Attention(Q,K,V) = softmax(QK^T/√d_k)V
package transformer

import (
	"math"
	"math/rand"

	"generator_core/pkg/consciousness"
	"generator_core/pkg/generators"
)

Exact original notation

Attention(Q,K,V) = softmax(QK^T/√d_k)V

Related builds

Related mathematical records

Query-key dot products are divided by √dₖ, row-wise softmax converts them into weights that sum to one, and those weights average the value vectors. The code implements these…