3 ms·Pixart-α: Fast Training of Diffusion Transformer for Text-to-Image Synthesis3 points by webmaven 3y ago