HiT: Hierarchical Diffusion Transformer for End-to-End Image Generation
This repository hosts the official PyTorch checkpoints of HiT. HiT is a pixel-space diffusion Transformer that learns to group related image patches into a visual hierarchy and runs most of its blocks on the compact groups.
- Paper: https://arxiv.org/abs/2611.22962
- Website: https://nicholas0228.github.io/blog/hit/
- Code: https://github.com/Nicholas0228/HiT-Released
| Checkpoint | Model | Should reproduce |
|---|---|---|
ImageNet256/HiT-XL16_ep600/model.safetensors |
HiT-XL/16, ImageNet 256×256, 600 epochs | FID-50K 1.48 |
T2I512/HiT-B16_sft40k/model.safetensors |
HiT-B/16, text-to-image 512×512, 40k fine-tuning updates | GenEval 87.78, DPG 85.33, PRISM 58.62, CoRe 45.22, Spatial 52.68 |
Text-to-image training inputs: Nicholas0228/minit2i-frozen-inputs-v1.