Silent NaN: the check that belongs in every diffusion pipeline

Automated excerpt

About 150 seconds per image in fp32 with model offload. The only change was the precision flag: Same seed, same prompt, same scheduler. fp32 gave me an image every time. fp16 gave me NaN tensors that survived all the way to image.save(). Two lines, run every image, and you never debug a silent NaN through a pinned file.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.