Alpha-VLLM / Lumina-mGPT

Official Implementation of "Lumina-mGPT: Illuminate Flexible Photorealistic Text-to-Image Generation with Multimodal Generative Pretraining"
https://arxiv.org/abs/2408.02657
493 stars 20 forks source link

cannot generate 256*256 images? #20

Closed zhangzheyu0 closed 2 months ago

zhangzheyu0 commented 2 months ago

Based on line119 of the readme.md, if we want to "generate an image of 256x256 according to the following prompt xxxx.", it doesn't work, why?

Sincerely look forward for your reply.

ChrisLiu6 commented 2 months ago

The model was not trained at 256x256. For more details, please refer to the FP-SFT paragraph in our paper.

1724932766545