llnl / llnl/GPLaSDI

Removal of multihead attention from activation

Open
#13 0 comments 0 reactions 1 assignee Claimed by @punkduckable View on GitHub
enhancement
Dominant language
Python
Stars
38
Forks
7
PR merge metrics
No merged PRs in 30d

Description

Per @punkduckable , multihead attention should be implemented as a layer, not an activation function. However, the current implementation simply uses multihead attention as an activation function, which also disrupts overall structure within `MultiLayerPerceptron`.

multihead attention should be removed from activation, and probably implemented as a derived class of latent space.

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.