Skip to content

❌ This version of handling ViT activation (through trucation) is problematic! The latest version will use the pytorch built-in method for saving and tracking activation and gradient.#2

Open
TyBruceChen wants to merge 10 commits intoold-versionfrom
main
Open

❌ This version of handling ViT activation (through trucation) is problematic! The latest version will use the pytorch built-in method for saving and tracking activation and gradient.#2
TyBruceChen wants to merge 10 commits intoold-versionfrom
main

Commits

Commits on Mar 22, 2026