Skip to content

Fix transpose in attention scores equation #59

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
shensquared merged 1 commit into from
Jun 5, 2026
Merged

Fix transpose in attention scores equation #59

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion transformers.qmd
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,7 +80,7 @@ $$
The attention mechanism calculates similarity using the dot product, which efficiently measures vector similarity:

$$
a_i = \text{softmax}\left(\frac{[q_i^T k_1, q_i^T k_2, \dots, q_i^T k_n]}{\sqrt{d_k}}\right)^T \in \mathbb{R}^{1\times n}
a_i = \text{softmax}\left(\frac{[q_i^T k_1, q_i^T k_2, \dots, q_i^T k_n]}{\sqrt{d_k}}\right) \in \mathbb{R}^{1\times n}
$$

The vector $a_i$ (softmax'd attention scores) quantifies how much attention token $q_i$ should pay to each token in the sequence, normalized so that elements sum to 1. Normalizing by $\sqrt{d_k}$ prevents large dot-product magnitudes, stabilizing training.
Expand Down