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

Training large DiT model on single GPU shows efficiency gains.

AI生成の要約です。詳細は元ソースで確認してください。