Training a 210M text-to-image DiT from scratch on one GPU: what I measured [P]

Training a 210M DiT text-to-image model from scratch on a single GPU.

本摘要由 AI 生成，请从原始来源核实细节。