Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.
Remember the eval mode of submodules when switching trainer stages (#18951)
Co-authored-by: Carlos Mocholí <carlossmocholi@gmail.com>
A
Adrian Wälchli committed
3d448ac48da257cab6bf74f258dd305ec868c107
Parent: 792cb73
Committed by GitHub <noreply@github.com>
on 11/16/2023, 9:32:27 PM