[Auto] Fix DreamBooth fp16 validation training crash (cluster-1246-28): merged 2 of 5 PRs#2

Open

evalstate wants to merge 4 commits into

mainfrom

merge-cluster-cluster-1246-28-20260427125508

Commits on Mar 16, 2026

fix: wrap validation inference with torch.no_grad() in dreambooth examples
gambletan
committed

Commits on Apr 20, 2026

Fix `train_dreambooth_lora.py` fp16 unscale error after validation
Ricardo-M-L
committed

Commits on Apr 27, 2026