Storia: How to Speed Up Transformer Training Using NVIDIA Apex (FusedAdam, FusedLayerNorm) and Native torch.amp — Warptech Lab News