NAMI: Efficient Image Generation via Bridged Progressive Rectified Flow Transformers
Paper • 2503.09242 • Published
import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("qihoo360/NAMI-T2I", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]NAMI: Efficient Image Generation via Bridged Progressive Rectified Flow Transformers
💡 Github
💡 Pretrained Models
We provide model weights for testing and deployment. Please download the files and place them in the weights directory.
For mclip tokenizer and text encoder weights can be found in https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/tree/main/t2i/clip_text_encoder.
For mt5 tokenizer and text encoder weights can be found in https://huggingface.co/google/mt5-xxl/tree/main.
💡 Update
[2026.03.06] Release the NAMI-2B inference code and weights.
[2026.03.01] Paper was accepted by CVPR2026.
💖 Citation
If you find this repository useful, please consider citing our paper:
@article{ma2025nami,
title={NAMI: Efficient Image Generation via Bridged Progressive Rectified Flow Transformers},
author={Ma, Yuhang and Cheng, Bo and Liu, Shanyuan and Zhou, Hongyi and Wu, Liebucha and Leng, Dawei and Yin, Yuhui},
journal={arXiv preprint arXiv:2503.09242},
year={2025}
}
- Downloads last month
- 9