Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding

arXiv:2405.08748v1 Announce Type: new
Abstract: We present Hunyuan-DiT, a text-to-image diffusion transformer with fine-grained understanding of both English and Chinese. To construct Hunyuan-DiT, we carefully design the transformer structure, text encoder, and positional encoding. We also build from scratch a whole data pipeline to update and evaluate data for iterative model optimization. For fine-grained language understanding, we train a Multimodal Large Language Model to refine the captions of the images. Finally, Hunyuan-DiT can perform multi-turn multimodal dialogue with users, generating …

abstract arxiv build chinese construct cs.cv data data pipeline design diffusion diffusion transformer encoder encoding english fine-grained image image diffusion iterative pipeline positional encoding resolution scratch text text-to-image transformer type understanding update