itsmag11/age-inclusive

These are the pretrained weights for <age-inclusive> token from AITTI: Learning Adaptive Inclusive Token for Text-to-Image Generation (IJCV, 2025).

Download the weights and load them following the AITTI inference script like:

...
embed_dim = 1024 if args.sd_model == "stabilityai/stable-diffusion-2-1" else 768
adaptive_mapping = AdaptiveTokenMapping_v1(embed_dim, embed_dim, embed_dim).to(dtype=torch.float16)
adaptive_mapping.load_state_dict(torch.load(os.path.join(path/to/download/dir, 'adaptive_mapping.safetensors'), map_location=torch.device('cpu')), strict=True)
pipe = StableDiffusionAdaptiveTokenPipeline.from_pretrained(args.sd_model, adaptive_mapping=adaptive_mapping, torch_dtype=torch.float16).to(device)
pipe.load_textual_inversion(path/to/download/dir, weight_name='learned_embeds.safetensors', token=args.token_name)
...

If you find our work helpful, please cite:

@inproceedings{hou2025aitti,
  title={AITTI: Learning Adaptive Inclusive Token for Text-to-Image Generation},
  author={Hou, Xinyu and Li, Xiaoming and Loy, Chen Change},
  booktitle={International Journal of Computer Vision (IJCV)},
  year={2025}
}