Lightning-AI / pytorch-lightning UNCLAIMED

Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.

0 0 104 Python

fix: restore cached AMP step context after no_grad workaround (#21616)

* fix: restore cached AMP step context after no_grad workaround

* chore: trigger ci

* chore: trigger ci

* test: add CUDA coverage for AMP no_grad cache handling

littlebullGit committed 5mo ago

4a548c96c3578c497a316f9451df2d0b8535164d

Parent: d3b25f8

Committed by GitHub <noreply@github.com> on 4/1/2026, 10:30:48 AM