Skip to content
GitLab
Explore
Sign in
This is an archived project. Repository and other project resources are read-only.
deep-learning-nlp
token-tricksters
Merge requests
!4
Feat/sophia
Code
Review changes
Check out branch
Download
Patches
Plain diff
Lars Benedikt Kaesberg
requested to merge
feat/sophia
into
main
Jun 17, 2023
Overview
10
Commits
25
Changes
7
Expand
Implemented the Sophia optimizer.
Added SophiaH (with the Hutchinson Estimator) as an option to the train loop.
Added gradient clipping to the train loop.
Added AMP Autocast to
bfloat16
.
Edited
Aug 21, 2023
by
Niklas Bauer
Merge request reports