PixArt-α

An efficient text-to-image generation model that significantly reduces training costs through an optimized training pipeline, maintaining generation quality while substantially decreasing computational resource consumption.