NieShenRuc

+1

+1

Thanks for your reply！

Due to the lack of Chinese in the training data, although unidiffuser can generate images based on Chinese text, using English will get higher image quality

We will release the code of Gibbs sampling and image interpolation soon. Please stay tuned :)

Thank you for your interest. Before training UViT, we first need to fine-tune GPT-2 for text reconstruction. For specific fine-tuning methods, please refer to Appendix E. During the UViT training...

Thank you for your interest. Our model was trained on large-scale image-text data pairs, but all these texts are in English. Based on your description, you have new data, and...